Improving Time-Series Momentum Strategies: The Role of Trading Signals and Volatility Estimators

Abstract

Motivated by studies of the impact of frictions on asset prices, we examine the effect of key components of time-series momentum strategies on their turnover and performance from 1974 until 2013. We show that more efficient volatility estimation and price trend detection significantly reduce portfolio turnover and therefore rebalancing costs. The poor performance of time-series momentum strategies during the post-2008 period is explained by an increased level of pairwise correlations. We propose a novel correlation-based leverage-adjustment to the strategy's weighting scheme and show that it improves performance by safeguarding against tail risk, even after accounting for realistic transaction costs. JEL CLASSIFICATION CODES: E37, G11, G15, F37

Asset Pricing, Investment, and Trading Strategies

“…Similarly to Elaut and Erdős (2019), we use a decay factor of 0.4 as a way to achieve an ex ante volatility of 40% per security, which can be expected to result risk factors with an ex post volatility of approximately 12% per year (which represents a typical CTA target volatility of around 12%). This model is based on the earlier works of Moskowitz et al (2012) and Baltas and Kosowski (2012). The term "time-series momentum" was first introduced by Moskowitz et al (2012) who documented a presence of persistent "trend" factors across a broad range of futures markets.…”

“…Their paper offers one of the most comprehensive time-series momentum studies across various futures markets (equity index, commodity, foreign exchange, and fixed income). Baltas and Kosowski (2012) contribute by suggesting several alternative estimates for a time series momentum: return sign, moving average, trend extraction, time-series t-statistics, and statistically meaningful trend, with the last alternative being referred to as the most efficient. We use the model proposed by Elaut and Erdős (2019), which is based on Moskowitz et al (2012), and allow for estimating and comparing time-series momentum signals over selected lookback periods.…”

Prediction of cyclic and trend frequencies in time series using the Hilbert-Huang transform

“…The authors estimate the 'true' trend by selecting the sequential IMF components which have a correlation coefficient above a given threshold and then summing the components. Baltas et al generate a trend predictor from the direction of change of the residual component of EMD and are able to show statistical significance for the predictor (Baltas and Kosowski, 2012). Moghtaderi et al propose extracting trend from EMD based on a change in the ratios of sequential IMF zero-crossing numbers and also an increase in the IMFs energy compared to the expected behaviour of broadband processes (Moghtaderi et al, 2013).…”

“…Believing that trends are the dominant feature in financial time series, it makes sense to incorporate this information our prediction. Baltas et al generate a trend predictor from the direction of change of the residual component of EMD (Baltas and Kosowski, 2012). Their predictor is scaled by the 30-day ex-ante market volatility (Yang and Zhang, 2000).…”

Partial Moment Momentum

“…The literature has further shown that momentum returns are related to partial moments, especially lower partial moments [see Menkhoff and Schmeling (2006), Baltas and Kosowski (2012), and Daniel, Jagannathan, and Kim (2012)]. We employ the idea of downside realized partial moment ( − ) and upside realized partial moment ( + ) in the context of momentum trading strategies.…”