We have spent a lot of time over many years seeking to employ technology to free up time to focus on “thinking big thoughts” and meeting interesting business people. It’s a continuous process.

This is an interesting piece that 120 percent confirms the obvious- to us -there is simply a limit on how to employ technology to factorize what is at the end of the day moving targets of human achievements and failures. And “machine learning” is simply a process that just tries to delay an inevitable. We don’t try to play a game in which we know we are the undermanned chump. Instead, we seek to focus on a curated set of business/value/people assumptions in which a blend of actual research and judgment may be highly rewarded, preferably within the lifetime of the portfolio manager…or client time horizon!

Original article posted at evidenceinvestor.com.

WHY AND HOW SYSTEMATIC STRATEGIES DECAY

By LARRY SWEDROE

In his 2014 study, Claude Erb asked the question, Has the Stock Market Been Overgrazed? He began by noting that, over time, the market beta, size and value premiums have all declined and were then at lower levels than their historical averages. He then asked, “What if too many investors are demanding too much from a possibly limited supply of opportunities?” Asked another way, are the “trades” too crowded?

Erb explained why he believed this had happened: “Empirical research over the last 50 years has produced much awareness of past asset returns.” He added: “Empirical academic research breeds familiarity with previously successful investment opportunities” and “familiarity breeds investment.”

The 2016 study Does Academic Research Destroy Stock Return Predictability, by David McLean and Jeffrey Pontiff, provided support for Erb’s thesis. The authors found that the average characteristic’s return has a “56% post-publication decay.” They also found that “strategies concentrated in stocks that are more costly to arbitrage have higher expected returns post-publication. Arbitrageurs should pursue trading strategies with the highest after-cost returns, so these results are consistent with the idea that publication attracts sophisticated investors.”

These studies also align with the results of the 2014 paper Have Capital Market Anomalies Attenuated in the Recent Era of High Liquidity and Trading Activity? by Tarun Chordia, Avanidhar Subrahmanyam and Qing Tong. The authors found that as trading has become cheaper and easier, the average excess return of 12 anomalies or factors had approximately halved. They “provide evidence that hedge fund assets under management, short interest and aggregate share turnover have led to the decline in anomaly-based trading strategy profits in recent years.”

All of this raises the question of whether or not the markets have been overgrazed. Declining premiums, at least, raise the suspicion that they have. What is the implication for investors? If a trade or strategy is going to get crowded, you want to be there before it happens because you will benefit from investors driving prices in your favor. However, there is a reason for the adage among investment professionals that you don’t want to be a member of a crowd. When an investment strategy gets “crowded” due to large inflows from investors chasing returns, it is time to exit. Think of the 2007 bubble in residential real estate, the 1990s tech bubble, the “tronics” bubble of the 1960s and all the other bubbles that have occurred.

Latest research
Antoine Falck, Adam Rej and David Thesmar contribute to the factor-based investing literature with their May 2021 study, Why and How Systematic Strategies Decay. Using a sample of 72 published investment strategies, they investigated the determinants of the performance decay of known investment strategies after their publication. They cited the 2016 study by McLean and Pontiff that two likely candidates for explaining decay are arbitrage and overfitting. Falck, Rej and Thesmar explained: “Arbitrage arises because, as soon as a new investment idea is disseminated, arbitrage capital moves in, reducing trading profits. This is a concern for investors trading on known ideas (published research), but much less so for investors using proprietary ideas that only slowly leak out of their organizations. Of much greater concern is overfitting, i.e., the notion that a portion of the risk-adjusted performance only happened by chance because the researcher has tested multiple hypotheses.”

They constructed 11 different variables that were expected to predict post-publication Sharpe ratio decay: four of them related to arbitrage and six pertaining to overfitting (for example, increasing the number of variables and sensitivity to a small subset of stocks). They also included the publication date. Their data sample covered the period January 1963 to December 2018. Following is a summary of their findings:

Out of the 11 variables investigated, six variables were found to predict out-of-sample Sharpe decay. The main predictor of decay was the date of publication.
More recently published factors performed less well out-of-sample. Every year the Sharpe ratio of newly published factors decreased by approximately 5 percentage points—consistent with the idea that recently published signals are more likely to have been mined and that arbitrage capital rushes in faster after publication in recent years.
International results (China, Hong Kong, Korea, Japan, Australia, Continental Europe, the UK and Canada) were similar.
The year of publication alone explained 30 percent of the variance of Sharpe decay across factors. The overfitting variables explained another 15 percent, while the additional contribution of arbitrage-related variables was negligible.
When the traded firms are large, the performance drops were more pronounced, as large companies in general are cheaper to trade and allow for larger trading capacity, thus allowing arbitrage capital to flow more aggressively into anomalies that have enough capacity. The decay was much lower for smaller stocks.
Their findings led Falck, Rej and Thesmar to conclude that both overfitting and arbitrage activity contribute to the decay in premiums and systematic strategies.

Investor takeaways
The important takeaway for investors is that crowding can lead to not only the shrinking of premiums but even their elimination. However, if markets are rational, risk-based premiums cannot disappear, though crowding can cause them to shrink. And even behavioral anomalies can persist because human behaviour tends not to change and there are limits to arbitrage.

It’s also important to note that we should expect that risk-based premiums are time-varying, being dependent on the economic regime. And research, including the study Value Return Predictability Across Asset Classes and Commonalities in Risk Premia published in the March 2021 issue of the Review of Finance, demonstrates that the spread in valuations between the long and the short side of factors provides important information in terms of future returns.