Rosa Aceves

The obsession with 2-year / 10-year inversion as a recession predictor has trained the market to read the curve for timing (when will the recession start?). That is not what the curve is good at. The curve is good at telling you what kind of cycle comes next: steep recovery, shallow recovery, or no recovery.

The 30-year / 2-year spread measures the market's expectation of the long-term growth path after the cycle turns. When 30yr / 2yr is steep (positive and widening), the market is pricing a full recovery: growth resumes, term premium rebuilds, long-duration assets reprice upward. When 30yr / 2yr is flat or inverted, the market is pricing a scarring event: growth does not resume at previous trend, term premium stays compressed, long-duration assets stay cheap.

The 2yr / 10yr inversion tells you the market expects the Fed to cut. The 30yr / 2yr slope tells you whether the market believes the cuts will work.

Right now (as of the last full read), the 2yr / 10yr has un-inverted but the 30yr / 2yr remains flatter than historical norms for this point in the hiking cycle. The read: the market expects cuts, but does not expect them to restore trend growth. This is the pricing of a shallow recovery or a growth regime shift.

The load-bearing question is whether this is a positioning effect (bond market participants are structurally long and cannot express optimism by selling duration) or a structural forecast (the market genuinely believes long-term growth has slowed). Palanor's read: it is structural. The 30yr yield has not responded to equity market strength the way it did in previous cycles, which means fixed-income investors are not pricing AI productivity gains or fiscal stimulus as durable growth drivers.

We cite the 30yr / 2yr spread as the regime slope indicator. When it steepens beyond 100 bps, the market has shifted to pricing durable expansion. When it stays below 75 bps, the market is pricing managed stagnation. The 2yr / 10yr tells you the Fed's next move. The 30yr / 2yr tells you whether that move matters.

The practitioner decision tree for which spread to cite is simpler than the literature implies:

• If the bond has no embedded optionality (straight bond, no call, no prepayment risk), cite the Z-spread. It is the constant spread over the Treasury spot curve that equates PV to market price. It is the purest measure of credit risk + liquidity premium.

• If the bond has optionality (callable, MBS, ABS with prepayment), cite the OAS. It is the spread after removing the option cost. The difference between Z-spread and OAS is the dollar value of the option you sold to the issuer.

The option cost is not a separate line item; it is embedded in the price. When you buy a callable bond, you are simultaneously buying the credit and selling a call option to the issuer. The Z-spread includes both. The OAS removes the call value and isolates the credit compensation.

This matters because the option cost is rate-path-dependent. In a falling-rate environment, the call becomes more valuable (the issuer is more likely to refi), so Z-spread > OAS by a wider margin. In a rising-rate environment, the call becomes worthless, so Z-spread ≈ OAS.

When Palanor cites spreads, we default to OAS for any security with optionality, because we want to isolate the credit signal from the rate-path bet. For straight corporates, we cite Z-spread. For MBS and callable munis, we cite OAS and note the option cost separately when it moves.

The confusion in the market comes from the fact that most corporate bonds are now callable, but many participants still cite nominal spread or Z-spread out of habit. This conflates two risks: the credit risk and the reinvestment risk of getting called. We do not conflate them.

The foundational mispricing puzzle in fixed income is not that spreads exist, but that what they compensate for differs by rating tier in ways structural models cannot reconcile.

Huang and Huang (2012) calibrated the full class of Merton-descendant structural models to historical default loss and found that for investment-grade bonds, credit risk accounts for only a small fraction of observed yield spreads. The residual—what you earn after subtracting actuarial default cost—is enormous and persistent. For speculative-grade bonds, the opposite holds: structural models can explain most of the spread through default risk alone.

This is the rating-dependent asymmetry that governs how we read corporate credit. In IG, the spread is a liquidity premium, a complexity premium, a balance-sheet-cost premium—not primarily compensation for loss-given-default. In HY, the spread converges toward actuarial truth: you are being paid to hold default risk, and the models work.

The implication for market reading: when IG spreads tighten or widen, you are observing a risk appetite shift or a balance-sheet capacity shift, not a re-rating of default probability. When HY spreads move, you are watching the market reprice survival.

DTS (Duration Times Spread) measures this correctly: it predicts credit volatility—how much the spread itself will move—not default risk. The spread is the instrument; default is one component of what the spread prices. In IG, it is the minor component. In HY, it is the major one.

This is why the IG market can appear calm (tight spreads, low DTS volatility) even when macro risk is rising, and why HY can price recession six months before IG acknowledges the turn. They are pricing different things.

Palanor's read: we cite the IG-HY dispersion as a regime indicator. When IG spreads ignore what HY is pricing, the market is segmented and one of them is wrong. When they move together, the regime is unified and the turn is confirmed.

The conventional read treats "tightening credit conditions" as a single variable. The reality arriving in 2025–2027 is bifurcated: credit is available, but the plumbing to distribute it is constrained.

Readings [13–16] establish that Treasury market intermediation is already under stress — repo rates persistently above IORB, dealer bid-ask spreads widening, auction bid-to-cover ratios weakening. The SLR binds where risk-weighted capital does not, meaning the constraint is balance-sheet capacity itself, not credit risk appetite. This matters because it means even high-quality collateral faces friction in the distribution chain.

Meanwhile, readings [9–12] show private credit markets operating in a covenant-void environment with structural yield concealment (PIK interest) and recovery-rate deterioration. BDCs are deploying capital up the risk curve into BSL exposure not because underwriting standards improved, but because deployment pressure from scale forces them there.

The operating stance this demands: watch the plumbing, not just the spreads. When CRE maturities peak in 2027 [1], the question will not be "is there capital for refinancing?" but "can that capital clear through constrained intermediaries at the debt yields borrowers can bear?" [3]. The same SLR constraint that weakens Treasury auctions will compress the bid side of distressed CRE sales when those assets need to move.

Dealer capacity is the first constraint to bind in a cycle turn. Credit spreads widen later, after the distribution problem has already raised the cost of capital for everyone downstream. Regional banks [6,8] face this from both sides: deposit betas that compress slowly on the way down, and loan portfolios that need to refinance through markets where the intermediation spread has already widened.

The 2027 maturity wall is not a credit event waiting to happen. It is a distribution event waiting to stress-test whether post-crisis capital regulation left enough capacity in the system to handle non-linear refinancing demand.

[21, 22, 23, 24] settle the question that dominated pre-2000 asset pricing: price variation is driven by time-varying discount rates, not by changing expectations of future cash flows. The Campbell-Shiller decomposition, implemented via VAR, shows that dividend yields forecast returns, not dividend growth. The cross-section of returns exploded from CAPM into a factor zoo, each factor representing a dimension of time-varying risk premia. [24] notes the tension with survey-based subjective expectations, which show cash-flow expectations mattering more — but subjective expectations are backward-looking and smooth through exactly the moments when discount rates move most.

This is the foundation under everything else in this tier. When [17, 18, 19] describe Treasury term premia moving with clientele demand shocks, they are describing discount-rate variation. When [25, 26, 27, 28] describe the monetary transmission channel working through credit spreads and balance-sheet constraints, they are describing discount-rate variation. When [5, 6, 7, 8] document that IG credit spreads are wider than default risk justifies, they are documenting discount-rate variation that structural models miss because those models solve for expected cash flows under a fixed risk-neutral measure.

The insight for capital-markets commentary: the correct frame is not what will earnings do but what discount rate will the market apply to those earnings. A firm with stable EBITDA can see its equity cut in half not because the cash flows deteriorated but because the market's required return rose 300 basis points. A Treasury with no credit risk can see its yield spike 100 basis points not because inflation expectations moved but because the preferred-habitat clientele shifted and arbitrageurs are capital-constrained.

The error mode is assigning all price moves to 'the story' (cash-flow news) when most large moves are 'the sentiment' (discount-rate news). Correcting this error requires instruments that separate the two — which is what [21, 22] built for equities and what the credit literature has not yet built for corporate bonds.

The cost of capital for a firm is not a single rate — it is a spread structure layered on top of a risk-free term structure, where both layers move independently and the interaction determines what can be financed.

The bottom layer is Treasury rates, which [1, 2, 3] model as mean-reverting processes with square-root diffusion keeping them non-negative. But the level and shape of the curve reflect [17, 18, 19] clientele demand shocks transmitted through risk-averse arbitrageurs. A flight-to-quality episode compresses the term premium; quantitative tightening stretches it. Neither is a 'fundamental' shift in the short rate — both are shifts in the price of duration risk itself.

The top layer is credit spreads. Here [5, 6, 7, 8] establish that structural models radically underpredict IG spreads while fitting HY reasonably well. The gap is not credit risk; it is liquidity premium, rollover risk, and covenant tightness — all of which vary with the [25, 26, 27, 28] transmission strength of monetary policy through bank balance sheets. When the external finance premium rises, IG spreads widen even when default probabilities have not moved.

The capital window opens and closes in the interaction. [29, 30] show that Treasury supply itself affects corporate spreads through the convenience yield. When the Treasury curve steepens and credit spreads tighten simultaneously — as in mid-2023 — the 10-year cost of capital can fall even as the Fed holds rates high. When both layers widen — Q4 2022 — the window slams shut for anything outside the megacap complex.

This is the operating stance: every financing event is priced on two independent curves, and the commentary that ignores one or conflates them is financing fanfiction.

Markets coverage at Palanor is a calibration discipline, not a forecasting exercise. The job is to give the steward a continuously updated read on the capital window — when it's open, when it's closing, when it's already shut.

Three commitments hold this work:

1. Range, not point. Every read carries a probability band. A single number pretending to be a forecast is a betrayal of the discipline.
2. Pricing, not narrative. I trust prices that have been argued for by people with skin in the game more than I trust narratives that haven't been priced anywhere.
3. The window, not the call. The macro call is rarely the action; the window for the action is. When the window closes, the action gets harder regardless of whether the call was right.

The Palanor Index is the canonical instrument. Custom indices and Currents extend it; nothing replaces it.

Three layers, every read:

Layer 1 — Term structure + volatility regime. The 2yr/30yr term structure tells me the shape of the cycle. Treasury vol tells me whether positioning is comfortable with that shape or fighting it.

Layer 2 — Spreads, with dispersion. IG aggregate spreads are the headline number. The dispersion within IG is the load-bearing read — when the BBB cohort decouples from the AA cohort, the window is closing on a name basis even when the aggregate looks fine.

Layer 3 — The refi wall. 2027–2029 maturities by sector, by vintage, by sponsor. The wall is the calendar. The window has to be open when the wall arrives, or the wall reshapes the cycle.

I cross-reference all three against prediction-market depth on Fed paths + recession probability. When the three macro layers and the prediction-market read disagree, the disagreement IS the read.

• The 2027 refinancing wall — sponsor finance + IG-sub + commercial real estate. The aggregate number is large; the names where it concentrates are the read.
• The Palanor Index print weekly — belief vs. uncertainty in the macro layer. The composite has been pricing in calmer water than the dispersion underneath supports.
• Prediction-market depth on Fed paths — Kalshi + Polymarket. The volume reads are more informative than the levels right now.
• The Capital Tightness Index — credit conditions composite. Watching for divergence against the term-structure read.

The Cox-Ingersoll-Ross model introduced square-root diffusion in 1985 specifically to ensure that interest rates remain positive in simulation. The stochastic differential equation governs the short rate r with volatility proportional to √r, which means as r approaches zero, volatility approaches zero, making further declines infinitesimally unlikely.

This was the solution to the Vasicek model's flaw: Vasicek allowed negative rates in theory because its diffusion term was constant (not state-dependent). CIR fixed this by making volatility shrink as rates fall, creating a boundary at zero that the process asymptotically approaches but never crosses.

Then the Bank of Japan, the ECB, and the Swiss National Bank took policy rates negative.

The machinery of CIR is now historically falsified in the domain it was designed to govern. Rates can and do go negative, not as a simulation artifact but as observed policy. The square-root boundary is not a mathematical law; it is an assumption about central bank behavior that held for 30 years and then broke.

The implication is not that CIR is useless—it still prices the positive-rate regime better than alternatives—but that it cannot be used to price the zero-bound and negative-rate regime without modification. The extensions that permit negative rates (shifted CIR, shadow rate models) are add-ons, not native to the affine framework.

Palanor's use of CIR: we cite it as the baseline diffusion process for positive-rate environments, knowing that if the Fed returns to zero (which is not our base case but is a tail scenario), the model will underestimate volatility near the bound. The square-root term is the feature that makes CIR tractable; it is also the feature that makes it regime-dependent.

The deeper lesson: models that assume structural boundaries (rates cannot go negative, spreads cannot go below zero, volatility cannot spike above X) are useful until the boundary breaks. The boundary is not in the mathematics; it is in the institutional behavior the mathematics encodes. When institutions change—when central banks cross zero, when credit markets stop differentiating, when volatility regimes shift—the models do not break, but their domain of validity shrinks.

We do not use CIR to forecast. We use it to describe the diffusion process in the regime we are in, and we flag when the regime assumptions (positive rates, mean reversion, affine structure) no longer hold.

The Vasicek and CIR models are affine term structure models: the bond price is an exponential-affine function of the state variable (the short rate). This property is what permits closed-form bond pricing without simulation. The yield curve is a deterministic transformation of the short rate process.

This holds for Treasury curves. It does not hold for corporate credit curves.

The reason is that corporate bond prices depend on at least two state variables: the risk-free rate (which follows an affine process) and the credit spread (which does not). The credit spread is not mean-reverting to a constant; it is regime-dependent, driven by:

• default probability (which is asset-value-dependent in structural models)
• recovery rate (which is collateral- and seniority-dependent)
• liquidity premium (which is balance-sheet- and funding-dependent)
• risk appetite (which is volatility- and correlation-dependent)

None of these are affine in the short rate. Therefore, the corporate credit curve is not an affine transformation of anything, and there is no closed-form solution for corporate bond prices except in the unrealistic case where spreads are constant.

The multifactor extensions (two-factor CIR, three-factor Vasicek) attempted to preserve affine structure by adding state variables, but they still assumed that all state variables follow affine processes. This works for term structure decomposition (level, slope, curvature) but not for credit pricing, because credit risk is not a diffusion process—it has jumps, contagion, and regime breaks.

The implication for practitioners: you cannot build a corporate bond pricing model by adding a spread to a risk-free curve and assuming the machinery still works. The spread itself is the unmodeled variable. The best you can do is:

• Use structural models (Merton, CreditGrades) to price the default component, knowing they will underestimate IG spreads and approximate HY spreads.
• Use reduced-form models (Jarrow-Turnbull, Duffie-Singleton) to fit the observed spread curve without claiming to explain it.
• Use empirical models (DTS, spread volatility regressions) to predict spread changes without claiming to price the level.

Palanor's approach: we do not attempt to price corporate bonds from first principles. We observe the spread, decompose it into components (credit, liquidity, convexity), and track which components are moving. When the credit component moves, we cite default probability or rating migration. When the liquidity component moves, we cite balance-sheet capacity or dealer inventory. When the convexity component moves, we cite rate volatility or duration exposure.

The affine models are foundational because they define what does and does not carry over from risk-free to risky pricing. The answer is: the math does not carry over. The intuition sometimes does.

The Fed hiked 525 bps between March 2022 and July 2023. Cumulative deposit beta—the percentage of the Fed hike that banks passed through to depositors—continued rising even after the final hike. This is the liability-side lag that defines the next phase of bank NIM compression.

Deposit beta measures how much of a rate move gets passed to depositors. In the hiking cycle, beta rises slowly at first (depositors are sticky), then accelerates as competitive pressure builds. But the acceleration does not stop when the Fed stops hiking. It continues for 6–12 months because:

1. Cumulative beta is a lagging indicator: depositors reprice their behavior slowly, shopping for yield after observing sustained higher rates elsewhere.
2. Competitive pressure persists: banks that held beta low early in the cycle face deposit outflows late in the cycle and must raise rates to retain funding.
3. The liability side reprices in waves, not in lockstep with Fed moves. Each wave is triggered by visible alternatives: money market funds, Treasuries, brokered CDs.

The implication for NIM: when the Fed eventually cuts, deposit betas will compress slowly. Banks will not immediately lower deposit rates in line with Fed cuts, because depositors will resist and switch to higher-yielding alternatives. The cumulative beta that rose to ~50% during the hiking cycle will fall to ~20–30% in the first year of cuts.

This is the asymmetry that favors depositors and compresses bank margins on both sides of the cycle:

• On the way up: asset yields (loans, securities) reprice faster than deposit costs, widening NIM early but compressing it late as cumulative beta catches up.
• On the way down: deposit costs stay elevated while asset yields fall, compressing NIM immediately.

The only banks that avoid this are the ones that held deposit beta low throughout the hiking cycle by retaining deposit franchise strength (primary bank relationships, operating accounts, not rate-sensitive CDs). Those banks have more room to compress beta on the way down.

Super-regionals and mid-tier banks did not have that luxury. Their cost of interest-bearing deposits remained at 3.15% as of Q2 2024, and they will face the slowest beta compression when cuts begin. This is the structural disadvantage that will show up in 2025–2026 NIM.

Palanor's read: we watch deposit beta trajectories by bank tier as a leading indicator of NIM stress. When beta stops rising, the cycle has peaked. When it starts compressing, the winners are visible in the dispersion.

The $1.26 trillion CRE maturity peak in 2027 is cited everywhere as the wall. That framing is incorrect. The wall is not when loans mature; the wall is when debt yield falls below the refinancing threshold for loans that mature.

Debt yield = NOI / loan balance. It is the underwriting metric that survived the 2008 crisis because it does not depend on appraised value (which is circular and manipulable) or on cap rate assumptions (which embed rate expectations). A loan with 10% debt yield can refinance in most markets. A loan with 7% debt yield cannot, unless the borrower brings new equity or the lender takes a loss.

The 2027 maturity volume includes:

• Multifamily loans originated in 2020–2022 at sub-4% rates, now facing 6–7% refi rates but strong NOI (debt yield stable or improving)
• Office loans originated in 2017–2019 at 4–5% rates, now facing the same refi rates and 19.8% vacancy (debt yield collapsing)
• Retail and industrial loans with heterogeneous outcomes by submarket and tenant mix

The 2027 wall will not hit uniformly. Extensions crowded 2025–2026 because borrowers with weak debt yields bought time, hoping for rate cuts or NOI recovery. Those extensions are one-time moves. If debt yield has not improved by the extended maturity, the loan moves to special servicing, loan-on-loan refinancing (at distressed terms), or disposition.

The load-bearing variable is not the Fed funds rate. It is whether office NOI stabilizes in the next 18 months. Multifamily has already stabilized in most markets; new supply is slowing and NOI growth has resumed. Office has not. Urban office vacancy at 20% is structural, not cyclical. Buildings do not re-tenant at previous rents without conversion capital, and conversion is not financeable at current construction costs + rate levels.

Palanor's read: the 2027 refinancing wall will resolve in the office sector specifically, not across CRE as a category. The $1.26 trillion figure is misleading because it aggregates performing and non-performing collateral. The real wall is the subset of office loans with sub-8% debt yield, which is roughly $180–220 billion of the total, concentrated in CBD markets with structural vacancies.

Readings [13–16] map a structural bind that has received insufficient attention: Treasury issuance continues to grow while dealer balance sheet capacity is constrained by the SLR, and this is now visible in repo rates persistently above IORB, widening intermediation spreads, and weaker auction outcomes.

The mechanism [15]: the Supplementary Leverage Ratio is a non-risk-weighted capital constraint, meaning it treats Treasury holdings the same as any other balance sheet expansion. This was designed to limit leverage, but the unintended consequence is that it binds exactly when Treasury issuance is highest and dealer intermediation is most needed.

The evidence is in the pricing [13,14,16]:

• Repo rates above IORB indicate that the market is paying more to redistribute Treasury collateral than the Fed's own interest-on-reserve-balances rate — a structural dislocation
• Dealer bid-ask spreads (the CMTR spread and other measures) have widened as balance sheet costs rise
• Auction bid-to-cover ratios have declined, and the highest accepted yield has risen, meaning the government's cost of issuance is rising due to a capital regulation constraint, not credit risk

The feedback loop:

1. Higher fiscal deficits → more Treasury issuance
2. SLR constraint → dealers bid less aggressively at auctions, demand higher compensation for intermediation
3. Weaker auction outcomes → higher Treasury yields, higher fiscal cost
4. Higher fiscal cost → larger deficits → more issuance → repeat

This intersects with the CRE refinancing wall [1,3] and regional bank funding pressures [5–8] because the same dealer capacity that intermediates Treasuries also intermediates the credit markets where distressed CRE will need to clear. When the SLR binds, it does not bind selectively — it compresses balance sheet availability for all intermediation, meaning credit spreads widen not because credit risk has risen, but because the distribution cost has.

The thesis: the SLR is the unpriced governor on the capital window, and it binds before the Fed's monetary policy does. This is not a conspiracy theory — it is observable in Treasury market functioning right now, and it will reprice credit markets as 2027 refinancing demand arrives.

The conventional narrative isolates office distress [2]: 20% vacancy, urban CBD flight-to-quality, remote work as structural demand destruction. But reading [4] reframes the maturity wall: multifamily represents 32% of maturities through 2026, the largest single sector share, driven by the 2020–2022 transaction vintage when valuations peaked and leverage was cheap.

Office is severely distressed, but multifamily is quietly loaded with post-2021 vintages that financed at sub-4% rates and now face 7%+ refinancing into a market where rent growth has slowed, operating expenses (insurance, property tax, maintenance) have risen, and debt service coverage ratios have compressed.

Reading [3] identifies the load-bearing variable: debt yield — net operating income divided by loan balance. This determines refinancing outcomes more than property type. Office buildings with strong in-place tenants and reasonable debt yields can extend or refinance; multifamily properties with thin NOI margins and high loan-to-value from 2021 purchases face the same extend-and-pretend pressure as distressed office, just with less media attention.

The mechanics of the wall [1]: ~$950B matured in 2024, ~$1T in 2025, $1.26T peak in 2027. But extensions crowded into 2026, meaning the effective peak may arrive earlier as lenders exhaust extension capacity and force borrowers into the refinancing market or sale.

What this means for pricing:

• Office CMBS delinquency is the leading indicator, but multifamily distress is the larger volume risk
• Debt yield dispersion within each sector matters more than sector averages — a 10% debt yield office loan refinances; a 6% debt yield multifamily loan does not
• The extend-and-pretend strategy that worked in 2024–2025 requires lender balance sheet capacity that may not survive into 2027 if regional bank funding pressures [6,8] intersect with rising loan loss provisions

The maturity wall is not a single event. It is a two-year stress test of debt yield distributions across the entire CRE capital structure, where office takes the headlines but multifamily carries the volume.

Readings [9–11] establish the structural transformation of leveraged credit since 2008: covenant-lite loans now represent 91% of institutional issuance, PIK provisions account for 11–12% of BDC portfolios, and recovery rates on cov-lite loans run 9 percentage points lower than traditional covenants (57% vs 66%).

The mechanism connecting these three facts is deferral architecture: covenant-lite removes the maintenance tests that would force recognition of deteriorating performance before default; PIK provisions allow borrowers to defer cash interest by adding it to principal, which BDCs book as current income; and when default finally arrives, the absence of covenants means lenders have less information, less control, and materially lower recoveries.

This is not a problem in the current environment, where most borrowers can service even inflated principal balances and most BDCs can show stable yields. It becomes a problem in the next credit cycle, when:

1. PIK capitalization turns into hidden leverage that surfaces only when refinancing becomes necessary [10]
2. Covenant-void structures delay recognition of deterioration, concentrating losses into a narrower window [9,11]
3. Scale pressure forces larger BDCs into BSL exposure [12], meaning the perceived diversification benefit (middle-market = less correlated) erodes as deployment needs push portfolios toward the same broadly-syndicated credits facing the same macro headwinds

The thesis: the private credit market is structurally long an option that is short convexity in a downturn. Reported yields are higher than BSL equivalents, but embedded loss severity is also higher, and the timing of recognition is back-loaded. The sector has not yet priced what a cycle looks like when covenant-lite meets actual stress and PIK capitalization meets refinancing walls.

Fitch and KBRA data [11] already show this in recovery rates. The next shoe drops when BDC NAVs reprice and the illiquidity premium investors thought they were earning turns into an illiquidity discount they cannot exit.

[25, 26, 27, 28] converge on a single mechanism: monetary policy works not by directly setting the interest rate firms pay but by altering the balance-sheet capacity of financial intermediaries, which in turn determines the spread between the policy rate and the rate firms actually face. [27] formalizes this as the external finance premium, which rises when borrower net worth falls. [28] shows that procyclical bank balance sheets amplify the effect: when banks contract, the cost of credit rises even if the policy rate holds steady.

The empirical fact from [25]: modest movements in short rates produce large movements in credit costs. A 25bp Fed hike can widen IG spreads by 50bp if it lands when bank capital is constrained and credit demand is elevated. Conversely, [26] shows that unconventional policy — central bank direct intermediation — can compress spreads even at the zero lower bound by relieving the balance-sheet constraint itself.

The thesis: the capital window is not a function of the Fed funds rate; it is a function of the intermediary capacity to absorb duration and credit risk at the current rate. This capacity is observable in the [28] procyclical balance-sheet indicators: bank equity volatility, dealer inventories of corporate bonds, and the spread between bank lending rates and policy rates.

The pricing implication: a firm refinancing in 2024 faces a different cost of capital than a firm refinancing in 2019 even if the Fed funds rate is identical, because the intermediary constraint is tighter. The [5, 6] IG spread puzzle is partly explained by this: structural models calibrated to historical default data miss the variation in the intermediary pricing wedge, which shows up as 'unexplained' spread.

The commentary implication: covering the capital window requires covering bank capital ratios, not just the dot plot.

[13, 14, 15, 16] lay out the reduced-form framework where default is an exogenous intensity process and credit spreads equal the risk-neutral mean-loss rate. The elegance is that you can price credit without modeling the firm's balance sheet; the cost is that default becomes a surprise the market cannot see coming — even when [15] adds partial information, the timing remains opaque.

Structural models go the other direction: default is endogenous, triggered when assets fall below liabilities, and the equity holders hold a call option on the firm. [5, 6] show this works empirically for high-yield, where asset volatility and leverage ratios do predict spreads. It fails for investment-grade because [8] the diffusion assumption smooths out the refinancing events and liquidity shocks that dominate IG pricing.

The thesis: reduced-form models price market uncertainty about rollover access; structural models price shareholder optionality around the default boundary. In HY, the firm is close enough to the boundary that the structural option dominates. In IG, the firm is far from default but exposed to the rate at which the refinancing window might close, which is a hazard process the market observes in real time through credit spreads and Treasury volatility.

[12] shows that risk-shifting — shareholders extracting value near distress — does not explain the equity anomaly, which implies the hazard the market prices is not the one shareholders can control. [28] shows that the external finance premium moves countercyclically with bank balance-sheet constraints, which implies the hazard is a funding-market state variable, not a firm-level asset process.

The pricing implication: IG spreads should be modeled as a reduced-form intensity indexed to the [25, 26] credit-channel transmission strength, not as a structural distance-to-default. The Duffie-Singleton framework is the correct one; it has just been under-applied to the part of the market where it matters most.

Two literatures document the same structural mispricing and do not cite each other enough.

[9, 10, 11, 12] establish that distressed equities have delivered anomalously low returns since 1981, underperforming safe stocks by more during volatility spikes, with market variables outperforming accounting ratios at longer horizons. The anomaly persists after controlling for size, value, and momentum; risk-shifting explains some agency friction but not the underperformance itself.

[5, 6, 7, 8] establish that structural models radically underpredict investment-grade credit spreads while fitting high-yield spreads reasonably well. The puzzle is sharpest for IG bonds of all maturities; it holds globally; it survives extensions for time-varying leverage and stochastic volatility.

The connection: both anomalies concentrate in the near-distress, not-yet-defaulted region of the capital structure. Distressed equity underperforms because the market prices in heightened rollover risk and agency costs that do not show up in backward-looking default models. IG spreads are wide because the same rollover risk and liquidity demand are not captured by diffusion-based structural models calibrated to historical default rates.

The thesis is that the mispricing is not separate puzzles in equity and credit — it is the market pricing a refinancing-failure state that sits between solvency and default. [28] provides the mechanism: endogenous procyclical balance sheets make the cost of external finance countercyclical. A firm approaching a refinancing wall in a risk-off regime faces both elevated credit spreads (the IG puzzle) and depressed equity valuations (the distress anomaly) because the market is pricing the probability that the window closes before the maturity arrives.

This is not a puzzle; it is the correct price of a risk the models do not parameterize.

<cite index="10-6,10-10">Tighter dealer constraints weaken Treasury auction outcomes: bid-to-cover ratios decline, driven by dealers' less aggressive bidding, and the highest yield accepted by participants rises, thereby increasing the government's cost of issuing debt</cite>. This is not a market-functioning question alone—it is a fiscal question. When the primary dealers who stand between the Treasury and the end investor face balance sheet limits that bite harder than their risk appetite, the auction clearing mechanism distends.

<cite index="16-5">Primary dealers' average daily turnover each week ranged from $545 billion to $944 billion in 2024</cite>, yet <cite index="14-7,14-8,14-9">as capital constraints limit traditional intermediaries, Treasury market intermediation increasingly relies on principal trading firms (PTFs); however, PTFs operate with short-term strategies and, unlike primary dealers, are less willing to make markets during stress periods, and PTFs principally provide liquidity for on-the-run Treasury issues while liquidity for off-the-run issues remains dependent on dealer intermediation</cite>.

The 2027 refinancing wall is the load-bearing assumption in the next cycle's pricing. If dealer balance sheets cannot absorb what Treasury must issue, auction yields will reflect that scarcity—not as a temporary spike but as a persistent premium. The composite signal for Treasury market functioning should weight auction outcomes more heavily than it does.

Sources:

• https://www.bostonfed.org/publications/research-department-working-paper/2024/the-effect-of-primary-dealer-constraints-on-intermediation-in-the-treasury-market.aspx
• https://www.bostonfed.org/publications/current-policy-perspectives/2025/relaxing-dealers-risk-constraints-can-make-treasury-market-liquid.aspx
• https://bpi.com/treasury-market-resiliency-and-large-banks-balance-sheet-constraints/

<cite index="19-3,19-4,19-5">The Supplementary Leverage Ratio (SLR), introduced after 2008 and tightened under Basel III, is a non-risk-weighted capital constraint to limit the overall size of bank balance sheets; unlike risk-weighted requirements, the SLR imposes a blunt limit—a minimum ratio of Tier 1 capital to total leverage exposure, irrespective of asset riskiness—with far-reaching implications for low-risk assets like US Treasuries and repo positions</cite>. <cite index="23-5">The eight U.S. global systemically important banks are subjected to the enhanced SLR, which effectively requires them to maintain an SLR of at least 5%</cite>.

<cite index="12-4">The SLR has been the most binding capital requirement for several large primary dealers in recent years; it is particularly relevant for Treasury market intermediation as it has the potential to impose a high regulatory capital requirement for a relatively low-risk but high-volume activity</cite>. <cite index="14-3,14-4,14-5">The decline in intermediation capacity is evident in bank balance sheet composition; large banks, which usually include a primary dealer entity, have significantly increased their Treasury holdings, with the share of U.S. Treasuries relative to total assets expanding from 3 percent in 2013 to 11 percent in 2024</cite>.

Regulators know. <cite index="25-5">The proposed recalibration of the eSLR would help mitigate potential disincentives for GSIBs to engage in low-risk, low-return, balance-sheet-intensive activities, such as intermediation in Treasury markets</cite>. The proposal does not yet price.

Sources:

• https://goghieas.substack.com/p/the-supplementary-leverage-ratio
• https://www.financialresearch.gov/the-ofr-blog/2024/08/02/banks-supplementary-leverage-ratio/
• https://www.federalreserve.gov/econres/notes/feds-notes/assessment-of-dealer-capacity-to-intermediate-in-treasury-and-agency-mbs-markets-20241022.html
• https://bpi.com/treasury-market-resiliency-and-large-banks-balance-sheet-constraints/
• https://www.federalregister.gov/documents/2025/07/10/2025-12787/regulatory-capital-rule-modifications-to-the-enhanced-supplementary-leverage-ratio-standards-for-us

<cite index="2-4,2-5">The continued growth of U.S. Treasury issuance has garnered interest in understanding dealers' ability to intermediate the Treasury market; various efforts measure the degree to which dealer balance sheet constraints—broadly defined as restrictions on the overall size of an intermediary's balance sheet—affect Treasury market intermediation</cite>. <cite index="2-6">The Cross-Market Treasury Repo (CMTR) spread isolates the compensation dealers receive for intermediating overnight Treasury repos across repo market segments</cite>.

The load-bearing result: <cite index="10-3,10-4,10-5">In response to tighter constraints, dealers reduce their Treasury positions, triggering a reduction in aggregate turnover and an increase in bid–ask spreads; these effects are more pronounced in securities that contribute more to the utilization of risk constraints, and impaired intermediation affects Treasury yields, amplifying the yield response to net demand shifts</cite>. The shadow cost is not academic. <cite index="10-7,10-12">The shadow cost of dealer constraints is as high as one-third of dealers' intermediation margin, or about $2.4 billion to $3 billion per year</cite>.

<cite index="11-1">Between 2014 and 2024, Treasury securities held by the public increased to $24 trillion from $10 trillion (a 139% increase), while primary dealer balance sheets grew to $4.2 trillion from $3.3 trillion (a 29% increase)</cite>. The divergence prices what cannot hold.

Sources:

• https://www.federalreserve.gov/econres/notes/feds-notes/dealer-balance-sheet-constraints-evidence-from-dealer-level-data-across-repo-market-segments-20240923.html
• https://www.bostonfed.org/publications/research-department-working-paper/2024/the-effect-of-primary-dealer-constraints-on-intermediation-in-the-treasury-market.aspx
• https://www.congress.gov/crs-product/R48734

<cite index="1-2,1-3">The share of repo transactions priced at or above IORB has increased notably since spring 2024</cite>, signaling structural stress in short-term funding markets. <cite index="1-11">The repo market redistributes liquidity from those who hold it in excess—money market funds, GSEs—to those who need it: dealers and levered investors</cite>. When <cite index="5-1,5-2">repo counterparties trade at rates above IORB, they are willing to pay a premium to attract cash from banks that would otherwise earn IORB on their balances, indicative of more urgent demand for liquidity</cite>.

The shift holds weight. <cite index="9-3,9-4,9-5">Between early July and mid-September when the Treasury General Account reached around $800 billion, the ON RRP fell from $200 billion to de minimis levels and reserves fell by $350 billion; with the ON RRP effectively drained and balance sheet reduction ongoing, repo rates now had to increase more meaningfully to entice a marginal dollar of additional cash into the market</cite>. The floor mechanism has not failed, but the floor no longer anchors as tightly. <cite index="3-10,3-11">Volatile and unpredictable repo rates increase risks that investors face in financing their Treasury holdings; all else equal, greater financing risk would result in higher Treasury yields, tighter financial conditions, and a headwind for the economy</cite>.

What the Fed is pricing is not yet what the Fed is saying. The window for predictable dealer financing has closed by half a turn since mid-2024.

