The solution to the forward-bias puzzle: Reply

My proposed econometric solution to the forward-bias puzzle provoked several comments. Those comments raise three primary objections to my solution. (1) It suffers from multicollinearity, miss-specification and other serious econometric problems. (2) My key test equation is a tautology or identity. (3) My econometric solution has nothing to do with either the forward-bias puzzle or uncovered interest parity. This is my reply to those objections.     

On the (in)feasibility of covered interest parity as a solution to the forward bias puzzle

This paper examines Pippenger's (2011) proposed solution to the forward bias puzzle, which is based on the covered interest parity (CIP) condition. It is argued that the CIP-based approach does not solve this well known and long-standing puzzle in international finance in a meaningful way. Moreover, it is shown that empirical results from such an approach follow mechanically from the identity-like nature of the theory of covered interest parity, which, aside from small deviations due to transaction costs, is assumed to hold in all periods (as if it were an identity). We show that rather than leading to new insights, the simple reconfiguration of CIP to solve for the time t + 1 spot exchange rate leads to tautological expressions that, when estimated, might appear to successfully explain the forward bias, but in actuality are trivial. Results from simple simulation exercises further illustrate the inconclusiveness of the proposed solution method.     

Habit-based asset pricing with limited participation consumption

We calibrate and estimate a consumption-based asset pricing model with habit formation using limited participation consumption data. Based on survey data of a representative sample of American households, we distinguish between assetholder and non-assetholder consumption, as well as the standard aggregate consumption series commonly used in the CCAPM literature. We show that assetholder consumption outperforms non-assetholder and aggregate consumption data in explaining bond returns, bond yields, and the volatility of bond yields. We further show that the high volatility of assetholder consumption enables the model to explain the equity premium puzzle and the risk-free rate puzzle simultaneously for a reasonable value of relative risk aversion.     

When can you immunize a bond portfolio?

This paper presents a condition equivalent to the existence of a Riskless Shadow Asset that guarantees a minimum return when the asset prices are convex functions of interest rates or other state variables. We apply this lemma to immunize default-free and option-free coupon bonds and reach three main conclusions. First, we give a solution to an old puzzle: why do simple duration matching portfolios work well in empirical studies of immunization even though they are derived in a model inconsistent with equilibrium and shifts on the term structure of interest rates are not parallel, as assumed? Second, we establish a clear distinction between the concepts of immunized and maxmin portfolios. Third, we develop a framework that includes the main results of this literature as special cases. Next, we present a new strategy of immunization that consists in matching duration and minimizing a new linear dispersion measure of immunization risk.     

Exchange rate dynamics under alternative optimal interest rate rules

We explore the role of interest rate policy in the exchange rate determination process. Specifically, we derive exchange rate equations from interest rate rules that are theoretically optimal under a few alternative settings. The exchange rate equation depends on its underlying interest rule and its performance could vary across evaluation criteria and sample periods. The exchange rate equation implied by the interest rate rule that allows for interest rate and inflation inertia under commitment offers some encouraging results — exchange rate changes “calibrated” from the equation have a positive and significant correlation with actual data, and offer good direction of change prediction. Our exercise also demonstrates the role of the foreign exchange risk premium in determining exchange rates and the difficulty of explaining exchange rate variability using only policy based fundamentals.     

Forward exchange market unbiasedness: the case of the Australian dollar since 1984

This paper implements a robust statistical approach to regression with non-stationary time series. The methods were recently developed in other work and are briefly exposited here. They allow us to perform regressions in levels with non-stationary time series data, they accommodate data distributions with heavy tails and they permit serial dependence and temporal heterogeneity of unknown form in the equation errors. With these features the methods are well suited to applications with frequently sampled exchange rate data, which generally display all of these empirical characteristics. Our application here is to daily data on spot and forward exchange rates between the Australian and US dollars over the period 1984–1991, following the deregulation of the Australian foreign exchange market. We find big differences between the robust and the non-robust regression outcomes and in the associated statistical tests of the hypothesis that the forward rate is an unbiased predictor of the future spot rate. The robust tests reject the unbiasedness hypothesis but still give the forward rate an important role as a predictor of the future spot rate.     

