Subscribe
Featured Image

How Much Alpha Can Be Derived from a Mean Reversion Strategy?

Companies and Markets

By Robert Zielinski  |  August 16, 2021

In March 2020, the longest bull market in financial history (which began in 2009) was briefly interrupted by the fastest market correction qualifying to be a bear market. In just 23 trading days, markets fell by more than 30% as the COVID-19 pandemic triggered a global lockdown. Global indices quickly met and then fell below their 200-week averages—the S&P 500 dropped 13% below the average in question.

The end of the last two speculative bubbles, the dot-com bubble in 2000 and the housing/credit market bubble in 2008, saw the market enter long-term secular bear markets characterized by steep dips in the S&P 500 below its 200-week moving averages. In October 2002, the index fell 37% below the average and in March 2009, it was 48% below. It remains to be seen whether the short dip below the average last spring qualifies as a full reset in today’s quantitative easing (QE) flooded markets or if we are headed for a secular bear market in the not-too-distant future. If the latter is the case, is it possible to profit from a change in direction?

Mean Reversion and the Winner-Loser Effect

As of July 2021, the S&P 500 was again 40% above its 200-week moving average, a simple measure which can indicate whether an asset is overvalued. Compared to the end of the dot-com bubble in March 2000, there is not much room left for the S&P 500—a mere five percentage points. On the other hand, the Nasdaq still has room to grow to match the record set 21 years ago when the index was 150% over its 200-week moving average (currently about 50%). Looking at cryptocurrencies such as bitcoin, the price was 1,500% above the 200-week average in December 2017 and 470% above the average in April 2021.

sp500-200-week-moving-average

For more than three decades, research was focused on evaluating the predictability of future security price returns based on historical price data. In 1985, Werner de Bondt and Richard Thaler identified a pattern called mean reversion, whereby securities and indices return to their longer-term mean values.

De Bondt and Thaler allocated the constituents of the NYSE into winner and loser portfolios and analyzed the performance of the securities between 1926 and 1982. Their research showed that the portfolio consisting of the previously underperforming securities achieved an excess return of 25% over a period of 36 months compared to the portfolio with previously outperforming securities.

This anomaly, known as the winner-loser effect, is used in investor behavior research within the field of behavioral finance. It can be explained through the availability bias whereby investors are inclined to choose the bellwethers with impressive returns as seen in the media and recalled from short-term memory. In addition, they tend to be too pessimistic about underperforming securities and too optimistic about outperforming ones. They overreact to information, which causes the securities to move away from their fundamental valuations as they decide to buy or sell based on the perceived value of a security.

However, over time, the exaggerations are slowly reduced. Accordingly, the weak form of the efficient market hypothesis in which excess returns would not be possible through knowledge of historic prices, cannot be confirmed. Consequently, one may be able to generate excess returns with historically underperforming securities compared to historically outperforming ones as they return to their long-term means.

How successful would a mean-reversion investment strategy be under current conditions? Is there a difference in performance based on the rebalancing period of the portfolios? What about volatility and potential drawdowns?

Building Loser and Winner Portfolios

To answer these questions, we have divided the constituents of the S&P 500 into outperforming and underperforming portfolios. The winning portfolio contains the 50 securities with the highest one-, two-, three-, and five-year returns, whereas the loser portfolio contains the securities that have performed worst over the same periods. The securities in the two portfolios were held over the same periods and then rebalanced using the same criteria as above.

The table below shows the 15 largest securities and their returns for the prior 12 months for their respective periods. The security selection is based on a three-year rebalancing strategy.

top-15-securities-for-loser-winner-portfolios

We then turned to a sector analysis, selecting March 2000 as the end of the dot.com bubble for comparison with the current environment. Whereas in March 2000, securities from the consumer staples (including Coca-Cola, Safeway, and Philip Morris) and industrials sector (including Lockheed Martin, US Airways, and Waste Management) represented 36% of portfolios holdings (and thus resulted in an overweighting of 12% vs. 9% compared to the S&P 500), currently the energy sector has the largest weight at 40%, resulting in an active sector weight of 37% (including Chevron, Exxon Mobil, and Conoco Phillips). This is not surprising as oil-producing companies were among the hardest hit in 2020 during the initial months of pandemic-induced lockdowns and subsequent economic downturn.