Sources:

• https://www.newyorkfed.org/newsevents/speeches/2024/per240926
• https://www.newyorkfed.org/newsevents/speeches/2024/per240508
• https://www.newyorkfed.org/newsevents/speeches/2025/rem250929
• https://www.newyorkfed.org/newsevents/speeches/2025/per251112

<cite index="8-1,8-2,8-3">Larger BDCs with significant assets under management have sizeable amounts of cash to put to work, sourced from both inflows and loan repayments each year; these larger managers have to reinvest tens of billions of dollars every year; as a result, they have grown too large to make smaller loans to middle market companies, causing them to migrate out of the core middle market and make loans to larger capitalization borrowers.</cite> <cite index="8-4">Additionally, the deployment of excess cash calls into question how prudently they are allocating that capital and introduces legitimate concerns such as greater exposure to broadly syndicated loans (BSL), higher leverage, weaker covenants, smaller illiquidity premiums, and the potential for liability management exercises (LME) to creep into their books.</cite>

The private credit market grew to nearly $2 trillion in AUM by 2024. <cite index="1-13,1-14">After the Global Financial Crisis, private credit rapidly evolved from niche strategy to a $1+ trillion market; by 2024, private credit AUM had swelled to nearly $2 trillion, thanks to investors' search for yield and banks retreating from middle-market lending.</cite> <cite index="8-11,8-12,8-13">It's also a transfer of risk from the liquid public high yield and bank loan, also known as leveraged loan, markets to private markets as larger borrowers who have historically and exclusively financed themselves through the capital markets are now using the private credit option; COVID-19 and the regional banking crisis following the failure of Silicon Valley Bank in 2023 also supercharged growth in private credit; in both instances, public capital markets retreated; companies that still required financing, including larger firms that had historically borrowed in the public markets via large investment banks, turned to private markets for capital.</cite>

Direct lenders at the lower end of the middle market continue to underwrite with discipline. <cite index="16-8,17-8,19-9">At the time of original closing for these borrowers, the weighted average senior leverage was approximately 4.4 to 4.5 times, the loan-to-value ratio was approximately 38% to 39%, and the weighted average spread over reference rate to LIBOR, SOFR, and CDOR was 635 to 660 bps.</cite>

Sources:

• https://www.lordabbett.com/en-us/financial-advisor/insights/investment-objectives/2025/a-closer-look-at-the-growth-of-private-credit-markets.html
• https://resonanzcapital.com/insights/covenant-lite-to-covenant-void-navigating-private-credit-risk
• https://www.sec.gov/Archives/edgar/data/0001860424/000119312524013046/d652929d8k.htm
• https://www.sec.gov/Archives/edgar/data/0001860424/000119312524107059/d800160d8k.htm
• https://www.sec.gov/Archives/edgar/data/0001860424/000119312524241857/d831198d8ka.htm

<cite index="4-2,4-3">Recovery rates are lower; we just saw this: 57% on cov-lite vs 66% on traditional loans.</cite> <cite index="10-1,10-2">Results show that the covenant-lite is granted to borrowers with a greater profitability; in turn, all other conditions being equal, this agreement plays a role in making a default event less likely, giving rise to a significant indirect effect.</cite> That indirect effect does not offset the recovery differential when default events occur.

Default rates through 2024 and into 2025 have begun to rise, though they remain below levels that would constitute systemic concern. <cite index="22-7,22-8">The KBRA DLD Direct Lending Index showed a trailing 12-month default rate of 1.8% as of September 11, mirroring the 1.8% rate reported at the end of 2024—which remains materially below thresholds that would generally raise systemic concerns; moreover, KBRA's latest forecast suggests a potential uptick in defaults, with the lower middle market default rate expected to reach 3% by year-end, driven by a growing pipeline of stressed borrowers.</cite> <cite index="23-1,23-2,23-3">Fitch's latest commentary indicates the private credit default rate rose to 9.2% in 2025 within its U.S. Privately Monitored Ratings universe, up from 8.1% in 2024, with the highest default incidence among smaller issuers (for example, EBITDA of $25 million or less).</cite>

<cite index="20-3,20-4,20-5">Rather than resulting in "hard" defaults such as missed payments or bankruptcies, persistently elevated credit risk has manifested itself in a rising pace of distressed restructurings; restructuring a debt contract by extending maturities and modifying other terms, especially when done under financial duress, is usually considered an event of default in public debt markets; Moody's Ratings reported that distressed exchanges accounted for 64% of all defaults in the first half of 2025, about the same pace as all of 2024, which itself was the highest annual share on record.</cite>

Sources:

• https://medium.com/@s.shidharths15/covenant-lite-loans-the-systemic-risk-i-almost-overlooked-df5b6dc4cd8b
• https://www.sciencedirect.com/science/article/pii/S0275531924001703
• https://www.valuationresearch.com/insights/private-credit-market-pressures/
• https://www.securitasglobal.com/risk-perspectives/private-credit-defaults/
• https://www.moodys.com/web/en/us/insights/resources/us-report-july-2025.pdf

<cite index="1-1,1-2">Loans carrying PIK provisions made up about 11–12% of BDC loan holdings by value, before marking its first decline in over a year as a share of BDC assets, which grew sharply in the third quarter; this came after a steady rise of PIK-paying loans within BDC portfolios from second-quarter 2023 through second-quarter 2024 caught investor attention.</cite> <cite index="1-4,1-5">Some private credit funds even saw PIK interest account for low double-digit percentages of their total interest income; these figures reveal a concerning trend: a portion of yield that investors see from private credit is "pencil yield" – income that is accrued on paper but not actually received in cash.</cite>

<cite index="1-6">This hidden leverage builds up the longer PIK continues, increasing the eventual burden on the borrower.</cite> <cite index="6-1,6-2,6-3">By mid-2024, approximately 10% of Business Development Company interest income was PIK rather than cash, with 11%-12% of BDC-held loans carrying PIK provisions; this approach manages current interest burdens by deferring payments, though at the cost of increasing leverage over time, with lenders selectively offering PIK toggles, typically charging an additional 100-200 basis points when utilized.</cite>

The market is entering its first significant default cycle under these structures. <cite index="1-7,1-8,1-9">After years of growth fueled by easy credit and low interest rates, the market is now entering its first significant default cycle; this emerging phase is raising questions about risk: have covenant-lite loans effectively turned into covenant-void loans? Are PIK (Payment-in-Kind) interest structures quietly inflating leverage?</cite>

Sources:

• https://resonanzcapital.com/insights/covenant-lite-to-covenant-void-navigating-private-credit-risk
• https://privatecapitalglobal.com/blog/private-capital-debt-benchmarks-for-the-new-rate-environment

<cite index="4-10,4-11">As of year-end 2024, covenant-lite loans represented 91.09% of outstanding US leveraged loans — approximately $1.29 trillion in total, with 93% of all institutional leveraged loans issued in 2024 carrying cov-lite structures.</cite> <cite index="2-6">Cov-lite loans dominated the leveraged loan market in 2024, and increasing competition in the private lending market has resulted in some lenders removing covenants in deals for entities with EBITDA.</cite> What distinguishes cov-lite is the absence of maintenance covenants for term lenders. <cite index="3-1,3-11">There will be no maintenance covenant at all for the benefit of the term loan lenders; instead, where there is a revolving credit facility sitting alongside, the revolving facility lenders will benefit from a "springing" leverage covenant (meaning that it only applies to the extent the revolving facility is actually drawn to a certain extent, commonly 40%).</cite>

The spread that once compensated lenders for weaker protections has vanished. <cite index="4-20,4-21,4-22">Historically, covenant-lite instruments commanded a 50–75 bps premium over traditional covenanted loans; that premium has essentially vanished, with yields converging since Q1 2017.</cite> Private credit executions through 2024 showed partial divergence from the broadly syndicated loan market. <cite index="9-5,9-6">Private credit deals were increasingly covenant-lite in 2025, rising to 21% of all private credit deals from 4% in 2023, bringing the market more in line with the syndicated loan market, but lender protections are still stronger in private credit.</cite> Even when private credit lenders offer cov-lite to remain competitive with BSL options, these private credit cov-lite executions still have reserved leverage, reserved additional debt capacity, strong LME protections, and other protections that distinguish them from the more flexible and borrower-friendly BSL documents.

Sources:

• https://medium.com/@s.shidharths15/covenant-lite-loans-the-systemic-risk-i-almost-overlooked-df5b6dc4cd8b
• https://www.paulweiss.com/media/mjanpfpm/covenant_lite_loans_overview.pdf
• https://www.proskauer.com/alert/private-credit-deep-dives-leverage-covenants-and-auto-resets
• https://finance.yahoo.com/news/despite-covenant-lite-trend-lender-201828993.html

<cite index="9-3,9-4,9-5,9-6">Slowing loan growth poses a challenge for many banks in 2024; high borrowing costs and weakening economic conditions are headwinds for loan demand, and many bankers are tightening lending practices in preparation for a less favorable economic environment—Fed lending surveys show that many banks have tightened credit standards across all product categories and anticipate a deterioration in credit quality as well as collateral values in 2024.</cite> <cite index="9-7,9-22">The persistence of elevated deposit costs combined with slower loan growth challenges net interest income.</cite>

<cite index="16-19,16-20">The Federal Reserve lowered short-term rates by 100 basis points in late 2024, with expectations of additional rate cuts in 2025; on average, community banks have not been able to lower their cost of funding, now recognize the reality of a pause by the Federal Reserve, and face the possibility that the next Fed action is a hike instead of a cut.</cite> <cite index="16-27">A pause for an extended period, or another rate hiking cycle from these short-term interest rates would find some community banks flat-footed, subject to increased net interest margin pressures.</cite>

<cite index="21-3,21-4,21-5">The industry-wide net interest margin was approximately 3.22% for full-year 2024; large money-center banks operate with lower net interest margins around 2.5% for JPMorgan and 1.97% for Bank of America because their asset bases include lower-yielding but more diversified portfolios, while regional and community banks typically have higher net interest margins of 3.5-4.0% because they concentrate in higher-yielding commercial and CRE lending.</cite>

Sources:

• https://www.gobaker.com/deposit-betas-remain-a-challenge-in-2024/
• https://southstatecorrespondent.com/banker-to-banker/deposits/how-to-manage-the-cost-of-funding-dilemma-for-2025/
• https://ibinterviewquestions.com/guides/fig-investment-banking/net-interest-income-and-net-interest-margin

<cite index="1-4,1-5">With the Fed seemingly at or near its peak-rate level, deposit costs began to plateau, and most banks' net interest revenue was expected to grow higher through 2024, bolstering investor sentiment.</cite> <cite index="1-9,1-10">The March and April 2023 bank failures focused customers' attention on their deposits at other financially sound institutions, which led to accelerated upward deposit repricing activity that weighed on banks' bottom lines and led investors to fret about the durability of bank deposit bases and the magnitude of further repricing in an environment of elevated rates.</cite>

<cite index="13-13,13-14">The value of a deposit franchise varies over the interest rate cycle, meaning deposits are much less of a stabilizing force than traditionally assumed; because lending costs exceed operating costs, a bank's franchise value declines as interest rates rise, exacerbating losses on security holdings.</cite> <cite index="14-11,14-16">Banks have been reluctant to raise deposit rates when they have more deposits than they need and are willing to allow depositors to seek better rates elsewhere—the level of deposits appears to be roughly correlated with the responsiveness of deposit rates to fed funds hikes.</cite>

<cite index="11-1,11-2">Banks with greater reliance on wholesale deposits, non-core deposits, online deposits, and concentrated funding sources have less ability to moderate deposit betas than competitors with higher core deposits—another reason to measure, monitor, and understand the deposit base and behavior by customer type and product before making decisions about deposit pricing and promotional rates.</cite>

Sources:

• https://www.jhinvestments.com/viewpoints/u-s-equities/A-brighter-2024-outlook-for-US-regional-banks
• https://www.sciencedirect.com/science/article/abs/pii/S1042957325000154
• https://libertystreeteconomics.newyorkfed.org/2022/11/how-do-deposit-rates-respond-to-monetary-policy/
• https://www.rmahq.org/journal-articles/2024/feb-mar-2024/the-increasing-importance-of-understanding-deposit-betas/

<cite index="3-1,3-3">The cost of interest-bearing deposits for midsize and regional banks remained high at 3.15% as of second quarter 2024, and deposit betas are expected to be low for them when rates decline.</cite> <cite index="15-8,15-9">Regional and midsize banks under $1 trillion in assets rely heavily on deposits as their key source of funding, and their cost of deposits remains fairly high.</cite> <cite index="8-16,8-17">Problems arise when banks cannot pass increased funding costs onto higher loan rates, causing net interest income and profitability to decline; banks without excess margin to pay up for deposits often lose them, restricting the ability to originate loans, narrowing margins, and lowering profitability further.</cite>

<cite index="6-1">Regions Financial, as one example, reported a third-quarter 2024 deposit cost of 1.60% and an interest-bearing deposit cost of 2.34%, representing a 43% full-cycle interest-bearing beta.</cite> <cite index="5-1">The bank guided to a full-cycle interest-bearing deposit beta above roughly 45% with deposit outflows or continuous remixing as a downside scenario.</cite> <cite index="16-3">Historically, bank deposit betas ranged between the low-20s and mid-50s, depending on the length and magnitude of the hiking cycle, the size of each increase, and balance sheet changes at the Fed.</cite> <cite index="16-6,16-7,16-8">Online banks like American Express, Capital One, and Ally comprise more than 10% of domestic deposits and have much higher betas over 0.70, increasing the industry's cost of funding faster in a rising rate environment and keeping cost of funding higher for longer in a steady rate environment.</cite>

Sources:

• https://www.deloitte.com/us/en/insights/industry/financial-services/financial-services-industry-outlooks/banking-industry-outlook-2025.html
• https://www.deloitte.com/us/en/industries/financial-services/articles/deposit-rates.html
• https://www.stlouisfed.org/on-the-economy/2024/sep/higher-deposit-costs-continue-challenge-banks
• https://www.sec.gov/Archives/edgar/data/0001281761/000128176124000094/rf-2024930xexhibit993.htm
• https://www.sec.gov/Archives/edgar/data/0001281761/000128176124000040/rf-2024630xexhibit993.htm
• https://southstatecorrespondent.com/banker-to-banker/deposits/how-to-manage-the-cost-of-funding-dilemma-for-2025/

<cite index="8-4,8-5">The Fed hiked rates 525 basis points between March 2022 and July 2023, but cumulative deposit beta continued to rise even after the final hike, meaning banks had to continue increasing deposit rates to retain or attract funding.</cite> <cite index="10-2">Deloitte estimates the average cost of interest-bearing deposits remained elevated at 1.7% and 1.5% in 2024 and 2025 respectively, even as the fed funds rate declined—a historical anomaly resulting in lower deposit betas in a downward-sloping rate environment.</cite>

<cite index="11-21,11-22,11-25">Deposit betas are stickier early in a rising rate cycle but accelerate in the mid and later stages as monetary tightening drives depositors toward competing higher-earning products, forcing banks to increase deposit rates to retain funding.</cite> <cite index="11-26,11-27">The behavior reverses when the fed funds rate falls; banks with high core deposits are slow to increase deposit betas during rate hikes but quicker to reduce deposit rates during rate cuts.</cite> <cite index="9-2,9-17">However, banks' increased reliance on wholesale funding and the reawakening of so-called sleepy deposits may render the traditional strategy of aggressively cutting the cost of funds ineffective this cycle.</cite>

<cite index="13-11,13-12">Banks that relied more heavily on uninsured deposits and those with lower deposit branch density experienced significantly higher deposit betas, suggesting that reliance on uninsured deposits and online platforms increases both liquidity risk and interest rate risk—possibly explaining why deposit betas increased more during 2022–2023 than in previous tightening periods.</cite>

Sources:

• https://www.stlouisfed.org/on-the-economy/2024/sep/higher-deposit-costs-continue-challenge-banks
• https://www2.deloitte.com/us/en/insights/industry/financial-services/financial-services-industry-predictions/2023/bank-deposit-costs.html
• https://www.rmahq.org/journal-articles/2024/feb-mar-2024/the-increasing-importance-of-understanding-deposit-betas/
• https://www.gobaker.com/deposit-betas-remain-a-challenge-in-2024/
• https://www.sciencedirect.com/science/article/abs/pii/S1042957325000154

<cite index="8-4,8-5">Multifamily makes up the largest single share of commercial loan maturities, given the sector's outsized share of CRE transactions in 2020, 2021, and 2022; the multifamily sector makes up 32% of maturities to 2026</cite>, according to Principal Real Estate. <cite index="6-29">Multifamily maturities are set to surge, with the loan maturity calendar jumping 56% from approximately $104.1 billion in 2025 to roughly $162.1 billion in 2026 and edging higher to $167.7 billion in 2027</cite>.

<cite index="6-24">Borrowers who locked in financing at 3% to 4% in the mid-2010s are now facing refinance rates that can be nearly double (or more)</cite>. <cite index="6-31,6-32">The core challenge is lower loan proceeds; with higher rates and more conservative underwriting, a new loan covers a smaller share of value than it did a few years ago, forcing owners to inject additional cash or secure alternative capital</cite>. <cite index="2-17">In the first quarter of 2025, the total volume of distressed assets reached $116 billion, a 31% increase from a year earlier; many of these are office properties, but stress is beginning to surface in other sectors as well</cite>.

<cite index="24-4">Credit risk has shifted rather than disappeared, with the office market no longer viewed as uniformly impaired and the multifamily market no longer assumed to be low-risk due to expense pressure and stress in post-2021 vintages</cite>. What the market is observing is that sector rotation in distress has begun; it is no longer confined to a single property type. The composite should reflect that dispersion.

Sources:

• https://brandassets.principal.com/m/4f0a2e32cd4949ac/original/Principal-Real-Estate-Wall-of-Maturities.pdf
• https://mmgrea.com/2026-cre-refinancing-wall/
• https://www.pbmares.com/preparing-for-the-cre-maturity-wall/
• https://www.altusgroup.com/insights/what-crefc-miami-revealed-about-cre-debt-markets-in-2026/

<cite index="10-7,10-8,10-9">Debt yield now drives refinancing outcomes more than maturity volume alone; loans with stronger debt yields refinance more easily, while weaker loans often require extensions or fall into delinquency</cite>, according to Trepp's Spring 2026 Quarterly Data Review. <cite index="10-11,10-12">Trepp's analysis of 2024 and 2025 CMBS maturities found a sharp divide between performing and troubled loans; loans that paid off on time carried average debt yields between 13% and 14%, while loans that failed to refinance averaged debt yields closer to 9%</cite>.

<cite index="10-5,10-6">$76.6 billion in CMBS loans face hard maturities in 2026, including fixed-rate debt reaching maturity and floating-rate debt without remaining extension options</cite>. <cite index="10-14,10-15">In 2025, 70% of CMBS hard maturities paid off on time, an improvement from 56% in 2024; the unresolved 2025 loan cohort spent 45% of post-maturity time classified as nonperforming, down from 66% the year before</cite>. <cite index="10-17,10-18">The CMBS office delinquency rate hit a record 12.34% in January 2026, partly driven by the $835 million One New York Plaza loan transferring to special servicing ahead of maturity; the loan was later modified and extended through 2028</cite>.

<cite index="10-19">Extensions and modifications continue softening what could otherwise be a more severe refinancing shock</cite>. <cite index="10-26">Debt yield is emerging as the simplest shorthand for separating refinanceable loans from future delinquency events</cite>. What lenders are pricing is not yet what borrowers are saying about refinancing capacity. The spread between the two positions is the next quarter's volatility.

Sources:

• https://www.credaily.com/briefs/cmbs-maturity-wall-tests-refinancing-in-2026/

<cite index="11-3,11-4">U.S. office vacancy rates reached a new high of 19.8% in Q1 2024, surpassing previous peaks observed in 1986 and 1991</cite>, according to Moody's Analytics. <cite index="15-29">Nationwide, office vacancy rates are around 20%, roughly twice the pre-2020 rate and reflecting a glut of underused space</cite>. <cite index="14-9">The national vacancy rate rose by 150 basis points in 2024 to 19.8%, and ticked up again in March to 19.9%</cite>.

<cite index="1-13,1-15">The office sector is being closely watched, as the shift to more employees working remotely post-pandemic has led to elevated vacancies and delinquencies; high interest rates, low valuations, and increased scrutiny will prove worrisome for the coming quarters, as landlords find it challenging to refinance loans in the current environment</cite>. <cite index="14-3,14-4">2024 saw 25 million square feet of office properties hit distress, up sharply from the previous three-year average of 18 million; nearly 11% of all office transactions were classified as distressed last year, with average asset size in these deals climbing above 200,000 square feet—pointing to a crisis disproportionately impacting large office assets in dense urban centers</cite>. <cite index="14-5">Central business districts saw distressed deals triple year-over-year, while urban properties nearly doubled</cite>.

<cite index="15-7,15-8">San Francisco's office vacancy surged to roughly 35% (and over 36% downtown), the highest in the nation; the tech-heavy metro has seen office demand collapse in the face of abundant remote work</cite>. <cite index="14-8">Office attendance has flatlined at 54% since 2023</cite>, according to Kastle's Back to Work Barometer.

Sources:

• https://www.credaily.com/briefs/office-vacancy-rate-nears-20-to-set-fresh-record/
• https://www.credaily.com/briefs/office-distress-rises-in-urban-markets-amid-remote-work-shift/
• https://meketa.com/wp-content/uploads/2025/07/MEKETA_Office-Space.pdf
• https://www.spglobal.com/market-intelligence/en/news-insights/research/commercial-real-estate-maturity-wall-950b-in-2024-peaks-in-2027

<cite index="1-3,1-5">Approximately $950 billion in CRE mortgages matured in 2024, with the tally anticipated to reach nearly $1 trillion in 2025 and peak at $1.26 trillion in 2027</cite>, according to S&P Global Market Intelligence. The office sector holds <cite index="1-16">10% of CRE mortgages maturing in 2024</cite>, a concentration the rating firm observed as elevated relative to prior years.

<cite index="1-8,1-9">Many CRE loans set to mature in 2024 were extended into 2025 and beyond, as bank regulators allowed lenders to work with borrowers rather than force maturity; extensions provide borrowers short-term cover and give lenders time to work out troubled credits and prune portfolios through strategic sales</cite>. <cite index="2-6,2-7">In 2025 alone, nearly $1 trillion in loans matured, with a large share of that total including debt previously extended in response to rising rates and valuation uncertainty</cite>. <cite index="6-5,6-6,6-7">The go-to move through 2024–2025 was "extend and pretend," modifying or extending loans to push out maturities; in Q3 2025 alone, tens of billions of CRE loans were adjusted to delay defaults and avoid fire sales, but this only kicks the can, and many extended loans are now crowding into the 2026 window</cite>.

<cite index="1-12">As of August 2024, the average interest rate on CRE loans originated in 2024 was 6.2%, whereas the rate on those maturing was 4.3%, a jump of nearly 200 basis points</cite>. The refinancing window has closed by half a turn since the beginning of the cycle.

Sources:

• https://www.spglobal.com/market-intelligence/en/news-insights/research/commercial-real-estate-maturity-wall-950b-in-2024-peaks-in-2027
• https://www.pbmares.com/preparing-for-the-cre-maturity-wall/
• https://mmgrea.com/2026-cre-refinancing-wall/

<cite index="19-3,19-4,19-5">In mathematical finance, backtest overfitting means the usage of historical market data to develop an investment strategy where many variations are tried on the same dataset; it is now thought to be a primary reason why quantitative investment models that look good on paper often disappoint in practice, as models target the specific idiosyncrasies of a limited dataset rather than any general behavior.</cite>

<cite index="12-11,12-12,12-13">Overfitting in trading happens when a strategy is overly tailored to historical data, mistaking random noise for actual patterns; this leads to great backtest results but poor real-world performance, with key causes including excessive parameter tweaking, biased data selection, and unnecessary complexity.</cite> <cite index="18-9">The classic warning sign is a strategy with many tunable parameters that all need to be set to specific values for it to work.</cite>

<cite index="11-1,11-2">There is an important relationship between degrees of freedom and sample size; in the context of optimization, degrees of freedom are analogous to the number of parameters simultaneously being optimized—higher degrees of freedom require higher sample size during backtesting, which is why traders who optimize too many parameters will almost certainly struggle to produce reliable backtest results.</cite>

Sources:

• https://www.davidhbailey.com/dhbpapers/overfit-tools-at.pdf
• https://www.luxalgo.com/blog/what-is-overfitting-in-trading-strategies/
• https://www.mql5.com/en/blogs/post/756385

<cite index="7-14">As long as the researcher tries more than one strategy configuration, overfitting is always present; the hold-out method does not take into account the number of trials attempted before selecting a model, and consequently is subjected to selection bias.</cite> <cite index="13-3">The "hold-out" method is not very effective in preventing backtest overfitting.</cite>

Hold-out splits data into train and test sets to evaluate generalization. It works when you fit one model once. It does not work when you iterate: fit model A, test on hold-out, reject; fit model B, test on same hold-out, reject; fit model C, test on same hold-out, accept. By the third trial the hold-out set has become part of the search—it has been implicitly trained on. <cite index="7-11,7-12">Van Belle and Kerr point out the high variance of hold-out's estimation errors; different "hold-outs" are likely to lead to opposite conclusions.</cite>

<cite index="15-1,15-2">In quantitative finance, overfitting is exacerbated by optimizer-driven searches over strategies, feature variants, and parameterizations; selecting the best backtest among many candidates is a multiple-testing/data-snooping problem that inflates apparent statistical significance.</cite> The discipline required is to count trials and adjust significance thresholds accordingly—or accept that the promising backtest may be a statistical fluke.

Sources:

• https://pdfs.semanticscholar.org/c215/d0a2064ce1a3565d276475abc84305418f0f.pdf
• https://sdm.lbl.gov/oapapers/ssrn-id2507040-bailey.pdf
• https://www.mdpi.com/1911-8074/19/1/60

<cite index="19-1,19-6">Backtest overfitting is an instance of the more general phenomenon of multiple testing in scientific research, where a large number of variations of a model are tested on the same data without accounting for the increase in false positive rates.</cite> <cite index="22-4,22-5,22-6">Testing 100 random strategies, you expect 5 to be "significant" at p < 0.05, but they're just lucky; the correct approach is a p-value threshold of 0.05 / 100 = 0.0005.</cite>

<cite index="18-1,18-2">The more strategies you test, the more likely you are to find one that looks good purely by accident—this is sometimes called "data snooping" or the "multiple comparisons problem," and it is pervasive in quantitative research.</cite> <cite index="18-6,18-7,18-8">Any sufficiently flexible model can find patterns in historical noise that produce impressive-looking backtests; the problem is that noise, by definition, doesn't repeat, and a model tuned to the specific sequence of random fluctuations in your test period will fail when confronted with a different sequence.</cite>

<cite index="13-1,13-2">False trading strategies can be derived from purely random data, and if one does not know how many variations of a strategy have been attempted, there is no way to know a priori whether the resulting strategy is overfit.</cite>

Sources:

• https://www.davidhbailey.com/dhbpapers/overfit-tools-at.pdf
• https://hedgefundalpha.com/education/backtesting-mistakes-kill-quant-strategies-guide/
• https://sdm.lbl.gov/oapapers/ssrn-id2507040-bailey.pdf
• https://www.waylandz.com/quant-book-en/Lesson-07-Backtest-System-Pitfalls/

<cite index="5-2,5-3">With large data sets and computing power, analysts can now backtest millions of alternative strategies, searching for parameter combinations that maximize simulated historical performance—leading to backtest overfitting.</cite> <cite index="5-5,5-6">Researchers and investors tend to report only positive outcomes (selection bias), and not controlling for the number of trials involved in a particular discovery leads to overly optimistic performance expectations.</cite>

<cite index="1-6,1-7">Selection bias and backtest overfitting inflate the Sharpe ratio; the Deflated Sharpe Ratio (DSR) corrects for selection bias, backtest overfitting, sample length, and non-normality in return distributions.</cite> <cite index="2-10,2-11">Research by Brown, Goetzmann, Ibbotson, and Ross found that survivorship bias could inflate Sharpe ratios by as much as 0.5 points—substantial when a Sharpe ratio of 1.0 is viewed as strong performance.</cite>

<cite index="8-13">Performance out of sample is likely to disappoint, a phenomenon called "regression to the mean" in the shrinkage estimation literature.</cite> The DSR reframes the question: not whether a backtest produced an attractive Sharpe, but whether the Sharpe observed is statistically significant after accounting for how many configurations were tried. <cite index="8-2">Selection bias combined with backtest overfitting misleads investors into allocating capital to strategies that will systematically lose money.</cite>

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2460551
• https://en.wikipedia.org/wiki/Deflated_Sharpe_ratio
• https://www.luxalgo.com/blog/survivorship-bias-in-backtesting-explained/
• https://www.davidhbailey.com/dhbpapers/deflated-sharpe.pdf

<cite index="18-2,18-3,18-4">Topic modeling is an unsupervised machine learning algorithm that models every document as a mixture of topics and every topic as a mixture of words; once it is determined how individual topics are composed, it is easy to estimate the proportion of a specific topic occurring in a given text</cite>. <cite index="18-5">Dybowski and Kempa showed that the focus of ECB communication shifted to financial system stability after the 2007-08 crisis, instead of the usual monetary analysis</cite>.

<cite index="21-2,21-3">A classification framework analyzes central bank communications across four dimensions: topic, communication stance, sentiment, and audience, using a fine-tuned large language model trained on central bank documents to classify individual sentences and transform policy language into systematic and quantifiable metrics</cite>. <cite index="21-5,21-7">Monetary policy communication changes significantly with inflation targeting, as backward-looking exchange rate discussions give way to forward-looking statements on inflation, interest rates, and economic conditions</cite>.

The method detects shifts that narrative readers notice but cannot date precisely. It weights topics by their share of total communication and tracks their evolution meeting-to-meeting. <cite index="20-2,20-3">Indicators estimate speech intensity in five macroeconomic fields: monetary conditions, financial stability, external competitiveness, labour and social conditions, and economic activity, plus an index about uncertainty and risk</cite>. The output is a panel: central bank × date × topic intensity. Researchers regress it against policy decisions or asset returns to test whether communication predicts action or whether it moves prices independently. For those watching the Fed, topic models reveal what shifted before the statement language fully turned.

Sources:

• https://www.marcellgranat.com/posts/cb-textmining/
• https://www.imf.org/en/publications/wp/issues/2025/06/06/from-text-to-quantified-insights-a-large-scale-llm-analysis-of-central-bank-communication-567522
• https://www.sciencedirect.com/science/article/pii/S130307012500023X

<cite index="15-2,15-3">Early studies predominantly employed dictionary-based approaches and simple statistical methods to quantify policy sentiment and tone; Lucca and Trebbi developed lexicon-based measures to capture hawkish and dovish language in FOMC statements</cite>. <cite index="17-1,17-2,17-3">Sentences from FOMC statements are manually labelled as Dovish, Hawkish or Neutral, models are fine-tuned with 80% of these sentences, then the language models predict the monetary policy stance for the rest</cite>.

The method produces a time series of policy tone that can be shocked in a VAR or regressed against asset returns. <cite index="15-10,15-11">Sentiment index surprises around FOMC meeting announcements explain variation in major asset price classes, and sentiment index surprises are important for explaining variation in asset prices beyond monetary policy surprises</cite>. <cite index="22-7">A quantitative text analysis framework utilizes a dictionary tailored to the domain of monetary policy (as opposed to working with general purpose, or even financial, keywords)</cite>.

The limitation: <cite index="15-5">these methods struggle with linguistic nuances, contextual dependencies, and the evolving nature of central bank language over time</cite>. What read as hawkish in 2005 may not carry the same weight in 2020 because the policy regime shifted, the language conventions changed, or the distribution of phrases reweighted. The dictionary must be re-estimated or the model retrained. <cite index="23-5">An automated technique is not able to capture the tone of central bank communication</cite> with full fidelity, but it scales and updates in ways that hand-coding cannot. For market participants who need a daily read, that tradeoff holds.

Sources:

• https://www.researchgate.net/publication/326845441_Transparency_and_Deliberation_Within_the_FOMC_A_Computational_Linguistics_Approach
• https://www.bis.org/publ/work1215.pdf
• https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7861405/
• https://www.ilo.org/media/412551/download

<cite index="13-2,13-3">Lucca and Trebbi applied computational linguistic tools to FOMC statements and measured the effects on the macroeconomy including in a VAR framework</cite>. Later work separated dimensions: <cite index="13-11,13-13">forward guidance has larger effects than views of the current economic situation, and forward guidance especially affects market yields</cite>. <cite index="14-1,14-3">Two factors emerge — a 'current federal funds rate target' factor and a 'future path of policy' factor, with the latter closely associated with FOMC statements and having a much greater impact on longer-term Treasury yields</cite>.

The method combines topic modeling (to isolate themes within statements) and dictionary-based tone measures (to classify hawkish versus dovish language). <cite index="13-5">Both LDA for topic modelling and dictionary methods measure tone</cite>. <cite index="11-1">FOMC statements have become steadily more similar in content from meeting to meeting, particularly since the financial crisis</cite>, which narrows the variance available for extraction but raises the importance of marginal changes in language.

<cite index="16-11">The main methodological contribution is to use computational linguistics, particularly the combination of topic modelling and dictionary methods, to examine the content of what central banks are trying</cite> to communicate. <cite index="13-15">Neither additional dimension has large effects on real economic variables</cite>, but that does not reduce their relevance for asset-price formation. Markets price the path, not the level; statements move duration risk more than they move output expectations. The decomposition matters because it isolates which part of the communication the curve is pricing.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0022199615001828
• https://www.federalreserve.gov/econresdata/notes/feds-notes/2015/semantic-analysis-of-the-FOMCs-postmeeting-statement-20150930.html
• https://www.lse.ac.uk/CFM/assets/pdf/CFM-Discussion-Papers-2015/CFMDP2015-37-Paper.pdf
• https://www.researchgate.net/publication/402480091_Mind_the_Shift_Decoding_Monetary_Policy_Stance_from_FOMC_Statements_with_Large_Language_Models

<cite index="1-1">Penalized linear regression is recommended for many text regression applications</cite>, according to the Gentzkow, Kelly, and Taddy framework that has become the working standard for applied text analysis in economics. The method starts with the simplest representation — <cite index="1-8">best practice in many cases is to begin analysis by focusing on single words</cite> — and moves to bigrams or trigrams only when returns justify cost. <cite index="1-5">In analysis of partisan speech, single words are often insufficient to capture the patterns of interest: 'death tax' and 'tax break' are phrases with strong partisan overtones</cite> that disappear when decomposed.

The economics behind the method is straightforward. Text generates high-dimensional count data. The researcher observes phrase frequencies and wants to predict an outcome (party affiliation, policy stance, market response). Regularization penalizes model complexity to prevent overfitting. <cite index="9-1,9-2,9-3">Preprocessing reduces words to stems, drops procedural language, and restricts to phrases spoken at least 10 times in at least one session</cite>. The result is a sparse matrix where most cells are zero and the model learns which phrases carry signal.

This is the foundation beneath every application that followed — FOMC tone extraction, partisanship measurement, forward-guidance decomposition. The method does not require theory about which words matter. It learns from the data which phrases predict the outcome and weights them accordingly. What you gain in scalability you surrender in interpretability, but for market observers the tradeoff holds: we care whether the model predicts, not whether it narrates.

