Performance assessment of volatility and distribution timing strategies

Resumen

As it si well known, an investor is more or less risk averse according to the economic and political circumstances. For instance, nowadays we are in a period in which even the most adventurous investor has had to reduce his optimistic expectations. Given that, it seems reasonable to model the risk aversion parameter so that it changes over time, in order to take into account the variability in agents' expectations. Thus, we aim to continue our study on the time-varying optimal protfolio context. According to the previous lines, there are some studies in financial literature that refer to time-varying risk aversion. For instance, Kim (2014) proposes a consistent indicator of conditional risk aversion in consumption-based CAMP. Other studies have differed widely in their estimates of time-varying risk aversion, such as Dionne (2014), who aim to extend the concept of orders of conditional risk aversion to orders of conditional dependent risk aversion. However, our motivation is in the line of the framework proposed by Cotter and Hanly (2010), which is based on estimating the risk aversion parameter as a derivation of the CRRA (This term refers to the changes in relative risk aversion, which is a way to express the risk aversion attitude of an investor through his utility function). Thus, we will estimate the risk aversion parameter through the conditional mean and variance. To bring out the last purpose, we will model these conditional moment through several GARCH schemes, such as the GARCH in mean (GARCH-M). Otherwise, creating an optimal portfolio is based on estimating its future risk and expected return as accurately as possible, based on a variety of inputs. The problem is how to estimate the last parameters. What goes into forecasting an optimal asset allocation has a big impact on what comes out. This often ends in quite inaccurate optimal allocations and may lead to counter-productive portfolio decisions. Given the above, we will try to answer the question: How best to forecast future portfolio weights?. To answer the question and to make a good forecasting of the future volatility and covariance matrix. A wide rnage of literature exists dealing with modelling and forecasting volatility. Most of this literature examines the relative performance of tompeting forecast in a generic statistical setting. For a broad overview of such literature see Poon and Granger (2003, 2005), Hansen and Lunde (2005), Becker and Clements (2008), Copeland (1999) or Fleming et al. (2001). However, we will follow the research line proposed by Clements and Silvennoinen (2013). According to these authors, we could forecast the covariance matrix through a recursive procedure based on the average of the time-varying covariance matrix calculated using the GARCH schemes. Finally, I would like to thank the opportunity to advance in this project following the line of the research team directed by the proposed teacher (Antonio Días Pérez). I consider the above a key element in my personal and professional development as a researcher.