Is Inflation a Driver of Value Stock Performance? Not So Much ...

By: John Alberg and Michael Seckler

Recently, we have heard several people state, almost reflexively, that value investing underperforms growth investing in periods of low inflation and that value outperforms in periods of high inflation. 

Figure 1

One source of this view is from a thoughtful piece by Ben Carlson, in which he plots inflation rates vs. the outperformance of value stocks (1). Ben’s chart is an update of an analysis performed by William Bernstein during the last period (the late 1990s) when growth stocks outperformed value stocks. The chart in Figure 1 is our reproduction of their analysis (2).

This perspective caught our attention in part because of questions we received following our last data post. There, we discussed the cycles of value vs. growth outperformance, and some of our investors were interested in discussing what caused them. Was it inflation? Or something else...?

As we reflected on the chart in Figure 1, we noted the limited number of data points as a consequence of the decade-long lens and that the apparent relationship hinged on a few outliers. Moreover, in the context of “normal inflation” between 0% and 5%, it seemed that there has been a very wide range of outcomes with regard to value vs. growth investments.

Figure 2

Clearly, if we looked at shorter time periods, we would have more data points. So, we asked the question: how does the relationship between inflation and value investing change as an individual’s time horizon grows from a short period (1 year) to a long period (10 years)? The results are plotted in Figure 2.

Figure 3

From these results, we can see that as one’s time horizon increases, we go from no relationship between inflation and value outperformance to a relationship that has a correlation of 0.7 – a number that is considered a moderate but not strong correlation (3). And, again, this number is based on a tenth of the number of data points that underlie the one year correlation. 

Given this, it seems important to ask: can you use the correlation that emerges over longer time horizons to make good investment decisions? That is, if inflation is high (low) for some reasonable period of time, can we use that information to say that it is likely to be a good (bad) time to invest in value stocks going forward? To untangle this question, we calculated the lagged correlation between inflation and value for increasing time horizons. For example, we split the last 90 years into sequential two-year periods and calculated the correlation between the inflation in the first half of each period and value’s performance in the second half of each period. Then we did this for 4-, 6-, 8-, and 10-year periods. The results are plotted in Figure 3.

As you can see, when we look at the data in this way, there is no correlation between current period inflation and how value performs in the subsequent period. Finally, Figure 4 shows the lagged relationship for all nine 10-year periods from 1926 to 2015.

Figure 4

Our view is that the primary driver of why value investing works over time is not the level inflation. Rather, we believe it works because investors are human and, as they search for a signal in the noise that surrounds the stock market, they often attribute too much meaning and overreact to information that proves to have little to do with the long-term value of their investments. This frenetic response to external developments sometimes causes prices to fluctuate widely around companies’ fundamental values and presents opportunities to buy good companies at favorable prices. Indeed, as we have seen during the past 15 years, investors can become both excessively optimistic and pessimistic at similar levels of inflation. 


Would you like to reproduce these results in the R programming language? Click here to get the code that generates the charts. Or clone the github repo to get the code from many of our posts.

Historical results represented herein are for illustrative purposes only and are not based on actual performance results. The hypothetical portfolio and the associated returns do not reflect the effect of transaction costs, bid/ask spreads, slippage, or management fees. Historical results are not indicative of future performance.

1) Kenneth French’s website archives and maintains many time series of returns related to fundamental investing. One such monthly time series splits the universe of equities into three groups representing the top 30%, middle 40%, and bottom 30% of the universe sorted by each stock’s book-equity-to-market-value ratio. The 30% with the highest ratios are called the High-Book-to-Market (HBM) group or VALUE stocks. And the 30% with the lowest ratios are called the Low-Book-to-Market (LBM) group or GROWTH stocks. You can think of the difference of these two time series (HBM minus LBM) as a hypothetical fund that goes long HBM stocks and short LBM stocks. Therefore, the returns of such a hypothetical fund will be positive when value stocks outperform growth stocks and negative when growth stocks outperform value stocks. This data series begin in June 1926 and ends August 2015, therefore 1926 and 2015 are partial years in all analysis.

2) To calculate the annualized change in CPI we started with the monthly CPI data series provided by Robert Shiller on his website. For every year we calculated the annual percentage change. Since the value factor series has partial years for 1926 and 2015, we used only partial years for 1926 and 2015 in the CPI series as well (see Footnote 1). Then, for each decade, we calculated that decade's annualized change in CPI by compounding the annual change in each year. The 1920s and 2010s are both partial decades. For the value factor series in the chart, we use the data provided by Kenneth French's website as described in Footnote 1. The value factor returns are compound annualized returns.

3) The correlation coefficient is a measure of the linear correlation between two variables. The value of the coefficient varies between -1 and +1, where a value of +1 indicates a perfect positive correlation, 0 indicates no correlation, and -1 indicates a perfect inverse correlation. For more detail, see the Wikipedia entry on Pearson product-moment correlation coefficient.