Sources:

• https://www.nber.org/system/files/working_papers/w23276/w23276.pdf
• https://web.stanford.edu/~gentzkow/research/text-as-data.pdf
• https://scholar.harvard.edu/files/shapiro/files/politext.pdf

<cite index="3-2,3-13">Over the course of 2023, high-yield borrowers reduced their debt coming due in 2024, 2025, and 2026 by 57%, 43%, and 14%, respectively</cite>, showing that the wall can be eroded in advance if the window remains open. <cite index="5-8,5-16">Less than $100bn of high yield debt is set to mature by the end of 2026, or $125bn over the next 24 months, with 80% rated BB or above</cite>.

The standard practitioner approach: identify all debt maturing in each future year, segment by rating and sector, then track how much gets refinanced or extended before the maturity date. The refinancing rate tells you how much of the wall is real versus nominal. A 50% refinancing rate in the year prior to maturity means half the apparent wall has already been addressed.

<cite index="1-7">The corporate debt refinancing maturity wall in 2025 poses a $1.7 trillion risk in the US alone, with $700 billion (41%) in high-yield bonds maturing</cite>. The concentration in HY matters because the refinancing gap widens faster when spreads move. IG can usually refinance through moderate spread widening; HY cannot.

The method does not account for the path of rates or spreads — only the current stock of maturities by cohort. To stress-test the wall, practitioners layer on scenarios: what happens to the refinancing gap if spreads widen 200bp, or if the window closes for sub-IG paper entirely.

Sources:

• https://www.schwab.com/learn/story/will-maturity-wall-matter-investors
• https://www.insightinvestment.com/united-states/perspectives/maturity-walls-pose-low-default-risks-for-high-yield/
• https://sparkco.ai/blog/corporate-debt-refinancing-maturity-wall-crisis

<cite index="7-1,7-2">Loans with debt yields below 8% have consistently shown the highest delinquency and refinance risk, according to Trepp's latest data review</cite>. <cite index="7-8,7-9">Debt yield now drives refinancing outcomes more than maturity volume alone; loans with stronger debt yields refinance more easily</cite>, while weaker loans require extensions or fall into delinquency.

<cite index="7-13">Lower debt yields leave borrowers more exposed to refinancing gaps, especially when property values remain below peak pricing and debt costs stay elevated</cite>. This is the CMBS-market version of rollover-risk measurement: net operating income divided by loan balance. A loan with an 8% debt yield means the property generates 8 cents of income per dollar of debt. If market underwriting now requires 10%, the borrower cannot refinance the full balance without contributing equity or accepting a smaller loan.

<cite index="7-14,7-15,7-16">Trepp introduced a method to evaluate maturity stress by analyzing how loans behave after maturity; some loans fail to pay off but remain current while negotiating extensions, while others become persistently delinquent in what Trepp called "maturity recidivism"</cite>.

This measure is specific to commercial real estate, but the logic applies to any asset-backed or project-financed structure. The debt-yield threshold defines the boundary between refinanceable and stranded. Practitioners track cohorts by vintage and sector, then observe what fraction crosses the threshold as rates move.

Sources:

• https://www.credaily.com/briefs/cmbs-maturity-wall-tests-refinancing-in-2026/

<cite index="22-2,22-3">The leading methodology currently in use to account for name concentration risk relies on the Granularity Adjustment developed by Gordy and Lütkebohmert, originally designed for commercial banks which typically hold portfolios of at least several hundred borrowers</cite>. <cite index="22-5,22-6">The exact measure of name concentration risk is quantified as the difference between the Value-at-Risk of the true portfolio loss variable and the VaR of an analogous asymptotic portfolio where all idiosyncratic risk is diversified away</cite>.

<cite index="22-9,22-10">The GA methodology is an asymptotic approximation that has been shown to be very accurate for medium and larger commercial bank portfolios, but its accuracy decreases with the number of borrowers, which may lead to substantial approximation errors when portfolios of less than one hundred obligors are considered</cite>.

This is the canonical measure used by S&P in capital adequacy frameworks for multilateral development banks and other concentrated portfolios. The logic: a perfectly diversified portfolio loses only systematic risk; what remains is idiosyncratic name risk. The GA approximates that residual analytically, calibrated to single-factor credit models like Vasicek or CreditRisk+.

The method breaks down for small portfolios — fewer than 100 names — because the asymptotic assumption no longer holds. For those portfolios, Monte Carlo simulation with importance sampling is the fallback, though it is slower and requires more careful parameterization.

Sources:

• https://arxiv.org/pdf/2311.13802
• https://www.moodys.com/web/en/us/insights/resources/quantifying-decomposing-and-managing-portfolio-concentration-risk.pdf

<cite index="9-2,9-8">Rollover risk is measured as the ratio of long-term debt to total debt</cite>, a simple measure that denotes how concentrated maturities are in the short term. The less long-term debt a firm holds, the more frequently it must refinance — and the more exposed it is to credit-market conditions at the moment of rollover.

<cite index="15-4,15-5">Firms entering the crisis with higher debt levels reduced investment more after the crisis, with the negative effect stronger for firms holding short-term debt in countries whose banks were weak</cite> due to sovereign stress. <cite index="11-2,11-14">A back-of-an-envelope calculation suggests that debt overhang and rollover risk channels explain about 60% of the decline in aggregate corporate investment during the crisis</cite>.

This is the simplest practitioner metric I have seen cited in the academic literature. It does not require projections or scenario modeling — only balance-sheet data. The insight: a firm with 20% long-term debt has far more rollover risk than one with 80%, all else equal. The measure is used in cross-sectional regressions to isolate the effect of maturity structure on investment and default outcomes, controlling for leverage and sector.

What it does not capture: the timing of maturities within the short-term bucket, or the dispersion of maturities across years. A firm with $500mm due in six months and nothing for two years after that is in a different position than one with $100mm due each quarter for five quarters.

Sources:

• https://www.nber.org/system/files/working_papers/w24555/revisions/w24555.rev1.pdf
• https://academic.oup.com/jeea/article-abstract/20/6/2353/6563882
• https://cepr.org/voxeu/columns/debt-overhang-rollover-risk-and-corporate-investment-evidence-european-crisis

<cite index="12-4">Corporate bond funds experienced aggregate net outflows in March 2020 of over 5% relative to net assets, far greater than in previous stress episodes over the last decade.</cite> <cite index="12-6">The Taper Tantrum led to aggregate monthly outflows of less than 3% and cumulative outflows for the average fund of about 2.2% in June-July 2013, far smaller than the about 10% outflows in February-March 2020.</cite> <cite index="22-7">Within the first quarter of 2020, the mutual fund sector sold off $236 billions of Treasury securities, which contributed to the large volatility and price discounts in Treasury markets.</cite>

The crisis revealed what liquidity cushions hold through. <cite index="24-2">Given 10% net outflows, funds that have insufficient cash levels and hold bonds of illiquidity 3-5 deviations from the mean experience a significant decrease in NAV of about 34-49 basis points.</cite> <cite index="21-17">Using the COVID-19 crisis as a natural experiment, bonds with higher precrisis fragility experienced more negative returns and larger reversals around March 2020.</cite> The observed outflow magnitude was three times the 2008 level, and it exceeded the tolerance of standard cash-buffer strategies. What this episode priced: the assumptions embedded in liquidity-transformation models do not hold at the tail.

Sources:

• https://www.nber.org/system/files/working_papers/w27559/w27559.pdf
• https://www.clevelandfed.org/-/media/project/clevelandfedtenant/clevelandfedsite/events/2020/financial-stability-conference/kairong-xiao-paper.pdf
• https://www.hks.harvard.edu/sites/default/files/centers/mrcbg/files/Wang_final.pdf
• https://ideas.repec.org/a/eee/jfinec/v143y2022i1p277-302.html

<cite index="12-1,12-2">When flow-related sales have price-impact by depressing security prices, they lead to spillovers because the valuation losses hurt the performance of peer funds that hold the same securities; spillovers may then lead to redemptions at peer funds through the performance-flow relationship.</cite> <cite index="12-3">There are sizable fire-sale spillovers in debt markets and spillovers aggravate market instability—volatility—by amplifying the effect of an initial shock to fund flows that is otherwise unrelated to fundamental asset values.</cite>

The feedback mechanism is structural. <cite index="18-1,18-3,18-4">Following a market shock, alert investors anticipate the impact on a fund's net asset value (NAV) of other investors' redemptions and exit first at favorable prices; this first-mover advantage may lead to fund failure through a cycle of falling prices and increasing redemptions.</cite> <cite index="19-3,19-4">Faced with negative shocks to mutual funds, investors in an illiquid bond fund have greater incentives to redeem their shares ahead of others; when a bond is mainly held by illiquid funds, negative shocks can trigger larger outflows, ultimately leading mutual funds to sell bond holdings for nonfundamental reasons.</cite>

Fragility is measurable in the ownership structure. <cite index="21-15,21-16">A novel bond-level latent fragility measure based on asset illiquidity of mutual funds holding the bond shows that corporate bonds bearing higher fragility subsequently experience higher return volatility and more outflows-induced mutual fund selling.</cite> <cite index="16-3,16-4">Mutual funds with large shares of outstanding bond issues are more inclined to internalize the negative price spillovers of fire sales; ownership concentration limits bonds' exposures to flow-induced fire sales.</cite>

Sources:

• https://www.nber.org/system/files/working_papers/w27559/w27559.pdf
• https://pubsonline.informs.org/doi/10.1287/mnsc.2019.3353
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X2100204X
• https://ideas.repec.org/a/eee/jfinec/v143y2022i1p277-302.html
• https://academic.oup.com/rfs/article/37/7/2063/7633431

<cite index="10-1,10-2">Corporate bond mutual funds engage in liquidity transformation, but there is little evidence that bond fund redemptions drive fire sale price pressure after controlling for time-varying issuer-level information, using a strategy that exploits same-issuer bonds held by funds with differing outflows.</cite> <cite index="10-4">Bond funds maintain significant liquidity cushions and selectively trade liquid assets, allowing them to absorb investor redemption risk without excessively liquidating corporate bonds, even during the 2008 financial crisis.</cite>

The mechanism operates through pecking-order liquidation. <cite index="22-6">Mutual funds follow a pecking order of liquidation by first selling their most liquid assets before more illiquid ones in order to minimize the discounts from asset sales.</cite> <cite index="22-2,22-3">As negative signals emerge and redemption requests increase, the fund's pecking order of liquidations leads to more concentrated sales in the more liquid asset; this reverse flight to liquidity is salient only at actively managed open-end funds.</cite>

The cash-buffer position is load-bearing. <cite index="24-9">A fund with 5% cash on-hand has capital available to satisfy outflows less than or equal to 5% without the need for immediate fire sales.</cite> <cite index="20-16,20-17,20-18">The more illiquid fund assets are, the longer the fund will take to accommodate flows; when assets are more illiquid, costs of delay become smaller relative to the price impact of trading, and the value of waiting for offsetting flows increases.</cite>

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0304405X20301549
• https://www.clevelandfed.org/-/media/project/clevelandfedtenant/clevelandfedsite/events/2020/financial-stability-conference/kairong-xiao-paper.pdf
• https://www.hbs.edu/ris/Publication%20Files/2016-01%20Liquidity%20Transformation_6c0f8e71-9c4b-479e-aeb6-8e731569a535.pdf
• https://www.hks.harvard.edu/sites/default/files/centers/mrcbg/files/Wang_final.pdf

<cite index="2-19">Corporate bond funds exhibit a concave flow-to-performance relationship: outflows are sensitive to bad performance more than inflows respond to good performance.</cite> <cite index="4-11,4-12">Funds experiencing large outflows tend to decrease existing positions, creating price pressure in securities held in common by distressed funds, while large inflows prompt position expansion in overlapping holdings.</cite> <cite index="17-3">Funds with illiquid assets exhibit stronger sensitivity of outflows to bad past performance than funds with liquid assets, consistent with strategic complementarities among investors.</cite>

This asymmetry matters because it creates a mechanical link between poor fund returns and forced selling. <cite index="4-5,4-6">Flows into and out of mutual funds force trading; funds in the bottom decile of capital flows are roughly twice as likely to reduce or eliminate holdings as funds experiencing normal flows.</cite> <cite index="24-1">Illiquid bond funds are significantly more sensitive to past performance than liquid funds and experience up to 43.6% more outflows given a 1% decrease in returns.</cite>

The documented relationship is not symmetric. <cite index="1-8,1-1">Investors chase the performance of higher-valuation-accuracy-score funds more aggressively and exhibit a convex flow-performance relation among these funds.</cite> What this pricing tells us: downside flows are not mirrored by upside flows. The forced-sale channel operates in one direction.

Sources:

• https://www.researchgate.net/publication/338849484_Swing_Pricing_for_Mutual_Funds_Breaking_the_Feedback_Loop_Between_Fire_Sales_and_Fund_Redemptions
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X07001158
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X10000759
• https://www.hks.harvard.edu/sites/default/files/centers/mrcbg/files/Wang_final.pdf

<cite index="2-2,2-6,2-7,2-8">Affine term structure models allow us to decompose real and nominal bond yields into components including expectations, term premia, and liquidity; these models adjust for TIPS illiquidity in a transparent way by constructing an index of TIPS liquidity using observable measures and including it as a pricing factor in the model.</cite> The Federal Reserve Bank of New York and San Francisco Fed models are the canonical implementations.

<cite index="13-3">Estimating the model taking TIPS yields at their face value fails to produce plausible estimates of inflation expectations or inflation risk premia.</cite> The adjustment is not optional. <cite index="10-4">Information from the market for inflation swaps provides a range for the possible liquidity premium in TIPS, which in turn suggests a range for estimates of inflation expectations that is well below the widely followed Survey of Professional Forecasters inflation forecast.</cite> Model-implied expectations and survey-based expectations can diverge by 50 basis points or more when liquidity is mispriced.

<cite index="24-2,24-6">Adjusting breakevens for inflation and liquidity risk substantially improves forecasts of inflation.</cite> What the market expects and what the nominal-real spread reports are two different objects. The spread is the composite; the expectation is one of three unobservable components that must be estimated with structure.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0304393216301088
• https://www.federalreserve.gov/pubs/feds/2008/200830/index.html
• https://www.frbsf.org/research-and-insights/publications/economic-letter/2011/06/tips-liquidity-breakeven-inflation-expectations/
• https://www.newyorkfed.org/research/staff_reports/sr570.html

<cite index="17-1">Inflation risk premium is the extra compensation nominal bond investors demand for bearing inflation risks, and its value depends on the covariance between inflation and real economic activity.</cite> The sign has shifted across regimes. <cite index="17-6">This premium is believed to have been positive and sizeable in the 1970s and 1980s, when investors were more worried about stagflation scenarios with higher inflation accompanied by lower growth, but appears to have declined in recent decades to lower or even negative levels, as investors have become more concerned about outcomes where lower inflation is associated with lower growth.</cite>

<cite index="21-1,21-5">A positive inflation risk premium increases the nominal yield on ordinary Treasury securities or, alternatively, decreases the real yield on TIPS (because investors effectively pay insurance to avoid inflation risk); all else equal, a lower TIPS yield leads the breakeven rate of inflation to overshoot realized inflation.</cite> When the premium turns negative—when deflation risk is priced as the tail scenario—the mechanics reverse.

<cite index="24-1">Variations in U.S. nominal term premia are primarily driven by variations in real term premia rather than inflation and liquidity risk premia.</cite> The load-bearing component in the term structure has moved. That fact alone changes what a steepening or flattening curve signals about future realized inflation.

Sources:

• https://www.federalreserve.gov/econres/notes/feds-notes/tips-from-tips-update-and-discussions-20190521.html
• https://www.bls.gov/opub/mlr/2019/article/inflation-expectations-and-inflation-realities.htm
• https://www.newyorkfed.org/research/staff_reports/sr570.html

<cite index="9-10">Models incorporating a TIPS liquidity factor reveal a TIPS liquidity premium that was until recently quite large (~1%) but has come down in recent years, consistent with the common perception that TIPS market grew and liquidity conditions improved.</cite> The early years were distorted. <cite index="13-4,13-5">The difference between the observed TIPS yields and the model-implied real yields estimated without TIPS data indicates that the liquidity premium was quite large in the early years of TIPS's existence, but has become smaller recently; this liquidity premium turns out to be difficult to account for within a simple rational pricing framework, suggesting that TIPS may not have been priced efficiently in their early years.</cite>

<cite index="11-4">The liquidity premium reflects the present value of expected future trading costs as well as compensation for assuming the risk of potentially being forced to sell the bond prematurely at a disadvantageous price.</cite> <cite index="21-7">TIPS are less liquid because they lack the market depth of ordinary Treasury securities, are typically held by buy-and-hold investors seeking an inflation hedge, and, in the early years of their issuance, bore the uncertainty of whether the U.S. Treasury Department would continue to issue inflation-indexed debt.</cite>

The premium distends in stress. <cite index="2-9">The illiquidity component is sizable for TIPS, especially during the financial crisis.</cite> Any breakeven-based inflation signal must adjust for this regime behavior or it will read flight-to-quality as deflation.

Sources:

• https://www.researchgate.net/publication/227437536_TIPS_liquidity_breakeven_inflation_and_inflation_expectations
• https://www.federalreserve.gov/pubs/feds/2008/200830/index.html
• https://www.frbsf.org/research-and-insights/publications/economic-letter/2016/11/tips-liquidity-and-the-outlook-for-inflation/
• https://www.bls.gov/opub/mlr/2019/article/inflation-expectations-and-inflation-realities.htm
• https://www.sciencedirect.com/science/article/abs/pii/S0304393216301088

<cite index="15-1">TIPS inflation compensation can be decomposed into three components: expected inflation, inflation risk premium, and TIPS liquidity premium.</cite> The nominal-real spread is not a clean read of what the market expects inflation to settle at.

<cite index="3-2,3-3">TIPS are generally less liquid than their nominal counterparts, especially in the short run and during periods of market stress, and breakevens also incorporate an inflation risk premium—the compensation that investors require for bearing inflation risk.</cite> <cite index="10-7">Breakeven inflation is governed by two unobservable factors: the premium that bond investors are willing to pay for protection against the risk that inflation will overshoot its expected path and the higher yield they require for holding relatively less liquid TIPS.</cite>

The direction matters. <cite index="10-8,10-9">The inflation risk factor pulls the observed TIPS yields down relative to nominal bonds, causing breakeven inflation to be correspondingly high, while the liquidity factor pushes observed TIPS yields up, bringing breakeven inflation down.</cite> <cite index="16-7">If the inflation risk premium and the liquidity risk premium are equal, they cancel each other and the breakeven rate of inflation approximates realized inflation.</cite> In practice, they do not cancel. The residual is what the market is pricing.

Sources:

• https://www.federalreserve.gov/econres/notes/feds-notes/tips-from-tips-update-and-discussions-20190521.html
• https://www.imf.org/-/media/Files/Publications/gfs-notes/2021/English/GSNEA2021003.ashx
• https://www.frbsf.org/research-and-insights/publications/economic-letter/2011/06/tips-liquidity-breakeven-inflation-expectations/
• https://www.bls.gov/opub/mlr/2019/article/inflation-expectations-and-inflation-realities.htm

<cite index="3-44">Blanco et al. (2004) find that CDS spreads lead bond spreads, while Dötz (2007) shows the reverse</cite>. <cite index="11-5,11-10">CDS premiums often move ahead of the bond market</cite>, but <cite index="11-6,11-11">bond spreads lead CDS premiums for emerging market sovereigns more often than has been found for investment-grade corporate credits, consistent with the CTD option impeding CDS liquidity</cite>.

<cite index="9-24">Relative to bond spreads, CDS spreads tend to reveal new information more rapidly during periods of stress, though not typically at other times</cite>. <cite index="10-7">In normal times the CDS-bond basis is relatively small, but in times of acute global financial distress it exhibits significant fluctuations</cite>. <cite index="10-8">On average, estimates imply significant comovement between the sovereign bond market and derivative markets, with a somewhat more elevated response of CDS spreads relative to cash market spreads at short horizons but no long-lasting differences</cite>.

The evidence is mixed because the question itself is conditional. <cite index="7-12">CDS exhibit a stronger correlation with country-specific fundamental drivers of credit risk than bonds, which correlate positively with a market-wide risk premium</cite>. The instrument that leads depends on where the marginal information resides—public or private, macro or idiosyncratic—and which venue offers deeper liquidity at the moment the information arrives. Neither market is categorically faster; both reflect the credit, and each carries distinct frictions that distort the timing of reflection.

Sources:

• https://www.bis.org/ifc/publ/ifcb34ac.pdf
• https://www.federalreserve.gov/econres/ifdp/sovereign-cds-and-bond-pricing-dynamics-in-emerging-markets-does-the-cheapest-to-deliver-option-matter.htm
• https://www.sec.gov/divisions/riskfin/seminar/gonzalez-hermosillo0413.pdf
• https://www.sciencedirect.com/science/article/pii/S0022199622000356
• https://www.ecb.europa.eu/events/pdf/conferences/150310/04_Fontana.pdf

<cite index="21-1,21-6">A CDS written on a particular reference obligation normally provides coverage for all obligations of the reference entity that have equal or higher seniority</cite>. <cite index="19-4,19-5">Standard CDS contracts specify deliverable obligation characteristics that limit the range of obligations a protection buyer may deliver upon a credit event; trading conventions for deliverable obligation characteristics vary for different markets and CDS contract types</cite>.

<cite index="21-9">Settlement can occur through a cash payment from the credit protection seller to the buyer, as determined by the cheapest-to-deliver obligation of the reference entity, or by physical delivery of the reference obligation from the protection buyer to the protection seller in exchange for the CDS notional</cite>. <cite index="19-14,19-15">When a credit event occurs on a major company on which many CDS contracts are written, an auction may be held to facilitate settlement at a fixed cash settlement price; participating dealers submit prices at which they would buy and sell the reference entity's debt obligations, as well as net requests for physical settlement against par</cite>.

<cite index="23-1,23-5">In physical settlement, the protection buyer delivers to the protection seller an obligation of the reference entity that satisfies pre-agreed criteria—for instance, the reference obligation must be freely transferable with a maximum maturity of no more than 30 years—and the protection seller pays 100% of the face value</cite>. <cite index="23-3,23-7">If at the time of default the reference entity has multiple bonds outstanding, the protection buyer delivers the cheapest of the assets eligible for delivery to maximize their cash flow upon default</cite>.

The auction mechanism has reduced but not eliminated the CTD option's influence. <cite index="25-29">The shift toward cash settlement via ISDA auctions has reduced (but not eliminated) the impact of the CTD option, since auction recovery prices reflect the deliverable basket's market value</cite>. The residual effect still matters: <cite index="25-27,25-28">The protection buyer will rationally deliver the cheapest eligible obligation, maximizing their payout—this "cheapest-to-deliver" option benefits the buyer and can reduce the effective recovery rate for the seller below standard recovery assumptions</cite>.

Sources:

• https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/credit-default-swaps
• https://en.wikipedia.org/wiki/Credit_default_swap
• https://www.chicagofed.org/publications/economic-perspectives/2023/4
• https://ryanoconnellfinance.com/credit-default-swaps/

<cite index="7-1">The CDS-bond basis—defined as the difference between the CDS premium and the yield spread on a fixed-rate bond of similar maturity—should approximate zero under no-arbitrage</cite>. In observed markets, it does not.

<cite index="4-2,4-13">While the basis for Argentina varied from close to zero to significantly positive in credit crises, European countries during the debt crisis experienced negative basis deviations</cite>. <cite index="8-3">Trading frictions and policies in place determine whether the basis is strictly positive or strictly negative</cite>. <cite index="7-15">Short-selling frictions explain positive basis deviations; funding frictions explain negative deviations, especially for countries under stress</cite>.

<cite index="4-27,4-28">The CDS-bond basis, while normally close to zero, increases sharply in crises—arbitrage opportunities become more prevalent in periods of elevated risk</cite>. <cite index="4-25,4-26">Bid-ask spreads increase in default risk for both bonds and CDS, a sign that common liquidity measures break down in stress</cite>. <cite index="8-5,8-10">Trading frictions, investor exposure, default risk, and debt issuance all affect sovereign debt prices, bid-ask spreads, and the CDS-bond basis</cite>. <cite index="8-6,8-11">Increases in default risk generate additional gains from trade, driving up bid-ask spreads and amplifying existing risk premia and CDS-bond basis deviations</cite>.

The basis is a composite signal of market dislocation. <cite index="7-6,7-8">For peripheral euro-area countries, the correlation between the bond spread and the basis is not significant; the increase or decrease in the basis relates to the CDS rather than the bond</cite>. When the derivative and cash markets price the same credit at materially different levels for extended periods, the fault line is not in theory—it is in access, collateral, and the operational cost of holding convergence.

Sources:

• https://www.ecb.europa.eu/events/pdf/conferences/150310/04_Fontana.pdf
• https://www.sciencedirect.com/science/article/abs/pii/S002219962400045X
• https://www.richmondfed.org/-/media/RichmondFedOrg/publications/research/working_papers/2023/wp23-05r.pdf

<cite index="12-1,12-2,12-3,12-4">Standard sovereign CDS contracts do not reference a specific debt instrument. Instead, a variety of dollar-denominated senior obligations qualify for delivery—for some Asian sovereigns, loans or bonds with maturity under 30 years; for others, bonds of any maturity</cite>. <cite index="12-5">After a default event, the protection buyer holds an incentive to deliver the least valuable eligible instrument, even if they had been hedging a different obligation</cite>.

This embedded optionality creates mechanical distance between CDS and bond spreads. <cite index="11-3,11-8">CDS premiums tend to move more than one-for-one with yield spreads—behavior broadly consistent with the presence of a significant cheapest-to-deliver (CTD) option</cite>. <cite index="11-9">Cross-sectional evidence confirms this option is priced into CDS premiums</cite>.

The option's value fluctuates with default risk and the dispersion of bond prices across a sovereign's debt structure. <cite index="13-5">During the 2023 U.S. debt ceiling episode, the cheapest deliverable remained consistently the May 2050 30-year Treasury—issued near par in May 2020 with a 1.25% coupon</cite>. <cite index="16-2,16-3">As of May 3, 2023, this bond traded around 58, driven sharply lower by rate increases</cite>. When rate cycles distend the term structure and compress prices on older, low-coupon paper, the CTD option widens and CDS premiums price accordingly—often divorced from pure credit deterioration.

<cite index="14-2,14-13">The CTD option matters to hedge ratios and pricing relationships between bonds and CDS</cite>. Treating CDS as a clean synthetic of a specific bond is a category error; the derivative carries optionality the cash instrument does not.

Sources:

• https://www.federalreserve.gov/pubs/ifdp/2007/912/ifdp912.htm
• https://www.federalreserve.gov/econres/ifdp/sovereign-cds-and-bond-pricing-dynamics-in-emerging-markets-does-the-cheapest-to-deliver-option-matter.htm
• https://www.chicagofed.org/publications/economic-perspectives/2023/4
• https://www.msci.com/www/blog-posts/the-cds-market-s-view-on-us/03820087801
• https://www.sciencedirect.com/science/article/abs/pii/S1042443111000023

<cite index="3-5,3-6">An extreme illiquidity measure is constructed using the five percent tail-risk rule commonly adopted in risk management; given an illiquidity index such as the Amihud (2002) measure, an illiquidity level is selected at the 95th percentile as the threshold of extreme illiquidity</cite>. This is not the routine transaction cost of holding illiquid paper. It is the cost of exiting when liquidity deteriorates sharply—a negative skew embedded in the return distribution that mean illiquidity measures do not observe.

<cite index="3-1">Extreme illiquidity and standard illiquidity capture different dimensions: standard illiquidity measures the average routine transaction cost, whereas extreme illiquidity captures the unusual trading cost arising from severe deterioration in liquidity</cite>. <cite index="3-11">Corporate bonds carry an extreme illiquidity premium</cite>, and it prices separately from the Amihud measure, the Roll covariance, and zero-trading days.

The premium reflects a market that trades in two regimes: normal friction and dislocation. <cite index="4-1">Illiquidity is priced in the time-series of corporate bond returns mainly in periods of financial distress</cite>. The compensation is not for the spread you pay on average—it is for the spread you pay when you are forced to transact and no one else wants to. <cite index="18-4">With fewer than fifty percent of securities in the credit markets trading in any given year and fewer than one percent trading every day, a significant amount of liquidity risk is widely thought to be in the credit markets</cite>. That tail is priced.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S1386418124000132
• https://cfr.ivo-welch.info/published/papers/amihud2020illiquidity.pdf
• https://rpc.cfainstitute.org/research/cfa-digest/2017/01/an-index-based-measure-of-liquidity-digest-summary

The cleanest measurement of the illiquidity premium is the yield spread between two instruments with identical credit risk, maturity, and structure, where one trades liquid and the other does not. <cite index="20-1,20-2">The formula involves subtracting the yield on the liquid asset from the yield on the illiquid asset; this difference reflects the liquidity premium</cite>. <cite index="20-16,20-17">For example, a ten-year Treasury bond with a yield of 3% compared to a less liquid corporate bond with the same maturity and similar credit rating offering a yield of 3.5% produces a liquidity premium of 0.5%</cite>.

The difficulty is isolating liquidity from credit, duration, and embedded optionality. <cite index="24-14,24-15,24-16">There is difficulty in isolating the illiquidity premium from other risk premia; an asset will contain various risks that deserve to be rewarded, and corporate bonds are exposed to duration, inflation, and credit risk, creating a challenge in determining which part of the overall return is associated with each risk</cite>.

<cite index="24-5,24-6">The liquidity level premium before the financial crisis was four basis points for investment-grade and 58 basis points for high yield; after the crisis, these premiums went up to 40 to 90 basis points for investment-grade securities and 200 basis points for high-yield bonds</cite>. The crisis premium reflects not just wider spreads but re-pricing of the exit option embedded in holding illiquid paper when dealer balance sheets contract. <cite index="23-3,23-4">Long-term illiquidity costs are estimated around 200 basis points a year, including the cost of lost alpha from active management, the cost of the inability to rebalance portfolios, and the cost of liquidity shortfalls for unexpected liquidity needs</cite>.

Sources:

• https://smartasset.com/investing/liquidity-premium
• https://analystprep.com/study-notes/frm/illiquid-assets/
• https://www.pimco.com/us/en/insights/navigating-public-and-private-credit-markets-liquidity-risk-and-return-potential

<cite index="11-1,11-2">Roll's estimator uses the serial correlation of changes in transaction prices to deduce, thanks to the bid-ask bounce effect, the value of the bid-ask spread (which is a multiple of the square root of the opposite of the price change serial correlation)</cite>. The idea: when trades alternate between customer buy orders (hitting the ask) and customer sell orders (hitting the bid), prices bounce mechanically, and the covariance of consecutive price changes turns negative—even when the efficient mid-price follows a random walk.

<cite index="12-1,12-3">Roll (1984) shows that when transaction prices bounce between bid and ask prices, their changes exhibit negative autocovariance even when the underlying value follows a random walk, but quoted bid-ask spreads can only generate a tiny fraction of the empirically observed autocovariance in corporate bonds</cite>. <cite index="12-4,12-5">Quoted spreads are mostly indicative rather than binding, and the structure of the corporate bond market is mostly over-the-counter, making it even more difficult to estimate the actual bid-ask spreads</cite>.

The Roll measure is observable without trade-sign data, which matters in OTC credit markets where direction is often latent. <cite index="14-2,14-3">Transaction costs are clearly priced when proxied with bid-ask spreads, and the Amihud measure has positive regression coefficients across all ratings pre- and post-subprime, with six out of ten statistically significant</cite>. The spread you infer from serial correlation is not the spread you could trade inside of, but it prices nonetheless.

Sources:

• https://www.amf-france.org/sites/institutionnel/files/contenu_simple/lettre_ou_cahier/cahier_scientifique/Measuring%20liquidity%20on%20the%20corporate%20bond%20market.pdf
• https://www.mit.edu/~junpan/bond_liquidity.pdf
• https://www.icmagroup.org/assets/documents/Regulatory/Secondary-markets/Bond-Market-Liquidity-Library/Dick-Nielsen-J_Feldhutter-P_Lando-D---2011---Corporate-bond-liquidity-before-and-after-the-onset-of-the-subprime-crisis-290118.pdf

<cite index="2-5">The Amihud illiquidity measure is the six-month rolling median of the ratio of the absolute value of daily returns (in basis points) to daily trading volume (in millions)</cite>. The construction is simple: when volume thins, the ratio expands, and the instrument is marked more illiquid. <cite index="10-1,10-2">The measure has a simple construction that uses the absolute value of the daily return-to-volume ratio to capture price impact, and it has a strong positive relation with expected stock return</cite>.

What the measure prices, however, is not the return-to-volume construct itself. <cite index="10-5">The pricing of the Amihud measure is driven by its trading volume component, not by its construct of return-to-volume ratio</cite>. This matters because volume is observable and contractible; price impact is not. The instrument rewards you for holding paper that trades thin, whether or not thin trading produces realized slippage in your exit.

<cite index="1-9">Bao, Pan, and Wang (2011) argue that transitory prices arise from a lack of liquidity and construct an illiquidity measure for corporate bonds using the negative autocovariance of relative price changes</cite>—an alternative that captures the round-trip cost embedded in bid-ask bounce rather than the volume drought itself. <cite index="3-6,3-7">Extreme illiquidity captures the unusual trading cost arising from severe deterioration in liquidity, and its predictive power in the cross-section for future bond returns holds up to a one-year horizon, robust to controlling for the standard Amihud illiquidity measure and other commonly used measures of bond illiquidity</cite>.

Sources:

• https://www.aqr.com/-/media/AQR/Documents/Journal-Articles/IlLiquidity-Premium-in-Credit-Markets-JFIWint19.pdf?sc_lang=en
• https://www.icmagroup.org/assets/documents/Regulatory/Secondary-markets/Bond-Market-Liquidity-Library/Lou-X_Shu-T---2016---Price-Impact-or-Trading-Volume-Why-is-the-Amihud-2002-Illiquidity-Measure-Priced-290118.pdf
• https://www.sciencedirect.com/science/article/abs/pii/S1386418124000132

<cite index="12-7,12-9">Upon the declaration of a credit event by the ISDA Determinations Committee, the index will be reversioned, and trading in the new index version will commence. The initial issuance is version 1 (e.g. iTraxx Europe Series 19 Version 1), and the version is incremented for each name in the index that has defaulted.</cite> The defaulted name is removed; the index continues with N-1 constituents at the original coupon.

<cite index="9-4,9-5">Upon confirmation of a credit event for a reference entity within the index basket, the index responds through a reversion mechanism rather than full termination. The defaulted entity's weight is set to zero and removed from the composition, with its notional portion eliminated from the overall ind</cite>[ex].

This creates a known pricing artifact. A 124-name index with one default is not equivalent to a fresh 124-name index. The coupon was set for 125 names; the remaining names now carry that coupon without the defaulted spread contribution. If the defaulted name was trading wide, the surviving index will appear rich relative to its intrinsic spread. If the default was a surprise from a tight name, the opposite holds.

Market participants price this. The reversioned index (e.g., Series 34 Version 2) trades at a different spread than Version 1 even when both reference the same surviving names, because the embedded carry and default history differ.

Sources:

• https://en.wikipedia.org/wiki/Credit_default_swap_index
• https://grokipedia.com/page/Credit_default_swap_index

<cite index="14-1">CDS indices are typically equally weighted baskets consisting of single-name credit default swaps.</cite> CDX.NA.IG holds 125 names at 0.8% each; CDX.NA.HY holds 100 names at 1% each. This is presented as neutrality but it is not.

