Essays on Financial Markets

My dissertation consists of three main chapters: one on portfolio optimization, and two on empirical asset pricing. The first chapter employs monthly and daily returns of US stocks to evaluate the out-of-sample performance of investment rules stemming from the mean-variance, Kelly and universal portfolio literature. We find that none of the strategies considered is significantly better or worse than all the others. Moreover, we show that the theoretical goal of the different strategies, be it either the maximization of the risk-adjusted portfolio return or the final wealth, is not related to their out-of-sample performance relative to the different measures adopted. Conversely, agents should take into account the properties (return, risk and correlation) of the set of stocks selected for investment when they are choosing the portfolio model to follow. The second chapter adopts the latent variables approach by Hwang & Salmon (2004) to analyse style herding in the value-growth and size dimensions of US domestic equity mutual funds. We document that mutual fund herding in styles is significant and persistent. Furthermore, the results show that mutual fund herding tends to increase after periods of high cumulative returns and market volatility. A higher sentiment is followed by an increase in mutual fund herding towards small stocks. Instead, mutual fund herding in value stocks significantly decreases after an improvement in economic conditions. We also observe that mutual fund herding in styles causes overpricing in the market portfolio, and SMB and HML factors. The third chapter incorporates a speculative bubble subject to a surviving and a collapsing regime into the present-value model by Binsbergen et al. (2010). To estimate this new high-dimensional model, we develop an efficient Markov chain Monte Carlo sampler to simulate from the joint posterior distribution. Our setup is able to correctly identify 92.27% of all the bubble collapsing dates in the artificial datasets, and it never signals a bubble when there is none in the data generating process. We then show the existence of significant Markov-switching structures in real-world stock price bubbles. Further, the results indicate that dividend growth rates are highly predictable. Finally, we find that in the surviving bubble regime, bubble variation accounts for most of the variation in the price-dividend ratio and unexpected return.