Blog - Investing Notes

December 11, 2007 - When Momentum Does and Doesn't Work

Does momentum investing have cycles, or at least better/worse times? In their December 2007 paper entitled "Winner-minus-Loser Return Spreads, Return Dispersion, and Changes in the Market State", Chris Stivers and Licheng Sun investigate the profitability of momentum investing in stocks and industries over symmetric ranking-holding periods of 6, 18 and 36 months relative to the state of the stock market as indicated by recent dispersion of returns. Using monthly return data for individual NYSE/AMEX stocks and for 48 value-weighted industries during 1962-2005, they conclude that:

• Generally confirming prior studies:
  • For 6-month ranking-holding periods, the average winner-minus-loser (WML) return for individual stocks is a little over 1% per month, with the return somewhat lower in the second half of the sample. An industry-level WML strategy generates a positive return less than half that found for individual stocks.
  • For 18-month ranking-holding periods, the average WML return for individual stocks is about 0.25% per month.
  • For 36-month ranking-holding periods, the average WML return for individual stocks is negative.
• Lagged RD reliably exhibits a negative relationship with subsequent WML profitability for both individual stocks and industry groups over all three tested horizons. This relationship is particularly strong for the 18-month and 36-month horizons. For individual stocks over the entire sample period, change in RD explains over 34% (41%) of the variability in the subsequent 18-month (36-month) WML return.
• RD is related both to business-cycle variables/crises and to the likelihood of major trend changes in the stock market. Months during which the economy contracts, according to the National Bureau of Economic Research, tend to be preceded by relatively high values of RD. And, a relatively high RD is associated with an elevated probability of a change in the market, especially a good-to-bad transition, during which the WML return is low.
• The RD-WML relationship is not substantially limited to small stocks or fringe stocks with more subjective valuations.

The following chart, taken from the paper, displays the behaviors of three variables over the 1962-2005 sample period:

1. Return for an 18-month ranking-holding WML strategy for individual stocks (top 10% winners minus bottom 10% losers).
2. Return for an 18-month ranking-holding WML strategy for value-weighted industry groups (top 25% winners minus bottom 25% losers).
3. The lagged large-firm RD (average standard deviation of monthly returns for the largest 10% of firms over the past three months relative to a longer-period baseline).

Visual inspection confirms that WML (momentum) returns tend to weaken when RD is high.

In summary, momentum investing strategies are generally more (less) successful when the dispersion of returns across individual stocks or industries is relatively low (high).

For related research, see Blog Synthesis: Momentum Investing/Trading.