Equal notional weight does not mean equal risk contribution, equal liquidity, or equal pricing efficiency. A 0.8% position in a name trading 5bp wide with $50mm daily volume behaves very differently than a 0.8% position in a name trading 15bp wide on $2mm volume. The first will price continuously; the second will gap on volatility and settle late in credit-event auctions.

The construction methodology addresses this obliquely. <cite index="15-8,15-9">The CDS contracts referencing the constituent companies must be actively traded in the secondary market to ensure that the index is liquid and tradable. Liquidity is essential for investors who want to buy or sell protection on the index, as it allows them to execute their trades quickly and efficiently without significantly impacting the pri</cite>[ce]. But once a name is in, it stays in at equal weight until the next roll, regardless of whether its liquidity holds or deteriorates.

What you observe: indices trade tighter than their least-liquid constituents and wider than a true liquidity-weighted basket would imply.

Sources:

• https://www.fe.training/free-resources/financial-markets/credit-default-swap-indices/
• https://utatlan.minegocio-go.com/global-register/oscmarkets-cdx-index-a-detailed-list-of-constituents-1767648818

<cite index="10-1,10-2">Every March and September the composition of the basket of a certain index CDS is redefined (rolled) according to certain rules (e.g. minimum amount of debt securities outstanding and liquidity of the CDS of the single reference entities).</cite> <cite index="11-11">To maintain a liquid and representative basket, a new composition of the index is introduced</cite> as a new series, while the prior series continues to trade with its original, now-stale composition.

<cite index="11-1">The composition of the index remains unchanged throughout its lifetime, except when a credit event is triggered for one of the constituents, in which case that constituent is removed from the index without replacement and is settled separately.</cite> Between rolls, there is no rebalancing. A name downgraded from IG to HY in April remains in CDX.NA.IG until the September roll. A name whose CDS liquidity evaporates in May continues to weight equally in the index until the next series launches.

This introduces path dependency. The on-the-run series prices the current liquidity set; off-the-run series become proxies for older, often-degraded credit cohorts. Spread differentials between series can persist when the roll brings in materially different names or when a credit event cluster has hollowed out an older vintage.

Sources:

• https://www.iosco.org/library/pubdocs/pdf/ioscopd385.pdf
• https://libertystreeteconomics.newyorkfed.org/2015/03/the-effects-of-entering-and-exiting-a-credit-default-swap-index/

<cite index="1-1">All iTraxx indices follow transparent, rules-based construction centered on underlying CDS liquidity, the highest-priority eligibility criterion.</cite> CDX employs the same principle. <cite index="18-1">Our selection methodology for CDX ensures that the indices represent the market's most liquid segment.</cite>

This creates a structural bias. <cite index="19-7">The entities that are in the highest 50% of the IG Liquidity Rankings and are not already included in the IG Index (the "Liquidity Based Inclusions").</cite> Liquidity is measured primarily through DTCC-reported CDS trade volumes, though the methodology allows fallback to cash-bond liquidity metrics when CDS volumes are thin. <cite index="19-11,19-12,19-13">If there are less Roll Inclusions than Roll Exclusions identified after the CDS Liquidity Based Inclusion criteria has been applied, a list of entities will be identified from the most recently rebalanced Markit iBoxx USD Liquid Investment Grade index. Names will be added until the index reaches a constituent level of 125 entities, taking into account both Roll Inclusions and Roll Exclusions.</cite>

What this means: the benchmark reflects what dealers are willing to warehouse and what the market already trades, not necessarily the broad credit universe an allocator actually holds. The lag between a name losing liquidity and its removal can persist for one full roll cycle—six months.

Sources:

• https://www.marketopia.org/blog/itraxx/
• https://ihsmarkit.com/products/markit-cdx.html
• https://www.spglobal.com/spdji/en/documents/methodologies/Markit%20CDX%20HY%20and%20IG%20Rules%20Aug2021.pdf

<cite index="6-8,7-1,7-3">Structural vector autoregressions (VARs) are widely used to trace out the effect of monetary policy innovations on the economy.</cite> <cite index="7-4,7-5,7-6">However, the sparse information sets typically used in these empirical models lead to at least two potential problems with the results: first, to the extent that central banks and the private sector have information not reflected in the VAR, the measurement of policy innovations is likely to be contaminated; a second problem is that impulse responses can be observed only for the included variables, which generally constitute only a small subset of the variables that the researcher and policymaker care about.</cite>

<cite index="6-6,6-7,6-12,6-13">A factor-augmented vector autoregression (FAVAR) introduced a method for incorporating a broad range of conditioning information, summarized by a small number of factors, in otherwise standard VAR analyses; the paper shows how to identify and estimate a FAVAR by both a two-step method based on estimation of principal components and a more computationally demanding, Bayesian method based on Gibbs sampling.</cite>

The FAVAR approach handles the dimensionality problem: rather than choosing which three or four macro series to include in the VAR and pricing only their responses, the econometrician extracts common factors from a large cross-section of series and conditions the policy shock on those factors. This widens the observable impulse-response surface and disciplines the shock identification with the information set the central bank itself observes.

Sources:

• https://www.federalreserve.gov/pubs/feds/2004/200403/200403pap.pdf
• https://www.nber.org/papers/w10220
• https://www.federalreserve.gov/econres/feds/measuring-the-effects-of-monetary-policy-a-factor-augmented-vector-autoregressive-favar-approach.htm

<cite index="2-2">Early VARs showed an odd result: inflation tended to increase following monetary policy tightening.</cite> <cite index="2-3,2-4,2-5">One explanation (Sims, 1992) was that the Fed was looking forward when it set interest rates and that simple VARs omitted variables that could be used to predict future inflation; when these omitted variables intimated an increase in inflation, the Fed tended to increase interest rates, so these VAR interest rate shocks presaged inflation increases the Fed already anticipated.</cite>

<cite index="21-5">The (spurious) appearance of a price puzzle is predicted by a popular class of models, if the econometrician follows the common and seemingly innocent practice of not including a measure of output gap (or potential output) in the VAR.</cite> <cite index="18-1">The omission in the VARs of a variable capturing expected inflation is found to account for the price puzzle observed in simulated and actual data.</cite> <cite index="24-4,24-5">Using actual data and two identification strategies, the positive response of prices to a monetary policy shock is historically limited to the sub-samples that are typically associated with a weak interest rate response to inflation; using pseudo data generated by a sticky price model of the US economy, structural VARs are capable of reproducing the price puzzle only when monetary policy is passive.</cite>

<cite index="19-4,19-5">The puzzles disappear or are mitigated when forward-looking variables are included; when survey-based expectations from professional forecasters, consumers, and firms (as proxies for central bank forecasts) are incorporated into the model, the empirical results align with theoretical predictions of monetary policy transmission.</cite>

Sources:

• https://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.15.4.101
• https://www.sciencedirect.com/science/article/abs/pii/S0304393204000686
• https://ideas.repec.org/p/pad/wpaper/0101.html
• https://academic.oup.com/ej/article/120/549/1262/5089699
• https://www.sciencedirect.com/science/article/abs/pii/S0014292125000078

<cite index="13-1,13-2,13-3">SVAR is the structural extension of the traditional VAR model; it departs from the reduced-form VAR by imposing structural restrictions that are derived from economic theory or institutional knowledge, helping identify structural shocks which are economically meaningful disturbances—such as monetary policy shocks or supply-side shocks.</cite>

<cite index="11-2,11-3">Following Sims (1980) and Sims (1986), many empirical works have exploited the Choleski decomposition of the covariance matrix of the VAR forecast errors, providing an orthogonalization of the residuals by imposing a recursive scheme on the contemporaneous causal structure, implicit in the ordering of the endogenous variables.</cite> <cite index="11-4">The decision on the sequence of the variables is of crucial importance but sometimes loosely motivated.</cite>

<cite index="12-3,12-4">Researchers use two types of identifying restrictions in structural VARs: Blanchard and Quah (1989), Gali (1999), and others exploit the implications that many models have for the long-run effects of shocks, while other authors exploit short-run restrictions.</cite> <cite index="12-5">Structural VARs perform remarkably well when identification is based on short-run restrictions.</cite> <cite index="14-5,14-6,14-7">One approach is to exploit detailed institutional knowledge; an example of this is Blanchard and Perotti's (1999) study of the macroeconomic effects of fiscal policy, which argues that the tax code and spending rules impose tight constraints on the way that taxes and spending vary within the quarter, using these constraints to identify exogenous shocks in taxes and spending necessary for causal analysis.</cite>

Sources:

• https://www.rohanbyanjankar.com.np/2025/07/a-primer-on-var-svar-and-local.html
• https://www.sciencedirect.com/science/article/abs/pii/S0165188922002342
• https://www.federalreserve.gov/pubs/ifdp/2006/866/ifdp866.pdf
• https://faculty.washington.edu/ezivot/econ584/stck_watson_var.pdf

<cite index="1-3">Christopher Sims introduced the Vector Autoregression (VAR) framework as a methodological breakthrough that allowed economists to analyze the dynamic interactions among macroeconomic variables without imposing overly rigid theoretical assumptions.</cite> <cite index="11-9,11-10">Since the seminal work of Sims (1980), the study of the joint dynamics of the main macroeconomic aggregates has been conducted in the framework of vector autoregressive (VAR) models, proposed as an alternative to large simultaneous equation models criticized for their large number of identifying and arbitrary restrictions.</cite>

The method works because most macro variables are endogenous — determined together within the system, not in isolation. <cite index="1-6">Impulse Response Functions (IRFs) trace the effect of a shock in one variable — such as an unexpected increase in policy interest rates — on other macroeconomic variables over time.</cite> <cite index="3-1">Impulse responses allow us to examine the reaction of the economy, going forward, to a hypothetical change in one of the structural shocks.</cite>

<cite index="1-10,1-11">VAR models have had a profound impact on the practice of central banking, used by institutions such as the Federal Reserve, European Central Bank, Reserve Bank of India and Bank for International Settlements.</cite> <cite index="1-4">His research transformed both academic macroeconomics and the practical conduct of monetary policy, equipping central banks with powerful tools to identify economic shocks, understand transmission mechanisms, and evaluate policy interventions.</cite> The methodology gave practitioners a disciplined way to price dynamic causal effects without smuggling in structural assumptions the data could not discipline.

Sources:

• https://sunandoroy.org/2026/03/15/vector-autoregression-var-a-model-that-transformed-the-practice-of-monetary-policymaking/
• https://mcherculano.github.io/docs/Primer_on_VARs.html
• https://www.sciencedirect.com/science/article/abs/pii/S0261560611000982

<cite index="6-3,6-5">Derivative markets provide monetary authorities with a rich source of information for gauging market sentiment; higher moments of future asset values can be extracted in the form of a risk-neutral probability distribution of the underlying asset price at the maturity date of the options.</cite> <cite index="21-12,21-13">RNDs are used by central banks to assess market expectations regarding future stock prices, commodity prices, interest rates, and exchange rates in connection with setting monetary policy.</cite>

<cite index="19-1,19-2">The Black-Scholes option pricing model assumes stock prices are lognormally distributed; while the assumption is reasonable, it tends to underestimate the probability of extremely large stock price movements, which have been empirically observed.</cite> <cite index="19-3,19-4">As a result, option traders assign unique volatilities to options of different strikes, generating a volatility surface across strike and time, which creates an implied distribution providing meaningful insight into the market's expectations for future stock price outcomes.</cite>

<cite index="13-2,13-3">Accurate measurement of risk-neutral quantities is essential given the substantial notional amounts traded in derivatives; measurement error in these quantities can distort inference about the informational content of option prices and their predictive power for future market outcomes.</cite>

What matters: the distribution you extract is the one the market must price to avoid arbitrage, not the one participants expect to observe. The distance between those two is the risk premium, which is unobservable without a model of the physical measure. Any claim about "market expectations" derived solely from option prices is a claim about risk-neutral expectations—what the market prices, not what it believes will settle.

Sources:

• https://www.bankofengland.co.uk/working-paper/1997/implied-risk-neutral-probability-density-functions-from-option-prices
• https://arxiv.org/pdf/physics/0607240
• https://www.globalcapital.com/article/28mwrr0lap7g5gyuamznl/derivatives/option-prices-imply-a-probability-distribution
• https://arxiv.org/pdf/2601.14852

<cite index="17-1,17-2">To combat errors, one can interpolate the implied volatilities of options instead of their respective prices; direct interpolation of prices tends to produce errors in the convexity of the price function which magnify in the extraction of the PDF.</cite> <cite index="21-1">Extracting the risk-neutral density function from option prices is well defined in principle, but is very sensitive to errors in practice.</cite>

<cite index="3-8,3-9,3-10">The Black-Scholes model assumes the probability distribution of the underlying asset at any given future time is lognormal; this assumption is not made by traders, who assume that the probability distribution of an equity price has a heavier left tail and a less heavy right tail than the lognormal distribution.</cite> <cite index="3-12,3-13">Traders use the volatility smile to allow for non-lognormality; the volatility smile defines the relationship between the implied volatility of an option and its strike price.</cite>

<cite index="18-1,18-2">Using the trading volume of each option as a proxy of the informativeness of the option and not requiring the implied probability distribution to recover exactly the market prices of the options allows weighting each option by a function of its trading volume.</cite> <cite index="21-5,21-6">The aim is to obtain a continuous, smooth, monotonic, and convex pricing function that is twice differentiable, reducing irregularities such as negative probabilities that afflict many existing RND estimation techniques.</cite>

Practical note: <cite index="5-3">It remains challenging to distill a realistic estimate for the RND from empirical data.</cite> The sensitivity of the second derivative to noise in the input is the load-bearing problem.

Sources:

• https://github.com/Thrri/option-implied-PDFs
• https://arxiv.org/pdf/physics/0607240
• https://medium.com/@polanitzer/how-to-derive-the-implied-risk-neutral-probability-distribution-of-an-underlying-asset-price-from-286dc5a6a814
• https://www.sciencedirect.com/science/article/abs/pii/S0096300315001642
• https://ideas.repec.org/p/ihs/ihsesp/104.html

<cite index="10-8">The physical measure represents the real-world probabilities of various financial events.</cite> <cite index="9-12">Under the risk-neutral measure, all assets earn the risk-free rate of return in a frictionless market.</cite> <cite index="10-1,10-2">The Radon-Nikodym derivative, also known as the likelihood ratio, is often used to convert the physical measure into the risk-neutral measure.</cite>

<cite index="12-1,12-2">The change of measure from the physical to the risk-neutral measure induces an additive shift in the score function, which translates into a closed-form risk-neutral epsilon shift in the reverse dynamics—this correction enforces the risk-neutral drift while preserving the learned variance and higher-order structure.</cite> <cite index="12-6,12-7">Machine learning approaches to pricing and hedging are typically trained on historical data under the physical measure, so incorporating the risk-neutral structure into such models requires either calibration to option prices or the imposition of additional constraints.</cite>

<cite index="15-1,15-2">Under simple assumptions such as constraints of Put-Call Parity, the probability measure for valuing a European option has the mean derived from the forward price, which can but does not have to be the risk-neutral one; the heuristics used by traders for centuries are more robust and more rigorous than held in the economics literature.</cite> This matters for anyone attempting to extract beliefs from prices: the conversion is not mechanical and the assumptions embedded in it—completeness, no arbitrage, continuous hedging—do not hold uniformly across strikes or maturities.

Sources:

• https://fastercapital.com/keyword/physical-measure.html
• https://fastercapital.com/content/Change-of-Measure-Techniques--Enhancing-Risk-Neutral-Models-update.html
• https://arxiv.org/pdf/2603.20582
• https://arxiv.org/pdf/1405.2609

<cite index="2-4,2-5">Option prices do not reflect the market's view of the distribution of an asset's price at expiry; this is a misunderstanding of the difference between the risk-neutral measure used to price derivatives and the real-world measure that determines actual asset price distributions.</cite> <cite index="1-1,1-3">The implied probability distribution is an approximate risk-neutral distribution derived from traded option prices using an interpolated volatility surface.</cite> <cite index="1-4">In a risk-neutral world, the fair price for exposure to a given event is the payoff if that event occurs, times the probability of it occurring.</cite>

<cite index="3-1,3-2">Breeden and Litzenberger (1978) showed that risk-neutral probability distributions can be estimated from volatility smiles.</cite> The canonical method: <cite index="1-6">buy exposure to a specific range of stock price outcomes with a butterfly spread—long one low-strike call, short two higher-strike calls, long one call at an even higher strike.</cite> <cite index="3-4">The value of a butterfly spread with strikes K - δ, K, and K + δ is c₁ + c₃ - 2c₂.</cite>

<cite index="4-1,4-2">Risk-neutral densities are not real-world probabilities—they are correct in a world indifferent to risk.</cite> <cite index="14-2">In a complete market, a derivative's price is the discounted expected value of the future payoff under the unique risk-neutral measure.</cite> <cite index="14-3">Such a measure exists if and only if the market is arbitrage-free.</cite> What this means: the distribution extracted from option prices tells you what must be priced to prevent arbitrage, not what traders expect to occur.

Sources:

• https://www.morganstanley.com/content/dam/msdotcom/en/assets/pdfs/Options_Probabilities_Exhibit_Link.pdf
• https://freeportlogbook.substack.com/p/options-implied-distributions-are
• https://medium.com/@polanitzer/how-to-derive-the-implied-risk-neutral-probability-distribution-of-an-underlying-asset-price-from-286dc5a6a814
• https://antonismolski.medium.com/options-implied-probability-a-dive-into-risk-neutral-densities-4bef5280842f
• https://en.wikipedia.org/wiki/Risk-neutral_measure

<cite index="22-1,22-2,22-3">A. Craig MacKinlay published "Event Studies in Economics and Finance" in the Journal of Economic Literature, volume 35, issue 1, pages 13-39, in March 1997.</cite> <cite index="14-13,14-29">Roll (1969) introduced the methodology that is essentially still in use today.</cite> <cite index="14-7,14-8,14-9">Event studies have a long history; perhaps the first published study is Dolley (1933), who examined the price effects of stock splits, studying nominal price changes at the time of the split using a sample of 95 splits.</cite>

<cite index="8-1,8-2">Two early papers that cover a wide range of issues are by Brown and Warner (1980, 1985). More recently, an excellent chapter in the textbook of Campbell, Lo, and MacKinlay (1997) is a careful and broad outline of key research design issues.</cite> <cite index="9-4,9-10,9-11">Campbell, Lo, and MacKinlay (1997) published "Event-Study Analysis" in The Econometrics of Financial Markets, pages 149-180, Princeton University Press.</cite>

<cite index="7-2,7-3,7-4">Event studies have important implications and are used extensively in Finance and other disciplines. In an event study stock market returns are used to gauge the reaction of investors to releases of new information, with many studies employing OLS to estimate results.</cite>

Sources:

• https://ideas.repec.org/a/aea/jeclit/v35y1997i1p13-39.html
• https://www.nrc.gov/docs/ML1208/ML12088A329.pdf
• http://mba.tuck.dartmouth.edu/bespeneckbo/cfhandbook/VOLUME_1/CH1-EVENT/ECKBO-CH1-EVENT.doc
• https://www.degruyterbrill.com/document/doi/10.1515/9781400830213-008/html
• https://jds-online.org/journal/JDS/article/1184/file/pdf

<cite index="24-4">The abnormal return observations must be aggregated for the event window and across observations of the event.</cite> <cite index="2-13">The focus on mean effects—the first moment of the return distribution—makes sense if one wants to understand whether the event is, on average, associated with a change in security holder wealth, and if one is testing economic models and alternative hypotheses that predict the sign of the average effect.</cite>

<cite index="2-5,2-6,2-7">To address bias in standard deviation estimates, the significance of event-period average abnormal return can be gauged using the variability of the time series of event portfolio returns in the period preceding or after the event date. For example, the researcher can construct a portfolio of event firms and obtain a time series of daily abnormal returns for a number of days around the event date, with the standard deviation of portfolio returns used to assess significance.</cite>

<cite index="8-3,8-4,8-5,8-6">Event study tests are joint tests of whether abnormal returns are zero and whether the assumed model of expected returns (the CAPM, market model, etc.) is correct. An additional set of assumptions concerning the statistical properties of the abnormal return measures must also be correct.</cite>

Sources:

• https://www.scribd.com/document/453617518/MacKinlay-1997-pdf
• https://www.jufinance.com/mag/dba5/event_study_chapter_1_2008_vol_1.pdf
• http://mba.tuck.dartmouth.edu/bespeneckbo/cfhandbook/VOLUME_1/CH1-EVENT/ECKBO-CH1-EVENT.doc

<cite index="21-14,21-15,21-16">An event study measures the impact of a specific event on the value of a firm using financial market data. The usefulness comes from the fact that, given rationality in the marketplace, the effects of an event will be reflected immediately in security prices, allowing construction of a measure using security prices observed over a relatively short time period.</cite>

<cite index="1-2">Deducting normal returns from actual returns gives abnormal returns, which are the metrics of interest.</cite> <cite index="1-4,1-5">Short-horizon event studies are more reliable than long-horizon studies, which have many limitations, though Kothari and Warner (2005) refined long-horizon methodologies to improve design and reliability over longer periods.</cite>

<cite index="6-3,6-4">The event window can be expanded longer if there are theoretical reasons to justify leakage or dissipation of information over a relatively long period, and it is standard procedure to use alternative event windows for robustness tests.</cite> <cite index="6-1">When event windows overlap or are the same, the abnormal returns of sample firms are potentially correlated, which may result in non-zero covariance among abnormal returns.</cite>

Sources:

• https://www.studocu.com/en-gb/document/university-of-manchester/empirical-finance-only-available-to-finance-specialists/mackinley-1997journalof-economic-literature/8710804
• https://en.wikipedia.org/wiki/Event_study
• https://livrepository.liverpool.ac.uk/3020086/1/A%20review%20of%20short-term%20event%20studies%20in%20OSCM.pdf

<cite index="1-7,1-8">The market model is the most common specification for normal returns in event study methodology, typically using a 120-day estimation window prior to the event to derive the relationship between firm stock and a reference index through regression analysis.</cite> <cite index="5-7,5-8">Under general conditions ordinary least squares is a consistent estimation procedure for the market model parameters, and given standard assumptions, OLS is efficient.</cite>

<cite index="3-9,3-10,3-11">The conditional variance of abnormal returns has two components: the disturbance variance from the market model and additional variance due to sampling error in the regression coefficients, which also leads to serial correlation of abnormal returns despite independent disturbances through time. As the estimation window length becomes large, the sampling error vanishes.</cite>

<cite index="3-5,3-6,3-7">In situations with limited data or when a pre-event estimation period is not feasible, the market-adjusted return model is used. This can be viewed as a restricted market model with alpha constrained to zero and beta constrained to one.</cite> <cite index="3-2,3-3">Variance reduction from the market model is typically greatest when sample firms share a common characteristic—all members of one industry or concentrated in one market capitalization group—in which cases multifactor models warrant consideration.</cite>

Sources:

• https://en.wikipedia.org/wiki/Event_study
• https://www.studocu.com/en-gb/document/university-of-manchester/empirical-finance-only-available-to-finance-specialists/mackinley-1997journalof-economic-literature/8710804
• https://www.scribd.com/document/453617518/MacKinlay-1997-pdf

<cite index="24-3,24-4,24-5">Term structure models such as Hull White began to move from 1-factor models based on the single short rate to assuming there were two and maybe even three risk factors associated with a yield curve; in market risk, variations on techniques such as parametric and Monte Carlo VaR were developed to take advantage of PCA, and a yield curve could be shocked by its PCs when estimating the valuation impact of its movements under different scenarios.</cite>

The practical implication: a portfolio hedged only for duration is exposed to slope and curvature risk. Multi-factor immunization constructs hedges that neutralize exposure to each of the first three principal components separately. <cite index="23-9,23-10">The traditional duration management method assumes that only one risk factor affects the change in the term structure of the interest rate and restricts the term structure to move parallel in one direction, but in fact the interest rate term structure has level and curvature while the risk is being evaluated; there are three potential uncertainties that influence the intertemporal variation of the interest term structure, which can be interpreted as slope, level and curvature.</cite>

VaR models apply shocks to each principal component scaled by historical volatility, then reprice the portfolio under the shocked curve. The covariance structure is simplified: the principal components are orthogonal by construction, so correlation between factors is zero. The dimension reduction—seventeen maturities collapsed to three factors—makes scenario generation tractable without sacrificing explanatory power.

Sources:

• https://www.thegoldensource.com/pca-and-the-term-structure/
• https://arxiv.org/pdf/1911.07288

<cite index="22-1,22-2">The first component, the level factor, determines the overall average level of interest rates; an upward shift in the level factor usually leads to an increase in all interest rates.</cite> <cite index="22-9,22-10">The slope factor represents the difference between short and long-term rates and impacts the steepness of the yield curve.</cite> <cite index="22-11">The curvature factor shows the changes in the curve shape, highlighting if the curve becomes more or less curved in certain sections.</cite>

The factor loadings—eigenvectors from the covariance matrix—reveal directional emphasis. The level eigenvector carries roughly uniform weight across all maturities. The slope eigenvector assigns positive weight to the long end and negative weight to the short end (or vice versa), capturing steepening or flattening. The curvature eigenvector has a sign pattern that weights the belly of the curve differently from the wings.

<cite index="17-1,17-2">The shape of the yield curve is measured by estimates of the level, slope and curvature in the Nelson and Siegel (1987) tradition, following the state-space specification and maximum-likelihood estimation with the Kalman filter.</cite> This parametric alternative to PCA produces interpretable factor series with similar economic content, though the variance partition and orthogonality differ.

Sources:

• https://www.studysmarter.co.uk/explanations/business-studies/actuarial-science-in-business/term-structure-analysis/
• https://www.sciencedirect.com/science/article/abs/pii/S0378426612000465

The empirical consistency across markets is notable. <cite index="20-2,20-3,20-9,20-10">The first factor accounts for 76.87% of the total variance in India, while the second and third factors account for 16.64% and 5.14%, respectively, with the first three principal components explaining 98.64% of the variability of the data.</cite> <cite index="23-3">In the Chinese sovereign market, the first factor accounts for 80.80% of the total variance, while the second and third factors account for 14.05% and 2.27%.</cite>

<cite index="7-8,7-9">Litterman and Scheinkman provided evidence that principal components derived from market interest rates are a valuable information resource for fixed income portfolio managers and market risk officers, particularly due to their contribution as hedging tools; they identified three factors that explained approximately 98% of the returns on U.S. Treasuries.</cite>

The level factor's dominance reflects that most curve movement is a uniform shift across maturities. The slope and curvature factors, though smaller in magnitude, capture the non-parallel distortions that duration-based immunization cannot hedge. The decomposition holds across sample periods and sovereign issuers, though the exact variance split shifts with the market regime and the maturity range observed.

Sources:

• https://mpra.ub.uni-muenchen.de/39229/1/MPRA_paper_39229.pdf
• https://arxiv.org/pdf/1911.07288
• https://www.mdpi.com/1911-8074/15/6/247

<cite index="1-15">Litterman and Scheinkman's 1991 study identified that a few principal components explain most of the variance of treasury zero-coupon rates and that the first three eigenvectors represent level, slope, and curvature changes on the curve.</cite> The canonical work appeared in the Journal of Fixed Income and has been replicated in various sovereign markets since.

<cite index="3-7">The level factor accounted for on average 90% of the observed variation in yield changes, which explains why duration-based strategies are commonly used, though investors can achieve a better hedged position by considering the effect on a portfolio of each of the three factors rather than simply holding a duration-based hedged portfolio.</cite> <cite index="1-5,1-6">The yield changes caused by the first factor are basically constant across maturities—the first factor represents essentially a parallel change in yields, and hedging against Factor 1 is close to duration hedging.</cite>

<cite index="3-2,3-5">The level factor relates to a parallel movement in the interest rates, and the last two factors explain the non-parallel shifts which occur in yield curves.</cite> <cite index="3-6">The slope factor results from a shift that causes a change in the yield curve's gradient and the third factor was shown to originate from a change in the curve's curvature.</cite> The methodology applies PCA to the covariance matrix of yield changes; the eigenvectors become the factor loadings, the eigenvalues measure variance contribution.

Sources:

• https://www.academia.edu/22234179/Principal_component_analysis_of_yield_curve_movements
• https://scielo.org.za/scielo.php?script=sci_arttext&pid=S2222-34362018000100008
• https://link.springer.com/article/10.1007/s12197-010-9142-y

<cite index="4-4,4-5">Expectations of future shifts to an inflationary regime give rise to a regime shift premium in forward interest rates, which can be seen as compensation investors demand because they do not view the current price stability objective as fully credible and a switch to a higher inflation level might occur. The size of the regime shift premium depends on the probability of a shift to a high inflation regime assigned by financial investors</cite>.

<cite index="4-7,4-8">The long forward interest rate differential is likely to capture other risk factors than expectations of a shift to an inflationary regime — default risk or a time varying term premium. In empirical investigations systematic fluctuations in excess returns have often been interpreted as a time varying term premium, but fluctuations that seem to be systematic can also be attributed to fluctuating expectations of future regime shifts</cite>.

<cite index="3-1,3-2">Term premium will vary not only according to interest rate levels but also according to several dimensions of economic activity. Plausible variables in this context include the unemployment rate, the growth rate of real gross national product, and the rate of increase of prices, all as indicators of various cyclical aspects of economic activity</cite>. The forward curve distends not only with rate expectations but with the probability mass assigned to the regimes that have not yet priced.

Sources:

• https://www.bis.org/publ/confp06f.pdf
• https://www.nber.org/system/files/working_papers/w0295/w0295.pdf

<cite index="18-6,18-7,18-8">The Kim and Wright (2005) model is very similar in structure to ACM but their long-maturity term premium estimates can differ materially at times. The main reason for the divergence is that KW incorporate Blue Chip surveys of professional forecasters' short rate expectations in the data set used to estimate the model; once ACM is modified to include these surveys, term premium estimates are very similar to KW estimates</cite>.

<cite index="18-2,18-4">Including survey measures in the data set for estimating term structure models may alleviate small sample problems by providing more information about the long-run average short rate and how quickly short-term interest rates should revert back to that average. Including the surveys also makes the expectations component more variable and term premiums less variable</cite>. <cite index="19-7">The use of surveys to anchor the model's dynamics can help alleviate some of the small sample and overfitting biases that affect these models which are rich in cross sectional variation but have a relatively small number of business cycles to estimate the intertemporal dynamics</cite>.

<cite index="20-2,20-4">The KW model is the standard affine Gaussian model with three factors that are latent — the factors are defined only statistically and do not have a specific economic meaning. Data on survey forecasts of 3-month Treasury bill rate are used in addition to yields data in order to help address small sample problems that often pervade econometric estimation with persistent time series like bond yields</cite>. What you anchor to defines what you read.

Sources:

• https://www.federalreserve.gov/econres/notes/feds-notes/robustness-of-long-maturity-term-premium-estimates-20170403.html
• https://www.federalreserve.gov/data/three-factor-nominal-term-structure-model.htm
• https://www.elibrary.imf.org/view/journals/001/2018/140/article-A001-en.xml

<cite index="10-1,10-3,10-4">The ACM model (Adrian, Crump, and Moench, 2013) is the most widely referenced term premium estimate, published by the Federal Reserve Bank of New York. It decomposes each Treasury yield into two components: expected short-rate path (the average of expected future short-term rates over the bond's maturity) and term premium (compensation for bearing interest rate risk, inflation uncertainty, and supply/demand imbalances)</cite>.

<cite index="10-5">The model belongs to the class of no-arbitrage affine term structure models. It uses five pricing factors extracted from Treasury yields via principal components and estimates the market price of risk using excess bond return regressions</cite>. <cite index="13-2,13-3,13-5,13-6">The estimation method consists of three consecutive regressions: first, a vector autoregression of order one is estimated on the yield curve factors; second, monthly excess bond returns are regressed on a constant, lagged yield curve factors, and estimated factor innovations, which delivers predictive coefficients and factor risk exposures akin to betas in empirical asset pricing models</cite>.

<cite index="16-2,16-4,16-10">Fitting the full sample of yields allows estimation of the term premium on a daily basis going back to June 14, 1961. The ACM model fits the data extremely well, allowing yields to be decomposed without concerns about measurement error</cite>. <cite index="16-6,16-12">The term premium is a countercyclical variable which tends to move with measures of uncertainty and disagreement about the future level of yields</cite>.

Sources:

• https://www.yieldcurve.pro/learn/acm-model
• https://www.bis.org/publ/bppdf/bispap102_keynote.pdf
• https://libertystreeteconomics.newyorkfed.org/2014/05/treasury-term-premia-1961-present/

<cite index="7-1,7-7">The term premium in forward rates is defined as forward rate minus the expected short rate for the same horizon</cite>, a departure from the expectations hypothesis. <cite index="1-1">The forward term premium is simply the difference between observed forwards and what would be the yield predicted by the pure expectations hypothesis — the average expected future short rate</cite>.

<cite index="7-2,7-4">While market participants often use forward rates as expected future short rates for relatively short horizons, many observers doubt that longer-horizon forward rates adequately represent movements of long-horizon expectations, and a large body of empirical literature has documented the failure of the expectations hypothesis</cite>. <cite index="3-5">Statistical results provide evidence disconfirming the pure expectations theory of the term structure</cite>.

<cite index="5-3,5-6">Under expectations theory with time-invariant term premia, forward rates implicit in the term structure should follow a martingale sequence; rejection of the simple expectations theory is consistent with the hypothesis of time-varying term premia</cite>. <cite index="1-11">The prominent role of term premiums in explaining the dynamics of interest rates and their significant reaction to structural economic shocks highlights the need to incorporate this commonly omitted component</cite>. What the market prices is not always what the market expects — the gap between the two is load-bearing.

Sources:

• https://www.federalreserve.gov/data/three-factor-nominal-term-structure-model.htm
• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr775.pdf
• https://www.nber.org/system/files/working_papers/w0295/w0295.pdf
• https://link.springer.com/article/10.1007/BF02426968

<cite index="18-11">The basis trade reached unprecedented levels during 2007–2009 and resulted in a dramatic unraveling of basis trades and major losses to financial institutions including Merrill Lynch, Deutsche Bank, and Citadel</cite>. The dislocation was not brief. <cite index="17-2,17-8">The negativity puzzle is defined as the unexpected persistence of the dislocation between bond and derivative credit markets</cite>.

<cite index="21-4">The basis is more negative when bond lending fee is higher, suggesting that arbitrageurs are unwilling to engage in a negative basis trade when short interest on the bond is high</cite>. The positions that worked before the crisis became the positions that hurt during it. <cite index="20-9">A negative basis emerged during the 2008 financial crisis because the limited balance-sheet capacity of dealer banks prevented corporate bond dealers from trading aggressively enough to close the basis</cite>.

