Does the size effect vary with the state of the economy? In his October 2010 paper entitled “The Behaviour of Small Cap vs. Large Cap Stocks in Recessions and Recoveries: Empirical Evidence for the United States and Canada”, Lorne Switzer examines the relative performance of small versus large capitalization stocks around economic peaks and troughs (per NBER business cycle data). Using monthly returns for U.S. (Canadian) stocks starting with January 1926 (1987), associated firm characteristics and contemporaneous economic and equity market benchmark data through August 2010, he finds that:
- During 1926-2009, the geometric mean gross annual return of U.S. small caps is 2.03% higher than that of the S&P 500 Index. This size effect is most pronounced during 1976-1982, disappears during 1983-2000 and reappears during 2001-2010.
- Small-cap value stocks exhibit significantly positive alpha for the overall sample period, 1976–1982 and 2001-2010.
- Small cap growth stocks exhibit significantly positive alpha only during 1976–1982.
- Canadian small caps outperform large caps during 2001-2010, but not during 1987–2000.
- Small caps tend to outperform (underperform) large caps in the year after (before) an economic trough (peak). Specifically:
- Small caps outperform during all recoveries except the one after the June 1938 trough, for which small caps still had a one-year return of 27%.
- While the average small-cap premium is positive (1.8%) during the year before onset of the 14 recessions from 1926 to 2007, it is negative for eight of 14.
- During recessions, neither small caps nor large caps consistently outperform the other.
- Default risk (long-term corporate minus government bond yield spread) is the economic indicator most closely aligned with the size effect. A high (low) default risk indicates a large (small) size effect.
In summary, evidence indicates that U.S. small caps, driven mostly by value stocks, tend to perform relatively well (poorly) in the year after (before) an economic trough (peak).
Cautions regarding findings include:
- The above summary incorporates clarifications/corrections from the author in response to questions.
- NBER business cycle peaks and troughs are not known until well after the fact, inhibiting exploitation of findings.
- Return calculations ignore trading frictions, which tend to be percentage-wise higher for small cap stocks. Including reasonable trading frictions for reforming style-based portfolios monthly would materially affect results.
- Statistical significance tests assume tame return distributions. To the extent actual return distributions are wild, these tests break down.
