Robert Shiller’s cyclically adjusted price-to-earnings (CAPE) ratio is the most-cited predictor of long-term equity returns. But new research shows that the “Buffett” indicator does a good job of forecasting, and both ratios predict subdued, long-term returns for stocks.
The ratio of market capitalization of equities to gross domestic product (MVE/GDP) is a long-term valuation indicator that has become popular in recent years, thanks to Warren Buffett. In 2001, he remarked in a Fortune Magazine interview that the ratio of market capitalization to economic output “is probably the best single measure of where valuations stand at any given moment.” The logic behind the MVE/GDP ratio (which became known as “the Buffett Indicator”) providing information as to future returns is straightforward: When the price of equities – which are a claim on future profits – goes up (down) without a commensurate increase in economic output, the forward earnings yield decreases, lowering (raising) the expected return.
Q1 2022 hedge fund letters, conferences and more
Laurens Swinkels and Thomas Umlauft, authors of the March 2022 study “The Buffett Indicator: International Evidence,” investigated the return-predictive characteristics of the market value of equity-to-GDP for a data set comprising 14 developed market countries over the period 1973-2019. They also compared the Buffett Indicator to other well-known stock market valuation signals and phenomena, such as the CAPE ratio and mean reversion.
They began by noting: “The main challenge with using the MVE/GDP ratio as a signal for long-horizon market timing purposes is the lack of a theoretical fundamental level for the ratio to converge to in equilibrium. This is one of the reasons why at each point in time, there exist substantial differences for the MVE/GDP ratio across countries. For example, by year-end 2019, the ratio stood at 148% in the U.S. and only 55% in Germany. The financial structure and the regulatory environment are important factors determining the size of the stock market in a country.”
For that reason, they chose to examine the time-series predictability of the MVE/GDP valuation ratio, not the cross-sectional predictability. They then used the predicted returns from the time-series analysis to determine which countries had the most highly valued stock markets. Thus, the trading strategy invested in countries for which the MVE/GDP ratio was low compared to its own history (which is not the same as investing in countries with low MVE/GDP ratio compared to other countries). Their forecast horizon was 10 years.
Following is a summary of their findings:
- MVE/GDP is highly variable over time.
- MVE/GDP possesses predictive forecast qualities over longer horizons.
- MVE/GDP showed substantially higher explanatory power (r-squared), on average, with regard to nominal forward returns (0.77) than with excess returns (0.19) and real returns (0.67) between 1983 and 2019.
- Mean annual total returns for the equal-weighted country aggregate were 18.3% for MVE/GDP starting values in the bottom quintile, with returns between 12.3% and 21.0%. At the other side of the spectrum, mean returns averaged 3.6% for MVE/GDP ratios in the top quintile, with a minimum annualized return of -0.4% and a maximum annualized return of 8.7% over the subsequent 10-year window. In addition, generally speaking, maximum returns exhibited quintile-based linearity comparable to mean and median returns, and minimum returns became lower as MVE/GDP increased.
- The results were statistically significant at the 5% confidence level.
- In a predictive regression that included MVE/GDP, naive mean reversion in returns and the cyclically adjusted price-earnings (CAPE 10) ratio, only MVE/GDP was statistically significant – the findings are not subsumed by other valuation signals.
- Compared to the CAPE 10, MVE/GDP was substantially more accurate for individual countries and for the pooled country aggregate, almost unexceptionally exhibiting materially higher t-values.
- Over the period 1985-2019, investing in countries with the highest model-predicted returns (top half) would have yielded statistically and economically significant alpha of 1% per annum (10.5% versus 9.5%) over a corresponding buy-and-hold benchmark and would do so with less risk, increasing the Sharpe ratio from 0.49 to 0.57 and reducing the maximum drawdown by 4 percentage points. Remarkably, all portfolio characteristics (i.e., returns, volatility and maximum drawdown) in each year were better for the strategy than for the corresponding benchmark. In fact, in only 129 out of the total 3,550 months, or 3.6%, did the strategy (of all 10 backtested strategies) trail the benchmark.
Read the full article here by Larry Swedroe, Advisor Perspectives.
