Swedroe: Past Performance Deceives

November 10, 2017

Despite most investors’ belief that past performance matters, a large body of academic research has found little-to-no evidence that more managers than would be randomly expected persistently outperform the market on a risk-adjusted basis.

While this outcome is disheartening to those engaged in manager selection, it shouldn’t be at all surprising. As my co-author Andrew Berkin and I explain in our book, “The Incredible Shrinking Alpha,” there is fierce competition among highly skilled managers that makes markets highly, though not perfectly, efficient. The result of this fierce competition is that there are few “free lunches” to be had—and certainly not enough of them after taking into account the costs of active management.

In addition, as Jonathan Berk pointed out in his paper, “Five Myths of Active Portfolio Management,” rational active managers extract what is referred to as the “economic rent” for their services, given that alpha skill is the scarce resource, while capital is plentiful.

While the common practice of hiring managers with superior recent performance and firing managers with poor recent performance may seem logical, and doing the opposite would seem counterintuitive, the evidence shows past performance is not predictive.

Yet investors still allocate about twice as much capital to active strategies as they do to passive ones. For retail investors, capital flows follow performance over relatively the short period of six to 12 months.

The research also shows that trillions of dollars in pension plan assets are delegated to actively managed funds through a “beauty contest” process that focuses substantially on performance over the slightly longer period of the past several years.

Sometimes past performance is measured against market benchmarks such as the S&P 500. In other cases, a more sophisticated factor analysis ensures performance is measured against appropriate risk-adjusted benchmarks.

Latest Research

Bradford Cornell, Jason Hsu and David Nanigian contribute to the literature on this topic with the study “Does Past Performance Matter in Investment Manager Selection?”, which was published in the Summer 2017 issue of the Journal of Portfolio Management.

Specifically, the authors examined “whether selecting managers based on recent simple outperformance against the stated benchmark (the dominant manager selection heuristic) can lead to subsequent simple outperformance against the benchmark (the desired outcome for most institutional investors).”

To simulate the impact of the popular manager selection heuristic, they compared the performance of hypothetical pension portfolios that follow policies that mandate investing in products based on recent benchmark-adjusted returns.

The authors start with the commonly employed “winner strategy,” defined as follows: At the beginning of each three-year period, investors purchase equal positions in products that rank in the top decile of benchmark-adjusted returns. At the end of three years, monies are reallocated to a new portfolio that is once again equal-weighted among the top-decile performers.

They then compared the investment results of their winner strategy with those of a “median strategy,” whose three-year asset allocation policy is to invest in products that rank between the 45th and 55th%ile of benchmark-adjusted returns.

 

Winners Vs. Losers

Cornell, Hsu and Nanigian also examined the counterintuitive “loser strategy,” which follows the same procedure but invests in products that rank in the bottom decile of benchmark-adjusted returns.

The winner-strategy bucket would generally consist of fund managers that investment consultants would recommend to their pension clients. Managers in the loser-strategy bucket would generally be put on a “watch list” and actively replaced in client portfolios by managers on the recommended list.

Finally, they also compared the investment performance produced by an unorthodox strategy of investing in products that underperformed their benchmarks by more than 1% per year, as well as the even more extreme case of investing in products that underperformed their benchmarks by more than 3% per year.

These portfolios provide insight into the impact of the common manager firing heuristic. Their sample excluded funds that did not have at least $1 billion in AUM and also those that ranked in the top decile of expense ratio (research shows that expensive mutual funds tend to be persistent underperformers because of costs). Their study covers the period 1994 through 2015.

Following is a summary of their findings:

  • The average benchmark-adjusted return for the median strategy beats that of the winner strategy by 1.32 percentage points per year (-1.07% versus -2.39%), and the loser strategy outperforms the median strategy by 0.96 percentage points per year (the loser strategy had a benchmark-adjusted return of -0.11). Thus, the loser strategy outperformed the winner strategy by 2.28 percentage points per year.
  • In addition to benchmark-adjusted returns, the median strategy outperforms the winner strategy across all other performance metrics commonly employed in academic studies, while the loser strategy outperforms the median strategy.
  • The Sharpe ratio of the median strategy is 0.42 versus 0.25 for the winner strategy, while the loser strategy produced a Sharpe ratio of 0.48. Thus, investors would have nearly doubled their mean-variance efficiency by switching from chasing winners to investing in loser funds.
  • The CAPM alpha generated by the median strategy beats that of the winner strategy by a statistically significant 2.76 percentage points per year (-0.85% versus -3.61%) and its Carhart four-factor (market beta, size, value and momentum) alpha beats that of the winner strategy by a statistically significant 1.03 percentage points a year (-2.16% versus -3.19%). The loser strategy managed to do even better, producing a CAPM alpha of just -0.11 and a Carhart four-factor alpha of -0.17, though neither was statistically significant.

Extreme Losers

Cornell, Hsu and Nanigian found very similar results when examining the performance of the extreme loser portfolios.

They write: “At the 3% threshold, the fired funds outperform the kept funds by over 1 percentage point per year based on benchmark-adjusted return, raw return, CAPM alpha, and Carhart four-factor model alpha. The Sharpe ratios indicate that the fired funds also exhibit greater mean-variance efficiency than their counterparts. The results are largely similar when we use a 1% threshold in place of the 3% threshold. Once again, the fired funds outperform the kept funds across all performance metrics.”

For example, they found the funds that underperformed by more than 1% (fired managers) produced four-factor alphas of -0.69%, which compares favorably to the -1.88% alpha of funds that did not underperform by more than 1% (managers who were retained). For funds that underperformed by more than 3%, the four-factor alpha was -0.48% versus -1.64% for those that did not underperform by more than 3%.

