Markov Switching Processes and Financial Applications

点击翻译按钮获取中文摘要

Since the seminal paper of Hamilton (1989), regime switching models have been used widely in diverse fields of finance. These models can capture abrupt and sudden changes often observed in the behavior of financial markets. Currency crises, stock market bubbles, or sharp changes in the volatility of asset prices or exchange rates are examples of such phenomena. Means and volatilities of financial returns can change between regimes, as well as the correlations between returns of several assets. In this project we intend to use such models in asset allocation and asset pricing. We also want to develop easy to implement tests for the number of regimes.******In asset allocation, regime switching models of asset returns are able, for example, to capture a higher correlation structure in bear periods or heteroscedasticity through distinct volatilities in different regimes. Our interest is to apply this class of models in a general investor's problem with finite horizon, in which the agent optimally chooses portfolio allocation and (possibly) consumption with stochastic differential utility preferences. The goal is to develop approximate analytical solutions for the optimal portfolio and consumption in a multiregime economy. ******In asset pricing, Bonomo, Garcia, Meddahi et Tedongap (2011) have modeled long-run risks in fundamentals with regimes to price assets in equilibrium with disappointment aversion recursive utility. The Markov structure leads to closed-form expressions for many statistics of interest such as means and volatilities of returns, and coefficients of predictability regressions. Our new contribution will extend such model to capture the term structure of equity returns. Traditional models have a hard time explaining the fact that short-term equity claims, or dividend strips, have higher average returns and Sharpe ratios than the aggregate stock market (see van Binsbergen and Koijen, 2016).******Our third line of investigation aims at developing a test to determine the number of regimes in a given set of data. The available tests (Garcia, 1998; Carrasco et al., 2014) are applicable to testing a two-regime model against a linear model. Their extension to higher numbers of regimes appears very challenging. We intend to examine the***size and power properties of two simple regression-based tests used in Garcia and Perron (1996).