Perils of past performance
GIVEN THE SIGNIFICANCE of underperformance of active managers between 2007 and mid-2008 it would hardly be surprising if clients were questioning whether managers are worth their fees. A few may be ready to ditch the pilot. But are clients in danger of jumping to hasty conclusions? Given the complexities of assessing managers’ performance sacking them may be satisfying but ultimately self-defeating. Academic studies suggest even experienced investors may not be very good at manager assessment.
Two American professors ‒ Amit Goyal and Sunil Wahal ‒ examined the selection and termination of managers for 3 700 US institutional funds over a 10-year period. They found investors’ tendency was to end the contracts of investment managers who had underperformed and take on managers who outperformed.
If, however, investors had stayed with their fired managers, outperformance would have been larger than that delivered by the newly hired managers. And that’s before taking into account the costs of changing managers, which can be hefty.
So how much weight really can be put on the short-term ‒ even medium-term ‒ performance of professional investors? The evidence suggests caution is needed. This paper looks at some of the issues involved in interpreting the past performance of active equity managers and suggests it’s essential to accompany quantitative analysis with qualitative assessment. The past as a poor guide to the future We looked at perfor- mance statistics for a large number of active managers recorded on the database of consulting firm Mercer. We tracked the performance of the best-performing 25% for every rolling three-year period between December 1985 and December 2005. For those managing US shares, only 42% of the previous top quartile turned in a performance above the median performance in the following three years. Managers of non-US shares did slightly better, with 47% performing above the median. But the key point is in neither case did the managers’ three-year track records provide a reliable guide to their future results.
Analyzing the problem
So given those limitations, how should you think about a manager’s recent performance?
First, it’s critical to remember a performance target represents only an average expected return. For example, take a manager claiming to offer a 3%/year return over rolling three-year periods. That means half the time the performance will range above that return and half the time it will range below. While, over time, the average return should centre on 3%, it can mean that over shorter periods performances can deviate sharply from that mean and still be in line with our 3%/year long-term returns.
A manager’s returns can’t be assessed without considering how much risk he’s taking. Although there are many ways of quantifying risk, perhaps the most relevant is tracking error: a measure of how far a portfolio is deviating from the chosen benchmark, whether due to stock selection, industry weights or other factors.
Tracking error measures the volatility of a portfolio’s returns relative to its benchmark: in other words, it measures the consistency a portfolio hits its return targets. In statistical terms, that’s the “standard deviation” of a portfolio’s relative returns. It’s normally used to measure how much risk a manager is currently taking. It can also be used to show how volatile the portfolio’s relative returns have actually been in the past.
The way the maths works means a tracking error is expected to encompass a little more than 66% of outcomes. So a manager with a 5% expected tracking error who aims to generate 2% more than the benchmark should expect to produce a performance of between +7% and -3% in roughly two years out of three. For most of the rest of the time statistical theory suggests the outcomes should be covered by returns within two standard deviations from the target, from +12% to -8%. But in one in 20 years it would be perfectly normal to experience returns beyond those limits. (Seegraph).
Put another way, in any one year of a rolling three-year performance period, statistical theory suggests even a skilled manager with a target premium of 2% and an expected tracking error of 5% has about a 7% chance of producing a performance worse than -8%, relative to the benchmark.
In reality, managers’ results only approximate that ideal picture, with extreme outperformance and underperformance happening more often than a statistical normal distribution would suggest.
The odds of any one manager beating the market consistently through luck may seem remote. Nonetheless, given the number of managers and strategies that have existed the odds of one or more