A new paper by two scholars affiliated with the University of New South Wales compares Australian institutional funds to U.S. mutual funds. Along the way, these two scholars illuminate the distinction between true alpha on the one hand and beta-posing-as-alpha on the other.
Their bottom line: Australian institutional funds generate positive alpha, even when alpha is defined in narrow and challenging ways. U.S. mutual funds don't. U.S. mutual funds can barely generate positive alpha when it is defined in such a way as to include the success of the simplest beta strategies, and even then they do so only when return is considered prior to fees.
Zhe Chen and David R. Gallagher co-authored the paper, "When Funds Diverge from Their Long-Term Factor Loadings." Their title refers to one of their critical findings: that the pay-off pattern of the Australian funds is convex to a baseline created from the returns that would be anticipated based on the funds' long-term factor loadings. Yes, though it sounds like quant-speak, but bear with me.
Convex and Concave
The convexity means two things: funds outperform the benchmarks when the factor loadings are particularly bad, and again when the factor loadings are particularly good. Both of these points are true whether the factor model involved is CAPM (one factor), Fama-French (three), or Carhart (four).
Chen and Gallagher see their findings as confirming a hypothesis published by Kay Mazuy and Jack Treynor in 1966, back in the early foundational days of the capital asset pricing model. Mazuy/Treynor asserted that funds that could outguess the market would be convex to the benchmark because they would reduce equity beta when returns were poor and they would load up on equity beta when the returns were good. Mazuy/Treynor found no such hopeful convexity in 1966, and Chen/Gallagher find none of it in the U.S. market now, either.
The hapless U.S. mutual funds Chen/Gallagher have sampled have a nominally positive pre-fee alpha only when measured against CAPM. Their alpha disappears against models that employ three or four factors, generating negative alpha even before fees: -0.30% alpha on the 3-factor model and -0.36% on the 4-factor model. Further, the pay-off patterns are concave against the baseline.
As we've discussed on this blog recently, 80% of respondents in a recent European survey on investors said they think the use of smart beta indexes allows for the capture of factor risk premia. If that is so, then managers who pursue alpha actively have to ask themselves: is what I am offering really alpha, or is it simply something that such an index could capture?
Meanwhile in Australia: roughly 80% of the institutional funds in the authors' database display convexity against their baseline. Their behavior is consistent with the presumption that those antipodean asset managers do have the ability to time market moves.
Reliance on Beta
The data suggest that mutual funds are relying especially on the size premium, one of the Fama-French factors, which is the one factor that seems to be most important in making their apparent alpha disappear into the statistical wind.
The Aussies lean the other way on the matter of size. Their portfolios are tilted toward larger stocks, as if assigning themselves a handicap for the comparison with the multi-factor models. Obviously that is a heck of an "as if." Handicapping themselves is surely the furthest thing from their minds! The tilt may simply be a consequence of the concentrated character of the Australian stock market, since the large cap stocks are where the liquidity is. Or it may arise for some more complicated reason. Still, given the empirical evidence supporting the Fama-French model, the Aussies' reliance on large cap firms does handicap them. Yet they show positive alpha against all the tested benchmarks. This is impressive.
Does all of this mean that Australia is a continent of great stock pickers? It could. Or it could be a reflection of the greater efficiency of U.S. stock markets, and so the greater challenge facing fund managers who seek alpha in that environment.