International evidence on misspecification of the standard money demand equation

This paper tests the proposition that the interest rate in the standard money demand equation acts as a proxy for the expected change in interest rates. If this is true, the interest elasticity in the standard equation will depend on the expectations process for interest rates. Using cross-country data, we find that there does exist a relationship between the observed interest elasticity in the standard equation and the predictability of future interest rates from current interest rates.     

Co-integrating currencies and yield differentials

This study investigates the relationship between currencies and interest rates of different maturity horizons. The real exchange rate is found to depend both on short-term real domestic and foreign interest rate difference and on long-term real domestic and foreign interest rate difference. Co-integrating regressions of contemporaneous currency rates generate negative and significant coefficients for long-term rate differentials, consistent with uncovered interest parity. Therefore, the expectations hypothesis holds for long horizons. On the other hand, positive coefficients for real short-term interest rate differentials reveal the forward premium puzzle: the failure of uncovered interest parity for short-horizons. Results are partly driven by the very different risk characteristics of short-term bonds and foreign bonds.     

Covered interest rate parity in emerging markets

This paper finds that while covered interest rate parity holds for large and small triple A rated economies, it holds for emerging markets only for a three-month maturity. For a five-year horizon the size and frequency of violations lead to the conclusion that covered interest rate parity does not hold for longer maturities for Brazil, Chile, Russia and South Korea. Overall this paper finds that aspects of credit risk are the source of violations in CIRP in the long-term capital markets rather than transactions costs or the size of the economy.     

Stock repurchases: How firms choose between a self tender offer and an open-market program

In practice, open-market stock repurchase programs outnumber self tender offers by approximately 10–1. This evidence is puzzling given that tender offers are more efficient in disbursing free cash and in signaling undervaluation – the two main motivations suggested in the literature for repurchasing shares. We provide a theoretical model to explore this puzzle. In the model, tender offers disburse free cash quickly but induce information asymmetry and hence require a price premium. Open-market programs disburse free cash slowly, and hence do not require a price premium, but because they are slow, result in partial free cash waste. The model predicts that the likelihood that a tender offer will be chosen over an open-market program increases with the agency costs of free cash and decreases with uncertainty (risk), information asymmetry, ownership concentration, and liquidity. These predictions are generally consistent with the empirical evidence.     

Incorporating the dynamics of leverage into default prediction

A firm’s current leverage ratio is one of the core characteristics of credit quality used in statistical default prediction models. Based on the capital structure literature, which shows that leverage is mean-reverting to a target leverage, we forecast future leverage ratios and include them in the set of default risk drivers. An out-of-sample analysis of default predictions from a hazard model reveals that the discriminative power increases substantially when leverage forecasts are included. We further document that credit ratings contain information beyond the one contained in standard variables but that this information is unrelated to forecasts of leverage ratios.     

On the relationship between exchange rates and equity returns: A new approach

This paper employs a new approach in order to investigate the underlying relationship between stock markets and exchange rates. Current approaches suggest that the relative equity market performance of two countries is linked to their exchange rate. In contrast, this study proposes an alternative approach where one global variable – global equity market returns – is believed to have an effect on exchange rates, with the relative interest rate level of a currency determining the sign of the relationship. Our empirical findings suggest that exchange rates and global stock market returns are strongly linked. The value of currencies with higher interest rates is positively related with global equity returns, whereas the value of currencies with lower interest rates is negatively related with global equity returns.     

