A Preferred-Habitat Model of Term Premia, Exchange Rates, and Monetary Policy Spillovers

Layer 1 — Core Argument

The paper develops a two-country preferred-habitat model in which currency and bond markets are populated by different investor clienteles — currency traders with price-elastic demand for foreign assets, and bond investors whose preferences are habitat-specific by country and maturity — with segmentation partly overcome by global arbitrageurs who have limited capital and bear mean-variance risk. Risk premia in the model are time-varying, connected across markets, and consistent with the empirical violations of Uncovered Interest Parity (UIP) and the Expectations Hypothesis (EH): in particular, currency carry trade (CCT) and bond carry trade (BCT) strategies earn abnormally high expected returns in ways that co-vary across the two markets in a manner the standard frictionless model cannot generate. Through these time-varying, connected risk premia, large-scale bond purchases (QE) lower domestic bond yields, lower foreign bond yields, and depreciate the purchasing country’s currency; short-rate cuts also lower foreign yields, but with smaller effects than bond purchases. A key structural finding, quantified in the estimated model calibrated to US and Eurozone data, is that currency returns are nearly uncorrelated with long-maturity bond returns — an exchange-rate disconnect — yet the currency market is instrumental in transmitting bond demand shocks across countries, because arbitrageurs hedge their cross-currency positions in bond markets and vice versa. Sterilized foreign-exchange interventions have strong effects on the exchange rate but weak effects on bond yields, while QE/QT has weak effects on the exchange rate but sizeable effects on foreign bond yields — a sharp asymmetry that follows directly from the disconnect.

Layer 2 — Q&A

Q1. Why do UIP and EH fail in the standard model, and what changes in this model?

In the standard model with perfect capital mobility, risk premia are constant, so the yield curve depends only on expectations of the domestic short rate and the exchange rate absorbs short-rate differentials exactly. In this model, arbitrageurs bear the residual risk when currency traders and bond clienteles are unwilling to absorb excess supply or demand at prevailing prices. Because arbitrageurs have limited capital (captured by a risk-aversion parameter a ≥ 0 that can also represent capital or Value-at-Risk constraints in reduced form), they demand compensation — time-varying risk premia — for holding currency and maturity risk. When a = 0, arbitrageurs are risk-neutral, UIP and EH both hold, and the model collapses to the standard frictionless benchmark.

Q2. What are the three types of agents and what does each do?

Currency traders hold foreign assets and have a demand that is downward-sloping (price-elastic, with slope coefficient αe ≥ 0) in the log exchange rate; their demand also shifts with a stochastic currency demand factor γt. They can be interpreted as households engaged in expenditure switching or central banks managing reserve levels. Bond investors form clienteles, each with a preferred-habitat demand for bonds of a specific country and maturity that is downward-sloping in the log bond price (slope αj(τ)) and shifts with a country-specific bond demand factor βjt; examples are pension funds and insurance companies whose liabilities are long-dated and denominated in their home currency. Global arbitrageurs trade the currency and all bonds of both countries, maximizing mean-variance utility over instantaneous wealth changes; they bridge the segmented markets and their positions pin down equilibrium risk premia.

Q3. What is the equilibrium structure and which factors drive prices?

The equilibrium exchange rate and bond prices are log-affine functions of five stochastic factors: the home short rate iHt, the foreign short rate iFt, the currency demand factor γt, and the two bond demand factors βHt and βFt. These factors follow a mean-reverting (Ornstein-Uhlenbeck) system. The equilibrium is characterized by a scalar nonlinear system (25 equations in the general case) whose solution pins down the loadings of prices on each factor. This affine structure means each asset’s risk premium is the product of the arbitrageur’s risk-aversion coefficient, the factor covariance matrix, and arbitrageur net positions, which are themselves determined by market-clearing.

Q4. How does a conventional short-rate cut transmit domestically and internationally in the model?

Following a home short-rate cut, arbitrageurs find it attractive to enter the CCT — borrow home currency, invest in foreign currency. If currency traders’ demand is price-elastic (αe > 0), arbitrageurs’ equilibrium foreign-currency holdings rise, and the expected return on the CCT rises too (arbitrageurs must be compensated for the increased risk). This attenuation effect means the foreign currency appreciates less than implied by UIP: the exchange rate response is dampened. Simultaneously, arbitrageurs enter the home BCT (borrow at the home short rate, invest in long home bonds); if home bond investors’ demand is price-elastic (αH(τ) > 0), arbitrageurs’ long-bond holdings rise and the BCT’s expected return rises, attenuating the transmission to domestic long-maturity yields (which fall less than EH would imply). A propagation effect to foreign bond yields arises through arbitrageur hedging: by taking long positions in foreign currency (CCT), arbitrageurs become exposed to the risk that the foreign short rate drops and the foreign currency depreciates; long-maturity foreign bonds provide a natural hedge (their price rises when the foreign short rate drops), so arbitrageurs increase foreign bond demand, depressing foreign yields. This international transmission of conventional policy is absent from the standard model.

