The Interaction Of Value And Momentum Strategies

Summary of “The Interaction of Value and Momentum Strategies”

“The interaction of value and momentum strategies” is an article which summarizes the results of a research made by Clifford S. Asness.
This research is about two effective and related strategies which can be used to predict stock returns. The first one, value, generally works and it is strong among loser stocks (weak momentum) andquite weak among winner stocks (strong momentum). The second one, momentum, is particularly strong among expensive stocks.
Also, they realized their results could be consistent with BV/MV being a noisy proxy for a distress factor in a rational asset pricing model.

Data and methodology

The research consisted of monthly data tests from July 1963 to December 1994 on all NYSE, Amex and Nasdaqfirms.
They constructed value-weighted portfolios based on some variables: PAST(2,12), industry-relative log(BV, MV) and industry-relative D/P. For each month, they sort all NYSE firms independently on each of the variables into five portfolios. Then, they placed each firm into one of the quintiles based on the NYSE breakpoints. Finally, they also constructed intersections of those quintileportfolios.

Univariate tests

First of all, they examined the properties of the value-weight portfolios, considering each variable individually and they concluded that returns were increasing in past recent returns which is a statistically significant relationship; beta showed a small U-shaped relation to PAST(2,12) and size had a corresponding inverted U-shape.
Then, they measured dividends andbook value over the prior fiscal year with a measurement lag from December until July. Thus, they deduced that returns had an statistically significant increasing in log(BV,MV); beta was U-shaped. Besides, log(BV, MV) and PAST(2,12) are positively and significant associated with future returns and they are negatively associated with each other.
At the same time, for the univariate sort of D/P,they perceived returns were increasing in D/P, but this increase was weak compare to the other variables. Thus, they realized that industry-relative dividend yield was a weaker predictable variable than was BV/MV. Beta was generally falling in D/P; firm size was a shaped like an inverted U.

Sorting on both variables

At this point, they tested whether the two strategies were each strongerholding to other constant; and whether the marginal relations did vary.
They took the intersection of the PAST(2,12) and log(BV,MV) quintiles to construct twenty-five value-weighed portfolios with monthly returns. Then, they reported four measures: the average monthly return, the average value-weighed log(BV,MV), the average value-weighed D/P and the average value-weighed PAST(2,12). These reportstested a value strategy across the past year’s loser firms, whereas they also tested it across all the firms.
As a result, they obtained that the spread in average return between log(BV,MV) and any of the past four quintiles was statistically more significant than the spread of all the firms. However, the one induced by log(BV,MV) is nonexistent for the quintiles of winner firms.
Furthermore, theyobtained that value should work better if momentum was relatively constant. Momentum strategies worked within all value quintiles, and that was stronger for the most expensive firms, even though the spread in actual momentum among low-value firms was smaller than the one among high-value firms.
As a consequence, they determined that value and momentum were generally good strategies. However,value was weak among firms with strong momentum and momentum was weak among firm with high value.
After that, they tested if previous results were statistically significant and they finally concluded that:
* Value strategy was statistically significantly stronger over loser firms.
* Momentum strategy was statistically significant stronger over expensive firms.
* Increasing both of them...