Thursday, June 20, 2019

Times series versus cross-sectional momentum - History is on the side of just following winners, trend-following


There have been some managers who have started to say that trend-following is dead. These pronouncements have always occurred when performance has been down or lackluster. Yet, history tells a different story. Yes, there are time when momentum works and when it does not, but the long-run is on the side of the trend-follower and momentum strategies. Perhaps there have been market changes. There will always be market change, but consistency has always been a key advantage for the trend-follower. 

The extensively researched paper, "The Enduring Effect of Time-Series Momentum on Stock Returns over nearly 100-Years" by Ian D’Souza, Voraphat Srichanachaichok, George Wang, and Yaqiong Yao, does a great job of sifting through many combinations and alternatives to reach some well-documented conclusions. While this work is a few years old, it makes a good case for simple time series momentum modeling. The odds are still in your favor by following price action.  

The choices for using price momentum are simple. A time series momentum approach will focus on some look-back return period and buy stocks that have had positive returns and sell those that have had negative returns. The cross-sectional approach will rank order all of the returns for a specific time period and buy the top performing and sell the bottom performing stocks regardless of whether the returns are positive or negative. Both momentum choices can be market neutral with longs matching shorts. 

After looking at decades of stock information both in the US and around the globe, the researchers were able to generate some strong conclusions.
  • Times series momentum performs well over the long run as measured in decades. There has been a fall-off in performance over the last two decades, but the overall results are still positive for momentum strategies.  
  • Times series momentum will do better than cross-sectional momentum and can capture cross-sectional stock momentum return patterns while cross-sectional does not seem to capture the time series effects.
  • Cross-sectional momentum does well in up markets, but time series does well in both up and down markets. Cross-sectional models have a seasonal effect in January while time series does not. 
  • The time series models cannot be explained by conventional factors like Fama-French. The time series momentum is not a proxy for some other risk premium.
  • The data supports the under-reaction hypothesis for why time series modeling is effective. The data do not support an over-reaction hypothesis for momentum. The data also support the "frog-in-the-pan" hypothesis that the slow diffusion of information leads to time series momentum. Discrete information shocks do not lead to price trends. This suggests that small cap stocks with smaller analyst following and the slow dissemination of information will do better with momentum strategies.
  • The researchers also found that blending time series and cross-sectional approaches will do best of all. In this case, rank order performance, but only take positive returns for the long side and negative performers for the short exposure.


This work is consistent with other research posted on this blog that shows that trend-following is superior to momentum (See my blog's futures trading and alternative risk premia categories.) Under-reaction and the slow dissemination which likely leads to positive autocorrelation can be picked up through trends and momentum. Investors can increase their odds of success through focusing on market segments that are less followed and less likely to have discrete jumps in new information.

My take-away for all portfolio decisions no matter what asset class is very straightforward; never forget to look at trends and momentum. You can use fundamentals or other criteria for decision-making, but measure your actions against the price trends. Think of your actions based on the price priors. If prices are moving higher, the bar is set higher for taking action against the trend. If price momentum is falling, buying should always be done with caution. If you are not strongly informed or skilled with fundamentals just focus on the times series. 

No comments: