Essays on Financial Econometrics and Forecasting

This thesis is comprised of three essays on the topics of financial econometrics and forecasting. In the first essay we examine whether speculative bubbles are present in the US and UK commercial, equity and residential real estate markets. The real estate indices are decomposed to fundamental and non-fundamental components using a wide set of economic indicators and penalized regressions. In order to determine whether the observed deviations of the actual price index from its fundamental value are due to the presence of bubbles, we use two complementary methodologies, the first based on right-side unit root tests for explosive behavior and the second on regime switching models for bubbles. The models using the alternative fundamental specifications are found to exhibit superior out-of-sample performance compared to the stylized alternative models.

In the second essay we set out to evaluate the benefits of integrating return forecasts from a variety of machine learning and forecast combination methods into an out-of-sample asset allocation framework. The performance of the portfolios consisting of stock, bond and commodity indices is evaluated for different levels of risk aversion and investment constraints, around business cycles and for different rebalancing frequencies. The mean-variance allocations are based on several estimates of the covariance matrix, while the effects of the return forecasts are also investigated when using the Conditional Value-at-Risk as an alternative risk measure in optimization. Comparing the multi-asset portfolios incorporating machine learning return forecasts, we find evidence of added economic value relative to the equally-weighted or the historical average benchmark portfolios.

In the final essay we propose a new approach to pairs trading, which takes advantage of the information in the conditional quantiles of the distribution of asset returns. In this framework the pairs are sorted and selected based on cointegration tests and during trading the trading signal is extracted using quantile regression. We apply the strategy to the S&P 100 constituents and evaluate the performance of the pairs trading strategy using a variety of economic and risk-adjusted metrics and under an asset pricing framework, in order to examine whether the profitability of the new strategy can be explained by various risk factors. Our findings suggest that the quantile regression pairs trading strategies based on the lower quantiles tend to outperform all other models.