Measuring the Size Effect with Capitalization-based ETFs
Posted in Size Effect
June 30, 2009
Do popular capitalization-based exchange-traded funds (ETF) confirm the existence of an exploitable size effect? To investigate, we compare the difference in returns (small minus large) for the following matched pair of small-large ETFs:
iShares Russell 2000 Index (Smallcap) Index (IWM)
iShares Russell 1000 (Largecap) Index (IWB)
Using monthly adjusted closing prices (incorporating dividends) for these ETFs during May 2000 (the earliest month available for both) through June 2009 (110 months), we find that:
The following table shows the average monthly small-minus-large return for the entire sample period and for two equal subperiods. Overall, the effect is large enough to be interesting. However, the average outperformance of small stocks concentrates in the first half of the sample.
Do outliers drive the size effect? Excluding the two best (worst) months for small stocks decreases (increases) the average small-minus-large return to 0.3% (0.6%) for the entire sample period.
Does the January effect drive the size effect? Removing Januaries from the sample reduces the average monthly small-minus-large return from 0.5% to 0.4% for the overall sample, so Januaries do not dominate the overall result.
For a closer look, we plot the variation in small-minus-large monthly returns.
The following chart plots the monthly small-minus-large return over the entire sample period. A best-fit line indicates that any size effect has disappeared over the sample period. It is possible that the introduction of capitalization-based ETFs has made it easy for investors to pursue, and extinguish, any size-related abnormal returns. Or, it may be that the sample is too short in duration for reliable measurement of a size effect (if, for example, the effect has some underlying multi-year dependency on some economic factor).
For a different perspective, we look at annual returns.
The final chart summarizes small-minus-large returns at the annual level (with 2009 a partial year), showing that results for 2001 and 2003 substantially determine the positive result for the overall sample period.
In summary, a simple test of capitalization-based ETFs with available data (about nine years) indicates a potentially interesting but disappearing (disappeared?) size effect.