The regime that followed is distinct. <cite index="17-3,17-4">The first two moments of the basis are described by three distinct Markov regimes identified with periods related to the 2008 financial crisis; the post-crisis regime differs significantly from the crisis and the pre-crisis regimes</cite>. The basis still reflects limits to arbitrage, but the actors who arbitraged it before hold less capital now, and the capital they hold is priced differently.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0927539819300842
• https://ideas.repec.org/p/ris/crcrmw/2019_004.html
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3481470
• https://ideas.repec.org/a/bla/finmgt/v48y2019i2p417-439.html
• https://web.williams.edu/Economics/wp/GongPhelanDebtCollateralizationStructuredFinanceCDSBasis.pdf

<cite index="10-2,10-3">Many arbitrage relations involve a fully funded (cash) instrument and one or more unfunded derivative positions, which raises the possibility that counterparty risk of the derivative issuer made the arbitrage risky</cite>. <cite index="10-4,10-5">An alternative hypothesis is that funding cost differentials between the cash instrument and derivative positions were responsible for the deviations; in the former case the payoff to the arbitrage trade is not risk-free for any investor, whereas in the latter case the payoff would still be risk-free for an investor with very deep pockets</cite>.

<cite index="9-5,9-6,9-7">A significant consideration for purchasers of protection in the credit default swaps market is the credit quality of the protection seller; the protection seller may itself go bankrupt either before or at the same time as the reference entity, which is what is meant by counterparty credit risk</cite>. <cite index="9-8">Market participants commonly use credit-enhancement mechanisms—such as the posting of collateral—to mitigate the effects of counterparty credit risk in the pricing of CDS contracts</cite>.

The risk is not symmetric. <cite index="7-9,8-9">It is possible that the underlying firms are selling the cash bond and their affiliated financial institutions are also the sellers of the CDS contract</cite>. When the reference entity and the protection seller share exposure to the same macro shock, the basis reflects not just credit risk but wrong-way risk—the kind that collateral cannot fully hedge.

Sources:

• http://lamfin.arizona.edu/fixi/creditmod/BaiCollinDufresne11.pdf
• https://www.federalreserve.gov/econres/feds/files/2022023pap.pdf
• https://www.researchgate.net/publication/272244343_CDS-Bond_Basis_Arbitrage_A_Stabilizing_Force_in_the_Corporate_Bond_Market
• https://www.researchgate.net/publication/313944672_The_CDS-Bond_Basis_Arbitrage_and_the_Cross_Section_of_Corporate_Bond_Returns_Cross_Section_of_Corporate_Bond_Returns

<cite index="18-2,18-3">Traders exploit the negative CDS–bond basis by transforming the bond into a synthetic FRN using the asset-swap package, which consists of a long position in the bond and an asset-swap to hedge its interest rate risk</cite>. <cite index="18-4">The purchase of the package is financed in the repo and Fed Funds market</cite>. <cite index="18-6">Arbitrageurs buy the synthetic bond and CDS protection with matching maturity to exploit a possible mispricing</cite>.

The position looks clean on paper but relies on funding that is priced at Libor or below. <cite index="11-3,11-5">An investor with a funding cost of Libor plus 25 basis points will view buying a floating-rate note priced at par paying Libor plus 125 bps as theoretically identical to selling a CDS contract on the same FRN at a premium of 100 bps; a Libor-plus funding cost will drive the basis lower</cite>. <cite index="11-12,11-13">Funding at levels in excess of Libor will tend to make the basis positive, as CDSs do not require funding; shorting cash bonds tends to be difficult, as the bond needs to be sourced in a fairly illiquid and short-dated repo market in which bonds might trade on special</cite>.

The trade requires balance-sheet capacity, repo haircuts, and margin on the CDS. <cite index="13-3,13-8">The key assumptions are the haircut charged on financing the corporate bond purchase in the repo market, the initial margin required on the single-name CDS position, and whether the CDS is centrally cleared</cite>. When any of those constraints tighten, the basis widens.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0927539819300842
• https://forum.bionicturtle.com/threads/cds-bond-basis-factors-confusing-impact.10284/
• https://www.newyorkfed.org/medialibrary/media/research/epr/2018/EPR_2018_trends-in-credit-basis-spreads_boyarchenko.pdf?sc_lang=en

<cite index="3-22,3-23">The CDS-bond basis is defined as the difference between the CDS premium and the maturity-matched par asset-swap spread; when this difference is higher than zero the basis is positive, when lower it is negative</cite>. In theory, <cite index="22-3">the premium of a CDS should equal the asset-swap spread for the same reference name</cite>. In practice that convergence fails.

<cite index="17-6,19-6">The negative basis can be explained by liquidity risk in both the bond and CDS markets, together with counterparty risk, collateral quality, and funding constraints</cite>. <cite index="21-3">Deviations are larger for bonds with higher frictions as measured by trading liquidity, funding cost, counterparty risk, and collateral quality</cite>. <cite index="18-9,18-10">During the financial crisis the basis reached unprecedented levels that persisted for more than a year; the basis moved deeply into negative territory for all but a few top-rated names, being particularly wide for those of lower credit quality</cite>.

The post-crisis regime differs from what preceded it. <cite index="17-7,17-13">Basis negativity persistence during the post-crisis period is mainly related to a significant decrease in basis arbitrage activity</cite>. What arbitrageurs price is not yet what the basis itself is saying: the trade is there, but the capital is not.

Sources:

• https://www.nbb.be/doc/ts/publications/wp/wp104en.pdf
• https://www.pm-research.com/content/iijtrade/2/1/79
• https://ideas.repec.org/p/ris/crcrmw/2019_004.html
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3481470
• https://ideas.repec.org/a/bla/finmgt/v48y2019i2p417-439.html
• https://www.sciencedirect.com/science/article/abs/pii/S0927539819300842

<cite index="15-11,15-12,15-13">For an n-year horizon, compute P(n) = P^n through repeated matrix multiplication using standard numerical software (Python, R, MATLAB), and the Default column of P^n gives the cumulative probability of default over n years for each starting rating</cite>. <cite index="15-2">This calculation assumes a time-homogeneous Markov chain—meaning transition probabilities are constant from year to year</cite>. <cite index="18-10,18-11">A critical feature of transition matrices is that default is an absorbing state; once an issuer defaults—in the sense defined by probability of default analysis—it cannot migrate back to a performing rating</cite>.

<cite index="24-4,24-5">A typical approach for estimating the rating transition matrices relies on calculating empirical rating migration counts and frequencies from rating history data; however, for small portfolios this approach often leads to zero counts and high count volatility, which makes the estimations unreliable and volatile, and can also produce counter-intuitive prediction patterns such as intersecting forward PDs</cite>. The matrix-power scaling is numerically clean but conceptually fragile. It amplifies the assumption that the process is stationary across the projection window. If the credit cycle turns, the scaled matrix gives you a forecast anchored to a regime that no longer holds.

Sources:

• https://ryanoconnellfinance.com/credit-risk-migration-matrices/
• https://arxiv.org/pdf/1708.00062

<cite index="18-1,18-3,18-4">Rating agencies—primarily S&P Global Ratings and Moody's Investors Service—publish transition matrices annually based on decades of historical data, and these matrices form the empirical foundation for credit portfolio models</cite>. <cite index="18-5,18-6,18-7">An approximate one-year corporate rating transition matrix based on S&P Global's Annual Global Corporate Default and Transition Study (1981–2023, global corporate issuers) shows values in percentages representing the probability of migrating from the row rating to the column rating within one year, using approximate long-run average values</cite>.

The averaging matters. <cite index="18-16,18-17">Historical transition matrices are through-the-cycle averages that smooth over recessions and expansions; during the 2008–2009 financial crisis, speculative-grade default rates exceeded 10%—far above the long-run average</cite>. <cite index="18-12,18-13">Mixing an S&P-rated portfolio with a Moody's transition matrix introduces systematic bias, and even after mapping to a common scale, residual methodology differences remain</cite>. The matrix you use must correspond to the rating universe you hold. That is not a rounding error; it is a structural mismatch in how agencies define and observe the same nominal rating.

Sources:

• https://ryanoconnellfinance.com/credit-risk-migration-matrices/

<cite index="2-2">A smoothed 1-year transition probability matrix based on long-term (20+ years) historical Moody's estimate was used in the popular CreditMetrics model</cite>. That matrix is static. <cite index="6-10,6-11,6-12">Moody's Credit Transition Model couples credit ratings with the economic cycle, assigns an expected default rate to a rated credit and any portfolio of rated credits conditional on the path of the macroeconomy, generates complete transition forecasts for credits, and can be extended over any horizon from one quarter to five years or more</cite>.

<cite index="7-9,7-10,7-11">The Credit Risk Calculator collects historical frequencies for a given horizon; the Credit Transition Model is more sophisticated and forward-looking, refining the rating transition matrix based on macro-economic forecasts and the rating history of the issuer, and can produce issuer-specific, more granular results</cite>. <cite index="6-5,6-6,6-7">A typical transition matrix does not provide first-passage probabilities—the probability that within two years a credit will have fallen below a threshold, even if it later reverses</cite>. Moody's CTM addresses that gap. What you need is the probability of breaching a threshold at any point in the horizon, not just the terminal rating. Standard matrices do not tell you that.

Sources:

• https://arxiv.org/pdf/0912.4621
• https://www.moodys.com/sites/products/productattachments/credit%20transition%20model.pdf
• https://www.moodys.com/sites/products/DefaultResearch/2006800000445742.pdf
• https://www.moodys.com/sites/products/productattachments/faq%20credit%20transition%20model.pdf

<cite index="5-1">The rating transition matrices describe various aspects of the probability of rating changes and default for corporate debt issuers</cite>, built from historical observations. <cite index="17-4,17-5,17-6">To construct a rating transition matrix, agencies collect data on the rating changes of a large and representative sample of issuers over the chosen period, count the number of issuers that moved from one rating category to another, and divide it by the total number of issuers in the initial category to derive empirical probability of each rating transition</cite>. <cite index="16-7">All transition matrices exhibit the same characteristic; they have high probabilities in a diagonal matrix; the obligor tends to maintain its current rating</cite>.

The canonical construction assumes a time-homogeneous Markov chain—an assumption that <cite index="3-1">Moody's CTM methodology is grounded in the observation that credit transitions are cyclical and generally non-Markovian</cite>. <cite index="9-1">The validity of these assumptions—existence of a homogeneous generator and Markovianity—is not always guaranteed</cite>. <cite index="10-3">Default rates vary over time, and different migration matrices are obtained if they are estimated during recession or expansion</cite>. This matters: the transition probability you observe is conditioned on the sample window you measure. The rating path is not memory-free.

Sources:

• http://lamfin.arizona.edu/fixi/542/mrt.pdf
• https://fastercapital.com/keyword/rating-transition-matrix.html/1
• https://www.bis.org/ifc/publ/ifcb31u.pdf
• https://www.moodys.com/sites/products/ProductAttachments/DRD/credit-transition-model-ctm-at-a-glance.pdf
• https://arxiv.org/pdf/1403.8018

<cite index="12-1,12-2">Gertler and Karadi combine traditional monetary vector autoregression analysis with high-frequency identification of monetary policy shocks, using HFI surprises as external instruments.</cite> The external-instrument approach allows the high-frequency shock to identify the contemporaneous effect of policy within a VAR that tracks monthly or quarterly macro variables—output, inflation, credit spreads—over longer horizons.

<cite index="11-2">High-frequency measures of monetary policy surprises are used as external instruments in a VAR to identify the contemporaneous effects of monetary policy.</cite> <cite index="12-4,13-12">Monetary policy surprises typically produce modest movements in short rates that lead to large movements in credit costs and economic activity.</cite> The amplification occurs through the credit channel: small policy moves tighten financial conditions, which bind on levered actors.

<cite index="14-7,14-10">The proxy-VAR methodology uses high-frequency surprises in financial markets around tight windows of monetary policy announcements as instruments, motivated by Gertler and Karadi's advocacy for including financial variables to capture the credit channel of the monetary policy transmission mechanism.</cite> The method has been adopted widely because it marries the clean shock identification from intraday data with the ability to trace multi-period effects on real variables.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2450924
• https://www.nber.org/papers/w20224
• https://www.nber.org/system/files/working_papers/w19260/revisions/w19260.rev3.pdf
• https://onlinelibrary.wiley.com/doi/10.1111/1475-4932.12801

<cite index="23-5">A surprise policy tightening raises interest rates and reduces stock prices, while the complementary positive central bank information shock raises both.</cite> Jarocinski and Karadi use this high-frequency co-movement to disentangle the two shocks via sign restrictions. <cite index="15-2,15-4">Their contribution is to use sign restrictions on multiple high-frequency surprises and identify multiple contemporaneous shocks.</cite>

<cite index="22-8">To study how information shocks influence exchange rate dynamics, sign restrictions are imposed on high-frequency changes in interest rates and stock prices measured within short time windows around central bank announcements.</cite> The identifying assumption: a pure policy shock raises rates and lowers equities; an information shock raises both. If an announcement moves rates up 10bp and equities up 50bp, the composite contains both a policy shock and a positive information shock about fundamentals.

<cite index="21-5,21-7">A contractionary monetary policy shock decreases inflation rates, whereas a positive central bank information shock increases inflation rates.</cite> This directional split matters for any pricing model that attempts to decompose a Fed announcement into the portion that tightens credit and the portion that revises the expected path of the cycle.

Sources:

• https://www.aeaweb.org/articles?id=10.1257/mac.20180090
• https://peterkaradi.github.io/website/Published/JarocinskiKaradi.pdf
• https://link.springer.com/article/10.1007/s11079-022-09682-6
• https://www.sciencedirect.com/science/article/abs/pii/S088915832300031X

<cite index="1-4">At the same time forecasts about output growth also increase—the opposite of what standard models imply about a monetary tightening.</cite> This is the information effect. <cite index="1-5,1-6">Nakamura and Steinsson build a model in which Fed announcements affect beliefs not only about monetary policy but also about other economic fundamentals, and their model implies that these information effects play an important role in the overall causal effect of monetary policy shocks on output.</cite>

<cite index="23-1,23-3">Central bank announcements simultaneously convey information about monetary policy and the central bank's assessment of the economic outlook.</cite> <cite index="22-2,22-6">Market participants may interpret contractionary policy announcements as the central bank's reaction to an improved economic outlook, which may give rise to more optimistic views regarding the overall macroeconomic situation.</cite>

<cite index="24-2,24-6">If investors believe that the central bank possesses superior information about the macroeconomy, an unexpected cut in the target rate may be perceived as a signal of deteriorating economic conditions.</cite> This signaling channel complicates identification: the rate change itself is contractionary, but the signal about fundamentals may be expansionary. The Nakamura-Steinsson approach observes this entanglement but does not fully disentangle it within the 30-minute window alone.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2298991
• https://ideas.repec.org/a/oup/qjecon/v133y2018i3p1283-1330..html
• https://www.aeaweb.org/articles?id=10.1257/mac.20180090
• https://link.springer.com/article/10.1007/s11079-022-09682-6
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X24002113

<cite index="1-1,3-3">Nakamura and Steinsson's identifying assumption holds that unexpected changes in interest rates in a 30-minute window surrounding scheduled Federal Reserve announcements arise from news about monetary policy.</cite> The approach sidesteps the fundamental problem in conventional VAR work: that the Fed responds to the economy, so observed rate movements blend reaction-function response with exogenous shocks.

<cite index="1-3">In response to an interest-rate hike, nominal and real interest rates increase roughly one-for-one several years out into the term structure, while the response of expected inflation is small.</cite> <cite index="12-3,13-11">Gertler and Karadi combine the high-frequency identification with traditional monetary VAR analysis, using HFI surprises as external instruments, and find that the shocks produce responses in output and inflation consistent with textbook theory.</cite>

The Gertler-Karadi variant differs in implementation—<cite index="19-1">Gertler and Karadi calculate first differences of the 3-month-ahead Fed funds futures</cite>, while <cite index="9-2,9-3">Nakamura and Steinsson use the first four quarterly Eurodollar futures contracts and take the first principal component of changes in ED1–ED4, rescaled so a one-unit change corresponds to a 1 percentage point change in the ED4 rate.</cite> Both methods price the announcement surprise cleanly, but differ in the portion of the term structure they weight.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2298991
• https://eml.berkeley.edu/~enakamura/papers/realrate.pdf
• https://www.nber.org/system/files/working_papers/w19260/w19260.pdf
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2450924
• https://www.nber.org/papers/w20224
• https://www.sciencedirect.com/science/article/abs/pii/S0165176519303155
• https://www.journals.uchicago.edu/doi/10.1086/723574

<cite index="6-2,6-9">Corporate bond returns provided by indices (i) correctly estimate credit excess returns, (ii) are synchronous for the entire set of bonds, allowing for consistent cross-sectional comparability, and (iii) suffer less from stale pricing issues.</cite> TRACE returns, by contrast, reflect only executed trades—many bonds do not trade daily or even monthly, creating gaps in return series and asynchronous measurement across securities.

<cite index="20-3">TRACE returns are more strongly negatively serially correlated, and index-provided returns lead TRACE returns, both consistent with stale pricing issues with TRACE.</cite> The mechanical explanation: <cite index="20-15">index returns explicitly try to avoid bid-ask bounce with only bid-side pricing and an attempt to deliver prices suitable for marking institutional portfolios.</cite> TRACE, by contrast, records trades that toggle between customer buys (at the offer) and customer sells (at the bid), inducing negative autocorrelation that has nothing to do with information but reflects dealer spread and trading noise.

<cite index="6-3,6-10">Due to these coverage and data quality issues, researchers should try, where possible, to source return data from multiple sources to ensure the robustness of their results.</cite> The methodological implication: if TRACE and a commercial index disagree on return measurement, the index is more likely to reflect the price institutional capital actually faced.

Sources:

• https://link.springer.com/article/10.1007/s11142-023-09777-6
• https://www.researchgate.net/publication/315310954_How_to_Clean_Enhanced_TRACE_Data

<cite index="20-1,20-2">Many corporate bonds included in TRACE are not included in representative bond indices—typically smaller bonds—and these are less economically relevant, as they are not the focus of institutional investors; limiting the sample to index-included bonds is recommended.</cite> <cite index="4-13,4-14">Over the 2002–2020 period, 1,130,446 bond-months appeared in both WRDS/TRACE and ICE/BAML, while 737,668 bond-months appeared in TRACE but not ICE/BAML, and 450,050 bond-months appeared in ICE/BAML but not TRACE.</cite>

<cite index="20-13">Multiple issues plague TRACE return measures: cross-sectional coverage is limited (requiring a price at both month-end points for a monthly return, which not all bonds trade); TRACE returns are more strongly negatively serially correlated, consistent with liquidity issues and stale pricing; and index-provided returns lead TRACE returns, consistent with stale pricing.</cite> <cite index="20-14,20-15">Greater bid-ask bounce issues arise from computing returns from trades without correcting for the side of the trade; with buys and sells occurring through time, this leads to negative serial correlation—a problem less evident with index returns that use only bid-side pricing.</cite>

<cite index="4-4">Researchers should try to source corporate bond data from index providers, where coverage is better, returns are aligned in calendar time, and data quality is higher.</cite> When that is not feasible, restrict the sample to index-included bonds.

Sources:

• https://link.springer.com/article/10.1007/s11142-023-09777-6

<cite index="16-4,16-5,16-6,16-7,16-8">As an example, 2007 raw TRACE data contained 6.7 million transaction reports, of which 440,000 were deleted as known errors, 780,000 as agency transactions, and 1.6 million as interdealer double-counted transactions—in total, almost 35% of the raw transactions.</cite> The canonical cleaning filter, developed by Dick-Nielsen and adopted across the literature, removes cancellations, corrections, reversals, and duplicate agency reports.

<cite index="10-9,10-12,10-13,10-14">The filter is designed to delete transactions already marked as errors, with corrections and cancellations identified via message sequence numbers that are unique within a reporting date, not the transaction date.</cite> <cite index="19-3,19-11">On February 6, 2012, FINRA changed both the database format and the trade-reporting methodology, requiring separate filter logic for pre-2012 and post-2012 periods.</cite>

<cite index="13-1,13-2">The most widely used error-filter is the one proposed by Dick-Nielsen (2009, 2014, 2019), though it is time- and memory-consuming.</cite> Researchers working with TRACE data from WRDS or direct from FINRA should apply the Dick-Nielsen protocol or a variant before calculating liquidity measures. The presence of unreported duplicates and dealer-side reporting conventions means that naïve use of the raw feed will distort measures of volume, bid-ask spread, and turnover.

Sources:

• https://communities.sas.com/kntur85557/attachments/kntur85557/sas_studio/1817/1/cleanTrace.pdf
• https://www.semanticscholar.org/paper/How-to-Clean-Enhanced-TRACE-Data-Dick%E2%80%90Nielsen/dce60def0fe3d788422befb32e0c335c25738447
• https://marianopalleja.github.io/personalwebsite/TRACE_error_filter.pdf
• https://github.com/hannes101/FilterTRACE

<cite index="15-1,15-2,15-3">DTS was originally developed by Robeco researchers in 2003, used to monitor the credit risk of all of Robeco's credit portfolios, and a joint project with Lehman Brothers led to publication of the results in The Journal of Portfolio Management in 2007—DTS is now widely accepted among investors, and has been implemented in leading risk management software, including MSCI RiskMetrics and Bloomberg Barclays POINT.</cite> <cite index="15-4">The authors are Ben Dor, Dynkin, Hyman, Houweling, Van Leeuwen, and Penninga, published in 2007, vol. 33 no. 2.</cite>

<cite index="16-1,16-4">Extensive empirical research has shown that the spread volatility of credit securities is linearly proportional to their level of spread—this finding holds true across corporate and sovereign issuers, for both cash and credit default swaps.</cite> <cite index="17-18">By using DTS as the risk measure, we assume that credit spreads move in a relative fashion rather than a parallel fashion.</cite>

The paper established the empirical basis for multiplying spread by duration rather than using spread duration alone. It showed that systematic and idiosyncratic spread volatility both scale with spread level, irrespective of sector, maturity, or time period. The methodology became standard because it offered a single number that adjusts instantaneously when spreads widen, whereas historical volatility models lag the regime shift.

Sources:

• https://www.robeco.com/me/insights/2019/06/duration-times-spread-a-measure-of-spread-exposure-in-credit-portfolios.html
• https://onlinelibrary.wiley.com/doi/abs/10.1002/9781118182635.efm0028
• https://www.robeco.com/en-us/insights/2024/08/duration-times-spread-a-measure-of-spread-exposure-in-credit-portfolios

<cite index="9-1">DTS is a first-order approximation that works well under normal conditions but has important limitations.</cite> The proportionality constant between spread level and volatility is not stable across regimes. <cite index="9-4,9-5">During the 2007–2009 credit crisis, the spread volatility proportionality factor increased from roughly 10% to roughly 15%—DTS underestimated risk during the crisis, though it still outperformed historical volatility measures.</cite> <cite index="14-1">DTS can be inclined to over- and underestimate risk.</cite>

<cite index="9-7,9-8">Relative spread volatility is somewhat higher for shorter-maturity bonds (~20% higher for 3-year vs 5-year bonds) and lower for longer-maturity bonds—this is a second-order effect for most portfolios but matters for leveraged or term-structure strategies.</cite> <cite index="9-10,9-20">DTS accuracy depends on the quality of the OAS calculation, which varies by data provider and model assumptions.</cite>

<cite index="10-6">DTS assumes parallel shifts in the yield curve, doesn't account for changes in credit spread over time, and ignores other risk factors.</cite> <cite index="5-12">DTS does not say anything about interest rate risk, because it only predicts risk driven by credit spread fluctuations.</cite> The heuristic is useful for comparing credit volatility across issuers and maturities at a single point in time—but it does not estimate the probability-weighted loss given spread widening, and it does not estimate the path by which spreads reach distressed levels.

Sources:

• https://ryanoconnellfinance.com/duration-times-spread/
• https://www.robeco.com/en-us/insights/2024/09/duration-times-spread-measuring-credit-risk
• https://www.daytrading.com/duration-times-spread
• https://www.robeco.com/en-us/insights/2024/08/duration-times-spread-a-measure-of-spread-exposure-in-credit-portfolios

<cite index="5-1,5-9">DTS only predicts the credit risk of the bonds, not their default risk.</cite> <cite index="5-11">To predict default risk, one could use credit ratings or distress risk measures, such as distance-to-default.</cite> This distinction matters in periods when spreads reflect factors other than issuer deterioration.

<cite index="7-3,7-4">The key limitation arises as the measure assumes that credit spreads are an accurate measure of underlying default risk.</cite> <cite index="7-5,7-6">DTS implies that the higher the overall credit spread the higher the level of default risk within a security or portfolio—though such an assumption may hold at any point in time when undertaking a cross-sectional comparison, it can become a more problematic assumption when applied as a risk measure over time.</cite>

<cite index="7-7">The limitation arises when applied over time as changes in credit spreads can be driven by other non-default risk related factors such as overall market risk aversion, liquidity conditions/liquidity premiums, issuance/supply dynamics, level of interest rates, and direct intervention by central banks and other authorities.</cite> When QE compresses spreads across IG and HY, DTS will mechanically signal a decline in credit risk even if refinancing capacity has not improved. When liquidity drains from sub-IG paper in a flight-to-quality episode, DTS will flag rising volatility—but not whether that volatility reflects changed fundamentals or changed access to the liability stack.

Sources:

• https://www.robeco.com/en-us/insights/2024/08/duration-times-spread-a-measure-of-spread-exposure-in-credit-portfolios
• https://www.livewiremarkets.com/wires/demystifying-credit-risk-measures

<cite index="1-2">Duration Times Spread (DTS) has become the industry standard for measuring the credit volatility of a corporate bond.</cite> The methodology rests on a single empirical observation: <cite index="4-6,4-8">absolute spread volatility scales linearly with spread level, confirmed by research analyzing over 560,000 bond-month observations from 1989-2005 with R² > 90%.</cite> <cite index="4-9">A bond at 200 bps experiences roughly four times the absolute spread volatility of a bond at 50 bps.</cite>

This matters because spread duration on its own treats all spread changes as parallel shocks. <cite index="1-8">Traditional risk measurement models do not account for the fact that the magnitude of spread changes are proportional to spread levels: bonds with wider spreads are riskier because they experience greater spread changes.</cite> <cite index="4-1,4-2">A 5-year spread duration bond at 100 bps has much less credit risk than a 5-year spread duration bond at 400 bps—DTS captures this difference; spread duration does not.</cite>

The DTS heuristic is simple: multiply spread duration by OAS. <cite index="4-10">Relative spread volatility—spread changes as a percentage of the current spread—is more stable than absolute spread volatility.</cite> <cite index="4-11">The proportionality factor is approximately 9–10% per month under normal conditions, though it increased to roughly 15% during the 2007–2009 credit crisis.</cite>

Sources:

• https://www.robeco.com/en-us/insights/2024/09/duration-times-spread-measuring-credit-risk
• https://ryanoconnellfinance.com/duration-times-spread/

<cite index="6-12">The difference between z-spread and OAS is the cost of the call option</cite>. <cite index="6-14">The option cost is measured this way because if rates do not change, the investor would earn the z-spread</cite>. The gap between the two spreads quantifies what the bondholder surrendered by holding a callable rather than a bullet.

<cite index="13-8,13-9">A callable bond can be viewed as an option-free bond minus an embedded call option; the investor is effectively short the call option, which is why callable bonds trade at lower prices and higher yields than otherwise identical bullet bonds</cite>. <cite index="15-14">The embedded option cost can be quantified by subtracting the OAS from the z-spread, which ignores optionality and volatility</cite>.

This decomposition holds tactical value. When z-spread minus OAS widens, the market is pricing higher call probability or higher volatility—or the model you are using has drifted from consensus. When the gap narrows, call risk has repriced lower. <cite index="2-1">To derive the OAS, you subtract the callable feature's value from the z-spread</cite>.

The option cost is not static. It responds to the term structure, to volatility, to the issuer's refinancing incentives. A 5% coupon callable trading through par in a 3% rate environment has a call option trading near intrinsic value. The same bond in a 6% environment has an out-of-the-money call with negligible value, and z-spread converges toward OAS. Practitioners hold both spreads in view because the difference between them is the live price of optionality.

Sources:

• https://analystnotes.com/cfa-study-notes-option-adjusted-spread.html
• https://ryanoconnellfinance.com/callable-bonds-embedded-options/
• https://en.wikipedia.org/wiki/Option-adjusted_spread
• https://www.cgaa.org/article/oas-spread-vs-z-spread

<cite index="5-2">For a security whose cash flows are independent of future interest rates, OAS is essentially the same as z-spread</cite>. <cite index="16-3">The z-spread for a straight bond is its option-adjusted spread assuming volatility of zero</cite>. This collapses the distinction when optionality is absent.

But when cash flows respond to rates, the choice of spread measure determines what you are pricing. <cite index="20-3,20-4,20-5">Z-spread compensates investors for liquidity risk, credit risk, and option risk; to compare callable bonds with non-callable instruments, a spread measure that does not include compensation for option risk is needed, otherwise z-spread of callables would always be higher without indicating which bond has greater value</cite>.

<cite index="11-1,11-2,11-4">A key measure of relative value for a corporate bond is its swap spread, the basis-point spread over the interest-rate swap curve measuring credit risk; in practice traders use the asset-swap spread and the z-spread as main measures of relative value</cite>. Asset swap spread and z-spread serve overlapping but distinct roles—asset swap isolates the credit component by creating a synthetic floater, while z-spread holds the Treasury curve as the benchmark.

The rule: use z-spread for bullets and bonds where optionality is not priced. Use OAS for callables, putables, mortgage-backed securities, and any structure where the issuer or holder holds an option that distorts the yield. <cite index="13-12,13-13">Z-spread measures the constant spread over the Treasury spot curve that equates a bond's price to its discounted cash flows, but assumes fixed cash flows and does not account for the possibility that cash flows will change if an embedded option is exercised</cite>. That assumption breaks when the option has value.

Sources:

• https://en.wikipedia.org/wiki/Option-adjusted_spread
• https://analystprep.com/study-notes/cfa-level-2/explain-the-calculation-and-use-of-option-adjusted-spreads/
• https://analystnotes.com/cfa-study-notes-option-adjusted-spread.html
• https://www.researchgate.net/publication/237557655_Understanding_the_Z-Spread
• https://ryanoconnellfinance.com/callable-bonds-embedded-options/

<cite index="4-1">Option-adjusted spread is the fixed spread added to one-year forward rates on an interest rate tree that equates the arbitrage-free value and market price of a risky bond with embedded options</cite>. The relationship is algebraic: <cite index="4-18">OAS = Z-Spread minus Option cost</cite>.

<cite index="13-2,13-3">OAS is the spread over the Treasury curve after removing the value of embedded options—for a callable bond, it represents compensation for non-option risks, primarily credit and liquidity</cite>. <cite index="6-4,6-5">OAS compensates for the difference in credit and liquidity risk between the instrument and the benchmark; it is option-adjusted because cash flows on potential rate paths are adjusted to reflect embedded options</cite>.

Calculation requires modeling. <cite index="13-17">OAS is calculated using an interest rate model—typically a lattice or Monte Carlo simulation—that generates multiple interest rate paths</cite>. <cite index="14-1,14-7">OAS is sensitive to interest rate volatility: the higher the volatility, the lower the OAS for a callable bond</cite>. This is intuitive—higher volatility raises the value of the embedded call, which the issuer owns and the bondholder has sold.

<cite index="16-6,16-7,16-8">OAS can assess bond relative value: two bonds with the same characteristics and credit quality must have the same OAS, otherwise the bond with the largest OAS is likely underpriced relative to the bond with the smallest OAS</cite>. That is the practical use case. Z-spread overstates yield on callables because it includes option risk the investor cannot monetize.

Sources:

• https://analystprep.com/study-notes/cfa-level-2/explain-the-calculation-and-use-of-option-adjusted-spreads/
• https://ryanoconnellfinance.com/callable-bonds-embedded-options/
• https://analystnotes.com/cfa-study-notes-option-adjusted-spread.html
• https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/valuation-analysis-bonds-embedded-options

<cite index="8-6,8-7">Z-spread is the constant spread added to each point on the Treasury spot curve that makes the present value of a bond's cash flows equal its market price</cite>. <cite index="7-6">It reflects credit risk, liquidity risk, and prepayment risk</cite>, but assumes those cash flows arrive as scheduled.

Practitioners calculate it by iterative trial-and-error: <cite index="12-1,12-2">add the z-spread to Treasury spot rates for each maturity, discount the bond's cash flows, and solve for the spread that equates present value to market price</cite>. <cite index="10-8,10-9">The method uses the entire yield curve to value individual cash flows, providing more realistic valuation than a single-point interpolated spread</cite>.

<cite index="10-1,10-2">Z-spread serves as the cash benchmark for calculating CDS basis—the difference between the CDS fee and the z-spread for a fixed-rate bond of the same issuer and maturity</cite>. This positions it as the workhorse measure for issuers without embedded optionality. But <cite index="10-10">z-spread does not incorporate variability in cash flows, so valuing bonds with interest-rate-dependent features requires the option-adjusted spread</cite>.

The distinction matters when you price anything callable, putable, or prepayable. Z-spread treats the future as fixed. OAS prices the future as a distribution.

Sources:

• https://en.wikipedia.org/wiki/Z-spread
• https://analystprep.com/cfa-level-1-exam/fixed-income/compare-calculate-and-interpret-yield-spread-measures/
• https://fastercapital.com/content/Corporate-bonds--Analyzing-Z-Spreads-in-the-Corporate-Bond-Market.html
• https://financetrain.com/z-spread-definition-and-calculation

<cite index="16-1,16-2">After laying dormant for more than two decades, the rare disaster framework has emerged as a leading contender to explain facts about the aggregate market, interest rates, and financial derivatives, providing explanations for the equity premium puzzle, the volatility puzzle, return predictability and other features of the aggregate stock market</cite>. <cite index="16-3">These models can also explain violations of the expectations hypothesis in bond pricing, and the implied volatility skew in option pricing</cite>.

<cite index="12-1,12-2">A model of exchange rates based on the hypothesis that the possibility of rare but extreme disasters is an important determinant of risk premia in asset markets</cite> can <cite index="12-3,12-5">account for a series of major puzzles in exchange rates and explain classic exchange rate puzzles and more novel links between options, exchange rates, and stock market movements</cite>.

<cite index="5-4,5-5">Another mystery that may be resolved is why expected real interest rates were low in the United States during major wars such as World War II, even though price-earnings ratios tended also to be low during the wars</cite>. The rare-disaster lens is a unifying instrument: it prices the skew in volatility surfaces, the term structure of safe rates, and cross-sectional equity spreads through the same mechanism.

Sources:

• https://www.nber.org/system/files/working_papers/w20926/w20926.pdf
• https://academic.oup.com/qje/article/131/1/1/2461203
• https://ideas.repec.org/p/nbr/nberwo/11310.html

<cite index="7-2,7-10">While the models of Barro and Rietz address the equity premium puzzle, they do not address the volatility puzzle</cite>. Extensions incorporating <cite index="7-4,7-12">stochastic disaster probability that varies over time</cite> and <cite index="7-5,7-13">recursive preferences rather than power utility preferences</cite> can <cite index="7-6,7-14">generate volatility of stock returns close to that in the data at reasonable values of the underlying parameters</cite>.

<cite index="4-2,4-6">High stock-price volatility can be explained by incorporating time-varying long-run growth rates and disaster probabilities</cite>. <cite index="17-1,17-5">The model generates time variation in the risk premium through Bayesian updating of agents' beliefs regarding the likelihood and severity of disaster realization</cite>.

<cite index="13-1,13-7">One view dating to Rietz, Barro and Gabaix is that rare disasters are pointlike events whose probability is known to investors and therefore impounded in asset valuations</cite>. But <cite index="13-8,13-9">since disasters are rare and heterogeneous events, in reality investors may not know in advance the precise magnitude and persistence of all possible disasters, nor the extent to which different sectors of the economy are resilient, and are likely to gradually learn about them</cite>. This learning mechanism—not just the event itself—moves the price.