The effect of monetary policy on real estate price growth in China

Using quarterly data from 1998:Q1 to 2009:Q4 and monthly data from July 2005 to February 2010, this paper examines the impact of key monetary policy variables, including long-term benchmark bank loan rate, money supply growth, and mortgage credit policy indicator, on the real estate price growth dynamics in China. Empirical results consistently demonstrate that expansionary monetary policy tends to accelerate the subsequent home price growth, while restrictive monetary policy tends to decelerate the subsequent home price growth. These results suggest that Chinese monetary policy actions are the key driving forces behind the change of real estate price growth in China. We also show that hot money flow does not have a significant impact on the change of home price growth after controlling for the money supply growth. Finally, a bullish stock market tends to accelerate subsequent home price growth.     

International transmission of information: evidence from the Euroyen and Eurodollar futures markets

This paper examines the information transmission between Japan and the US by using the Tokyo Euroyen and Chicago Eurodollar futures. These two interest rate futures markets provide a better understanding of international information transmission than stock markets, which have been shown to exhibit nonsynchronous trading and market segmentation. The results show that traders in Tokyo (Chicago) use information that is revealed overnight in Chicago (Tokyo). The bivariate EGARCH-t model provides no evidence of volatility spillovers in either direction, suggesting that the opening price rapidly reflects foreign information. The overall results support the hypothesis that the domestic market efficiently adjusts to foreign news. The results are also broadly consistent with the covered interest arbitrage effects.     

The Forward Exchange Rate Bias Puzzle Is Persistent: Evidence from Stochastic and Nonparametric Cointegration Tests

An important puzzle in international finance is the failure of the forward exchange rate to be a rational forecast of the future spot rate. We document that even after accounting for nonstationarity, nonnormality, and heteroskedasticity using parametric and nonparametric tests on data for over a quarter century, U.S. dollar forward rates for the major currencies (the British pound, Japanese yen, Swiss franc, and the German mark) are generally not rational forecasts of future spot rates. These findings deepen the forward exchange rate bias puzzle, especially as these markets are the most liquid foreign exchange markets with very low trading costs.     

Taylor rules, McCallum rules and the term structure of interest rates

Recent empirical research shows that a reasonable characterization of federal-funds-rate targeting behavior is that the change in the target rate depends on the maturity structure of interest rates and exhibits little dependence on lagged target rates. See, for example, Cochrane and Piazzesi [2002. The Fed and interest rates—a high-frequency identification. American Economic Review 92, 90-95.]. The result echoes the policy rule used by McCallum [1994a. Monetary policy and the term structure of interest rates. NBER Working Paper No. 4938.] to rationalize the empirical failure of the ‘expectations hypothesis’ applied to the term structure of interest rates. That is, rather than forward rates acting as unbiased predictors of future short rates, the historical evidence suggests that the correlation between forward rates and future short rates is surprisingly low. McCallum showed that a desire by the monetary authority to adjust short rates in response to exogenous shocks to the term premiums imbedded in long rates (i.e. “yield-curve smoothing”), along with a desire for smoothing interest rates across time, can generate term structures that account for the puzzling regression results of Fama and Bliss [1987. The information in long-maturity forward rates. The American Economic Review 77, 680-392.]. McCallum also clearly pointed out that this reduced-form approach to the policy rule, although naturally forward looking, needed to be studied further in the context of other response functions such as the now standard Taylor [1993. Discretion versus policy rules in practice. Carnegie-Rochester Conference Series on Public Policy 39, 195-214.] Rule. We explore both the robustness of McCallum's result to endogenous models of the term premium and also its connections to the Taylor Rule. We model the term premium endogenously using two different models in the class of affine term-structure models studied in Duffie and Kan [1996. A yield-factor model of interest rates. Mathematical Finance 57, 405-443.]: a stochastic volatility model and a stochastic price-of-risk model. We then solve for equilibrium term structures in environments in which interest rate targeting follows a rule such as the one suggested by McCallum (i.e., the “McCallum Rule”). We demonstrate that McCallum's original result generalizes in a natural way to this broader class of models. To understand the connection to the Taylor Rule, we then consider two structural macroeconomic models which have reduced forms that correspond to the two affine models and provide a macroeconomic interpretation of abstract state variables (as in Ang and Piazzessi [2003. A no-arbitrage vector autoregression of term structure dynamics with macroeconomic and latent variables. Journal of Monetary Economics 50, 745-787.]). Moreover, such structural models allow us to interpret the parameters of the term-structure model in terms of the parameters governing preferences, technologies, and policy rules. We show how a monetary policy rule will manifest itself in the equilibrium asset-pricing kernel and, hence, the equilibrium term structure. We then show how this policy can be implemented with an interest-rate targeting rule. This provides us with a set of restrictions under which the Taylor and McCallum Rules are equivalent in the sense if implementing the same monetary policy. We conclude with some numerical examples that explore the quantitative link between these two models of monetary policy.     

