Beta estimation on different trading periods

The purpose of this thesis is to investigate on beta estimation considering different trading periods. Starting from a literature review, given in the first chapter, we explain the meaning of beta and the different methodologies used for its estimation. We find that there is poor literature evidence on the differences between daily, intraday and overnight betas, so we decide to focus on this topic. In order to understand if beta estimation on different trading periods matter, in the first chapter, we divide the total daily return in intraday and overnight return and evaluate daily, intraday and overnight betas using two different models. Starting by the classical Capital Asset Pricing Model (CAPM), and assuming a constant systematic risk, i.e. a constant beta over time, we estimate our three betas. Subsequently, we consider a nonparametric method for time-varying conditional betas, proposed by Ang and Kristensen (2012) and Li and Yang (2011). By using this model we compute time-varying betas in conditional factor models which are conditional on the realized betas. For both these models we estimate daily, intraday and overnight betas considering US stocks traded on the NYSE, AMEX, and NASDAQ markets. Furthermore, we try to understand if there is some relation between the beta parameter and the stock size. Taking into account the differences in pattern between the daily intraday and overnight betas, found in the second chapter, we decide to investigate if it is possible to take advantage of the different behaviors in a trading strategy based on the beta estimation. In particular, we consider the statistical arbitrage strategy, proposed by Frazzini and Pedersen (2014), Betting Against Beta (BAB) and we adapt it constructing three different portfolios ranked and organized by daily, intraday and overnight betas.