Q5. How does unconventional policy (QE/QT) transmit domestically and to the exchange rate and foreign yields?

Following QE purchases of home bonds, their prices rise; arbitrageurs accommodate by holding fewer home bonds, which reduces their exposure to home short-rate risk. With less home-rate risk, arbitrageurs become more willing to hold foreign currency (which depreciates when the home short rate rises, offering a natural hedge against the home rate risk they have shed). The increased foreign-currency position in turn makes arbitrageurs more willing to hold foreign bonds (which hedge the foreign-currency position against foreign rate changes). The net result in the model is: QE lowers domestic bond yields, lowers foreign bond yields, and depreciates the home currency. The quantitative finding from the estimated model is that QE/QT effects on foreign bond yields are sizeable and stronger than those of conventional short-rate policy.

Q6. What explains the exchange-rate disconnect, and how can the currency market still transmit bond demand shocks?

In the estimated model, variance decompositions reveal that long-maturity bond yields in each country are driven primarily by bond demand factors (βHt and βFt), while the exchange rate is driven primarily by the currency demand factor (γt); short rates account for a small fraction of movements in both, and each factor type accounts for negligible variation in the other asset class’s price. The disconnect between bond yields and the exchange rate arises because bond demand shocks in the two countries move the exchange rate in opposite directions — a home bond demand shock that lowers home yields also raises the exchange rate via arbitrageur hedging, while a foreign bond demand shock moves the exchange rate in the opposite direction. These offsetting effects make the exchange rate nearly uncorrelated with long-maturity bond yields. However, bond demand shocks in one country are transmitted to bond yields in the other country through the currency market: arbitrageurs hedge their bond positions using the currency, so a shock to home bond demand moves arbitrageurs’ currency positions, which in turn affects their willingness to hold foreign bonds. Cross-country bond yield comovement is therefore positive and sizeable, despite the exchange-rate disconnect.

Q7. What are the model’s implications for foreign exchange intervention?

A sterilized purchase of foreign currency by the home or foreign central bank — which shifts the currency demand factor — has strong effects on the exchange rate but weak effects on bond yields. This follows directly from the variance decomposition: the exchange rate loads heavily on the currency demand factor and bond yields load lightly on it. The asymmetry mirrors the QE result in reverse: QE shifts bond demand factors, which load heavily onto bond yields and lightly onto the exchange rate; FX intervention shifts the currency demand factor, which loads heavily onto the exchange rate and lightly onto bond yields. The model thus delivers a sharp policy instrument separation between QE/QT (primarily a bond yield tool) and FX intervention (primarily an exchange-rate tool), with each having spillovers in the other dimension that are quantitatively weaker.

Q8. How is the relationship between currency risk premia and bond risk premia captured, and what empirical regularities does the model match?

The model’s risk premia are linked through the shared arbitrageur portfolio: the price of each risk factor is proportional to the covariance between that factor and the arbitrageur’s overall portfolio return, so a shock that changes arbitrageurs’ currency positions also changes the compensation required for bond positions, and vice versa. The estimated model is reported to match closely the violations of UIP (CCT profitability) and EH (BCT profitability) documented in the literature, and the ways in which these violations are connected — including findings that yield-curve slope differentials predict CCT profitability, and that CCT profitability declines when carried out with long-maturity rather than short-maturity bonds. These matches are described as consistent with the empirical regularities, not structural identification of the underlying causes.

Q9. What is the role of segmented versus global arbitrage, and why does the distinction matter?

The paper considers both cases. Under segmented arbitrage, separate arbitrageur pools operate in the currency market (risk aversion ae), home bond market (aH), and foreign bond market (aF); first-order conditions for each pool reflect only their own portfolio risk, so the prices of risk factors differ across markets. Under global arbitrage, a single pool of arbitrageurs trades all assets, and their shared portfolio means the price of each risk factor is the same across currency and bond markets — this is the mechanism through which bond demand shocks in one country propagate through the currency market to bond yields in the other. Global arbitrage is the primary specification; segmented arbitrage serves as a benchmark to isolate the hedging-based transmission channel that requires global positions.

Q10. How does the model relate to and extend predecessor frameworks?

