I was recently asked, “How can you be confident that value investing will work again, when the historical results look so skewed to growth?” It’s a fair question. Looking at the Russell 1000 Growth and Value indices, growth is ahead on a one-, three- five-, 10-, and 30-year basis, and indeed since 1978, when data is first available. Among the most striking results are the one-year returns (37.5% for growth and -5.0% for value) and the 10-year annualized returns (17.3% for growth and 9.9% for value).1
However, these numbers mask how quickly the picture has shifted. As recently as February, value was beating growth since inception. And before the global financial crisis, value was ahead by more than 2% annually over almost three decades since 1978. Perhaps more importantly, there has been a strong cyclicality to the performance of growth and value that makes some “extreme” periods seem a bit more ordinary.
Haven’t we been here before?
In 1999 and 2000, growth was beating value since inception — and on a trailing one-, three-, five-, 10- and 20-year basis. Then, too, investors were asking, “Is value dead?” But value was about to turn the corner; it beat growth by 6.5% per year in the decade ended December 2009.
While this example may be extreme, it seems representative of the longer-term history. Figure 1 compares the trailing 10-year relative performance of value and growth to the forward 10-year relative performance using data back to 1978. A 10-year period of growth outperformance has never been followed by a second one (i.e., no data points in the lower left quadrant.) In fact, following such a period, the worst value has ever done in the subsequent decade is outperform growth by 3.3% a year (the best was 7.7%). Fama-French data, which goes back to 1926 but uses a simpler definition of value, shows almost the same result: The worst value did after a decade of lagging growth was outperform by 3.6% a year.1
The correlation of the trailing and forward 10-year difference in returns shown in Figure 1 is -0.89 — such a strong (inverse) relationship that in a regression, the explanatory power (R-squared) of this single factor (trailing 10-year difference in returns) is sufficient to explain almost 80% of the forward variability of growth versus value. Admittedly, 42 years of data is not an overwhelming amount with which to test decade-long predictions. But I believe it is sufficient to reinforce the cyclicality of the value/growth relationship.
Are we nearing a turn in the cycle?Unfortunately, this type of analysis doesn’t tell us much about the near term. The correlation of past results to future returns is close to zero when looking at a one-year horizon. (Indeed, the correlation between trailing one-year and forward one-year returns is +0.03, meaning that if growth has done well, it may continue to do well.) Many have been waiting for a catalyst that will help value pull ahead, and perhaps the recent positive news on a COVID-19 vaccine will be just that. But even if it’s not, in thinking about long-term policy and the decade ahead, I believe value will outperform, potentially by a healthy margin.
A final thought on the dataIn looking at the correlation analysis, quants will (rightly) point out that 1) the history is quite limited and 2) I am using overlapping data. However, the correlations are still very negative if we attempt to account for that. For example, using annual data (not monthly) generates correlations of between -0.47 and -0.73 (depending on the start/end points). Moreover, non-overlapping analysis (using the Fama-French book-to-price data for a longer history) also yields negative correlations. Thanks to my colleagues Ben Cooper, Danny Cook, and Steven Ye for flagging the statistical challenges.
1Sources: Russell, Datastream, Fama/French data library, Wellington Management. Based on total returns as of September 2020. Past results are not indicative of future results. Investments cannot be made directly into an index.