A long-standing challenge in the asset pricing literature is to understand why asset prices sometimes underreact and sometimes overreact to news. We seek to address this challenge in the context of currency markets. We construct a model of exchange rate determination disciplined by survey data, where short-lived investors each (1) receive noisy private signals about the future path of interest rate differentials between the US and other countries and (2) overestimate the persistence of interest rate differentials. The model is able to qualitatively and quantitatively match patterns of underreaction and overreaction of exchange rates in response to news. The model also matches the failure of uncovered interest rate parity (UIP), capturing the return predictability of interest rate differentials for the returns to borrowing in USD and lending in foreign currency, as well as the fact that this return predictability is declining in the maturity of bonds used to borrow and lend. Finally, we use the model to help understand the reversal of the failure of UIP in recent years, the role of higher-order uncertainty, and the persistence of subjective beliefs. Our results highlight the important role that investors' beliefs may play in asset price behavior.

This paper examines tail risk in the Brazilian hedge fund industry. We rely on a unique data set of daily returns for every hedge fund in Brazil, dead or alive. By employing the universe of hedge funds, we ensure the absence of selection, survivorship, and instant history biases. We estimate tail risk measures based on the cross-section of both equity and hedge-fund returns. We consider three tail risk measures. The first extracts a tail risk measure assuming that the tail of the cross-section distribution has a power law representation, whereas the second and third rely on the expected shortfall of the cross-section distribution under the physical and risk-neutral measures, respectively. We find that the tail risk estimates are very different not only across asset classes (equity vs hedge fund), but also across probability measures (physical vs risk neutral). More interestingly, we also show that, although the hedge fund industry in Brazil seems to exhibit more exposure to equity tail risk, hedge fund tail risk entails higher predictive ability to performance both over time and cross-sectionally.

We propose a conditional factor model for currency returns where the risk exposure evolves with the country's macroeconomic environment. We use Instrumented Principal Component Analysis (IPCA) to estimate the common risk structure of currencies and disentangle which macro characteristics are relevant for the cross-section of currency returns. The model is successful in explaining both realized variation in returns and differences in average returns across currencies. Estimated risk factors are correlated with the Dollar and Carry factors, but they still bring new information to the description of common risks. Interest rate level, the importance of the U.S. on a country's trade, GDP per capita, and commodities' share on exports are the most relevant macro variables for explaining the loadings of currencies on these risk factors.