PENSION FIDUCIARIES DON’T TALK MUCH ABOUT THEIR investment beliefs. So one is often left having to infer investment beliefs from investment actions.
For example, if fiduciaries maintain or raise their target weights to equities during an extended bull market, what does that imply about their investment beliefs? Or, if active managers choose to stay close to their benchmarks, what does that say about the managers’ investment beliefs?
I had recent occasion to articulate my own investment beliefs. This exercise took me back to a 1972 Financial Analysts Journal article where I posed this question: what should be the relationship between how we manage money and how good we think we are at making futurereturn predictions?
Logically, the better people think we are at making such predictions, the more justified they would be in holding concentrated portfolios, and engaging in market- timing activity. This raises an interesting question: what are realistic assumptions about the quality of market and individual securities return forecasts?
In the 1970s, I coined a term representing a forecasting- ability metric that has become part of the institutional lexicon: the “information coefficient” or IC. Specifically, ICs measured the correlation between forecasts and outcomes. So ICs of zero would be consistent with the Efficient Markets Hypothesis that all information is already included in securities prices.
That is not what I found. Instead, I discovered that most sets of forecasts monitored during the 1970s had low positive predictive ability(IC=0.15), peaking about six months after the initial forecasts. With large universes and multiple forecasts, such low-grade predictions were capable of generating juicy portfolio “information ratios”(net excess return per unit of excess risk)equal to 1.0.
What about the accuracy of stock and bond market predictions? If six-month market predictions have no greater accuracy than what research found to be the case for individual securities predictions, then the market predictions have no more potential value than individual securities return predictions.
However, how would this conclusion change if stock and bond market predictions had far greater accuracy over long-horizon periods such as 10 years? The ‘investment beliefs’ question is this: are long-horizon capital markets returns more predictable than their short-horizon counterparts? Logic tells us that through the seven different investment eras of the 20th century, there is a strong association between prior return expectations and subsequent return realizations.
Researchers Rob Arnott and Peter Bernstein used a simple model to predict the long-horizon equity risk premium(ERP)in the U.S. from the early 1800s to the late 1900s. The IC of these predictions with subsequent 10-year outcomes was an astounding 0.70. Long-horizon stock and bond market returns do seem to be highly predictive.
How should beliefs that short-horizon security returns are modestly predictive, and that long-horizon capital markets returns are highly predictive, impact investment policy decisions?
Keith Ambachtsheer is president of KPA Advisory Services Ltd., a strategic advisor to major pension plans around the world, based in Toronto. firstname.lastname@example.org