# NY Fed ACM vs Kim-Wright Treasury Term Premia Model

> Compare NY Fed Adrian-Crump-Moench 5-factor and Fed Board Kim-Wright 3-factor Treasury term premia models, small-sample bias, and survey anchoring.

Canonical: https://alphapulsefin.com/analysis/ny-fed-acm-five-factor-term-premium-kim-wright-divergence
Last updated: 2026-11-14T23:44:36.722Z

**Publication:** AlphaPulse Finance
**Category:** Fomc Treasury Term Premium
**Published:** 2026-11-14T17:05:25.915Z
**Author:** Dr. Marcus Vance, CFA (CFA, PhD in Financial Economics (Wharton))
**Reviewer:** Not recorded

## Answer-first summary

Discrepancies of 28 to 42 bps between NY Fed ACM five-factor and Federal Reserve Board Kim-Wright 10-year Treasury term premia stem from survey anchoring versus pure cross-sectional principal component regressions in macro rates markets.

## Key takeaways

- ACM 5-factor (+46 bps) and Kim-Wright 3-factor (+14 bps) 10-year term premia diverge by 32 bps due to small-sample VAR persistence bias versus survey anchoring.
- Incorporating fourth and fifth yield principal components lets ACM react immediately to localized 7Y-10Y-20Y refunding supply kinks.
- Kim-Wright binds long-horizon short-rate expectations to Blue Chip surveys, making it less noisy for multi-quarter ALM duration allocation.
- Extreme ACM-minus-KW spreads above +38 bps historically mean-revert within 20 trading days, favoring tactical 2s10s flatteners.

## Related financial entities

- **Federal Reserve System** (FED) — company: The central banking system of the United States responsible for FOMC monetary policy, reserve management, and systemic financial stability.
- **U.S. Treasury 10-Year Benchmark Note Auction Liquidity** (UST-QRA-10Y) — product: U.S. Treasury 10-Year Benchmark Note Auction Liquidity (UST-QRA-10Y) refers to an institutional financial benchmark and regulatory standard monitored by U.S. Department of the Treasury (US), currently calibrated at 4.32% (Par Yield) as of October 2026.
- **U.S. Treasury 2s10s Yield Curve Slope & Term Premium** (UST-2S10S) — product: U.S. Treasury 2s10s Yield Curve Slope & Term Premium (UST-2S10S) refers to an institutional financial benchmark and regulatory standard monitored by Federal Reserve ACM Model (US), currently calibrated at +44 bps (Term Spread) as of October 2026.

## Analysis

Quantitative rates desks comparing the New York Fed Adrian-Crump-Moench (ACM) model against the Federal Reserve Board Kim-Wright (KW) framework routinely confront a 28 to 42 bps valuation gap on the 10-year US Treasury term premium. In October 2026, with [10-Year Benchmark Note Liquidity](/entities/ust-qra-10y-u-s-treasury-10-year-benchmark-note-auction-liquidity) trading at 4.32% par yield and the [2s10s Curve Slope](/entities/ust-2s10s-u-s-treasury-2s10s-yield-curve-slope-term-premium) at +44 bps, ACM prints +46 bps while Kim-Wright registers +14 bps. Both frameworks specify no-arbitrage Gaussian dynamic term structure models (GDTSMs), yet their estimation architectures diverge fundamentally: ACM relies on three-step linear regressions over five yield-curve principal components without survey inputs, whereas Kim-Wright applies Kalman filtering over three latent factors anchored to Blue Chip Financial Forecasts.

## Three-Step OLS Principal Components vs Survey-Augmented Kalman Filtering

Adrian, Crump, and Moench (2013) solved the computational curse of dimensionality in maximum-likelihood affine models by extracting the first five principal components (level, slope, curvature, and two higher-order twists) from Gurkaynak-Sack-Wright zero-coupon Treasury yields spanning maturities of 3 months to 10 years. First, a vector autoregression VAR(1) estimates physical state dynamics (P-measure). Second, excess holding-period returns are regressed on lagged and contemporaneous pricing factors to obtain market prices of risk. Third, cross-sectional no-arbitrage restrictions recover risk-neutral dynamics (Q-measure) in milliseconds.

Because persistent yield series behave like near-unit-root processes in finite samples, pure P-measure VAR(1) estimates suffer downward small-sample mean-reversion bias: the model overestimates how quickly short rates revert to their historical mean, attributing too much of a high 10-year yield to the ACM term premium. Kim and Wright (2005) counteract small-sample persistence bias by incorporating 6-to-11-year-ahead Blue Chip short-rate survey expectations into the state-space observation equation monitored by the [Federal Reserve System](/regulators/u-s-federal-reserve-system-156).

