Faulty study design accounts for much of the evidence that risk-factor investing is superior to asset class models, according to two prominent managers.
(July 11, 2013) - Risk-factor models routinely outperform asset class-based portfolios in academic studies, but according to a new article, superior optimization often isn't the reason why.
Rather, these theoretical factor-based portfolios may win out thanks to investment constraints hindering model asset class portfolios. Thomas Idzorek, CIO of Morningstar Investment Management, and Maciej Kowara, research director at Transamerica Asset Management, contended that studies often compare simple, long-only strategies to risk-based models that are allowed to go short and pursue niche assets.
"Most such [pro-risk factor] papers use apples-to-oranges comparisons that lead all but the most careful reader to believe that risk-factor-based asset allocation is inherently superior to allocation based on asset classes," Kowara and Idzorek wrote in the Financial Analysts Journal article. "In some cases, the apparent superiority of the risk factors is a simple result of the fact that the risk factors are, in a guise, a set of asset classes with the long-only constraint removed."
One of the studies cited by the asset managers—"Portfolio of Risk Premia: A New Approach to Diversification," a 2009 MSCI paper—uses a two-index 60/40 portfolio of developed market equities and US investment-grade bonds as its traditional strategy. The risk premia model includes 15 factors represented by 11 indices across long and short positions.
Between May 1995 and October 2008, the asset class portfolio would have earned 2.2% annualized above the risk-free rate, with a maximum drawdown of -30.6%. The risk premia approach would have gained 2.7% in excess returns while its steepest drawdown was -11.8%.
Amid their critique of the field's research methodologies, Kowara and Idzorek are adamant that they are not against risk-aware investing. "We applaud the innovation of using risk factors in the asset allocation process," they wrote. "Positive messages from these papers with which we agree are that relaxing the long-only investment constraint and expanding one's opportunity set to include more potential exposures are often good things."
Based on their own self-proclaimed "apples-to-apples" study of the two strategies over the 1979 to 2011 period, risk factors are neither a better nor worse way to structure a portfolio than asset classes: "If one truly creates an even playing field, there is no gain in efficiency."
The industry has no standard set of portfolio risk factors. Still, based on the perspective of one leader in the field, Columbia Business School's Andrew Ang, the long-only 60/40 "asset class" model described above is also a factor portfolio—just a very, very simple one: "Some risk factors are themselves asset classes, like plain-vanilla stocks and bonds," Ang said in a working paper published last month. "These are passive factors earned by simply going long the equity or bond market."
Thomas Idzorek and Maciej Kowara's "Factor-Based Asset Allocation vs. Asset-Class-Based Allocation" is available in its entirety here.