When Equity Beta Correlations Collapse

Topics: , ,

Published on April 13, 2020

| 6 min read

Richard Yasenchak, CFA, Head of Client Portfolio Management


This blog is the first of a two-part series examining the anomalous dislocation of equity betas we observed in the drawdown during March 2020. Part 1 looks at the sharp decline in the stock beta correlations and Part 2 examines the narrowing of the overall beta spread.

A stock’s beta is a reasonably stable attribute and typically only changes very slowly over time. Low-beta stocks generally remain low beta, while high-beta stocks tend to remain high beta. Last month, this historical relationship temporarily unraveled, regardless of how one measures beta: e.g., month-over-month versus year-over-year, or realized versus predicted (forward-looking) beta.

Month-over-Month View

The historical persistency of beta on a month-over-month basis is easy to understand. The correlation of a stock’s beta at the beginning of a month with its beta at the end of the month tends to be high, since the data used to calculate each month’s beta have a lot in common. For example, if you compute a stock’s beta by using the trailing year’s daily returns, then the betas at the beginning and at the end of the month are based on the same returns for 11 out of 12 months. However, in March 2020, correlations of month-over-month stock betas significantly declined even though over 90% of the data used in the beta computations are the same (Figure 1)!


Figure 1 Correlation of Month-over-Month Stock Betas for the Russell 1000 Index


Figure 1 shows the correlation of stock betas at the end of each month, computed for all stocks in the Russell 1000 Index. You can see that this relationship is stable most of the time, with correlations of month-apart betas generally close to 1.0, and rarely dropping below 0.95.1 As expected, a stock’s beta is reasonably persistent over time.

The picture changed dramatically in March of this year as the correlation between stock betas plummeted from 0.97 to 0.59 in a single month. Many stocks that were previously low beta became high beta, and vice versa.

A rapid change of this magnitude has only occurred once before – in October-November 1987 – when the month-over-month beta correlation dropped from 0.97 to 0.68.

Year-over-Year View

Interestingly, the correlation of stock betas from one year to the next is still typically high, even though there are no common returns used to compute the two betas for each stock. Figure 2 uses the same index to illustrate the same phenomenon as Figure 1, but this time shows the correlation year-over-year with no data overlap. Using this measure of beta, the correlation is currently at an all-time low of around 0.25.


Figure 2 Correlation of Year-over-Year Stock Betas for the Russell 1000 Index


Forward-looking View

So far, we’ve looked at simple calculations of beta using historical data, but third-party risk management tools, such as Barra,2 provide forward-looking estimates of stock betas that are intended by design to be timelier. These estimates are intended to be an improvement, but they still necessarily use backward-looking data and likely rely on some degree of beta stability. Indeed, they too tell the same story about last month: using Barra’s predicted beta, we find that correlations fell from 0.89 to 0.51 between February 29, 2020 and March 31, 2020.

Implications for Portfolios

What does this mean for your portfolio? Investors and consultants will no doubt look closely at how their portfolios responded in this unique environment. This can indeed offer some insights into the benefits of diversification at the overall portfolio level. However, we caution against drawing any firm conclusions based on a short-term anomalous period. We expect markets to normalize.

If you’d like to gain fuller insight into the unusual shift in equity betas last month and the implications for your portfolio, we invite you to download our paper, March Mayhem: Was Your Portfolio Betrayed by Beta?”

1 Source for returns: CRSP. For each month between February 1980 and March 2020, initial beta is computed for each constituent in the Russell 1000 Index at the beginning of that month by regressing 253 days of arithmetic stock returns against the corresponding index returns, ending on the last trading day of the previous month. Final beta is computed for the same population of stocks in the same manner, except the 253 days of returns end on the last day of the month in question. Missing data, and returns greater than 638.9% or less than -86.5% (i.e., logarithmic return > 200% or < -200%), are replaced with index returns. The correlation between initial and final beta for the given month across all constituents is computed and depicted in the Figure. The same methodology is used to compute stock betas in the results shown in Figures 2, 3 and 4.
2 MSCI’s Barra predicted beta is a “fundamental” beta based on a multi-factor model, which regresses historical company returns against the returns of a market index using company-risk and industry-risk factors, re-estimated on a monthly basis, within the regression equation.
The information expressed herein is subject to change based on market and other conditions. The views presented are for general informational purposes only and are not intended as investment advice, as an offer or solicitation of an offer to sell or buy, or as an endorsement, recommendation, or sponsorship of any company, security, advisory service, or fund nor do they purport to address the financial objectives or specific investment needs of any individual reader, investor, or organization. This information should not be used as the sole basis for investment decisions. All content is presented by the date(s) published or indicated only, and may be superseded by subsequent market events or other reasons. Past performance is no guarantee of future results. Investing involves risk, including possible loss of principal and fluctuation of value. Hypothetical performance results presented are for illustrative purposes only. Hypothetical performance is not real and has many inherent limitations. It does not reflect the results or risks associated with actual trading or the actual performance of any portfolio and has been prepared with the benefit of hindsight. Therefore, there is no guarantee that an actual portfolio would have achieved the results shown. In fact, there will be differences between hypothetical and actual results. No investor should assume that future performance will be profitable, or equal to the results shown. Hypothetical results do not reflect the deduction of advisory fees and other expenses incurred in the management of a portfolio.