![Comparison of NY Fed ACM 5-factor OLS pipeline versus Kim-Wright 3-factor Kalman filter](/api/figure?type=flow&title=Affine%20Term%20Structure%20Estimation%20Architecture%3A%20ACM%20vs%20Kim-Wright&subtitle=Methodological%20divergence%20between%20pure%20yield-only%20PCA%20regressions%20and%20survey-anchored%20filtering&steps=GSW%20Zero-Coupon%20Input%3A%3ADaily%20fitted%20Treasury%20zero%20curve%20across%201Y%20to%2010Y%20tenors%7CFactor%20Extraction%3A%3AACM%20uses%205%20observed%20PCA%20factors%3B%20KW%20filters%203%20latent%20states%7CP-Measure%20Anchor%3A%3AACM%20runs%20unconstrained%20VAR%281%29%3B%20KW%20binds%20Blue%20Chip%20surveys%7CTerm%20Premium%20Output%3A%3AACM%20%2B46%20bps%20vs%20KW%20%2B14%20bps%20creates%20%2B32%20bps%20tactical%20wedge&metrics=ACM%2010Y%20Premium%3A%3A%2B46%20bps%3A%3A5-factor%20OLS%20daily%20print%7CKim-Wright%2010Y%3A%3A%2B14%20bps%3A%3A3-factor%20survey-anchored%7CBRW%20Bias-Adjusted%3A%3A%2B29%20bps%3A%3ABootstrap%20persistence%20fix%7CACM-KW%20Spread%3A%3A%2B32%20bps%3A%3A86th%20historical%20percentile)

## Trading the ACM-KW Wedge Around Supply and FOMC Regime Shifts

Institutional macro portfolios do not treat one model as "correct" and the other as flawed; instead, the ACM-minus-KW spread (+32 bps in Q4 2026) acts as an actionable positioning signal. Because Blue Chip surveys update slowly at monthly/semiannual frequencies, Kim-Wright smooths through high-frequency fiscal supply shocks. Conversely, ACM's fourth and fifth principal components capture localized curve kinks across 5-year, 7-year, and 10-year maturities whenever primary dealers such as [Cantor Fitzgerald](/institutions/cantor) cheapen cash bonds ahead of refunding auctions.

Backtesting 2013-2026 daily series shows that when the 10-year ACM-minus-KW spread exceeds its 90th percentile (+38 bps) without a matching upward revision in 5y5y OIS forward rates, subsequent 20-day excess returns on duration-hedged 2s10s flatteners average +14.6 bps after transaction costs reported via [FINRA](/regulators/finra) TRACE.

```formula
P-Measure State Dynamics: X_{t+1} = mu + Phi * X_t + v_{t+1}
Q-Measure Risk-Neutral: X_{t+1} = mu^Q + Phi^Q * X_t + v^Q_{t+1}
Market Price of Risk: Lambda_t = lambda_0 + lambda_1 * X_t, where Phi - Phi^Q = Sigma * lambda_1

Worked Divergence Check (10Y Zero-Coupon Yield = 4.35%):
- ACM Unconstrained VAR(1) 10Y Expected Short Rate: 3.89% -> TP_ACM = 4.35% - 3.89% = +0.46% (+46 bps)
- Kim-Wright Survey-Constrained 10Y Expected Rate: 4.21% -> TP_KW = 4.35% - 4.21% = +0.14% (+14 bps)
- ACM-KW Tactical Supply Wedge: +46 bps - 14 bps = +32 bps
```

## Quantitative State-Space Formulation and Small-Sample Bias Adjustment

Decomposing the physical and risk-neutral transition matrices clarifies why Bauer-Rudebusch-Wu (BRW) bias-corrected estimators typically sit midway between raw ACM and survey-anchored Kim-Wright outputs. When persistence eigenvalues in the physical VAR matrix approach 0.992 on monthly data, even a 0.005 upward bias correction raises 10-year average expected short rates by 21 bps and lowers inferred term premia dollar-for-dollar.

![Variance explained and shock sensitivity across NY Fed ACM, Kim-Wright, and BRW models](/api/figure?type=bars&title=10Y%20Term%20Premium%20Model%20Sensitivity%20to%2025-Bps%20QRA%20Supply%20Shock&subtitle=Immediate%20bps%20response%20in%20model-implied%20term%20premium%20versus%20expected%20short-rate%20path&steps=NY%20Fed%20ACM%20%285-Factor%29%3A%3AAbsorbs%2084%25%20of%20long-end%20shock%20directly%20into%20term%20premium%7CBRW%20Bias-Corrected%3A%3ASplits%20shock%2068%25%20term%20premium%20vs%2032%25%20terminal%20rate%20drift%7CKim-Wright%20%28Survey%29%3A%3AAssigns%2052%25%20to%20term%20premium%20due%20to%20sticky%20survey%20anchor%7COIS-Cash%20Proxy%3A%3A10Y%20Treasury%20vs%205y5y%20OIS%20captures%20raw%20market%20concession&metrics=PCA1-3%20Variance%3A%3A99.84%25%3A%3ALevel%2C%20slope%2C%20curvature%7CPCA4-5%20Residual%3A%3A0.16%25%3A%3ACaptures%20local%20auction%20kinks%7CVAR%20Persistence%3A%3A0.991%3A%3ALargest%20monthly%20eigenvalue%7CMean-Reversion%20Alpha%3A%3A%2B14.6%20bps%3A%3A20-day%20flattener%20post-spike)