A stubborn persistence: Is the stability of leverage ratios determined by the stability of the economy?

The choice of capital structure firms make is a fundamental issue in the financial literature. According to a recent finding, the capital structure of firms remains almost unchanged during their lives. This stability of leverage ratios is mainly generated by an unobserved firm-specific effect that is liable for the majority of the variation in capital structure. We demonstrate that even substantial changes in the economic environment do not affect the stability of firms' leverage due to the presence of credit constraints. Financially unconstrained firms are more responsive to economic changes and adjust to the target substantially faster than constrained firms. Moreover, accounting for the ownership structure of firms boosts the explanatory power of the model in the subsample of unconstrained firms, suggesting that annual information on ownership and ownership changes together with financial constraints have the potential to be an answer to the puzzle of stability in capital structure.     

Macroeconomic risk and the cross-section of stock returns

We develop a conditional version of the consumption capital asset pricing model (CCAPM) using the conditioning variable from the cointegrating relation among macroeconomic variables (dividend yield, term spread, default spread, and short-term interest rate). Our conditioning variable has a strong power to predict market excess returns in the presence of competing predictive variables. In addition, our conditional CCAPM performs approximately as well as Fama and French’s (1993) three-factor model in explaining the cross-section of the Fama and French 25 size and book-to-market sorted portfolios. Our specification shows that value stocks are riskier than growth stocks in bad times, supporting the risk-based story.     

In- and out-of-sample specification analysis of spot rate models: Further evidence for the period 1982–2008

We review and construct consistent in-sample specification and out-of-sample model selection tests on conditional distributions and predictive densities associated with continuous multifactor (possibly with jumps) and (non)linear discrete models of the short term interest rate. The results of our empirical analysis are used to carry out a “horse-race” comparing discrete and continuous models across multiple sample periods, forecast horizons, and evaluation intervals. Our evaluation involves comparing models during two distinct historical periods, as well as across our entire weekly sample of Eurodollar deposit rates from 1982 to 2008. Interestingly, when our entire sample of data is used to estimate competing models, the “best” performer in terms of distributional “fit” as well as predictive density accuracy, both in-sample and out-of-sample, is the three factor Chen (Chen, 1996) model examined by Andersen, Benzoni and Lund (2004). Just as interestingly, a logistic type discrete smooth transition autoregression (STAR) model is preferred to the “best” continuous model (i.e. the one factor Cox, Ingersoll, and Ross (CIR: 1985) model) when comparing predictive accuracy for the “Stable 1990s” period that we examine. Moreover, an analogous result holds for the “Post 1990s” period that we examine, where the STAR model is preferred to a two factor stochastic mean model. Thus, when the STAR model is parameterized using only data corresponding to a particular sub-sample, it outperforms the “best” continuous alternative during that period. However, when models are estimated using the entire dataset, the continuous CHEN model is preferred, regardless of the variety of model specification (selection) test that is carried out. Given that it is very difficult to ascertain the particular future regime that will ensue when constructing ex ante predictions, thus, the CHEN model is our overall “winning” model, regardless of sample period.