# Preferred Habitat: How Bond Demand Moves the Yield Curve

Cite as: Gorak, R. (2026). Preferred Habitat: How Bond Demand Moves the Yield Curve. Tradicted. https://www.tradicted.com/research/vayanos-preferred-2021/
Paper: Dimitri Vayanos and Jean-Luc Vila — *A Preferred-Habitat Model of the Term Structure of Interest Rates*
Published in: Econometrica (2021)
Original: https://doi.org/10.2139/ssrn.971439
DOI: 10.2139/ssrn.971439

Key finding: In the one-factor preferred-habitat model the Fama-Bliss slope coefficient equals (κ*r − κr)/κ*r > 0, rising from zero when arbitrageurs are risk neutral to one when they are infinitely risk averse, and adding a demand factor pushes it from zero to a value larger than one.

---
Preferred-habitat theory is the view that investor clienteles favor specific bond maturities, so demand shocks local to one maturity move its interest rate. When the US Treasury announced long-bond buybacks in January 2000, thirty-year yields fell 58bps within three weeks while two-year yields rose 9bps. In Vayanos and Vila's "A Preferred-Habitat Model of the Term Structure of Interest Rates" (2021), the Fama-Bliss coefficient spans zero to one. It is zero when arbitrageurs are risk neutral and one when they are infinitely risk averse, in the one-factor model.

## What the Study Found

The one-factor Fama-Bliss coefficient is γFB = (κ*r − κr)/κ*r > 0, and adding a demand factor lets it exceed one. A unit short-rate shock raises the forward rate for maturity τ by e^(−κ*r τ). The expected short rate at that horizon rises by the larger e^(−κr τ). In a numerical two-factor example, the first principal component explains 94.61% of return variation and is flat across maturities. Short-rate shocks explain 99.92% of variation for the one-year bond but only 4.79% for the thirty-year bond.

> "We show that the preferred-habitat view of the term structure generates a rich set of implications for bond risk premia, the effects of demand shocks and of shocks to short-rate expectations, the economic role of carry trades, and the transmission of monetary policy."
>
> Vayanos and Vila (2021), *A Preferred-Habitat Model of the Term Structure of Interest Rates*, abstract of the 2009 NBER working-paper version.

## Methodology

The paper is a theoretical model, not an empirical study, so it uses no dataset, sample, or time period. The short rate follows an exogenous Ornstein-Uhlenbeck process, and each clientele demands only the zero-coupon bond matching its maturity. Arbitrageurs trade every maturity and maximize a mean-variance objective over instantaneous wealth changes, with risk-aversion coefficient a. This page summarizes the November 2009 NBER working-paper version (No. 15487), the full text on file.

## Key Statistics

| Metric | Finding | Context |
|---|---|---|
| Thirty-year yield change, 2000 buyback | −58bps | Within three weeks of the January 2000 announcement |
| Five-year and two-year yield change | −9bps and +9bps | Same window; effect confined to long maturities |
| Buyback program scale | 45 reverse auctions, average 14% of each targeted issue | March 2000 to April 2002, maturities 10 to 27 years |
| Fama-Bliss coefficient (one-factor) | γFB = (κ*r − κr)/κ*r > 0 | Proposition 2, Δτ → 0 |
| Forward-rate response to short-rate shock | e^(−κ*r τ) < e^(−κr τ) | Proposition 3; forward rates under-react |
| First principal component, two-factor example | 94.61% of variation | Flat across maturities; equals 0.44rt + βt |
| Short-rate share of return variation | 99.92%, 92.20%, 65.22%, 18.50%, 4.79% | One-, five-, ten-, twenty- and thirty-year bonds |
| Demand decrease, three-factor example | Forward rates rise up to seven years (ks), three years and longer (kℓ) | Shocks at five and twenty years; effect more local than in two-factor model |
| Preferred-habitat demand | yt,τ = α(τ)τ(Rt,τ − βt,τ) | Eq. (4); demand rises linearly in the bond's yield |
| Risk-neutral mean reversion | κ*r = κr + aσr² ∫ α(τ)Ar(τ)² dτ | Proposition 1; κ*r > κr whenever a > 0 |

## Low vs High Arbitrageur Risk Aversion

| Measure | Low risk aversion (a ≈ 0) | High risk aversion (a ≈ ∞) |
|---|---|---|
| Demand-shock location affects | Magnitude only | Magnitude and relative effect across maturities |
| Short-rate shock, long forward rates | Same direction, A′r(τ) ≈ e^(−κr τ) | Opposite direction (Corollary 2) |
| Campbell-Shiller γCS | Close to one, increasing in τ | Can be negative, decreasing in τ |
| One-factor Fama-Bliss γFB | Zero when risk neutral | One when infinitely risk averse |

## Why This Matters

The model explains how central-bank bond purchases or Treasury issuance can move long yields without changing expected short rates. Short-rate policy reaches long maturities through arbitrageurs' carry trades, so its pass-through weakens when arbitrageur capital is scarce. In those conditions, buying long-term bonds directly may influence long rates more than moving the short rate. A steeper curve caused by weaker bond demand predicts larger premia than one caused by a falling short rate.

## FAQ

### What is the preferred-habitat theory of interest rates?

Under preferred-habitat theory, a supply shock at one maturity moves that maturity most, as when thirty-year Treasury yields fell 58bps after the 2000 buyback announcement. Within the same three weeks, five-year yields dropped only 9bps and two-year yields rose 9bps. Vayanos and Vila (2021) note these effects are hard to rationalize within a representative-agent model.

### Do bond risk premia depend on the slope of the yield curve?

In the Vayanos-Vila one-factor model, the Fama-Bliss coefficient equals (κ*r − κr)/κ*r, ranging from zero to one as arbitrageur risk aversion rises. Bond premia are positive when the curve slopes up and negative when it slopes down. Fama and Bliss estimated the coefficient at approximately one, with a standard deviation of predicted premia of about 1-1.5% per year.

### Can one principal component hide two yield-curve factors?

In a Vayanos-Vila numerical example, one flat principal component explains 94.61% of bond-return variation even though two independent factors drive yields. Short-rate shocks explain 92.20% of five-year bond variation but only 18.50% at twenty years. The component is the weighted average 0.44rt + βt of short rate and demand.

## Source

Dimitri Vayanos and Jean-Luc Vila. (2021). A Preferred-Habitat Model of the Term Structure of Interest Rates. Econometrica, 89(1), 77–112.

[Read the full paper →](https://doi.org/10.2139/ssrn.971439)