What is the best way to do asset price trend analysis? Two recent papers address this question. In the May 2015 version of their paper entitled “Which Trend is Your Friend?”, Ari Levine and Lasse Pedersen compare time series (intrinsic or absolute) momentum, moving average (fast and slow) crossovers and other trend indicators to determine the best way to identify a price trend. In the May 2015 version of their paper entitled “Uncovering Trend Rules”, Paul Beekhuizen and Winfried Hallerbach describe how to determine the underlying historical weighting schemes (a combination of continuation and reversion) of price moving averages and combinations of price moving averages. Using both theoretical analyses and examples, these papers conclude that:
From “Which Trend is Your Friend?”:
- Both empirically and theoretically, time series momentum and and moving average crossovers are closely related and, in their most general forms, equivalent. They also capture other types of linear filters, such as the Hodrick-Prescott filter and the Kalman filter.
- Past prices and past returns are equally effective for conventional trend identification.
- An empirical test of three time series momentum signals and three similar moving average crossover signals on daily data for 58 liquid futures/forwards contract series suggests that longer measurement intervals (12 months) work better than short ones (1 or 3 months) and perhaps that times series momentum is more flexible than moving average crossovers.
- Trend-following investors and managers should probably focus on portfolio implementation issues (transaction costs, position sizing and risk management) rather than which type of trend-following algorithm to use.
From “Uncovering Trend Rules”:
- Trend rules based on returns are more precise than those based on prices for exploration of weights assigned according to age of inputs, including adjustment for autocorrelation.
- Rules that combine different moving averages may inherently weight old returns more than recent ones and include hidden mean reversion contributions.
- The moving average convergence/divergence (MACD) indicator is both a trend-following and mean reversion measurement tool.
In summary, both theory and evidence suggest that trend-following indicators are not as diverse as they appear, and returns-based indicators accommodate precise empirical weighting of inputs by age.
Cautions regarding findings include:
- Accounting for implementation frictions is important in comparing alternative trend-following rules. Rules based on short measurement intervals generate more signals, and incur greater frictions, than those based on long measurement intervals.
- In general, exploration of alternative trend-following rules invites data snooping bias, such that the best-performing rules overstate expectations.
- Fine tuning of weights for trend indicator inputs by age, especially with noisy daily data, would tend to convey large snooping bias.