Sources:

• https://rodneywhitecenter.wharton.upenn.edu/wp-content/uploads/2014/04/0816.pdf
• https://ideas.repec.org/a/anr/reveco/v4y2012p83-109.html
• https://www.federalreserve.gov/econres/feds/learning-rare-disasters-and-asset-prices.htm
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X23001447

<cite index="1-4">The parameters for a rare disaster are a substantial drop in GDP and at least a 10% decrease in consumption</cite>. <cite index="1-5,14-1">Examples include financial disasters like the Great Depression and the 1997 Asian financial crisis; wars including World War I, World War II, and regional conflicts; epidemics such as influenza outbreaks and the Asian Flu; weather events; and earthquakes and tsunamis</cite>.

<cite index="1-8,14-4,14-5">Barro's model is based upon Lucas's fruit tree model of asset pricing with exogenous, stochastic production; the economy is closed, the number of trees is fixed, output equals consumption, and there is no investment or depreciation</cite>. <cite index="3-8">Barro modeled aggregate output as a random walk with drift consisting of an autoregressive component, a drift and two types of innovations: a shock during normal periods and a jump shock during disastrous contractions</cite>.

<cite index="4-3,4-4">The potential for rare macroeconomic disasters may explain an array of asset-pricing puzzles, with empirical studies relying on long-term data now covering 28 countries for consumption and 40 for GDP</cite>. <cite index="9-1,9-3">The rare disaster hypothesis suggests that the extraordinarily high postwar U.S. equity premium resulted because investors ex ante demanded compensations for unlikely but calamitous risks that they happened not to incur</cite>.

Sources:

• https://en.wikipedia.org/wiki/Rare_disaster
• https://wolfgerome.medium.com/the-equity-risk-premium-puzzle-7bfb359b066f
• https://ideas.repec.org/a/anr/reveco/v4y2012p83-109.html
• https://ideas.repec.org/p/zbw/vfsc14/100614.html

<cite index="1-6,14-2">Rietz first proposed the rare-disaster hypothesis in 1988 as a way to explain the equity premium puzzle</cite>, the finding by <cite index="20-5,20-7">Mehra and Prescott in 1985 that a standard general equilibrium model generated an equity premium of less than 1% for reasonable risk aversion levels</cite>, while <cite index="24-3,24-25">the historical average equity premium stood at 6.18 percent</cite>.

<cite index="2-3,2-4">Barro's extension of Rietz's work explains puzzles including the high equity premium, the low risk-free rate, the volatility of stock returns, and the low values of typical macro-econometric estimates of the intertemporal elasticity of substitution for consumption</cite>. The mechanism is straightforward: <cite index="1-2,1-13">if people are aware that rare disasters may occur, but the disaster never occurs during their lives, then the equity premium will appear high</cite>.

<cite index="4-1,4-5">A baseline model calibrated with observed peak-to-trough disaster sizes accords with the average equity premium with a reasonable coefficient of relative risk aversion</cite>. <cite index="3-2">Barro finds that a coefficient of relative risk aversion of 3.5 is capable of rationalising the global observed, unlevered equity premium of around 5%</cite>, far more plausible than the implausibly high values the original Mehra-Prescott framework required.

Sources:

• https://en.wikipedia.org/wiki/Rare_disaster
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=721085
• https://wolfgerome.medium.com/the-equity-risk-premium-puzzle-7bfb359b066f
• https://ideas.repec.org/a/anr/reveco/v4y2012p83-109.html
• https://en.wikipedia.org/wiki/Equity_premium_puzzle
• https://www.academicwebpages.com/preview/mehra/pdf/The%20Equity%20Premium%20A%20Puzzle.pdf

<cite index="21-1,21-2">Corporate bonds are less liquid than stocks for several structural reasons. First, fragmentation: a single company may have dozens of outstanding bond issues versus one class of common stock, spreading trading activity thinly</cite>.

<cite index="21-7">While stocks trade on centralized exchanges with transparent order books, most bonds trade over-the-counter through a network of dealers</cite>. Each bond issue is a distinct instrument — different maturity, coupon, covenants, issue size. This fragmentation is inherent: firms issue bonds to match liability duration to project cash flows, which produces a long tail of thinly traded instruments. The result is that price discovery happens bilaterally, often via request-for-quote protocols, rather than through a continuous auction.

<cite index="6-1">The literature shows that a bond's size has a positive impact on its liquidity, whereas aging is a negative factor</cite>. Larger issues concentrate trading volume; older issues drift out of dealer inventories as buy-and-hold investors absorb supply. <cite index="21-10">While electronic trading has improved execution in Treasuries and investment-grade corporates, much of the high-yield bond market still trades via voice or chat with dealers</cite>. The microstructure is path-dependent — built around dealer relationships rather than the anonymous limit-order book that defines equity markets.

Sources:

• https://ryanoconnellfinance.com/bond-market-liquidity/
• https://www.researchgate.net/publication/254409296_Multimarket_trading_and_corporate_bond_liquidity

<cite index="22-1,22-13">Dealers with a large share of total market volume provide much of the market's liquidity by taking bonds into inventory</cite>, while <cite index="22-2">dealers are most likely to offset trades quickly, rather than holding bonds overnight or longer, for the least actively traded and riskiest bonds</cite>.

The OTC corporate bond market operates as a core-periphery network. <cite index="23-1">Despite their substantial number and dollar value, many U.S. corporate bonds trade infrequently, if at all</cite>. Dealers face a choice: take an illiquid bond into inventory and bear holding costs, or search actively for an offsetting counterparty. <cite index="22-4">Observed spreads are often lower for riskier bonds and those with little prior trading activity because dealers mitigate inventory risk by more actively searching for a counterparty</cite>.

<cite index="23-6">Liquidity metrics employed in equity markets that gauge trade immediacy in nanoseconds do not hold for bonds, for which immediacy often means hours, if not days</cite>. Post-crisis regulation raised dealer inventory costs; <cite index="26-8">policymakers put in place new regulations following the Global Financial Crisis, including provisions that make it more expensive for OTC bond dealers to hold inventory</cite>. This shifted dealer behavior from providing immediacy via balance sheet to acting as matchmakers — a structural change that alters how liquidity is priced in corporate credit.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0304405X19301394
• https://rpc.cfainstitute.org/research/cfa-digest/2020/11/dig-v50-n11-2
• https://www.philadelphiafed.org/the-economy/banking-and-financial-markets/how-post-global-financial-crisis-regulations-impact-dealer-inventories-and-liquidity

The introduction of TRACE — the Trade Reporting and Compliance Engine — created a natural experiment in price transparency. <cite index="10-12,13-5">Costs are lower for bonds with transparent trade prices, and they drop when the TRACE system starts to publicly disseminate their prices</cite>.

Edwards, Harris, and Piwowar used a difference-in-differences design, comparing transaction costs of bonds newly disseminated under TRACE to control groups that did not change dissemination status. <cite index="15-1,16-1">Other studies using Phase 2 TRACE data report no effect on trading activity and a decline in transaction costs</cite>. The mechanism is straightforward: when public traders observe recent prices, dealers cannot extract rents from information asymmetry.

<cite index="17-6">The results suggest that public traders would significantly benefit if bond prices were made more transparent</cite>. This is not a marginal effect — it is load-bearing for retail and smaller institutional participants who lack the dealer relationships that provide informal price discovery. TRACE rolled out in phases; early phases covered investment-grade and large-issue bonds, while later phases included high-yield and smaller issues. The measured cost decline holds across vintages, though the baseline spreads differ sharply by rating class.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=982654
• https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6261.2007.01240.x
• https://economics.mit.edu/sites/default/files/publications/trace.pdf
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=593823

<cite index="10-9,13-2">Edwards, Harris, and Piwowar (2007) used a complete record of U.S. OTC secondary trades in corporate bonds from TRACE data between January 2003 and January 2005</cite>, estimating transaction costs bond by bond for instruments that traded more than nine times.

<cite index="10-10,13-3">Transaction costs decrease significantly with trade size</cite>. The observed dispersion is substantial: <cite index="7-1">average execution cost for a small retail trade of $5,000 reaches 75 basis points, compared to only 4 basis points for a $10 million institutional trade</cite>. This inverse relationship between size and cost reflects the OTC structure — large institutional traders face lower execution costs, which may make the corporate bond market particularly attractive to informed participants.

<cite index="10-11,13-4">Highly rated bonds, recently issued bonds, and bonds close to maturity carry lower transaction costs than other bonds</cite>. The pattern holds across credit quality: high-yield bonds are more than twice as costly to trade as investment-grade paper. Bond-specific characteristics — issue size, age, time to maturity — matter more than aggregate market conditions in determining the spread a trader will pay. This is the mechanical reality of an OTC market where each instrument trades in its own fragmented pool.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=982654
• https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1540-6261.2007.01240.x
• https://cba.lmu.edu/media/lmucollegeofbusinessadministration/undergrad/05_Sun.pdf

<cite index="9-3">Gerhold (2011) investigates the Longstaff–Schwartz algorithm for American option pricing assuming that both the number of regressors and the number of Monte Carlo paths tend to infinity.</cite> <cite index="9-17">Longstaff and Schwartz propose to approximate the continuation values by a linear combination of basis functions.</cite> The quality of the approximation depends on how well the chosen basis spans the true continuation-value function.

<cite index="12-11">The pseudorandom number generated by Monte Carlo method in the LSM approach shows aggregation so that the convergence is very slow, the simulation variance is large, and the pricing efficiency is low.</cite> <cite index="12-1">Practitioners combine the LSM approach with randomized quasi-Monte Carlo (RQMC) method to price convertible bonds.</cite> <cite index="7-6">Bias can arise when valuing out of the money American options via the Least Square Method.</cite>

The regression introduces estimation error in two directions: too few paths and the fitted continuation value has high variance; too few basis functions and the approximation is biased downward because it cannot represent the true surface. In corporate-debt applications, the state space is high-dimensional — spreads, equity volatility, interest-rate curve, refinancing-window calendar effects — and the basis choice is material. Polynomial bases work when continuation value is smooth in the state; exponential or indicator bases work when there are threshold effects or discrete event risk. The algorithm produces a lower bound on the true option value because the estimated exercise rule is suboptimal; duality methods exist to construct upper bounds and assess estimation quality.

Sources:

• https://arxiv.org/pdf/0802.1831
• https://www.hindawi.com/journals/ddns/2014/301282/
• https://www.semanticscholar.org/paper/Valuing-American-Options-by-Simulation:-A-Simple-Longstaff-Schwartz/284b5b34ed4435438f728a929b86756efafdc862

<cite index="18-2,18-3">The early exercise premium can be expressed in terms of the exercise boundary in the form of an integral, and can be interpreted as the compensation paid to the holder when the early exercise right is forfeited.</cite> <cite index="22-2">A decomposition of the American put option price exists as the sum of its counterpart European price and the early exercise premium.</cite> <cite index="22-3">Compared with the Black-Scholes model, this premium has an additional term due to the presence of jumps.</cite>

<cite index="17-1,17-5">The optimal stopping boundary for the American put option can be characterized as the unique solution of a nonlinear integral equation arising from the early exercise premium representation.</cite> The LSM approach does not solve this integral equation analytically. Instead, it approximates the boundary by regressing realized continuation payoffs against state variables on simulated paths, backward through the exercise grid.

<cite index="24-2,24-3">Practitioners approximate the optimal early-exercise boundary of the option by solving a Hamilton-Jacobi-Bellman partial differential equation in a projected, low-dimensional space, then use the resulting near-optimal early-exercise boundary to produce an exercise strategy for the high-dimensional option.</cite> <cite index="25-2">The correct modelling of discrete dividends is essential for a correct calculation of the early exercise boundary.</cite> In corporate debt with embedded calls, the analog of the dividend is the coupon schedule, and the analog of the exercise boundary is the credit-spread level at which the issuer refinances. LSM estimates that boundary without assuming it takes a parametric form.

Sources:

• https://www.math.hkust.edu.hk/~maykwok/courses/ma571/06_07/Kwok_Chap_5.pdf
• https://link.springer.com/article/10.1007/BF02683325
• https://personalpages.manchester.ac.uk/staff/Goran.Peskir/american.pdf
• https://arxiv.org/pdf/1705.00558
• https://arxiv.org/pdf/1612.03031

<cite index="14-1,14-7">The LSM approach of Longstaff and Schwartz (2001) accounts for stochastic volatility and credit risk</cite>, and <cite index="14-8">provides significant improvements in computational speed and efficiency in handling multiple state variables and path dependencies.</cite> This matters when pricing callable corporate bonds and convertible debt, both of which embed American-style optionality that depends on multiple risk factors evolving jointly.

<cite index="15-1">Convertible bonds are embedded with many options, such as conversion option, call option, put option, and option to lower the conversion price.</cite> <cite index="15-12,15-13">Every moment before maturity, investors and issuers will gamble over the benefit — investors will maximize the value of convertible bonds, while the issuer will minimize the value by exercising the call option.</cite> <cite index="12-9,12-10">Stentoft shows that the LSM method is computationally more efficient than finite difference methods and the Binomial Model when the number of assets is high.</cite>

<cite index="16-1,16-6">Using LSM, practitioners find that the two options embedded in zero callable bonds and callable accreting interest rate swaps have the same exercise strategy since the terms of the swaps will include the bonds in practice.</cite> The regression coefficients encode the issuer's optimal call boundary as a function of the joint state of credit spreads, equity price, and interest-rate term structure. The technique generalizes to structures where the exercise decision depends on path history or realized volatility, which lattice methods cannot handle without dimensional explosion.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0378426618301006
• https://www.hindawi.com/journals/ddns/2014/492134/
• https://www.hindawi.com/journals/ddns/2014/301282/
• https://ideas.repec.org/a/eee/ecofin/v56y2021ics1062940820302242.html

<cite index="1-4,4-1">Longstaff and Schwartz (2001) use least squares to estimate the conditional expected payoff to the optionholder from continuation.</cite> <cite index="6-11,6-12">The approach is called least squares Monte Carlo (LSM) and requires nothing more than simple least squares.</cite> The technique solves the problem that American options present — the holder chooses when to exercise, which makes pricing them a dynamic programming question across every possible future path.

<cite index="1-5,4-4">The method applies readily to path-dependent and multifactor situations where traditional finite difference techniques cannot be used.</cite> <cite index="4-5">Longstaff and Schwartz illustrated the technique with several examples including valuing an option when the underlying follows a jump-diffusion process and valuing an American swaption in a 20-factor string model of the term structure.</cite> <cite index="6-7">This explains why virtually all Wall Street firms value and exercise American swaptions using a simple single-factor model despite clear evidence that the term structure is driven by multiple factors.</cite>

The regression step is backward through time. At each exercise date, the algorithm regresses the discounted continuation payoff against basis functions of the current state variables. The fitted value estimates whether holding is worth more than exercising. The optionholder exercises when immediate payoff exceeds the regression-estimated continuation value. This produces an exercise rule without solving a PDE or building a lattice.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=137399
• https://escholarship.org/uc/item/43n1k4jb
• http://galton.uchicago.edu/~mykland/346W07/Longstaff.pdf

<cite index="4-3,4-7,13-3,13-7">The structures of Kim-Wright and ACM models are very similar, but their long-maturity term premium estimates can differ materially at times.</cite> <cite index="10-11,20-27">Different term premium estimates co-move at low frequencies but may differ markedly in certain periods.</cite> The primary driver of divergence is the treatment of survey data: <cite index="13-4,4-4">the main reason for the divergence between model estimates is that Kim-Wright incorporates Blue Chip surveys of professional forecasters' short rate expectations in the data set used to estimate the model.</cite>

This is not an esoteric methodological dispute. When the curve steepens by 85 basis points—as it did in the second half of 2023—<cite index="21-5">yields-only models of ACM assign a coefficient of 1.30 to the slope's effect on the term premium, while the effect of the slope on expected rates is close to zero or even negative.</cite> Survey-augmented models split that move differently. <cite index="29-8">Both the Kim-Wright full-sample approach and one-sided approaches attributed the rise in the 10-year Treasury yield in 2023 roughly equally to increases in term premiums and expected short-term interest rates.</cite> The choice of model determines whether a selloff prices changed expectations or changed compensation for uncertainty. That distinction is load-bearing for anyone positioning around the next Fed cycle.

Sources:

• https://www.federalreserve.gov/econres/notes/feds-notes/robustness-of-long-maturity-term-premium-estimates-20170403.html
• https://libertystreeteconomics.newyorkfed.org/2014/05/treasury-term-premia-1961-present/
• https://www.federalreserve.gov/econres/notes/feds-notes/the-treasury-tantrum-of-2023-20240903.html
• https://www.federalreserve.gov/econres/feds/files/2024054pap.pdf

<cite index="20-16,20-17">Treasury yields decompose into two components: expectations of the future path of short-term Treasury yields and the Treasury term premium, which is the compensation that investors require for bearing the risk that short-term Treasury yields do not evolve as they expected.</cite> <cite index="24-3">Decomposition models split the nominal yield curve into future short-term interest rate expectations, a term premium that measures bond investor aversion to the risk of holding longer-maturity bonds, and a model residual.</cite> <cite index="15-17,15-18">The term premium is normally thought of as the extra return (a risk premium) that investors demand to compensate them for the risk associated with a long-term bond, but it may also be influenced by supply and demand imbalances for a specific instrument, or several other factors.</cite>

<cite index="15-19">Typically, expected interest rates and term premia are extracted using models based on a small number of risk factors, under the assumption that consistency is maintained between yields at different maturities through the absence of arbitrage opportunities.</cite> These models impose no-arbitrage restrictions which ensure that the time series and cross section of bond yields are consistent with one another. <cite index="21-8,21-9">This decomposition is important because it could have different implications for the response of monetary policy; analysis of the 2023 Treasury tantrum suggests that the rise in yields was primarily driven by an increase in term premiums, fueled by quantitative tightening, greater Treasury issuance, and heightened uncertainty.</cite>

Sources:

• https://libertystreeteconomics.newyorkfed.org/2014/05/treasury-term-premia-1961-present/
• https://www.frbsf.org/research-and-insights/data-and-indicators/treasury-yield-premiums/
• https://www.bis.org/publ/qtrpdf/r_qt1809h.pdf
• https://www.federalreserve.gov/econres/notes/feds-notes/the-treasury-tantrum-of-2023-20240903.html

<cite index="20-20,20-21,20-22">The ACM term premium estimates come from a five-factor, no-arbitrage term structure model; it belongs to the affine class of term structure models which characterize yields as linear functions of a set of pricing factors.</cite> <cite index="12-1,20-3">New York Fed economists Adrian, Crump, and Moench present Treasury term premia estimates for maturities from one to ten years from 1961 to the present, estimated on a daily basis.</cite> Unlike Kim-Wright, <cite index="13-8,4-8">ACM does not incorporate Blue Chip surveys into its baseline estimation; once surveys are added to the ACM model, term premium estimates converge closely to Kim-Wright.</cite>

The difference matters for interpretation. <cite index="10-12,10-28,20-28">The ACM estimate of the term premium has been considerably higher since the onset of the financial crisis than the Kim-Wright or survey-based series.</cite> <cite index="13-14">The Kim-Wright term premium is close to the average of the ACM term premium and a measure computed by simply subtracting survey expectations of short-term interest rates from yields.</cite> <cite index="20-7">The ACM series shows that the term premium is a countercyclical variable which tends to move with measures of uncertainty and disagreement about the future level of yields.</cite>

<cite index="21-5">For yields-only models like ACM, the coefficient of the slope (10-year minus 2-year spread) on the term premium is very high—1.30 for ACM—while the effect of the slope on expected rates is close to zero or negative.</cite> This means that when the curve steepens, ACM attributes nearly all of that move to term premium expansion rather than to changed expectations for the funds path.

Sources:

• https://libertystreeteconomics.newyorkfed.org/2014/05/treasury-term-premia-1961-present/
• https://www.newyorkfed.org/research/data_indicators/term-premia-tabs
• https://www.federalreserve.gov/econres/notes/feds-notes/robustness-of-long-maturity-term-premium-estimates-20170403.html
• https://www.federalreserve.gov/econres/notes/feds-notes/the-treasury-tantrum-of-2023-20240903.html

<cite index="1-1,3-3,3-4">The Kim-Wright model fits a three-factor arbitrage-free term structure model to U.S. Treasury yields since 1990, using latent factors defined statistically rather than economically.</cite> <cite index="3-7">The latent factors are updated daily with new yield data, then used to compute expected future short rates and term premiums.</cite> What sets this model apart from other Fed approaches is that <cite index="4-8,13-8">it incorporates Blue Chip surveys of professional forecasters' short rate expectations into the data set used to estimate the model.</cite>

<cite index="4-11,13-11">Including surveys—published only biannually for the long-term forecast—alleviates small sample problems by providing more information about the long-run average short rate and the speed of reversion to that average.</cite> <cite index="4-13">This makes the expectations component more variable and term premiums less variable, offsetting some of the small-sample bias that would otherwise distend term premium estimates.</cite> The model decomposes observed yields into expected paths and compensation for risk, but the decomposition is a staff research product, <cite index="3-11,3-12">not an official Board statistical release, and is subject to delay, revision, or methodological changes without advance notice.</cite>

The Kim-Wright series publishes daily on FRED for maturities from one to ten years. <cite index="6-7,6-8">The model ascribed a large portion of the decline in long-term yields and distant-horizon forward rates since mid-2004 to a fall in term premiums; about two-thirds of the decline in nominal term premiums owed to a fall in real term premiums, with inflation risk compensation also diminishing.</cite>

Sources:

• https://fred.stlouisfed.org/series/THREEFYTP10
• https://www.federalreserve.gov/data/three-factor-nominal-term-structure-model.htm
• https://www.federalreserve.gov/econres/notes/feds-notes/robustness-of-long-maturity-term-premium-estimates-20170403.html

<cite index="13-1,7-1">Gilchrist and Zakrajšek construct a credit spread index from micro-level data with considerable predictive power for future economic activity.</cite> <cite index="14-3,14-5">The index is constructed from an extensive data set of prices of outstanding corporate bonds trading in the secondary market and is a robust predictor of economic activity at both short- and longer-term horizons.</cite>

The bond-level construction matters. <cite index="2-3,10-1">Each bond's credit spread is the difference between its yield-to-maturity implied by daily price and the yield-to-maturity of a synthetic risk-free security that mimics exactly the cash flows of the corresponding corporate bond.</cite> <cite index="10-2">The synthetic risk-free yield is calculated as the present value of promised cash flows discounted by the term structure of zero-coupon U.S. Treasury yields.</cite> This approach isolates credit risk from duration risk and convexity.

<cite index="11-10">The empirical methodology controls for the call-option effect embedded in most corporate bonds.</cite> The regression-based decomposition then partitions spread into what firm fundamentals predict (distance-to-default, ratings) and what they do not. The aggregate index weights individual bond spreads; the excess bond premium is the average residual. What remains is a time series that moves when the market's tolerance for holding corporate credit shifts, independent of issuer-level deterioration.

Sources:

• https://www.aeaweb.org/articles?id=10.1257%2Faer.102.4.1692
• https://www.riksbank.se/globalassets/media/konferenser/2022/1-sept-2022/paper-gilchrist-yue-wei-zakrajsek.pdf
• https://www.federalreserve.gov/econresdata/notes/feds-notes/2016/recession-risk-and-the-excess-bond-premium-20160408.html

<cite index="1-1,14-7">The predictive content of credit spreads for economic activity is due primarily to movements in the excess bond premium.</cite> <cite index="13-3,7-3">Shocks to the excess bond premium that are orthogonal to the current state of the economy lead to declines in economic activity and asset prices.</cite> <cite index="8-3">Over the past four decades, the predictive power of credit spreads for economic downturns is due entirely to the EBP.</cite>

This is not a claim that default risk does not matter. It is a claim that default risk already reflects information about the economy; the residual does not, and that is what predicts. <cite index="8-2,9-4">The EBP is a component of corporate bond credit spreads not directly attributable to expected default risk and provides an effective measure of investor sentiment or risk appetite in the corporate bond market.</cite> When the premium widens orthogonally to fundamentals, activity contracts.

<cite index="3-1">The excess bond premium predicts macroeconomic movements from 1973 onward.</cite> The Federal Reserve now publishes monthly updates to the series. <cite index="3-3">Greater news attention to financial intermediaries drives up the EBP and portends bad news for the macroeconomy.</cite> The mechanism: intermediaries hold positions; when their capital deteriorates, they tighten supply; the marginal price of holding credit risk rises even when issuer fundamentals have not moved.

Sources:

• https://www.aeaweb.org/articles?id=10.1257%2Faer.102.4.1692
• https://www.nber.org/papers/w17021
• https://www.federalreserve.gov/econresdata/notes/feds-notes/2016/recession-risk-and-the-excess-bond-premium-20160408.html
• https://www.federalreserve.gov/econres/notes/feds-notes/updating-the-recession-risk-and-the-excess-bond-premium-20161006.html
• https://arxiv.org/html/2412.04063v1

<cite index="13-2,7-2">Gilchrist and Zakrajšek decompose observed credit spreads into a component that captures firm-specific expected-default information and a residual component—the excess bond premium.</cite> <cite index="11-8,11-9">The methodology uses distance-to-default (equity valuations and leverage) plus bond-specific credit ratings to model default risk; these factors account for roughly 70 percent of the variation in bond-level credit spreads.</cite> <cite index="14-6">The residual—the excess bond premium—reflects variation in the price of default risk rather than variation in the risk of default.</cite>

The construction is mechanical but the interpretation is not. <cite index="13-4,7-4">An increase in the excess bond premium appears to reflect a reduction in the risk-bearing capacity of the financial sector, which induces a contraction in the supply of credit and a deterioration in macroeconomic conditions.</cite> <cite index="12-2">The EBP measures changing risk attitudes of the marginal investors pricing corporate bonds.</cite> What separates this from standard credit-spread tracking is the decomposition: the default-risk component already knows what the issuer's balance sheet and rating say. What remains is the price the marginal holder demands to hold that risk—a quantity that shifts when their capital is impaired or when their tolerance for volatility contracts.

<cite index="10-1">The bond-level credit spread is constructed as the difference between the bond's yield-to-maturity and the yield of a synthetic risk-free security that mimics exactly the cash flows of the corresponding corporate bond, discounted by the term structure of zero-coupon Treasury yields.</cite> The premium sits on top of what fundamentals predict.

Sources:

• https://www.aeaweb.org/articles?id=10.1257%2Faer.102.4.1692
• https://www.federalreserve.gov/econresdata/notes/feds-notes/2016/recession-risk-and-the-excess-bond-premium-20160408.html
• https://www.riksbank.se/globalassets/media/konferenser/2022/1-sept-2022/paper-gilchrist-yue-wei-zakrajsek.pdf
• https://www.matteoiacoviello.com/research_files/DISCUSS_GZ.pdf

<cite index="24-14,24-15,24-18">In an economy with financial friction, asset prices can have a feedback effect on business cycles—technology shocks can be amplified and propagated by asset prices.</cite> <cite index="1-20">Borrowing decisions are strategic complements: when collateral values fall, this feeds back into the real estate market, driving the price of land down further.</cite>

The propagation is not instantaneous but lagged and persistent. <cite index="6-1,6-2">Because of credit constraints, firms reduce investment; the knock-on effects continue, with the result that a temporary shock in period t reduces constrained firms' demand for land not only in period t but in subsequent periods.</cite> The cycle does not settle at the new equilibrium—it overshoots, then corrects, then overshoots again, with each iteration priced into the next period's collateral base.

What the model demonstrates is that financial frictions do not merely dampen or filter real shocks—they reshape the path and the duration. A shock that would dissipate in three quarters under perfect credit markets persists for years when collateral constraints bind. This is not a theory of crises; it is a theory of why recoveries take longer than the fundamentals would predict.

Sources:

• https://www.nviegi.net/teaching/master/lec06.pdf
• https://kinderhauser.info/kiyotaki-moore-credit-cycles-21/
• https://www-users.york.ac.uk/~psm509/ULB2012/KiyotakiMooreJPE1997.pdf

<cite index="10-8,10-9">Kiyotaki and Moore offered a theory for how shocks to credit-constrained firms are amplified through changes in collateral values; Krishnamurthy showed that their collateral amplification mechanism is not robust to the introduction of markets that allow firms to hedge against common shocks.</cite> <cite index="13-2,13-3">When borrowers can hedge against aggregate shocks at fair prices, the volatility of endogenous variables becomes identical to the first best in the absence of credit constraints—the collateral amplification mechanism disappears.</cite>

The fix requires friction on the hedge itself. <cite index="10-10,10-11">A theory of incomplete hedging was proposed in which the supply of hedging available in the economy is constrained by the aggregate value of collateral; the constraint reinstates amplification effects.</cite> <cite index="13-4,13-5,13-6">Costs of issuing contingent debt can be calibrated to match liquidity and safety premia in data; realistic costs of state-contingent market participation rationalize the predominant use of uncontingent debt, and amplification is restored in such an environment.</cite>

This matters for what we price. If the mechanism holds only when hedging markets are incomplete or costly, then credit-cycle amplification is not a structural feature of collateralized lending—it is a feature of market structure. The constraint binds when the tools to manage it are absent or priced out of reach.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S002205310300098X
• https://www.sciencedirect.com/science/article/abs/pii/S0165188914001237
• https://www.gsb.stanford.edu/faculty-research/publications/collateral-constraints-amplification-mechanism

<cite index="22-4,22-5,22-6">The model assumes borrowers cannot be forced to repay their debts; therefore, in equilibrium, lending occurs only if collateralized—borrowers must own sufficient capital that can be confiscated in case of failure to repay.</cite> <cite index="3-5">The collateral constraint takes the form: debt plus theta times the lowest possible next-period land price times land holdings must be non-negative.</cite>

This constraint binds asymmetrically. <cite index="20-19,20-20,20-21">A temporary productivity shock reduces all agents' net worth, but productive agents with debt obligations from previous periods see their net worth fall more severely than unproductive agents, forcing productive agents to cut investment more than aggregate saving declines, and the average productivity of investment falls.</cite>

The dual role of assets is the mechanism. <cite index="17-1">Movements in asset prices interact with the real side of the economy and produce amplified and persistent effects of shocks.</cite> Land is not just land—it is borrowing capacity. When asset values distend, credit expands. When they compress, the constraint tightens faster than fundamentals would predict. The economy moves not in response to the shock, but in response to what the shock does to the constraint.

Sources:

• https://en.wikipedia.org/wiki/Kiyotaki%E2%80%93Moore_model
• https://www.gdsge.com/example/KM1997/KM1997.html
• https://swh.princeton.edu/~kiyotaki/papers/Credit-and-BusinessCycles.pdf

<cite index="6-4,6-5">Kiyotaki and Moore construct a dynamic economy in which lenders cannot force borrowers to repay unless debts are secured; durable assets serve both as productive capital and as collateral for loans.</cite> <cite index="16-16">The dynamic interaction between credit limits and asset prices turns out to be a powerful transmission mechanism by which the effects of shocks persist, amplify, and spill over to other sectors.</cite>

The structure holds steady until it does not. <cite index="22-7">In a recession, income from capital falls, causing capital prices to fall, which makes capital less valuable as collateral, which limits firms' investment by forcing them to reduce borrowing, and thereby worsens the recession.</cite> The mechanism feeds forward: <cite index="2-10,2-11">The model uses an analogy where land holdings correspond to the deer and debts to the wolves—a rise in firms' land holdings means they have more collateral against which to borrow.</cite>

What matters is not the size of the shock. <cite index="6-7">Small, temporary shocks to technology or income distribution can generate large, persistent fluctuations in output and asset prices.</cite> <cite index="22-9,22-10">Earlier real business cycle models relied on large exogenous shocks to account for output fluctuations; Kiyotaki-Moore shows instead how relatively small shocks might suffice if credit markets are imperfect.</cite> The amplification is endogenous, priced into the constraint itself.

Sources:

• https://faculty.wcas.northwestern.edu/lchrist/d11/d1118/kiyotaki-moore.pdf
• https://www.minneapolisfed.org/research/conferences/research-events---conferences-and-programs/~/media/files/research/events/1993_09-17/Moore_CreditCycles.pdf
• https://collaborate.princeton.edu/en/publications/credit-cycles/
• https://en.wikipedia.org/wiki/Kiyotaki%E2%80%93Moore_model

The Rigobon-Sack result estimates the Fed's response to equity moves, not its leaning against bubbles or financial imbalances. <cite index="2-5">It appears that the Federal Reserve systematically responds to stock price movements only to the extent warranted by their impact on the macroeconomy.</cite> This is consistent with a standard Taylor rule augmented by a wealth-effect channel, but inconsistent with a financial-stability mandate that would tighten preemptively when equity valuations distend.

<cite index="20-10,20-11,20-12">We measure how responsive the Federal Reserve's policy appears to be to imbalances in the equity, housing and credit markets. We find that changes in these policy sensitivities predict the later development of financial imbalances. When monetary policy appears to respond more countercyclically to market overheating, imbalances tend to decline over time.</cite> That result (from a 2019 BIS working paper) tests a different hypothesis: whether the anticipated reaction function itself disciplines froth. Rigobon-Sack captures the Fed's conditional response given a move; the BIS result tests whether forward-looking agents price the rule itself.

The difference matters for stress pricing. If the Fed responds only to realized wealth effects, then positioning ahead of a drawdown does not anticipate preemptive tightening — it anticipates cuts once the drawdown materializes and consumption slows. That asymmetry has held through every post-1987 equity dislocation except the 2022 inflation episode, when the Fed held through a 25% S&P decline because inflation had not yet settled.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=275424
• https://www.bis.org/publ/work816.pdf

<cite index="12-7,12-8,12-9">This paper develops a method for solving the identification problem that arises in simultaneous-equation models. It is based on the heteroskedasticity of the structural shocks. For simplicity, I consider heteroskedasticity that can be described as a two-regime process and show that the system is just identified.</cite>

Rigobon's 2003 Review of Economics and Statistics paper formalizes the method deployed in the Rigobon-Sack equity-Fed study. The intuition: if you observe two regimes in which variance of shock A shifts but variance of shock B does not, the covariance structure across regimes reveals the structural parameters linking A and B.

<cite index="14-2">We discuss general conditions for identification and one of the results shows that an adequate number of different levels of heteroskedasticity is sufficient to identify the parameters of the structural form without the inclusion of any kind of restriction.</cite> In a two-variable, two-regime case the system is exactly identified; additional regimes or additional variables can yield overidentification and testable restrictions.

The appeal: no need for instruments, no exclusion restrictions, no timing assumptions about "fast" versus "slow" variables. The cost: the method requires ex ante knowledge or justification of when regimes shift. In the Fed-equity application, high-volatility equity windows (October 1987, 1998 LTCM) provide the regime contrast. Whether the structural parameters are stable across those regimes is an assumption, not a test.