The model extends Vayanos and Vila (2021) — a closed-economy preferred-habitat yield curve model — to two countries by adding a currency market and a second country’s bond market, with arbitrageurs who are global rather than country-specific. In the currency dimension, the attenuation of UIP deviations parallels Gabaix and Maggiori (2015), which models exchange-rate dynamics with financially constrained intermediaries but without a yield curve. The two-country structure allows the paper to simultaneously study term premia (EH violations), exchange rate dynamics (UIP violations), and their connection, and to quantify the effects of QE, conventional monetary policy, and FX intervention within a single internally consistent framework estimated on US-Eurozone data.

Key Concepts

Preferred-habitat demand: A bond investor’s demand for bonds of a specific country and maturity that does not arise from portfolio optimization over the full menu of available assets, but rather from institutional constraints or liability-matching motives (e.g., pension funds matching long-dated domestic liabilities). In the model, preferred-habitat demand is price-elastic with slope αj(τ) and shifts with a country-specific bond demand factor βjt; the elastic component means that as bond prices rise, clientele demand falls, so arbitrageurs must absorb the residual supply and require a risk premium to do so.

Global arbitrageur: An investor who trades the currency and bonds of both countries simultaneously, bridging the segmented currency and bond markets. In the model, global arbitrageurs maximize mean-variance utility over instantaneous wealth changes; their shared portfolio across all asset classes is the mechanism through which shocks in one market create hedging-driven demand in other markets, generating the cross-market linkages in risk premia and monetary policy transmission.

Currency carry trade (CCT): A strategy that borrows at the home short rate and invests at the foreign short rate, profiting when the foreign currency does not depreciate enough to offset the interest rate differential. Under UIP, the CCT earns zero expected return; the model generates a positive expected CCT return — a currency risk premium — when arbitrageurs are risk-averse and currency traders’ demand is price-elastic. In the paper’s notation, the CCT return is det/et + (iFt − iHt)dt.

Bond carry trade (BCT): A strategy that borrows at the short rate and invests in long-maturity bonds of the same country, profiting when long yields fall or when expected short rates are below current long yields. Under EH, the BCT earns zero expected return; the model generates a positive expected BCT return — a term premium — when arbitrageurs are risk-averse and bond clientele demand is price-elastic.

Exchange-rate disconnect: The empirical and model finding that movements in the exchange rate are nearly uncorrelated with movements in long-maturity bond yields, even though both are endogenously determined in the same model. The disconnect arises in the estimated model because long bond yields are driven primarily by bond demand factors, while the exchange rate is driven primarily by the currency demand factor, and the two sets of factors move the exchange rate in offsetting directions so that their net effect on bond yield-exchange rate covariance is approximately zero.

Attenuation effect: The dampening of monetary policy transmission to asset prices caused by the need to compensate risk-averse arbitrageurs for the increased risk they bear when accommodating the policy-induced excess demand. In the currency market, a home short-rate cut causes the CCT’s expected return to rise (arbitrageurs must be paid more to hold foreign currency), which means the foreign currency appreciates less than UIP predicts. In the bond market, a short-rate cut causes the BCT’s expected return to rise (term premia increase), so long yields fall less than EH predicts.

Propagation effect: The international transmission of a domestic monetary policy shock to foreign asset prices through arbitrageur hedging. A home short-rate cut causes arbitrageurs to increase their foreign-currency position (CCT); this exposes them to the risk of foreign short-rate declines (which depreciate the foreign currency), and long-maturity foreign bonds hedge this risk; so arbitrageurs increase foreign bond demand, depressing foreign yields. This channel is absent from the standard model where risk premia are constant.

Log-affine equilibrium: The conjectured and verified form of the equilibrium in which the log exchange rate and log bond prices are affine (linear plus constant) functions of the five state factors (iHt, iFt, γt, βHt, βFt). This structure allows the model to be solved as a system of ordinary differential equations and scalar equations, and enables closed-form or numerically tractable characterization of risk premia, variance decompositions, and policy effects.

Bond demand factor (βjt): A stochastic variable that shifts the intercept of bond clientele demand in country j, independent of maturity τ. A positive shock to βjt increases desired bond holdings of country-j clienteles at any given price, forcing arbitrageurs to shed country-j bonds, which lowers bond yields. The factor follows a mean-reverting process and in the estimated model is found to be the primary driver of long-maturity yields in both countries.

Currency demand factor (γt): A stochastic variable that shifts the intercept of currency traders’ demand for foreign assets, independent of the exchange rate level. A positive shock to γt increases desired foreign asset holdings of currency traders, so arbitrageurs reduce their foreign-currency position, which affects their bond positions through hedging. In the estimated model, γt is the primary driver of exchange-rate movements.

Summary based on LSE Research Online accepted version (accepted manuscript). AI-assisted, human review pending.