Momentum : autocovariances, cross-covariances or unconditional expected returns?
1University of Oulu, Oulu Business School, Department of Finance, Finance
|Online Access:||PDF Full Text (PDF, 1.2 MB)|
|Persistent link:|| http://urn.fi/URN:NBN:fi:oulu-202001241079
Oulu : L. Heikka,
|Publish Date:|| 2020-01-24
|Thesis type:||Master's thesis
Returns to momentum strategies can be decomposed into three sources of return: positive autocovariances in returns, negative cross-covariances in returns and cross-sectional variation in unconditional expected returns across assets. While theoretical literature on momentum generally assumes that positive autocovariances drive momentum returns, empirical literature presents inconsistent evidence on the importance of each component in explaining momentum returns. However, this prior literature ignores a key empirical issue related to short return histories and extreme return observations. These extreme returns cause a negative bias to sample estimates of return autocovariances and cross-covariances, and a positive bias to the cross-sectional variation in unconditional expected returns. Furthermore, prior literature focuses on portfolios of stocks instead of individual stocks, because the decomposition requires estimating a cross-covariance matrix, which is difficult for individual stocks with short and non-overlapping return histories.
I propose a novel, strategy-based decomposition, which allows for estimating the contribution of each component to momentum without bias in the presence of extreme returns and non-overlapping return histories. Empirical evidence from the strategy-based decomposition in a sample of US individual stocks and portfolios of US stocks extending from 1926 to 2018 suggests, that positive return autocovariances are the most important driver of momentum. This evidence is consistent with most behavioral theories on momentum but does not preclude a rational interpretation. When the contributions are allowed to vary over time, positive autocovariances consistently remain the most important driver of momentum returns throughout the sample period. However, the time-series predictability in individual stock returns slowly erodes after momentum is initially documented in the literature in 1993, potentially explaining the poor recent performance of momentum.
© Lauri Heikka, 2020. This publication is copyrighted. You may download, display and print it for your own personal use. Commercial use is prohibited.