Giuseppe Sandro Mela.
Theory & Data.
The Buffett Indicator is the ratio of total US stock market valuation to GDP. Named after Warren Buffett, who called the ratio “the best single measure of where valuations stand at any given moment”. (Buffett later walked back those comments, hesitating to endorse any single measure as either comprehensive or consistent over time, but this ratio remains credited to his name). To calculate the ratio, we need to get data for both metrics: Total Market Value and GDP.
Total Market Value
The most common measurement of the aggregate value of the US stock market is the Wilshire 5000. This is available directly from Wilshire (links to all data sources below), with monthly data starting in 1971, and daily measures beginning in 1980. The Wilshire index was created such that a 1-point increase in the index corresponds to a $1 billion increase in US market cap. Since inception that 1:1 ratio has drifted, and per Wilshire, as of Dec 2013 a 1-point increase in the index corresponded to a $1.15 billion dollar increase. We adjust the data back to inception (and projected going forward) on a straight-line basis to compensate for this drift. For example, the Sep 2020 Wilshire Index of 35,807 corresponds to a total real market cap value of $42.27T USD.
For data prior to 1970 (where Wilshire data is not available) we use the Z.1 Financial Account – Nonfinancial corporate business; corporate equities; liability, Level, published by the Federal Reserve, which provides a quarterly estimate of total market value back to 1945. In order to integrate the datasets, we index the Z.1 data to match up to the 1970 Wilshire starting point.
Combined, these data make our Composite US Stock Market Value data series, shown below. Our estimate of current composite US stock market value is $51.8T.
The Gross Domestic Product (GDP) represents the total production of the US economy. This is measured quarterly by the US Government’s Bureau of Economic Analysis. The GDP is a static measurement of prior economic activity – it does not forecast the future or include any expectation or valuation of future economic activity or economic growth. The GDP is calculated and published quarterly, several months in arrears, such that by the time the data is published it is several months old. In order to provide updated data for the most recent quarter we use the most recent GDPNow estimate published by the Federal Reserve Bank of Atlanta. The GDP data is all nominal and not inflation adjusted. Our estimate of current (annualized) GDP is $22.6T. A historical chart of GDP is shown below.
The Ratio of the Two
Given that the stock market represents primarily expectations of future economic activity, and the GDP is a measure of most recent economic activity, the ratio of these two data series represents expected future growth relative to current performance. This is similar in nature to how we think about the PE ratio of a particular stock. It stands to reason that this ratio would remain relatively stable over time, and increase slowly over time as technology allows for the same labor and capital to be used ever more efficiently.
Predictive Value of the Model
What can the Buffett Indicator tell us about future stock market returns? This is probably the most relevant question investors have when considering valuation models. While it is not possible to predict the future, it is easy enough to look at historical data to see how the market has performed after periods of high and low valuation per the model.
The above chart shows monthly datapoints from 1950 to 2016, mapping the relative value of our Buffett Indicator model (x-axis) against the subsequent 5 year S&P500 returns (y-axis). The colored, dashed vertical lines indicate the same under/overvaluation bands as shown in previous models (i.e., values to the right of the dark red line indicate datapoints that were > 2 standard deviations above the trendline, indicating the market was ‘Strongly Overvalued’).
As an example, the rightmost point on the chart (labeled as Example 2 in the chart) corresponds with March 2000. At this point the US aggregate stock market value was $15.5T, vs a GDP of $10.0T, giving a raw BI ratio of 155%. At the time, this value was 67% (2.2 standard deviations) above the long term BI trend. In March 2000 the S&P500 was at $1,499. Five years later, in March 2005, the S&P00 had fallen to $1,181. This was a 21% nominal decline, and a 30% ‘real’ decline after adjusting for inflation during the 5 years.
Overall, Figure 5 above shows that there is a slight correlation between BI valuation and subsequent S&P00 returns. The highest stock market returns tend to come after periods of undervaluation (left side of the chart). Periods of overvaluation (right side of the chart), particularly at the extreme, tend to be followed by negative S&P00 returns.
A few final comments on this:
– The current BI value is 2.7 standard deviations above the trend, putting us way over on the right side of the chart. (Obviously, we won’t know exactly where on the y-axis we’ll fall for another 5 years…)
– This chart shows correlation, but not causation. The trend here could be pretty quickly undone if the stock market continues its recent strong performance despite being so far above trend line BI values.
-We show the regression line in the chart above to indicate the general trend of the data, but not as a claim of statistical significance. The r-squared value here is low, and time series data like this is vulnerable to autocorrelation, making results seem more significant than they really are.