By LARRY SWEDROE
Momentum, the tendency of past winner stocks to outperform past loser stocks over the next several months, is one of the most well-documented and well-researched asset pricing anomalies. In our book, Your Complete Guide to Factor-Based Investing, Andrew Berkin and I present the evidence of a premium that has been persistent across long periods of time, pervasive around the globe and across asset classes, robust to various definitions and survives transactions costs.
Erik Theissen and Can Yilanci contribute to the momentum literature with their January 2021 paper, Momentum? What Momentum? They began by noting: “Previous papers usually estimate risk-adjusted momentum returns by sorting stocks into a long-short portfolio based on their prior return. The portfolio is rebalanced monthly. The returns of the momentum portfolio are then regressed on a set of factors in a full-sample regression. This methodology, which we denote , implicitly assumes constant factor exposure of the momentum portfolio. However, momentum portfolios are characterised by high turnover which results in strongly time-varying factor exposure.”
To determine if momentum’s excess returns could be explained by time-varying factor exposures (in effect, is momentum really factor momentum?), Theissen and Yilanci estimated factor sensitivities at the stock level using a rolling window approach. For each month , they estimated the factor exposure for the stocks in the winner and the loser portfolio using data up to month - 1. They then estimated the expected return in month for each stock. The momentum profit in month is then the actual return of the long-short portfolio minus the weighted average of the expected returns of the individual stocks.
Theissen and Yilanci's “stock-level risk adjustment” accounts for the turnover in the momentum portfolio because, in each month, the factor exposure of the momentum portfolio is based on the actual composition of the winner and loser portfolios. Their data sample included NYSE, Nasdaq and AMEX stocks covering the period 1963-2018. Each month they sorted stocks based on their prior period returns into decile portfolios. They constructed zero net investment portfolios by investing into the winner stocks (decile 10) and shorting the loser stocks (decile 1). They tested formation and holding periods of three, six, nine and 12 months, creating 16 strategies. Stocks below $3 were excluded. Following is a summary of their findings:
Without their risk adjustment, 15 of the 16 strategies delivered returns that were positive and significantly different from zero. Returns were also economically large, ranging from 0.18 percent to 0.85 percent per month.
Accounting for risk using portfolio-level risk adjustment based on the Fama-French five-factor (beta, size, value, profitability and investment) model, again, 15 of the 16 strategies delivered significant abnormal returns.
When implementing their stock-level risk adjustment procedure, profitability largely disappeared. The adjustment, on average, captured 94 percent of the momentum returns that remained after portfolio-level risk adjustment, and none of the 16 strategies delivered returns that were significantly different from zero.
There were no significant momentum returns for any size category (micro, small and large-cap stocks) when risk was adjusted at the stock level.
When they considered sub-periods, they found that the momentum strategy earned significant abnormal returns after stock-level risk adjustment in the first part of the sample period (1963-1979) but not thereafter. However, even during this sub-period, momentum returns were roughly 44 percent smaller if risk was adjusted at the stock level rather than at the portfolio level. And when transaction costs were considered, momentum profits became negative even for the first part of the sample period.
In a test of pervasiveness, in an international sample covering 20 developed countries, without risk adjustment there was a significant momentum effect (at the 5 percent level or better) in 19 <16> countries. With stock-level risk-adjustment, this number dropped to just three.
Stock-level risk adjustment (which captures the time variation in the market exposure of the strategy) reduces momentum profits significantly (or even eliminates them), while portfolio-level risk adjustment does not because it assumes constant factor exposures of the strategy under investigation.
The stock-level adjustment procedure largely explained the return of a volatility-scaled momentum strategy—the monthly stock-level adjusted mean return was 0.44 percent and was not significantly different from zero (t-statistic = 1.55).
Their findings led Theissen and Yilanci to conclude: “In contrast to the prior literature, we find that the Fama and French (2015) 5-factor model explains the profitability of momentum strategies.” They added that while the CAPM is unable to explain momentum returns even with stock-level risk adjustment, and the Fama-French three-factor (beta, size and value) model significantly reduces but does not eliminate momentum returns, both the Fama-French and q-factor (beta, size, investment and profitability) models are able to explain momentum returns. The authors concluded: “Thus, the profitability and investment factors appear to be necessary to explain momentum returns.” They also noted that their findings are consistent with prior research, which has found time-varying factor exposures of momentum strategies.
Theissen and Yilanci noted that their findings have important implications, as they document that the apparent profitability of momentum strategies is, to a large extent, compensation for factor exposures (or risk). These strategies may thus be delivering risk premiums rather than abnormal returns. Their findings are also consistent with those of Tarun Gupta and Bryan Kelly, authors of the 2019 paper Factor Momentum Everywhere, and those of Sina Ehsani and Juhani Linnainmaa, authors of the 2020 paper Factor Momentum and the Momentum Factor, who found that momentum in individual stock returns emanates from momentum in factor returns — a factor’s prior returns are informative about its future returns.
That momentum is found in factors should not come as a surprise, as the research continues to find that momentum exists wherever we look: in stocks, bonds, commodities, currencies, sectors, countries and regions. For example, Christopher Geczy and Mikhail Samonov, authors of the 2015 study 215 Years of Global Multi-Asset Momentum: 1800-2014 (Equities, Sectors, Currencies, Bonds, Commodities and Stocks), examined the evidence from 47 country equity indices, 48 currencies (including the euro), 43 government bond indices, 76 commodities, 301 global sectors and 34,795 U.S. stocks and found that over this 215-year history, the momentum return was consistently significant within each asset class and across six of them (country equities, currencies, country government bonds, commodities, global sectors and U.S. stocks).