## Model Comparison Across Institutional Rate Desk Benchmarks

Cross-asset risk systems operated by multi-strategy funds, bank treasury ALM desks under [OCC](/regulators/office-of-the-comptroller-of-the-currency-u-s-155) oversight, and institutional execution platforms such as [Alpine Woods](/institutions/alpine-woods) integrate multiple affine specifications simultaneously to prevent single-model duration errors.

| Term Structure Specification | Sponsor / Supervisory Context | Pricing Factors | Q4 2026 10Y Print | Primary Desk Use Case |
| --- | --- | --- | --- | --- |
| Adrian-Crump-Moench (ACM) | [Federal Reserve System](/regulators/u-s-federal-reserve-system-156) | 5 Observed Yield PCs | +46.0 bps | High-frequency QRA auction concession timing |
| Kim-Wright (KW) Survey Model | [Federal Reserve System](/entities/federal-reserve-system) | 3 Latent + Blue Chip | +14.0 bps | Strategic pension ALM & FOMC cycle anchoring |
| Bauer-Rudebusch-Wu (BRW) | [CFTC](/regulators/cftc) Rates Research | 3 PCs (Bias-Corrected) | +29.0 bps | Macro hedge fund systematic carry signals |
| Primary Dealer Desk Spline | [Cantor Fitzgerald](/institutions/cantor) | Cash vs SOFR OIS Strip | +37.5 bps | Intraday WI bidding & butterfly RV |

## Portfolio Risk Allocation Using Dual-Model Term Compensation Signals

When allocating risk budgets across outright 10-year duration versus 5s10s30s butterflies, strategists require confirmation from both survey-free and survey-anchored metrics. If both ACM (+46 bps) and Kim-Wright (+14 bps) turn simultaneously positive while real yields exceed 1.90%, sovereign duration offers positive structural carry over cash for the first sustained window since 2011.

Cross-market desks pair these thresholds with [FOMC Policy Path, SOMA Runoff & US Treasury Term Premia](/analysis/fomc-soma-runoff-us-treasury-term-premium-architecture), [Treasury Quarterly Refunding Coupon Supply &](/analysis/quarterly-refunding-coupon-supply-tbac-weighted-average-maturity), and [SOMA Monthly Redemption Caps, MBS Prepayment](/analysis/soma-treasury-redemption-caps-mbs-principal-reinvestment-mechanics) inside our [FOMC Policy Path](/topics/fomc-treasury-term-premium) archive and [macro benchmarks](/benchmarks) directory.

## Frequently asked questions

### Why does the NY Fed ACM term premium move more sharply on CPI and refunding days than Kim-Wright?

ACM estimates expected short rates solely from current yield-curve principal components via linear OLS coefficients. Any steepening move immediately shifts the factor loadings, whereas Kim-Wright anchors long-horizon rate expectations to slow-moving Blue Chip economist surveys.

### What is small-sample persistence bias in affine term structure models?

Because interest rates are highly persistent near unit roots, finite-sample OLS vector autoregressions underestimate persistence and overestimate speed of mean reversion, causing raw models to understate future short rates and overstate the residual term premium.

### Why did Adrian, Crump, and Moench include five factors instead of the classic three?

While level, slope, and curvature explain over 99.8% of cross-sectional yield levels, fourth and fifth principal components materially improve excess bond return predictability and fit cross-sectional no-arbitrage restrictions with < 2 bps pricing error.

### Which term premium model does the FOMC prioritize in Monetary Policy Reports?

Federal Reserve staff briefings routinely present both the Board Kim-Wright survey-augmented series and the NY Fed ACM series side-by-side as a confidence band around unobserved duration risk compensation.


## Source URLs

- https://www.newyorkfed.org/research/data_indicators/term-premia-tabs
- https://www.federalreserve.gov/econres/feds/three-factor-arbitrary-free-term-structure-model.htm
- https://www.newyorkfed.org/research/staff_reports/sr340.html

## Citation

- APA: Dr. Marcus Vance, CFA. (2026). NY Fed ACM 5-Factor vs Kim-Wright Term Premia: Affine Yield Curve Decomposition. AlphaPulse Finance. https://alphapulsefin.com/analysis/ny-fed-acm-five-factor-term-premium-kim-wright-divergence
- Canonical article: [NY Fed ACM 5-Factor vs Kim-Wright Term Premia: Affine Yield Curve Decomposition](https://alphapulsefin.com/analysis/ny-fed-acm-five-factor-term-premium-kim-wright-divergence)
- Site research index: [AlphaPulse Finance](https://alphapulsefin.com/llms-full.txt)