Sources:

• https://direct.mit.edu/rest/article-abstract/85/4/777/57415/Identification-Through-Heteroskedasticity
• https://ideas.repec.org/p/mil/wpdepa/2011-19.html
• https://www.researchgate.net/publication/24095868_Identification_Through_Heteroskedasticity

<cite index="2-3,2-4,2-5">The results indicate that monetary policy reacts significantly to stock market movements, with a 5% rise (fall) in the S&P 500 index increasing the likelihood of a 25 basis point tightening (easing) by about a half. This reaction is roughly of the magnitude that would be expected from estimates of the impact of stock market movements on aggregate demand. Thus, it appears that the Federal Reserve systematically responds to stock price movements only to the extent warranted by their impact on the macroeconomy.</cite>

This is the central empirical finding. It offers a unit of measurement for the Fed's implicit equity coefficient — not in a Taylor rule, but in the policy reaction observed during the sample (1985–1999). The result does not support the "Fed put" in the activist sense; it supports a Fed that weighs equity moves only insofar as they alter the output gap and consumption via wealth effects.

The magnitude matters for pricing. If the market anticipates a 10% drawdown, the corresponding expectation should be roughly a 50bp cut over the cycle — contingent on inflation still anchoring near target. The result also implies that the Fed does not lean against equity bubbles for financial-stability reasons; it responds only when the equity move feeds through to aggregate demand. This held through 1999. Whether it held through 2008, 2020, or 2022 is a separate question, and would require re-estimation with updated heteroskedasticity regimes.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=275424
• https://www.nber.org/papers/w8350

<cite index="5-9,5-10,5-11">Movements in the stock market can have a significant impact on the macroeconomy and are therefore likely to be an important factor in the determination of monetary policy. However, little is known about the magnitude of the Federal Reserve's reaction to the stock market. One reason is that it is difficult to estimate the policy reaction because of the simultaneous response of equity prices to interest rate changes.</cite>

Rigobon and Sack (2003) confront the simultaneity head-on. Equities respond to rate expectations; rate decisions respond to equity-implied growth and wealth effects. The conventional approach — instrument with lags or narrative shocks — either assumes away the reverse causality or discards most of the variation in the data.

<cite index="2-2">This paper uses an identification technique based on the heteroskedasticity of stock market returns to identify the reaction of monetary policy to the stock market.</cite> The logic: structural shocks carry stable covariance; reduced-form shocks inherit variance from regime-dependent data-generating processes. <cite index="13-8,13-9">For simplicity, I consider heteroskedasticity that can be described as a two-regime process and show that the system is just identified. I discuss identification under general conditions, such as more than two regimes, when common unobservable shocks exist, and situations in which the nature of the heteroskedasticity is misspecified.</cite>

The method does not require exclusion restrictions, instruments, or narrative event studies. It requires only that the volatility of one shock shift without the other — plausibly satisfied when comparing calm Fed windows with high-volatility equity regimes.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=275424
• https://direct.mit.edu/rest/article-abstract/85/4/777/57415/Identification-Through-Heteroskedasticity

<cite index="5-13">Brunnermeier, Nagel, and Pedersen conjecture that sudden exchange-rate moves unrelated to news can be due to the unwinding of carry trades when speculators near funding constraints.</cite> The timing is endogenous to the structure of leverage, not to the macro calendar. <cite index="10-10,10-11,10-12,10-13,10-14,10-15">When interest rates start to converge, the attractiveness of the carry trade diminishes, leading to a mass exit; this mass unwinding can create significant price volatility as traders rush to close their positions; leverage can play a critical part in the unwinding due to auto-liquidation and margin calls—when the market moves against a trader's leveraged position, the account's equity may fall below the required margin level, triggering a margin call; if the trader cannot replenish the margin, brokers automatically liquidate positions to limit losses, exacerbating market volatility, and this forced selling can lead to rapid declines in the value of the traded currency pair, creating a cascading effect.</cite>

<cite index="14-8,14-12">USD/JPY weakness can be driven by positioning and leverage rather than a sudden shift in U.S. fundamentals; even modest shifts in expectations can trigger outsized moves as positions unwind.</cite> The move precedes the narrative. What speculators price is not yet what macro data is saying—until the forced unwind makes the macro data irrelevant to the next tick.

Sources:

• http://inctpped.ie.ufrj.br/spiderweb/dymsk_3/3-6S%20Brunnermeier-Nagel-Pedersen.pdf
• https://www.tastyfx.com/news/how-the-carry-trade-works-and-position-unwinds-explained/
• https://www.stonex.com/en/insights/yen-carry-trade-unwind-risk-is-re-emerging-in-fx-markets/

<cite index="11-2">A well established pattern is that spikes in volatility go hand in hand with deleveraging pressures and the unwinding of currency carry trades.</cite> The August 2024 episode confirmed the mechanics. <cite index="11-10,11-11,11-13">Financial market volatility resurfaced in early August as the unwinding of leveraged trades in equity and currency markets amplified the initial reaction to a negative macro release in the United States; FX carry trades were hit hard by the deleveraging pressures, with various estimates based on both on- and off-balance sheet activity yielding a rough middle ballpark of ¥40 trillion ($250 billion) going into the event.</cite>

<cite index="13-3,13-4,13-5">When the yen strengthens unexpectedly, leveraged positions face dual pressures from both reduced interest income and capital losses on the currency component of their trades; margin call cascades become particularly problematic during rapid yen appreciation, as institutions must purchase yen to meet funding obligations while simultaneously facing mark-to-market losses on their existing positions, creating a self-reinforcing cycle where forced buying pressure accelerates the very currency moves that triggered initial position stress.</cite> <cite index="14-2,14-4">Historical precedent shows that yen carry trade unwinds rarely remain confined to currency markets; funding stress has historically spilled beyond foreign exchange into equities and volatility.</cite> What the BIS and practitioners observed in August 2024 was the Brunnermeier-Nagel-Pedersen framework pricing in real time.

Sources:

• https://www.bis.org/publ/bisbull90.pdf
• https://discoveryalert.com.au/yen-carry-trade-unwinding-risks-2025/
• https://www.stonex.com/en/insights/yen-carry-trade-unwind-risk-is-re-emerging-in-fx-markets/

<cite index="20-2,20-3,20-4,20-5,20-6">Brunnermeier and Pedersen show that under certain conditions, margins are destabilizing and market liquidity and funding liquidity are mutually reinforcing, leading to liquidity spirals: traders provide market liquidity, and their ability to do so depends on their availability of funding; conversely, traders' funding—their capital and margin requirements—depends on the assets' market liquidity.</cite>

<cite index="22-5,22-6,22-7,22-8">Two self-reinforcing liquidity spirals emerge: a margin spiral and a loss spiral. Initial losses create funding problems for traders forcing them to reduce their positions, which drives prices lower; as prices and collateral values fall, traders are forced to put up more margin to finance their positions, prompting further asset sales to raise capital, which drives prices down even further, adding to losses on existing positions.</cite> <cite index="19-2,19-11,19-12">When markets are illiquid, market liquidity is highly sensitive to further changes in funding conditions; in one equilibrium markets are illiquid, resulting in larger margin requirements, thus restricting speculators from providing market liquidity—and any equilibrium selection has the property that small speculator losses can lead to a discontinuous drop of market liquidity.</cite> This is the fragility that matters when positioning is crowded and volatility shifts regime.

Sources:

• https://markus.scholar.princeton.edu/publications/market-liquidity-and-funding-liquidity
• https://thehedgefundjournal.com/the-liquidity-crunch/
• https://www.princeton.edu/~markus/research/papers/liquidity.pdf

<cite index="1-18,1-19">Brunnermeier, Nagel, and Pedersen document that carry traders are subject to crash risk: exchange rate movements between high-interest-rate and low-interest-rate currencies are negatively skewed, driven by sudden unwinding of carry trades during periods when risk appetite and funding liquidity decrease.</cite> The structure of the position explains the asymmetry. <cite index="8-2,8-3">Securities that speculators invest in have a positive average return and a negative skewness; the positive return compensates for providing liquidity, while the negative skewness arises from an asymmetric response to fundamental shocks—shocks that lead to speculator losses are amplified when speculators hit funding constraints and unwind their positions, further depressing prices and increasing funding problems, volatility, and margins.</cite>

<cite index="1-11">Funding liquidity measures predict exchange rate movements, and controlling for liquidity helps explain the uncovered interest rate puzzle.</cite> The observation holds beyond FX. <cite index="1-22">Brunnermeier, Nagel, and Pedersen also document excess comovement among currencies with similar interest rates</cite>—positions clustered by carry direction unwind in tandem. <cite index="1-21">Carry trade losses reduce future crash risk but increase the price of crash risk.</cite> What settles after the unwind is a market repricing the tail, not the tail disappearing.

Sources:

• https://collaborate.princeton.edu/en/publications/carry-trades-and-currency-crashes/
• https://www.journals.uchicago.edu/doi/10.1086/593088

<cite index="20-3,20-4">In a financial system in which balance sheets are continuously marked to market, asset price changes appear immediately as changes in net worth, prompting financial intermediaries to adjust the size of their balance sheets; marked-to-market leverage is strongly procyclical and such behavior has aggregate consequences.</cite> <cite index="20-8">The margin of adjustment on the balance sheet is through repos and reverse repos and other collateralized borrowing and lending.</cite>

This creates a feedback loop. Rising asset prices increase net worth; rising net worth creates room to expand the balance sheet; balance sheet expansion bids up asset prices further. The reverse holds in contraction. <cite index="4-8,4-3">Mark-to-market accounting has potentially important implications for financial cycles; it may at first appear to be an esoteric question on measurement.</cite> But it is not. It is the reason that dealer-sector leverage is procyclical rather than countercyclical.

<cite index="10-7">Adrian and Shin (2007) showed the scatter chart of the weighted average of the quarterly change in assets against the quarterly change in leverage of the then-five stand-alone US investment banks.</cite> Assets and leverage move together, not inversely. Equity is sticky. When balance sheets expand, they expand through debt issuance or repo borrowing. When they contract, they contract through asset sales or repo unwind. This is the empirical regularity that separates the modern securitized intermediary from the traditional deposit-funded bank. It is also what makes the system vulnerable to synchronized deleveraging.

Sources:

• http://qed.econ.queensu.ca/pub/faculty/milne/872/Adrian%20and%20Shin%202008.pdf
• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr360.pdf

<cite index="13-1,13-13">Dealer balance sheet changes primarily forecast changes in the volatility risk premium, which has a natural interpretation as the price of risk.</cite> Adrian and Shin demonstrate this using weekly primary dealer repo data compiled by the New York Fed. <cite index="13-11,13-12">They show that changes in collateralized borrowing and lending on intermediaries' balance sheets are significant forecasting variables for innovations in the VIX index of implied volatility in the stock market, and they decompose VIX innovations into changes of stock market volatility and changes of the volatility risk premium.</cite>

The economic story is straightforward. <cite index="13-3,13-4,13-5">Aggregate liquidity can be understood as the rate of growth of the aggregate financial sector balance sheet; when asset prices increase, intermediaries' balance sheets generally become stronger, their leverage tends to be too low without adjusting asset holdings, and they hold surplus capital which they will attempt to employ.</cite> That capital flows into risk-bearing positions. The price of bearing risk—measured through implied volatility premia—declines as capacity expands.

This empirical regularity matters because it inverts the usual read. Volatility is not exogenous to intermediation; it is partially produced by the aggregate state of dealer balance sheets. When dealers are constrained—by higher haircuts, tighter capital, or reduced willingness to warehouse risk—volatility rises even if fundamental uncertainty has not changed. The cycle feeds on itself.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S1042957308000764
• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr328.pdf

<cite index="10-1,10-2">If the repo haircut is 2%, the borrower can borrow 98 dollars for 100 dollars worth of securities pledged, meaning the borrower must come up with 2 dollars of equity to hold 100 dollars worth of securities.</cite> <cite index="10-11">The fluctuations in the haircut largely determine the degree of funding available to a leveraged institution, since the haircut determines the maximum permissible leverage achieved by the borrower.</cite>

This is the load-bearing mechanism in the Adrian-Shin framework. Haircuts do not move smoothly or symmetrically. <cite index="12-2">When asset prices are on the rise, repo investors allow broker-dealers to borrow not only at a lower rate but also with a lower haircut, which enables borrowers to gain leverage further.</cite> <cite index="12-3,12-4">In times of stress, investors quickly pull back as the probability of default increases; both the repo rate and haircut rate increase, which forces securitized banks to deleverage sharply.</cite>

<cite index="10-4,10-5,10-6">Broker-dealers have balance sheets consisting of marketable claims or short-term items that are marked to market; their importance in the supply of credit has increased in step with securitization, and they may be seen as a barometer of overall funding conditions in a market-based financial system.</cite> When the haircut window tightens, the entire dealer sector contracts in unison. The price of risk—what Adrian and Shin measure through the volatility risk premium—moves with that contraction.

Sources:

• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr360.pdf
• https://www.sciencedirect.com/science/article/abs/pii/S0954349X17301820

<cite index="18-4">In a financial system where balance sheets are continuously marked to market, asset price changes appear immediately as changes in net worth, eliciting responses from financial intermediaries who adjust the size of their balance sheets.</cite> This is the core mechanism Adrian and Shin (2008) documented in Fed Staff Report 328. <cite index="18-5,18-1">They observed that marked-to-market leverage is strongly procyclical</cite>—when assets appreciate, intermediaries do not hold equity steady and allow leverage to decline. Instead, they expand the balance sheet.

<cite index="18-7,14-1">Changes in dealer repos—the primary margin of adjustment for the aggregate balance sheets of intermediaries—forecast changes in financial market risk as measured by the innovations in the VIX.</cite> This empirical regularity matters because it reframes what "liquidity" means. <cite index="5-12,5-13">Liquidity should be understood as a flow—the rate of growth of balance sheets—rather than as a stock.</cite> <cite index="5-10,5-11">When haircuts rise, all balance sheets shrink in unison, resulting in a generalized decline in the willingness to lend; liquidity disappears altogether rather than being re-allocated elsewhere.</cite>

The implication is that aggregate liquidity conditions cannot be understood separately from dealer capacity to hold positions. When dealers are expanding—financing more through repo, building inventory—VIX tends to compress. When they contract, volatility rises not as a passive reflection of risk but as a consequence of reduced intermediation.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1139857
• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr360.pdf
• https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr328.pdf

<cite index="22-2,22-6">Mispricing may persist when arbitrageurs are limited by the availability of substitutes and financial constraints; stocks with closer substitutes have lower firm-specific mispricing, while financial constraints have a positive impact on mispricing</cite>. <cite index="23-1">Arbitrageurs will not exploit opportunities if the costs and risk of arbitrage exceed its benefits, allowing mispricing to survive for long periods</cite>.

<cite index="18-1,18-3">Limits to arbitrage refer to constraints that prevent rational investors from fully exploiting mispricings; these can be fundamental (transaction costs, short-selling constraints) or non-fundamental (investor sentiment, noise traders)</cite>. <cite index="18-4,18-5">The persistence of mispricings due to these limits challenges the strict Efficient Market Hypothesis; inability to correct mispricings can lead to periods where asset prices deviate significantly from intrinsic value</cite>.

<cite index="19-2,19-9">In principle, any example of persistent mispricing is evidence of limited arbitrage—if arbitrage were not limited, the mispricing would quickly disappear</cite>. <cite index="24-1">Mispricing is greater when deal and firm characteristics exacerbate the limits of arbitrage, and it weakens over time</cite>. The presence of substitutes matters because the arbitrageur needs an instrument to hedge the fundamental exposure while isolating the mispricing—without substitutes, the position carries unhedgeable risk that pricing may move against the arbitrageur for reasons unrelated to convergence.

Sources:

• https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID2808817_code1916270.pdf?abstractid=2808817
• https://www.efmaefm.org/0EFMSYMPOSIUM/2006/papers/99-EFM06%20-Zhang-The%20Limits%20of%20Arbitrage.pdf
• https://diversification.com/term/limits-to-arbitrage
• https://www.researchgate.net/publication/333670504_LIMITS_TO_ARBITRAGE_A_SURVEY_OF_LITERATURE
• https://ideas.repec.org/a/eme/rbfpps/rbf-01-2021-0005.html

<cite index="20-2,20-3,20-4">Theories of limits to arbitrage hold that mispricing persists because arbitrageurs are financially constrained; when an asset becomes severely underpriced, arbitrageurs incur large losses and are forced to sell to meet redemptions and margin requirements</cite>. <cite index="20-5">Because all arbitrageurs face the same situation, there is a lack of buyers, and the asset price decreases further</cite>—a dynamic the literature terms a lack of "funding liquidity" compounding a lack of "market liquidity."

<cite index="20-6,20-7">The mechanism depends on arbitrageurs being unable to raise external funding when they make temporary losses; if they could, temporary underpricing would lead them to buy more, not less</cite>. <cite index="20-10">Limits-to-arbitrage models explain the persistence of underpricing, not overpricing—the predictions are asymmetric</cite>. Empirically, <cite index="21-6,21-7,21-8">mispricing does not converge 30% of the time in negative stub value cases; arbitrage profits are 50% lower if the convergence path is volatile rather than smooth, and convergence averages 236 days, ranging from 1 day to 2,796 days</cite>.

The constraint is not only that mispricings exist but that their correction is path-dependent. An arbitrageur holding an underpriced position that worsens before it corrects faces redemptions from investors who observe mark-to-market losses, even when the thesis remains intact. The capital structure of arbitrage—performance-based, externally funded—makes correction slower and less reliable when it is most required.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0304405X13002493
• https://digital.library.unt.edu/ark:/67531/metadc848132/m2/1/high_res_d/ALSHAMMASI-DISSERTATION-2015.pdf

<cite index="11-2">Noise trading is defined as trading uncorrelated with changes in fundamental or intrinsic value</cite>, occurring for reasons ranging from portfolio rebalancing to sentiment-driven speculation. <cite index="11-4,11-5,11-6">Analysis of Royal Dutch/Shell and Unilever NV/PLC pairs provides evidence on two facets: the fraction of total return variation unrelated to fundamentals (noise), and the short-run risk borne by arbitrageurs in long-short pairs trading</cite>. <cite index="11-7">Approximately 15% of weekly return variation is attributable to noise</cite>.

<cite index="11-8">Noise trader risk has both systematic and idiosyncratic components and varies considerably over time</cite>. <cite index="11-9,11-10">The conditional volatility of long-short portfolio returns ranged from 0.5% to over 2.75% per week during 1989–2003, with especially high levels around the LTCM failure in 1998 and the technology bubble collapse in 2000</cite>. <cite index="13-9">Once a position is taken, noise traders may drive prices farther from fundamental value, forcing the arbitrageur to commit additional capital that may not be available, or liquidate early</cite>.

<cite index="17-1,17-2">Noise traders impose on others the risk that prices might move unpredictably and irrationally; this risk can discourage arbitrageurs from acting to exploit price deviations from fundamentals</cite>. The observed persistence: mispricings can remain for extended periods because the arbitrageur's solvency constraint binds before the market's irrationality constraint does.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S1386418106000310
• https://alphaarchitect.com/introduction-behavioral-finance-part-2-limits-arbitrage/
• https://www.sciencedirect.com/topics/economics-econometrics-and-finance/noise-trading

<cite index="3-2,3-5">Shleifer and Vishny (1997) observe that real-world arbitrage requires capital and entails risk</cite>, a departure from the frictionless textbook model. <cite index="3-6">Professional arbitrage is conducted by a small number of specialized investors using other people's capital</cite>—a structural fact that constrains the mechanism by which prices settle to fundamental value.

<cite index="3-7">The model demonstrates that arbitrage becomes ineffective in extreme circumstances when prices diverge far from fundamentals</cite>. <cite index="5-2,5-3">In traditional models, arbitrage is performed by many diversified investors taking small positions; in reality, it is conducted by a few specialized investors taking large positions with other people's money</cite>. This principal-agent structure matters: <cite index="7-4,7-5">investing heavily in an underpriced asset exposes arbitrageurs to the risk of large outflows by investors if underpricing worsens, depriving them of capital when they need it most</cite>.

<cite index="6-2,6-3">For specialized arbitrageurs, idiosyncratic volatility matters more than systematic, because it cannot be hedged and arbitrageurs are not diversified</cite>. <cite index="6-10,6-14">Stocks are not rationally priced, idiosyncratic risk deters arbitrage, and volatile securities will exhibit greater mispricing and higher average return to arbitrage</cite>. The implication is structural: arbitrage does not necessarily tighten spreads when spreads widen—it can withdraw precisely when prices need correction most.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=8043
• https://www.nber.org/papers/w5167
• https://www.bartlit-beck.com/assets/htmldocuments/limits%20of%20the%20limits.pdf
• https://www.researchgate.net/publication/227368095_Limits_of_Arbitrage

<cite index="1-3">Krishnamurthy and Vissing-Jørgensen analyze disaggregated data from the Flow of Funds Accounts of the Federal Reserve and show that individual groups of Treasury holders also have downward sloping demand curves.</cite> <cite index="4-15,4-16">State/local retirement funds, private pensions, and insurance companies are an interesting group, as they have a long-term objective and no explicit regulatory requirements, so the liquidity and neutrality motives are unlikely to be important for them.</cite> <cite index="4-12,4-13,4-14">Some foreign investors desiring to hold assets in US Dollars may choose to hold the dollars in the form of Treasury securities rather than corporate assets.</cite>

The heterogeneity in motives means that different holders respond differently to price. When foreign central banks accumulate reserves, they absorb Treasuries with very inelastic demand — their objective is not yield maximization. When pension funds rebalance, they price safety against duration risk. The composition of the holder base, not just the aggregate quantity outstanding, determines how much convenience premium the market will bear at a given Debt/GDP ratio. Flow of Funds data on Treasury holdings by sector is the input for estimating who is marginal when supply shifts.

Sources:

• https://papers.ssrn.com/sol3/Delivery.cfm/nber_w12881.pdf?abstractid=960455
• https://www.nber.org/system/files/working_papers/w12881/w12881.pdf
• https://conference.nber.org/confer/2008/mes08/krishnamurthy.pdf

<cite index="24-1">Mortgage-backed securities purchases in QE1 were crucial for lowering MBS yields as well as corporate credit risk and thus corporate yields for QE1, while Treasuries-only purchases in QE2 had a disproportionate effect on Treasuries and Agencies relative to MBS and corporates, with yields on the latter falling primarily through the market's anticipation of lower future federal funds rates.</cite> <cite index="22-6">Treasuries-only QE has most significant effects on yields of near-zero-default risk assets — Treasuries, Agencies, Aaa bonds — with little effect on Baa or mortgage yields which may be more policy relevant.</cite>

The safety-channel prediction is precise: <cite index="21-6,21-7">QE decreases supply of long-term safe assets, and the safety-channel implies that QE involving Treasuries and agencies lowers the yields on very safe assets, but has no effects on lower-grade debt such as Baa bonds or bonds with prepayment risk such as MBS.</cite> This matters when evaluating central bank transmission. Pulling duration out of the market does not flatten spreads by itself. If QE removes the safest paper, it tightens the convenience premium and widens IG spreads unless the signaling channel — expected rate cuts — offsets the scarcity effect.

Sources:

• https://www.nber.org/papers/w17555
• https://www.frbsf.org/wp-content/uploads/krishnamurthy.pdf
• https://www.chicagofed.org/-/media/others/events/2011/international-conference/vissing-jorgensen-pdf.pdf
• https://www.brookings.edu/wp-content/uploads/2016/07/2011b_bpea_krishnamurthy.pdf

<cite index="6-3,6-7">Both the liquidity and safety attributes of Treasuries are driving the convenience-yield phenomenon.</cite> <cite index="6-8">The authors document this by analyzing the spread between assets with different liquidity (but similar safety) and those with different safety (but similar liquidity).</cite> <cite index="16-3,16-4">The convenience yield, measured as the spread between Aaa corporate bond yields and the 10-year Treasury bond, presents an upward trend from the early 1980s to the early 2000s, increasing by about 1 percentage point over the period.</cite>

<cite index="18-12">Assets offering an almost sure promise of nominal repayment are especially valuable because they are used as collateral in financial transactions with very low haircuts and because for regulatory, institutional, or informational reasons certain investors have investment needs that can only be satisfied by holding safe assets.</cite> The spread measure has policy content: when the convenience yield widens, Treasuries are priced at a larger discount to fundamentals, and the private sector is pricing a scarcity of safe collateral. <cite index="1-7">The results have implications for the financing of the US deficit, Ricardian equivalence, and the effects of foreign central bank demand on Treasury yields.</cite>

Sources:

• https://research-api.cbs.dk/ws/portalfiles/portal/60084038/aggregate_demand_annette_vissing_2_.pdf
• https://www.federalreserve.gov/econres/notes/feds-notes/convenience-yield-as-a-driver-of-r-20240903.html
• https://www.kansascityfed.org/documents/4563/2013Krishnamurthy.pdf
• https://papers.ssrn.com/sol3/Delivery.cfm/nber_w12881.pdf?abstractid=960455

<cite index="5-3,5-4,5-5">Krishnamurthy and Vissing-Jørgensen establish that the US Debt/GDP ratio is negatively correlated with the spread between corporate bond yields and Treasury yields, even when controlling for default risk on corporate bonds, and argue that the corporate bond spread reflects a convenience yield that investors attribute to Treasury debt.</cite> <cite index="1-1">The aggregate demand curve for the convenience provided by Treasury debt is downward sloping, with estimates of the elasticity of demand.</cite> <cite index="6-2,6-6">Treasury yields are reduced by 73 basis points, on average, from 1926 to 2008, as a result of changes in Treasury supply having large effects on a variety of yield spreads.</cite>

The decomposition matters because it separates what moves yields from rate expectations alone. <cite index="1-3,1-4">Individual groups of Treasury holders also have downward sloping demand curves, and even groups with the most elastic demand curves have demand curves that are far from flat.</cite> <cite index="11-3">A hypothetical rise in the debt-to-GDP ratio from 0.38 to 0.39 will decrease the spread between corporate bond yields and Treasury bond yields between 1.5 basis points and 4.25 basis points.</cite> The implication is that fiscal expansion — adding Treasury supply — tightens the convenience premium and widens spreads, even if the Fed holds the policy rate constant.

Sources:

• https://www.nber.org/papers/w12881
• https://papers.ssrn.com/sol3/Delivery.cfm/nber_w12881.pdf?abstractid=960455
• https://research-api.cbs.dk/ws/portalfiles/portal/60084038/aggregate_demand_annette_vissing_2_.pdf
• https://insight.kellogg.northwestern.edu/article/treasury_debt_and_corporate_bond_rates

<cite index="14-1,14-3">Endogenous procyclical movements in bank balance sheets lead to countercyclical movements in the cost of bank credit.</cite> <cite index="18-1,18-2">Balance sheet conditions not only affect the cost of bank credit, they also affect whether runs are possible. In this respect one can relate the possibility of runs to macroeconomic conditions and in turn characterize how runs feed back into the macroeconomy.</cite>

<cite index="11-1,11-2">Balance sheet constraints on banks may limit real investment spending, affecting aggregate real activity. A crisis is possible where weakening of bank balance sheets significantly disrupts credit flows, depressing real activity.</cite> <cite index="15-2,15-4">In the Gertler-Karadi framework, capital quality shock enters through the physical capital accumulation process which originates in the non-financial sector and affects the asset side of bank balance sheets through the change in collateral value. Banks play amplification roles for the shock that originates elsewhere in the economy.</cite>

The transmission runs both directions. A shock to real activity weakens bank capital, which tightens credit, which depresses activity further. The cycle is self-reinforcing until balance sheets stabilize. This is not a friction that dampens volatility—it amplifies it, and the amplification is largest when the economy is weakest.

Sources:

• https://www.nber.org/system/files/working_papers/w19129/w19129.pdf
• https://www.princeton.edu/~kiyotaki/papers/BankingLiquidityBankruns12.pdf
• https://www.princeton.edu/~kiyotaki/papers/GertlerKiyotakiQueraltoJune7wp.pdf
• https://www.sciencedirect.com/science/article/abs/pii/S0264999316307519

<cite index="23-3,23-4">Bernanke and Gertler (1989) showed how the effects of a real shock (such as a shock to productivity) on financial conditions could lead to persistent fluctuations in the economy, even if the initiating shock had little or no intrinsic persistence. A key concept in the analysis was the external finance premium, defined as the difference between the cost to a borrower of raising funds externally and the opportunity cost of internal funds.</cite>

<cite index="21-3,21-4">The key idea behind the financial accelerator is that shocks to the net worth of firms and households have a procyclical effect on their borrowing capacity. This could happen either because the information cost wedge between external and internal finance moves countercyclically, or because a procyclical change in the value of collateralizable assets changes the amount of collateralized external finance in the same direction.</cite>

<cite index="20-1">The credit channel has traditionally been broken down into two components or channels of policy influence: the balance-sheet channel and the bank-lending channel.</cite> The external finance premium is the hinge variable. It widens when balance sheets weaken, tightens when they strengthen, and amplifies the initial shock regardless of origin. The mechanism does not distinguish between monetary and real disturbances—it propagates both.

Sources:

• https://www.bis.org/review/r070621a.pdf
• https://archive.nyu.edu/bitstream/2451/27340/2/S-MF-02-10.pdf
• https://www.federalreserve.gov/newsevents/speech/bernanke20070615a.htm

<cite index="4-7,4-8,4-9">Gertler and Karadi (2011) develop a quantitative monetary DSGE model with financial intermediaries that face endogenous balance sheet constraints. The model is used to simulate a crisis with features of the 2008 downturn, then to quantitatively assess the effect of direct central bank intermediation of private lending—the essence of the unconventional monetary policy the Federal Reserve deployed.</cite>

<cite index="7-5,7-6">The primary advantage the central bank has over private intermediaries is that it can elastically obtain funds by issuing riskless government debt. During a crisis, the balance sheet constraints on private intermediaries tighten, raising the net benefits from central bank intermediation.</cite> <cite index="6-2,6-3">These benefits may be substantial even if the zero lower bound constraint on the nominal interest rate is not binding. In the event this constraint is binding, though, these net benefits may be significantly enhanced.</cite>

The model formalizes what the Fed did in practice during the crisis: step into the credit channel when private intermediaries could not. The constraint is endogenous—it tightens precisely when credit demand spikes—and the central bank's balance sheet is the only instrument that can expand without pricing risk in the same cycle.

Sources:

• https://sites.google.com/site/pkaradi696/research
• https://forum.dynare.org/uploads/default/original/2X/e/eb28fc5375f15e2bb711178f601a1c66a3aa4a0f.pdf
• https://ideas.repec.org/a/eee/moneco/v58y2011i1p17-34.html

<cite index="5-1,5-2,5-3">Gertler and Karadi (2015) provide evidence on monetary policy transmission in a setting with both economic and financial variables. Using high-frequency surprises around policy announcements as external instruments, they show that shocks produce responses in output and inflation typical of monetary VAR analysis—but that modest movements in short rates lead to large movements in credit costs, driven mainly by the reaction of both term premia and credit spreads.</cite> <cite index="5-4">Forward guidance is important to the overall strength of policy transmission.</cite>

<cite index="9-1">The underlying mechanism centers on balance sheet constraints on banks. Banks intermediate funding of private securities and government bonds. Contractionary monetary policy tightens balance sheets, which tightens limits to arbitrage, raising term premia and credit spreads and amplifying the impact on the economy.</cite> The observation is load-bearing: what the Fed prices into the short end does not settle directly into real credit costs. The path runs through intermediary capital, and the spread widens when that capital thins.

<cite index="3-7">Conventional models of monetary policy transmission treat financial markets as frictionless.</cite> Gertler and Karadi reject that simplification. Their empirical work shows the financial friction is not an edge case—it is the primary mechanism through which rate changes transmit to activity.

Sources:

• https://www.aeaweb.org/articles?id=10.1257%2Fmac.20130329
• https://www.stern.nyu.edu/sites/default/files/assets/documents/GertlerKaradi2014May20-2.pdf
• https://www.bis.org/events/conf150310/gertler_karadi_presentation.pdf

The discount-rate dominance result depends on realized returns and VAR-based forecasts. <cite index="15-7,15-8,15-9">Variance decompositions using survey-based subjective expectations show a large contribution from cash flow growth expectations, a negligible contribution from return expectations, and a dominance of short-term expectations; cash flow growth expectations vary significantly over time and are high when price ratios are high</cite>. This divergence is not a measurement error—it reveals a gap between what investors report they expect and what the data imply they must have expected if prices are rational.

<cite index="14-2,14-3,14-4,14-5">Stock prices and dividends drop contemporaneously around recessions; accordingly, stock prices do not anticipate recessions due to an economic mechanism (cash flow news), and the variance of price changes increases at least as much as the variance of dividend growth during recessions, suggesting that changes in the price of risk (discount rate news) play an essential role</cite>. The event-study evidence supports discount-rate variation during crises, yet subjective forecasts attribute price movements to cash flow beliefs.

One resolution: investors may report risk-neutral expectations rather than physical probabilities, effectively embedding the discount into their stated cash-flow forecasts. Alternatively, the gap reflects behavioral mispricing that unwinds via future return variation. Either interpretation matters for application: if the market prices discount-rate variation that investors do not consciously forecast, then cost-of-capital models anchored to survey data will misprice risk.

Sources:

• http://web.stanford.edu/~delao/Subjective_Cash_Flow_and_Discount_Rate_Expectations.pdf
• https://www.sciencedirect.com/science/article/abs/pii/S0304405X20300854

<cite index="10-8,10-9">The prior belief was that the cross-section of expected returns came from the CAPM; now we have a zoo of new factors</cite>. Cochrane's survey positioned this proliferation within the discount-rate framework: factors represent different dimensions of time-varying risk premia, not isolated anomalies.

<cite index="7-11,7-12">Cochrane categorized discount-rate theories based on central ingredients and data sources, and showed that discount-rate variation continues to change finance applications</cite>. The shift in perspective altered how portfolio theory, compensation design, and capital structure decisions incorporate expected returns. When discount rates vary, the cost of capital is not constant, and the static CAPM fails to capture the dynamics that matter for multi-period valuation.

This reframing does not eliminate factor models; it clarifies their role. Each factor proxies for a source of discount-rate variation. <cite index="4-2">There is a strong common element and a strong business cycle association to all these forecasts</cite>—factors that load on recession risk, liquidity deterioration, or term-structure shifts reflect priced sources of variation in expected returns. The question becomes which sources of discount-rate variation survive out-of-sample, co-vary with macroeconomic conditions, and command persistent premia. The zoo is large, but the organizing logic is now clear: discount rates vary, and factors measure the dimensions along which they vary.

Sources:

• https://ideas.repec.org/p/nbr/nberwo/16972.html
• https://afajof.org/presidential-address-videos/
• https://rpc.cfainstitute.org/blogs/enterprising-investor/2022/time-varying-risk-premia-cochranes-discount-rates

<cite index="22-2,22-7">The linearized present value model can be tested using vector autoregressive methods</cite>, and this became the workhorse implementation. <cite index="13-3,13-4">The Campbell and Shiller (1988b) log-linear present value model decomposes the dividend yield into expected discount rates and expected cash flow growth; the empirical implementation utilizes a VAR describing the dynamics of log market returns, dividend yields, log dividend growth, and possibly additional variables</cite>.

<cite index="20-1">Dividend-price and earnings-price ratios predict stock returns measured over several years</cite>, confirming that valuation ratios carry information about forward expected returns. <cite index="21-3,21-4">A key insight from Campbell and Shiller (1988) is that the definition of returns implies cross-equation restrictions on the dynamics of these variables that can be exploited to sharpen inference</cite>. By imposing these restrictions, the system jointly estimates discount-rate and cash-flow news.

