Research Review: Momentum – Jegadeesh & Titman, 1993
posted on October 15th, 2015 by Bellmont Research Team
Investment styles that rely on analysis of historical share price data (as opposed to historical fundamental data such as earnings, dividends, book value etc) have long been pilloried by many in the investment community. Whilst some of these approaches may indeed bear more than a passing resemblance to ‘snake oil’, there is in fact a substantial body of rigorous academic research, published in top rated peer reviewed academic journals, that suggest that momentum – one of the central tenets of this approach, is very much worthy of consideration. In fact, ‘momentum’ ranks alongside ‘value’ as one of the two most widely researched, and broadly accepted factors that drive equity market returns.
One of the groundbreaking works in this area was undertaken by Jegadeesh & Titman from the Anderson Graduate School of Management at UCLA, and published in the Journal of Finance in March 1993. Their paper, entitled “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency” sought to compare the performance of stocks that have performed well in the past with those that have performed poorly. Their theory was that if stock prices either overreact or underreact to information consistently (as suggested by behavioural finance theorists), then profitable trading strategies that select stocks based on their past returns will exist. Previous studies in this area had shown that over a 3 to 5 year horizon, stocks that had performed poorly over the previous 3 to 5 years achieved higher returns than stocks that performed well over the same period. That is, over this timeframe there is a ‘reversal’ of returns (a major contributing factor to the ‘value effect’). What Jagadeesh & Titman were interested in though was shorter timeframes of both investment and ‘look-backs’ of 3 to 12 months.
Their study, which covered the period from 1965 to 1989, involved ranking each company at the beginning of each month by its returns over the last ‘J’ months. Based on these rankings, ten equally weighted decile portfolios are constructed, with the portfolio comprised of those companies with the strongest historical returns dubbed the ‘winners’, and the portfolio with the lowest historical returns dubbed the ‘losers’. They then tracked the performance of each of these portfolios over the following ‘K’ months, as well as a ‘zero cost’ portfolio that was long the ‘winners’, and short the ‘losers’. The timeframes for both their look-back (‘J’) and investment (‘K’) horizons were 3, 6, 9 and 12 months. That is, for the 3 month look-back period for example, they constructed portfolios based on companies’ 3 month historical returns, and looked at their performance over the subsequent 3, 6, 9 and 12 months.
As the single most descriptive figure, the paper focused largely on the ‘zero cost’ portfolios, representing the difference in performance between the ‘winners’ (highest past returns) and ‘losers’ (lowest past returns) deciles. Incredibly, the returns of all 16 of these ‘zero cost’ portfolios were positive, with the 6 month look-back / 6 month investment portfolio (the median timeframe strategy that the authors studied in most depth for the paper) generating annualised returns of 12.01% per annum, and a very impressive 15.7% average return for the most successful timeframe – the 12 month / 3 month portfolio; very impressive results indeed for a portfolio that would theoretically require no capital to implement, and that has no net market exposure! Whilst the high turnover of the strategy clearly detracts somewhat from the excess returns, even under draconian assumptions of 0.5% per side transaction fees, the 6 month / 6 month portfolio still generates excess returns of 9.29% per annum.
What is particularly interesting about these results though is how they fit with other previous research into over and underreaction, and the ‘value effect’. Whilst the returns over the 3 to 12 month investment horizon were universally positive, with positive average returns for the zero cost portfolio in all but the first month in year 1, average returns in year 2 are negative in every month, as well as in the first half of year 3 – and virtually zero thereafter. This suggests that the momentum strategy hasn’t enabled the selection of stocks generating higher long term returns, but instead the ‘momentum’ over this 3 to 12 month time horizon appears to be a temporary effect.
“Trading strategies that buy past winners and sell past losers realise significant abnormal returns over the 1965 to 1989 period. For example, the strategy we examine in most detail, which selects stocks based on their past 6 month returns, and holds them for 6 months, realises a compound excess return of 12.01% per year on average”.
Whilst initially hard to reconcile with prior research into the ‘value’ effect, this paper in fact shows exactly how momentum dove-tails with value in the markets. Over short time-frames of 3 to 12 months, stock prices show continuation – with past winners continuing to perform strongly, and vice-versa (consistent with theories relating to overreaction). This pushes the price of these past winners (losers) above (below) fair value, which leads to lower (higher) medium to long term returns – entirely consistent with the ‘value’ effect, and past research showing reversal of prices over a 3 to 5 year horizon.
The key point then in reconciling the two apparently conflicting approaches appears to be timing. The potential to combine these two most robust anomalies appears to be a very interesting and appealing area of future research.