Residual time series momentum : An empirical asset price model based on a unique momentum effect

Absztrakt (kivonat)

I construct a model for asset price prediction that is based on two well-known methods in the literature, that were not yet applied jointly in the literature. These are the time series momentum that predicts an asset's next month's return with its own past return and residual momentum that is a method to reduce portfolio exposure to risk factors using residual returns. I use monthly returns of 55 liquid futures from 4 di�erent asset classes between 1985 and 2015 to analyze the model. To test its predictability, I implement univariate and pooled regressions and find stronger predicting power than that of time series momentum. I also implement a comprehensive analysis of the model-based trading strategy, finding that it is more profitable than buy-and-hold and time series momentum strategy with the same assets during the same period, and exhibit higher risk-adjusted returns than the most momentum strategy in the literature. The strategy performance is consistent and robust for several model parameters. I also examine the source of this anomaly and find that it is stronger in market states when volatility and uncertainty are low, investor sentiment is high and the yield curve is inverted. The residual time series momentum e�ect seems to disappear after 2005.