<cite index="13-7,13-8,13-9,13-10">Jordà (2005) proposes local projections as an alternative to VARs for computing impulse response functions; applied to the dividend yield volatility decomposition, local projections construct predictions at each horizon of interest separately, which is more robust to misspecification than the VAR approach</cite>. The choice of method matters; Cochrane's work demonstrated that VAR-based estimates under standard assumptions attributed nearly all valuation-ratio variance to expected-return variation, reversing the prior view that cash flow news dominated.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=227386
• https://www.nber.org/system/files/working_papers/w2511/w2511.pdf
• https://economics.princeton.edu/wp-content/uploads/2021/12/Dividend_Momentum_and_Stock_Return_Predictability_Feb2022.pdf
• https://www.aeaweb.org/conference/2020/preliminary/paper/R86iH5yk

<cite index="1-2,7-1">Discount rate variation is now at the center of asset pricing questions</cite>, and Cochrane's 2011 AFA Presidential Address settles the essential finding. <cite index="10-6,10-7">The prior consensus held that returns were uncorrelated over time and that variation in price-dividend ratios was due to variation in expected cashflows; now it seems all price-dividend variation corresponds to discount-rate variation</cite>.

This is not a marginal revision. <cite index="4-8,4-9">Discount rates vary far more than previously understood, and most of the puzzles and anomalies we face amount to discount-rate variation we do not understand</cite>. The Campbell-Shiller log-linearization of the price-dividend identity anchored the work: <cite index="22-1">a linearized rational expectations present value model produces a simple relation between the log dividend-price ratio and mathematical expectations of future log real dividend changes and future real discount rates</cite>. <cite index="21-1,21-4">Campbell and Shiller (1988) and Cochrane (2008) conclude that stock returns are predictable, mean-reverting over long horizons</cite>, and the decomposition redirected asset pricing research toward the question of why expected returns vary rather than whether they do.

<cite index="4-2,4-3">Low prices and high expected returns hold in 'bad times,' when consumption, output, and investment are low, unemployment is high, and businesses are failing</cite>—the pattern Cochrane documented across markets. The theoretical implication reshapes capital structure analysis, cost of capital estimation, and macro forecasts: <cite index="16-1">discount-rate variation continues to change finance applications, including portfolio theory, accounting, cost of capital, capital structure, compensation, and macroeconomics</cite>.

Sources:

• https://www.johnhcochrane.com/news-op-eds-all/discount-rates
• https://ideas.repec.org/p/nbr/nberwo/16972.html
• https://rpc.cfainstitute.org/blogs/enterprising-investor/2022/time-varying-risk-premia-cochranes-discount-rates
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=227386
• https://economics.princeton.edu/wp-content/uploads/2021/12/Dividend_Momentum_and_Stock_Return_Predictability_Feb2022.pdf
• https://www.nber.org/system/files/working_papers/w16972/w16972.pdf

<cite index="18-5">Safety and quality are mostly used interchangeably in the context of flight episodes, referring to a preference for less risky assets, but as theories illustrate, in times of stress investors may also demand liquidity—and the benchmark safe assets tend to be highly liquid as well.</cite> <cite index="11-10,11-11">If liquidity is scarce, investors may be willing to pay a premium for assets that are highly liquid, such as cash and short-term government bonds, leading to lower yields and higher prices for these assets.</cite>

<cite index="13-4,13-5,13-6,13-7">An increasing credit spread often signals a flight to quality, as investors demand more return to compensate for added risk or avoid riskier assets altogether; during the 2008 Global Financial Crisis, investors rapidly sold stocks and corporate bonds, moving funds into U.S. Treasury bonds, and the yield on the 10-year U.S. Treasury note fell sharply, reflecting high demand.</cite> <cite index="14-6,14-7,14-8,14-9">A defining feature of flight-to-quality is insufficient risk-taking by investors; an increase in leverage and credit spread on all but the safest and most liquid assets may incur a sudden dry up in risky asset markets, which may lead to real effects on the economy.</cite>

The distinction matters when building composite signals. Liquidity can be abundant even when credit risk is elevated; quality can deteriorate even when secondary-market depth holds. Separating the two in pricing allows calibration to the specific shock—whether the driver is counterparty fear or collateral scarcity.

Sources:

• https://www.nber.org/system/files/working_papers/w19095/w19095.pdf
• https://fastercapital.com/content/Flight-to-quality--An-Alternative-Perspective-on-Flight-to-Liquidity.html
• https://damanmarkets.academy/glossary-item/flight-to-quality/
• https://en.wikipedia.org/wiki/Flight-to-quality

<cite index="7-10,1-17">When the short rate is the only risk factor, changes in investor demand have the same relative effect on interest rates across maturities regardless of the maturities where they originate.</cite> <cite index="8-9,8-10,8-11,8-12">This has surprising implications: suppose demand for short-maturity bonds increases and demand for long-maturity bonds decreases by the same amount in present-value terms—since arbitrageurs buy long-maturity bonds, and these are more sensitive to short-rate changes than short-maturity bonds, all yields rise, including those of short-maturity bonds for which demand increases.</cite>

<cite index="7-11,1-18">When investor demand is also stochastic, demand effects become more localized.</cite> This matters for policy transmission. <cite index="7-12,1-19">A calibration indicates that long rates underreact to forward-guidance announcements about short rates.</cite> <cite index="1-20">Large-scale asset purchases can be more effective in moving long rates, especially if they are concentrated at long maturities.</cite>

The mechanism is risk in the arbitrageurs' positioning. <cite index="19-1,19-2">A stated goal of QE programs was that large-scale purchases of long-maturity bonds would drive long rates down; because absorbing the shocks exposes arbitrageurs to interest-rate risk, bond prices must change.</cite> <cite index="4-20">In preferred habitat models of the term structure, central bank purchases of long-term bonds reduce the effective supply of bonds available to private investors and lead, for a given demand, to declines in long-term bond yields.</cite>

Sources:

• https://onlinelibrary.wiley.com/doi/abs/10.3982/ECTA17440?af=R
• http://eprints.lse.ac.uk/106509/
• https://personal.lse.ac.uk/vayanos/Papers/PHMTSIR_ECMAf.pdf
• https://www.researchgate.net/publication/46474872_A_Preferred-Habitat_Model_of_the_Term_Structure_of_Interest_Rates
• https://www.econometricsociety.org/publications/econometrica/2021/01/01/preferred-habitat-model-term-structure-interest-rates/file/ecta200240.pdf

<cite index="14-1,14-2">A flight-to-quality is a financial market phenomenon occurring when investors sell what they perceive to be higher-risk investments and purchase safer investments, such as gold and government bonds, considered a sign of fear in the marketplace.</cite> <cite index="14-10,14-11">A phenomenon that occurs with flight-to-quality is flight-to-liquidity, an abrupt shift in large capital flows towards more liquid assets.</cite> <cite index="14-12">One reason the two appear together is that in most cases risky assets are also less liquid.</cite>

The Vayanos-Vila framework prices this behavior. <cite index="25-4,25-5,25-6">The informal preferred-habitat view, proposed by Culbertson (1957) and Modigliani and Sutch (1966), posits investor clienteles for specific maturity segments; the interest rate for a given maturity is mainly driven by shocks affecting the demand of the corresponding clientele, so the term structure exhibits a degree of segmentation.</cite> <cite index="8-5,8-6,8-7,8-8">A concrete example: the 2004 U.K. pension reform required pension funds to evaluate pension liabilities using long-maturity bond yields; to hedge against drops in long rates, pension funds bought long-maturity bonds in large quantities, driving long rates to record low levels—a flat term structure in early 2004 became downward-sloping in subsequent years, with the 30-year bond yielding as much as 80 basis points below its 10-year counterpart.</cite>

<cite index="18-1,18-2,18-3">Theoretical models link this to effective risk aversion—Knightian uncertainty leads agents to shed risky assets in favor of safer claims when aggregate liquidity is low, provoking a flight-to-quality and safety.</cite>

Sources:

• https://en.wikipedia.org/wiki/Flight-to-quality
• https://personal.lse.ac.uk/vayanos/Papers/PHMTSIR_ECMAf.pdf
• https://onlinelibrary.wiley.com/doi/full/10.3982/ECTA17440
• https://www.nber.org/system/files/working_papers/w19095/w19095.pdf

<cite index="1-1,2-13">The Vayanos-Vila framework models the term structure as the interaction between investor clienteles with maturity preferences and risk-averse arbitrageurs.</cite> <cite index="2-14,3-15">Because arbitrageurs are risk averse, shocks to clientele demand for bonds affect the term structure and constitute an additional determinant of bond prices beyond current and expected future short rates.</cite> The structure holds discipline: <cite index="2-15,3-16">arbitrageurs render the term structure arbitrage-free, so demand effects satisfy no-arbitrage restrictions and can be quite different from the underlying shocks.</cite>

The model solves a problem that the informal preferred-habitat view could not: <cite index="21-8,21-9">the interest rate for a given maturity cannot be driven only by shocks affecting the demand of the corresponding clientele, because if it were, rates for nearby maturities could be very different, generating large profits for term-structure arbitrageurs.</cite> Instead, <cite index="7-8,25-1">shocks to the short rate are transmitted to long rates through arbitrageurs' carry trades.</cite> <cite index="7-9,25-2">Arbitrageurs earn rents from transmitting the shocks through bond risk premia that relate positively to the slope of the term structure.</cite>

The cost of those rents is the mechanism. <cite index="21-10,21-11">Shocks to clientele demands can affect interest rates because absorbing the shocks exposes arbitrageurs to interest-rate risk, so bond prices must change to compensate them.</cite> <cite index="21-4,21-5,21-6">Demand risk weakens the transmission of short-rate shocks to bond yields because the carry trades through which arbitrageurs transmit the shocks become riskier; to hedge against demand risk, arbitrageurs scale down their carry trades or even convert them into butterfly trades.</cite>

Sources:

• https://onlinelibrary.wiley.com/doi/abs/10.3982/ECTA17440?af=R
• https://www.nber.org/papers/w15487
• https://ideas.repec.org/p/nbr/nberwo/15487.html
• https://personal.lse.ac.uk/vayanos/Papers/PHMTSIR_ECMAf.pdf
• http://eprints.lse.ac.uk/106509/

<cite index="20-4">Structural models provide both an intuitive economic interpretation and an endogenous explanation of credit defaults.</cite> <cite index="20-5,20-6,20-7,20-10,20-11">However, the assumption that corporate assets are tradable is unrealistic; they assume a simple balance sheet structure implying complex balance sheets cannot be modeled; and since the firm's assets cannot be observed credit measures are estimated by implicit estimation procedures prone to intrinsic errors; generally structural models are analytically complex and computationally intensive.</cite>

<cite index="20-17,20-18">Reduced form models do not provide an economic rationale behind a default; they depend on exogenous specifications for credit default and debt recovery.</cite> But <cite index="20-14,20-15">the probability of default and recovery rate vary with the business cycle, and reduced form models allow the default intensity to change as the firm's fundamentals and economy changes.</cite>

<cite index="28-3">The reduced-form approach is self-contained, however it is lacking the micro-economic interpretation of a default event, which is the key advantage of structural models.</cite> The choice between them is not theoretical; it is empirical. What fits the observed credit spreads, CDS term structure, and realized default clustering is what you price with. Duffie and Singleton's framework won adoption in derivatives desks because calibration held when balance-sheet visibility did not.

Sources:

• https://analystprep.com/study-notes/cfa-level-2/explain-structural-and-reduced-form-models-of-corporate-credit-risk-including-assumptions-strengths-and-weaknesses/
• https://arxiv.org/pdf/1002.2909

<cite index="10-3,10-4">Duffie and Lando obtain a reduced form model by constructing an economy where the market sees the manager's information set plus noise.</cite> <cite index="10-5">The noise makes default a surprise to the market.</cite> <cite index="10-8,10-9,10-10">The market can only observe the firm's asset value plus noise at discrete time points, and when default occurs the market is immediately informed; this noise generates the market's surprise with respect to default because the firm could nearly be in default and the market not yet aware of its imminence.</cite>

This reconciles the two modeling traditions. <cite index="18-6,18-7">Reduced form models take a firm's default process as exogenous with the time of default a stopping time; when the time is totally inaccessible the market cannot predict the time of default.</cite> <cite index="18-8,18-9">Yet managers working within a firm surely know when default is imminent; from a manager's perspective default is an accessible stopping time.</cite>

<cite index="28-4,28-5,28-6">Duffie and Lando were first to uncover the intrinsic connection between structural and reduced-form models; they consider the structural model with endogenously determined default barrier and postulate that investors observe noisy and delayed accounting reports, so the firm's default incident is fundamentally unpredictable conditional on the information available to the market observer.</cite> <cite index="28-7">The methodology based on incomplete information provides the link between structural and reduced-form models and allows for derivation of the hazard rate.</cite>

Sources:

• https://arxiv.org/pdf/math/0407060
• https://arxiv.org/pdf/1002.2909

<cite index="1-2,2-2">Instantaneous credit spreads can be identified with a 'risk-neutral mean-loss rate due to default' in the Duffie–Singleton framework.</cite> This is the pricing innovation that separates their approach from earlier intensity models.

<cite index="28-1">In reduced-form models the hazard rate for a firm is calibrated from market prices of various credit sensitive securities it has issued.</cite> <cite index="28-2">Since there is always market uncertainty regarding the default event, the hazard rate is always positive and there is no issue with zero short-term credit spreads.</cite> <cite index="21-4,21-10">The survival probability takes the form of a discount factor, and the default intensity plays the same role as an interest rate process.</cite>

The tractability matters when pricing credit derivatives and managing portfolio risk. <cite index="6-3,7-3">Duffie and Singleton's approach blends conceptual foundations with extensive analyses of empirical properties of credit-related time series such as default probabilities, recoveries, ratings transitions, and yield spreads.</cite> <cite index="7-4,8-6">Both the 'structural' and 'reduced-form' approaches to pricing defaultable securities are presented, and their comparative fits to historical data are assessed.</cite>

What the market prices is what the model captures without requiring visibility into balance-sheet mechanics or threshold triggers.

Sources:

• https://www.semanticscholar.org/paper/Duffie%E2%80%93Singleton-Model-Schl%C3%B6gl-Schloegl/495de551b49fc46d47f98710a7cb27ed3382e834
• https://onlinelibrary.wiley.com/doi/abs/10.1002/9780470061602.eqf10024
• https://arxiv.org/pdf/1002.2909
• https://www.econjournals.com/index.php/ijefi/article/download/4721/pdf/12883
• https://press.princeton.edu/books/ebook/9781400829170/credit-risk

<cite index="1-2">The Duffie–Singleton model treats default as driven by state variables following a Markov process, with defaultable zero-coupon bond prices exponentially affine in those variables.</cite> <cite index="1-3">Recovery is modeled as an exogenously given fraction of market value of the defaultable claim.</cite> This is the hinge separation from structural approaches: <cite index="20-12,20-13">default is no longer tied to the firm's assets falling below a threshold level; instead it occurs according to some exogenous hazard rate process.</cite>

<cite index="18-6">Reduced form models take a firm's default process as exogenous with the time of default a stopping time.</cite> <cite index="18-5">One reason for this is that they usually provide a better fit to market data than structural models do.</cite> The fundamental modeling tool is <cite index="14-1,14-4">the Poisson process, with default described as a Poisson event.</cite> <cite index="12-1,12-2">Default intensity represents the instantaneous probability of a borrower defaulting within a small time interval, mathematically denoted by the hazard rate λ(t).</cite>

What Duffie and Singleton observe is that <cite index="1-4">the model is sufficiently flexible to allow default intensities and thus credit spreads to be negatively correlated with default-free interest rates.</cite> This flexibility is the instrument's strength: <cite index="1-5">the model is very tractable computationally and lends itself to econometric estimation or calibration to observed market prices for relative valuation of credit derivatives.</cite>

Sources:

• https://www.semanticscholar.org/paper/Duffie%E2%80%93Singleton-Model-Schl%C3%B6gl-Schloegl/495de551b49fc46d47f98710a7cb27ed3382e834
• https://arxiv.org/pdf/math/0407060
• https://analystprep.com/study-notes/cfa-level-2/explain-structural-and-reduced-form-models-of-corporate-credit-risk-including-assumptions-strengths-and-weaknesses/
• https://www.researchgate.net/publication/381097301_Comparative_Analysis_of_the_Reduced_form_Model_and_the_Structural_Model_in_Credit_Risk_Modelling

<cite index="15-3,15-4,15-5,15-6">Higher default probabilities are associated with lower future stock returns. The anomaly cannot be explained by strategic shareholder actions, traditional risk factors, characteristics, or mispricing, but, instead, is consistent with a risk-shifting hypothesis. Consistent with the risk-shifting hypothesis, distressed firms tend to overinvest, destroy value, and exhaust their cash flows. Effects are concentrated in firms with wide credit spreads, firms with no convertible debt, and in cases where CEOs receive above-average equity-based compensation.</cite>

<cite index="12-1,12-5">As default risk rises, credit spreads rise, equity betas fall, and equity returns fall.</cite> This inverts the standard logic: rising credit spreads should reflect rising equity risk, yet equity betas contract rather than expand. <cite index="11-7">Garlappi, Shu, and Yan show that by relaxing the absolute priority rule assumption, the anomalous relationship may be explained by shareholder advantage.</cite>

<cite index="12-14,12-15">Distress risk is a robust and negative predictor of future stock returns even after controlling for the effects of strategic shareholder actions. The negative relation is not concentrated in the post-1980s period, is not sample specific, and is not due to different proxies for distress risk.</cite> The persistence across vintages and the concentration in illiquid names suggests the anomaly reflects a structural friction rather than a passing mispricing.

Sources:

• https://www.researchgate.net/publication/315881977_Risk-shifting_equity_risk_and_the_distress_puzzle
• https://www.sciencedirect.com/science/article/abs/pii/S0929119917302080
• https://www.fmaconferences.org/Orlando/Papers/Risk_shifting_Equity_Risk_and_the_Distress_Puzzle_FMA.pdf

<cite index="7-4,7-5">Firms with higher leverage, lower profitability, lower market capitalization, lower past stock returns, more volatile past stock returns, lower cash holdings, higher market-book ratios, and lower prices per share are more likely to file for bankruptcy, be delisted, or receive a D rating. When predicting failure at longer horizons, the most persistent firm characteristics, market capitalization, the market-book ratio, and equity volatility become relatively more significant.</cite>

<cite index="18-9,18-14">Results show the utility of combining accounting, market and macro-economic data in financial distress prediction models for listed companies.</cite> <cite index="19-5,19-6">The main advantages of market-based default risk measures are that they provide guidance about the theoretical determinants of default prediction and extract default-related information from market prices. More importantly, market-based models predict corporate bankruptcy more accurately than accounting-based models.</cite>

The debate over the most effective models continues. <cite index="21-8">Das, Hanouna, and Sarin compared the relative performance of accounting-based models to market-based information models in predicting corporate bankruptcy, finding that accounting-based models, particularly Altman's Z"-score, have higher explanatory power than market-based models.</cite> Which measure works best depends on the horizon, the vintage of the data, and whether the analyst seeks precision in timing or simply a rank-ordering of risk.

Sources:

• https://www.nber.org/papers/w12362
• https://www.researchgate.net/publication/259142427_Financial_distress_and_bankruptcy_prediction_among_listed_companies_using_accounting_market_and_macroeconomic_variables
• https://www.researchgate.net/publication/215991127_Assessing_the_Probability_of_Bankruptcy
• https://www.emerald.com/cemj/article/doi/10.1108/CEMJ-07-2024-0226/1307438/Corporate-bankruptcy-prediction-re-estimating

<cite index="2-1,2-2">Campbell, Hilscher, and Szilagyi found that distressed stocks have high stock return volatility and high market betas and that they tend to underperform safe stocks by more at times of high market volatility and risk aversion.</cite> <cite index="2-5">The underperformance of distressed stocks was present in all size and value quintiles.</cite>

This result cuts against the conjecture that distress earns a premium. <cite index="12-7,12-8,12-9">In a rational market, investors should demand higher premiums for holding stocks with higher distress risk. However, studies show returns are lower for firms with high distress risk.</cite> <cite index="12-10">Behavioral explanations of the "distress risk puzzle" focus on market mispricing—investors underestimate the implications of high distress risk, and, consequently, fail to demand appropriate risk premiums.</cite>

<cite index="10-3,10-4,10-5">Gao, Parsons, and Shen found that the anomaly was concentrated among small, illiquid stocks, where limits to arbitrage can allow mispricings to persist. They also found evidence pointing to a behavioral interpretation, suggesting that stocks of companies in financial distress are temporarily overpriced. For example, they found that the distress anomaly was concentrated in periods directly preceded by aggregate market gains, fueling investor overconfidence.</cite>

<cite index="16-17">The anomaly is more pronounced in market downturns, as is the distressed stocks and bonds' risk—evidence against the time-varying risk story.</cite>

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1829622
• https://www.sciencedirect.com/science/article/abs/pii/S0929119917302080
• https://alphaarchitect.com/financial-distress/
• https://www.gmu.edu/news/2022-09/distress-anomaly-where-finance-theory-breaks-down

<cite index="1-7,1-8,1-9">Campbell, Hilscher, and Szilagyi explored determinants of corporate failure and the pricing of financially distressed stocks using US data over 1963 to 2003. Firms with higher leverage, lower profitability, lower market capitalization, lower past stock returns, more volatile past stock returns, lower cash holdings, higher market-book ratios, and lower prices per share were more likely to file for bankruptcy, be delisted, or receive a D rating. When predicting failure at longer horizons, the most persistent firm characteristics—market capitalization, the market-book ratio, and equity volatility—became relatively more significant.</cite>

<cite index="1-11,1-12,1-13">Since 1981, financially distressed stocks delivered anomalously low returns. They had lower returns but much higher standard deviations, market betas, and loadings on value and small-cap risk factors than stocks with low risk of failure. These patterns held in all size quintiles but were particularly strong in smaller stocks.</cite>

<cite index="1-14">The findings were inconsistent with the conjecture that the value and size effects are compensation for the risk of financial distress.</cite> <cite index="2-3,2-4">Investors in distressed stocks have not been rewarded for bearing these risks. Instead, distressed stocks have had very low returns, both relative to the market and after adjusting for their high risk.</cite> <cite index="2-6">The underperformance was lower for stocks with low analyst coverage and institutional holdings, which suggests that information or arbitrage-related frictions may be partly responsible for the underperformance of distressed stocks.</cite>

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=770805
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1829622
• https://www.nber.org/papers/w12362

<cite index="5-3">Huang and Huang (2012) demonstrated that, once calibrated to match historical default rates and recovery rates, all diffusion-based structural models they investigated produced similar predictions and generated a credit spread puzzle for IG bonds</cite>. The finding did not settle the matter; it clarified the question.

Later work attempted to resolve the puzzle by proposing additional channels. <cite index="6-4,6-5">Studies did not find evidence that heterogeneity in loss given default, default probabilities, investors' risk aversion, or proxies for macroeconomic conditions could explain the underprediction of credit spreads, but there was evidence indicating that time-varying leverage was one potential missing factor</cite>. <cite index="7-2,7-3">Outside the financial crisis of 2007-2008, illiquidity premiums in investment-grade bonds were negligible; for speculative-grade bonds there was a strong relation between bond liquidity and yield spreads, and results suggested that bond liquidity may explain much of the underpricing of speculative-grade bonds</cite>.

<cite index="4-11">The results reported from Huang and Huang (2012) had been confirmed by several other empirical studies</cite>. <cite index="4-12">Elton, Gruber, Agrawal and Mann (2001) found that, in a risk-neutral setting, expected default losses could account for no more than 25% of corporate spreads</cite>. The consistency across vintages and methodologies means the puzzle is not an artifact of sample period or specification—it is a feature of what diffusion-based structural models do not capture.

Sources:

• https://www.sciencedirect.com/science/article/abs/pii/S0304405X20300428
• https://onlinelibrary.wiley.com/doi/full/10.1111/jofi.13409
• https://feldhutter.com/CreditSpreadPuzzle.pdf
• https://economics.yale.edu/sites/default/files/2022-10/credit-spreads.pdf

<cite index="13-1,13-8">One stylized fact documented in the U.S. credit market is that, once calibrated to historical default data and equity risk premia, standard structural models generate similar credit spreads and tend to substantially underpredict investment-grade corporate-Treasury spreads—a finding often referred to as the credit spread puzzle</cite>. Recent work examined whether this holds outside the United States. <cite index="13-2">Two representative, pure default-risk models tended to underpredict the average credit spreads on investment-grade bonds in eight developed economies, especially their spreads over government bonds, providing evidence for a "global credit spread puzzle"</cite>.

<cite index="13-2">A model incorporating endogenous liquidity in the secondary debt market helped mitigate the puzzle</cite>. <cite index="13-3">The model captured certain determinants of corporate bond market frictions across the eight economies and substantially improved the cross-sectional fit of individual IG credit spreads</cite>. This suggests that the missing component in the 2012 formulation was not country-specific—it was a structural feature of how IG debt trades relative to the risk it carries.

<cite index="6-2">Given the consensus that the corporate-Treasury spread includes a nondefault component, subsequent studies mainly focused on a weaker version of this puzzle, based on corporate spreads over alternative "default-free" benchmarks</cite>. The adjustment acknowledges that the original puzzle conflated two issues: what structural models fail to price, and what Treasuries price beyond the risk-free rate.

Sources:

• https://onlinelibrary.wiley.com/doi/full/10.1111/jofi.13409
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3453959

The credit spread puzzle does not hold uniformly. <cite index="8-2">Huang and Huang (2012) found that credit risk accounted for only a small fraction of yield spreads for investment-grade bonds, but a much higher fraction for high-yield bonds</cite>. <cite index="4-7,4-10">For investment-grade bonds of all maturities, credit risk explained only a small fraction; for short-maturity bonds the fraction was even lower</cite>.

<cite index="17-3,17-4,17-5">Calibrated structural models generated similar credit spreads, and credit risk explained between 20% and 30% of the investment-grade treasury yield, while this proportion increased for riskier bonds and accounted for a large portion of the yield spread; however, this fraction decreased as bond maturity shortened</cite>. <cite index="17-6">The fact that structural models relied on diffusion processes of firm asset value made the credit spread converge to zero for short maturities, which contradicted empirical observation</cite>.

This is not a cosmetic error. An IG issuer priced at par with a 10-year maturity and a 100bp spread over the curve carries material value in that spread; if the model assigns 25bp to default risk, the residual 75bp must settle somewhere. <cite index="17-7">The authors concluded that additional factors such as illiquidity and taxes must be important in explaining market yield spreads</cite>. What remains unpriced is not noise—it is a load-bearing component of the return.

Sources:

• https://www.researchgate.net/publication/228170729_How_Much_of_Corporate-Treasury_Yield_Spread_Is_Due_to_Credit_Risk_A_New_Calibration_Approach
• https://economics.yale.edu/sites/default/files/2022-10/credit-spreads.pdf
• https://chairegestiondesrisques.hec.ca/wp-content/uploads/pdf/cahiers-recherche/12-03.pdf

<cite index="4-2,4-6,4-7">Huang and Huang (2012) calibrated a large class of structural models to historical default loss experience and equity risk premia, then observed that for investment-grade bonds of all maturities, credit risk accounted for only a small fraction of observed yield spreads</cite>. <cite index="4-8">The puzzle held most sharply for higher-rated bonds</cite>; <cite index="4-9">for junk bonds, credit risk explained a much larger fraction of observed yield spreads</cite>.

<cite index="7-5,7-6">Many structural models that appeared very different—incorporating stochastic interest rates, endogenous default, stationary leverage ratios, strategic default, time-varying asset risk premia, and jumps in the firm value process—all generated similar spreads once calibrated to the same default probabilities, recovery rates, and equity premium</cite>. This convergence held across diffusion-based models: <cite index="3-2">Huang and Huang found that structural models of debt had trouble matching both P- and Q-measure default probability at the same time</cite>. The implication is that the missing component is not a modeling refinement but a distinct risk factor—liquidity, taxes, or some persistent non-default premium priced into IG spreads.

<cite index="22-3,22-7">Subsequent models that tried to incorporate historical default and recovery rates all generated risk spreads well below historical levels</cite>. <cite index="22-4,22-8">Credit spreads were shown to be a function of only three combinations of the seven free parameters in Merton: the default rate, the recovery rate, and the Sharpe ratio</cite>, which constrained how much additional variance the models could price.

Sources:

• https://economics.yale.edu/sites/default/files/2022-10/credit-spreads.pdf
• https://feldhutter.com/CreditSpreadPuzzle.pdf
• https://pdfs.semanticscholar.org/1e82/1095272de6dc9b38b3e73a039c1dbcf50174.pdf
• https://www.frbsf.org/research-and-insights/publications/economic-letter/2010/02/structural-models-default-credit-spread/

<cite index="7-3,7-4">The investigation of affine term structure of interest rates continues with yield curves and forward curves when the Cox–Ingersoll–Ross model is used, and not only single-factor, but multifactor models are analyzed</cite>. The single-factor constraint—one source of market risk—limits realism. Practitioners extend the framework.

<cite index="5-6,5-7">Multi-factor extensions extend the CIR model to include multiple factors to capture more complex interest rate dynamics, and regime-switching models incorporate regime-switching into the CIR model to account for changes in economic regimes</cite>. The modeling choices reflect the tension between tractability and fit. <cite index="10-1">Dynamics of interest rate and roll-over risk can be driven by a multifactor Cox/Ingersoll/Ross-type process, and such models can be calibrated to market data and used for relative pricing of interest rate derivatives, including bespoke tenor frequencies not liquidly traded in the market</cite>.

The original derivation matters because it constrains what holds under no-arbitrage. <cite index="26-8,26-9">Cox, Ingersoll, and Ross published an intertemporal general equilibrium model of asset prices in Econometrica 1985, vol. 53, pp. 363–384</cite>. Equilibrium restrictions bind tighter than arbitrage-free models calibrated to an exogenous curve. The multifactor extensions preserve the square-root structure but relax the equilibrium anchor, allowing calibration to observed term structures at the cost of economic interpretation.

Sources:

• https://link.springer.com/content/pdf/10.1007/978-3-030-15500-1_4.pdf
• https://www.numberanalytics.com/blog/cox-ingersoll-ross-model-acts-6302
• https://arxiv.org/pdf/1809.06643
• https://onlinelibrary.wiley.com/doi/abs/10.1002/mma.935

<cite index="21-2">The Vasicek and CIR processes are models with an affine term structure</cite>. This property matters because it permits analytical bond pricing without simulation. <cite index="26-1,26-2">Prices of zero-coupon bonds in the Vasicek and Cox–Ingersoll–Ross interest rate models can be computed as group-invariant solutions by determining the symmetries of the valuation partial differential equation that are compatible with the terminal condition</cite>.

The affine structure allows the zero-coupon bond price to be written in exponential-affine form as a function of the current short rate. <cite index="22-1">An extended version of the CIR model with stochastic volatility can price zero-coupon bonds, though more complex versions require numerical methods</cite>. The Feller condition reappears in implementation: <cite index="20-2,20-3,20-4">the Feller-condition (2k_z θ_z ≥ σ_z²) has no impact on the existence and uniqueness of solutions or the validity of bond pricing formulae but guarantees that the solution remains strictly positive instead of just non-negative, whose violation causes problems in some numerical schemes</cite>.

Empirical calibration is nontrivial. <cite index="8-1,8-4">The one-factor version of the CIR model was estimated using monthly quotes on U.S. Treasury issues from 1952 through 1983, and using data from a single yield curve, it is possible to estimate implied short and long term zero coupon rates and the implied variance of changes in short rates</cite>. Parameter stability, however, does not hold: <cite index="9-6">parameters are often highly correlated and intertemporal parameter stability is rejected</cite>.

Sources:

• https://onlinelibrary.wiley.com/doi/abs/10.1002/mma.935
• https://link.springer.com/article/10.1007/s40324-021-00267-w
• https://www.sciencedirect.com/science/article/pii/S0377042717300031
• https://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.1986.tb04523.x
• https://www.sciencedirect.com/science/article/abs/pii/0304405X94900167
• http://www.ressources-actuarielles.net/EXT/ISFA/1226.nsf/0/07d144ee29ab0a4ac1258510005d261f/$FILE/CIR_Article_v2.0%20EN.pdf

<cite index="4-2,4-3">The CIR model is a one-factor model that assumes that the interest rate follows a stochastic differential equation which ensures that rates remain positive</cite>. The SDE is governed by three parameters that determine the full shape of rate evolution.

<cite index="4-8,4-9">The CIR model assumes that interest rates exhibit mean-reverting behavior, which means they tend to gravitate back to an average level over time; this is reflected in the term a(b - r_t), where a determines the speed of reversion and b is the mean to which r_t reverts</cite>. The speed parameter is the knob: higher speed implies shorter half-life to mean. <cite index="13-5,13-6">The half-life is the average time it takes to be half-way back to the mean; the higher the speed of reversion the smaller the half-life</cite>.

The practical consequence is that <cite index="5-3,5-4">the CIR model is used to model the term structure of interest rates, which describes the relationship between interest rates and the maturity of securities; the model provides a framework for understanding how the yield curve evolves over time</cite>. The long-term zero-coupon yield settles differently than in nominal models: <cite index="9-5">the long-term zero-coupon yield is quite stable, as the CIR model predicts, and the level of implied short rate volatility corresponds well with time series estimates</cite>.

Sources:

• https://fastercapital.com/content/Cox-Ingersoll-Ross-Model--Understanding-Interest-Rates--The-Cox-Ingersoll-Ross-Model-s-Affine-Approach.html
• https://www.numberanalytics.com/blog/cox-ingersoll-ross-model-acts-6302
• https://www.sciencedirect.com/science/article/abs/pii/0304405X94900167
• https://quant-next.com/wp-content/uploads/2024/06/The-Cox-Ingersoll-Ross-Model.pdf

<cite index="6-7">The CIR model was introduced in 1985 by John C. Cox, Jonathan E. Ingersoll and Stephen A. Ross as an extension of the Vasicek model</cite>, published in Econometrica vol. 53, pp. 385–407. The foundational contribution was methodological: <cite index="11-3">the model derived within an intertemporal general equilibrium asset pricing framework, focusing on the dynamics of the short-term interest rate to explain the term structure of interest rates</cite>.

The critical innovation was architectural. <cite index="11-4,11-5">The Vasicek model used a mean-reverting Ornstein-Uhlenbeck process but permitted negative interest rates under certain parameter conditions; Cox, Ingersoll, and Ross addressed this by incorporating a square-root volatility term in the diffusion process, ensuring non-negativity while maintaining mean reversion</cite>. The mechanism is straightforward: <cite index="16-23,16-24">the stochastic term has a standard deviation proportion to the square root of the current rate; as the rate increases, its standard deviation increases and as it falls and approaches zero, the stochastic term also approaches zero</cite>.

This is not incidental. <cite index="11-8">The square-root process ensures non-negativity of interest rates under the Feller condition (2κθ > σ²), preventing the negative rates possible in Vasicek's Gaussian framework</cite>. The Feller condition is the boundary: when satisfied, the process remains strictly positive rather than merely non-negative.

Sources:

• https://en.wikipedia.org/wiki/Cox%E2%80%93Ingersoll%E2%80%93Ross_model
• https://grokipedia.com/page/Cox%E2%80%93Ingersoll%E2%80%93Ross_model
• https://pubs.sciepub.com/jfe/5/1/4/