sector-representation-loser-portfolio

Looking at the winner portfolios, the sector representation in March 2000 vs. June 2021 is notably different. Whereas 70% of portfolio holdings originated from the technology sector, resulting in an overweighting of 40% versus the S&P 500 in 2000, today the same 70% share includes four different sectors. This illustrates that the current rally is based on the broader participation of different sectors instead of the predominant technology companies in 2000.

sector-representation-winner-portfolio

Analyzing the Results

Our results show that a mean-reversion strategy could yield positive results. As such, we could even argue that rebalancing every two or three years is as good as it can be as the annual universe return of formerly underperforming securities is between 13.56% and 14.53% (see Figure 5, column B, lines 9-18).

However, the point of differentiation comes when we look at the associated risk needed to achieve these respective portfolio returns. The portfolios with a two-year rebalancing come among the highest annual portfolio returns within the 14% mark. The annualized volatility in column H shows that the risk associated with the slightly higher returns than the portfolios with a three-year rebalancing is correspondingly higher as well.

The higher the portfolio return, the more risk will be accepted. This becomes evident when we look at the Sharpe ratio, transforming the returns into risk-adjusted returns. With a value of 0.46 for the selected portfolio in line 12, it is among the highest ratios for all portfolios, even though the standard deviation is lower than for the portfolios with a two-year rebalancing.

Based on this, the portfolio in line 12 generates a maximum annual excess return vs. the winner portfolio of 3.73%. The formerly outperforming securities now underperform the winner portfolio by -0.77%, while outperforming against the benchmark (11.03%) as well as the competing loser portfolio (2.96%).

Looking at very long- and very short-term rebalancing of five- and one-year, the portfolios hardly generate excess returns worth pursuing (lines 5, 6, 21-24). In fact, the loser portfolios rebalanced every year keep underperforming the winner portfolios, especially if the selection of securities is based on a short-term performance horizon such as one or two years (line 23 and 24). Here a short-term long or short momentum strategy is likely the best choice.

overview-risk-return-profile-of-loser-and-winner-portfolios-vs-benchmark

Let´s have a closer look at the strategy in line 12 that rebalances the winner and loser portfolio every three years based on the last 12 months’ stock returns. The first 50 securities with the lowest 12-month returns are added to the loser portfolio and correspondingly, the first 50 securities with the highest 12 months are added to the winner portfolio. Looking at the cumulative excess return vs. the benchmark over 30 years, previously underperforming securities contribute returns of up to 3,000%, whereas the outperforming securities lose about 500% in the same period compared to the benchmark.

excess-return-of-loser-vs-winner-portfolio

Figure 7 illustrates this as well—focusing on constantly well-performing securities in the long run would leave you behind the benchmark.

cumulative-returns-loser-winner-portfolios-benchmark

The reason for the underperformance is exactly the phenomenon we are looking at—long term reversion to the mean where outperforming securities experience strong drawdowns during market corrections (see Figure 8). The winner portfolio experiences two strong drawdowns (one after the burst of the dot-com bubble in 2000 and the second with the housing/credit crisis in 2008).

maximum-drawdown-loser-and-winner-portfolios

Loser portfolios experience notable drawdowns as well (see 2008 and the COVID-pandemic-induced market sell-off in 2020), yet they tend to be recouped faster and can be less intense as was the case after the dot.com bubble. Does this provide us a clue for the next downturn?

Máté Facsar also contributed to this article.

The information contained in this article is not investment advice. FactSet does not endorse or recommend any investments and assumes no liability for any consequence relating directly or indirectly to any action or inaction taken based on the information contained in this article.

factset-qre-ondemand-webcast

Robert Zielinski

VP, Principal Sales Specialist, EMEA Analytics

Mr. Robert Zielinski is VP, Principal Sales Specialist for EMEA Analytics at FactSet. In this role, he focuses on providing performance, risk, reporting, and quantitative solutions to clients across equity, fixed-income, and multi-asset-class strategies across Austria, Germany, and Switzerland (DACH). Beginning as a consultant in 2010, he spent three years working with some of FactSet’s largest buy-side clients in Germany and Switzerland and later focused on performance, quant, and risk workflows. Mr. Zielinski earned a master’s degree in business administration and finance from the University of Augsburg.

Comments