As a test of robustness, Cornell, Hsu and Nanigian found similar results using a two-year evaluation period instead of three years. They also found similar results when they eliminated the $1 billion AUM requirement, and when they looked at only institutional share classes (which have lower costs).

They even found the same results when looking at funds by benchmark performance decile (in general, moving from the funds in the best-performing deciles to the worst, performance became worse on both a raw and risk-adjusted basis).

Such outcomes help to explain the often-reported “performance gap”—the finding that, on average, performance-chasing behavior can cause investors to underperform the very funds in which they invest.

 

Cornell, Hsu and Nanigian concluded: “We find that the common selection methodology turns out to be a detriment to performance.”

They added: “The greater benchmark-adjusted return to investing in ‘loser funds’ over ‘winner funds’ is statistically and economically large and is robust to reasonable variations in the evaluation and holding periods, as well as to standard risk adjustments. We also found that the standard practice of firing managers who have recently underperformed actually eliminates those managers that are more likely to outperform in the future.”

These findings are entirely consistent with previous research. Take note of the publication dates of the studies cited below. You’ll see that the evidence on using past performance to select actively managed funds has been there for a long time for all to see. Yet the evidence continues to be ignored by a large majority of investors, both institutional and individual.

Supporting Evidence

Rob Bauer, Rik Frehen, Hurber Lum and Roger Otten, authors of the 2008 study “The Performance of U.S. Pension Plans,” examined the performance of 716 defined benefit plans (over the period 1992 through 2004) and 238 defined contribution plans (over the period 1997 through 2004). They found that returns relative to benchmarks were close to zero. They also found no persistence in performance.

Importantly, the authors also found neither fund size, degree of outsourcing or company stock holdings were factors driving performance. This refutes the claim that large pension plans are handicapped by their size. Smaller plans did no better.

They concluded: “We show striking similarities in net performance patterns over time, which makes skill differences highly unlikely.”

Bauer, Frehen, Lum and Otten also studied the performance of mutual funds, adding to our body of evidence on them. As you should expect, the news for individual investors is even worse.

While pension plans failed to outperform market benchmarks, on a risk-adjusted basis, mutual funds underperformed pension plans by about 2% per year. Pension plans are able to use their size (negotiating power) to minimize costs and reduce the risks of any conflicts of interest between fund managers and investors.

The authors attributed the underperformance to the incremental costs incurred by mutual fund investors.

Counterproductive Activity

The 2008 study “The Selection and Termination of Investment Management Firms by Plan Sponsors” by Amit Goyal and Sunil Wahal provide even further evidence on the inability of plan sponsors to identify investment management firms that will outperform the market after they are hired.

In their study, Goyal and Wahal examined the hiring and firing of investment management firms by plan sponsors (public and corporate pension plans, union pension plans, foundations and endowments). They built a data set of the selection and termination decisions of about 3,400 plan sponsors from 1994 to 2003. The data represented the allocation of over $627 billion in mandates.

Following is a summary of their findings:

 

  • Plan sponsors hire investment managers after large, positive excess returns coming up to three years prior to hiring.
  • Such return-chasing behavior does not deliver positive excess returns thereafter.
  • Post-hiring, excess returns are indistinguishable from zero.
  • Plan sponsors terminate investment managers after underperformance, but the excess returns of these managers after being fired are frequently positive.
  • If plan sponsors had stayed with the fired investment managers, their returns would have been larger than those actually delivered by the newly hired managers.

It is important to note that the above results did not include any of the trading costs that would have accompanied transitioning a portfolio from one manager’s holdings to the holdings preferred by the new manager. The bottom line: All of the activity was counterproductive.

Another study of pension fund managers found the same results. T. Daniel Coggin and Charles Trzcinka, authors of the study “A Panel Study of U.S. Equity Pension Fund Manager Style Performance,” which was published in the Summer 2000 issue of the Journal of Investing examined the performance of 292 pension plans with 12 quarters of data up to the second quarter of 1993.

The following summarizes their findings:

  • It is very difficult to find investment managers who consistently add value relative to appropriate benchmarks.
  • There was no correlation found between relative performance in one period and future periods.
  • There was no evidence the number of managers who beat their benchmarks was greater than that attributable to pure chance.

The authors concluded: “Those who rely solely on historical style alphas to predict future style alphas are likely to be disappointed.”

Summary

There’s an axiom in finance that when the evidence conflicts with the theory, no matter how logical and intuitive it may be, throw out the theory.

Unfortunately, most investors continue to ignore the evidence that makes it clear a policy of hiring recently outperforming managers and firing recently underperforming managers is a losing strategy—one that can even be said to be 180 degrees wrong.

The results you have seen in the research pose a significant challenge for plan sponsors and investors in general who continue to base decisions on beliefs that run counter to the evidence.

The bottom line is that, in general, plan sponsors and other investors are doing the same thing over and over again and expecting a different outcome.

Most seem to never stop and ask the question: If the managers we hired based on their past outperformance have underperformed after being hired, why do we think the new managers we hire to replace them will outperform if we are using the very same criteria that has repeatedly failed? And, if we aren’t doing anything different, why should we expect a different outcome?

I’ve asked these very questions of many plan sponsors and never once received an answer—just blank stares.

The practical implication is that asset owners should change the criteria they use to select managers.

Instead of relying mainly, if not solely, on past performance, they should use criteria such as fund expenses and the fund’s amount of exposure to well-documented factors (e.g., size, value, momentum, profitability and quality) that have been shown to have provided premiums. These premiums should be: persistent, pervasive, robust to various definitions, implementable (they survive transaction costs) and have intuitive explanations for why they should persist.

A set of criteria like that will almost certainly lead investors to avoid actively managed funds and increase their likelihood of superior results.

Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.

Find your next ETF

CLEAR FILTER