Trend-following and market inefficiency - It is in the mean reversion

 

A lot can be learned about trend-following by using some simple time series processes to describe the behavior of markets. Some key insights on trend-following can be derived using the simple Ornstein-Uhlenbeck (O-U) process. This stationary Gaussian Markov process can help explain trends through variations in a mean reversion parameters, speed of adjustments. Mean reversion will create trends as price return to equilibrium, mean values, after shocks. A short paper, "To Be Or Not To Be a Trend Follower" by Andreas Junge of Methodica Ventures attracted my interest in this topic.

The O-U process states that tomorrow's price is just today's price minus an adjustment of today's price to a mean value plus an error term. The random walk, market efficiency, is just a special case where the mean reversion parameter is zero. This process is not new and has been a workhorse with many derivative pricing models, but it provides a framework on how markets may work. 

If there is a positive or negative shock to the market, the mean reversion term means there will be a slow trend adjustment back to a long-term mean. Market efficiency as defined by a random walk says there is no adjustment to a mean or a very low adjustment factor. The O-U process also tells us that volatility will be dampened relative to an efficient market environment (no mean reversion). Differences in mean reversion will reflect the potential for trend profits. Unfortunately, while a mean reversion parameter will help generate a trend, future shocks, especially in the opposite direction will create noise that will mask the first or primary trend. It is notable that trend changes will likely be centered on information dates when there is a greater chance of a shock.

A framework for how markets operate when they are efficient and a framework for when behavior will differ from efficiency is useful when looking for market opportunities. A framework describing the evolution of price behavior helps to form a prior for trend expectations.