US large cap equity valuations have moved back to their most extreme levels on record barring the 1990-2000 and the 2021 peaks. Based on an average of the most reliable valuation measures going back to 1991, the S&P500 trades at multiples consistent with 10-year real returns of around zero. With real annual growth in fundamentals likely to be at least a percentage point slower than they have been historically due to slower real GDP growth, investors should expect real total returns of -2.8% to -6.4%.
The -2.8% figure is based on John Hussman’s method of looking at the historical correlation between a particular valuation metric and subsequent returns. I have used the price to book ratio based on data since 1991 as it has shown the strongest correlation. The -6.4% figure is based on the decline required to restore historically normal future returns. In this case I have used the dividend yield and assumed a 5.0% required rate of return of 5.0% and real dividend growth of 1.0%.
Historical Valuation-Return Correlation (Hussman Methodology): -2.8%
The chart below shows the strength of the correlation between various valuation metrics and subsequent S&P500 returns over different time periods. The first thing to note is that equity index valuations have very little impact on short term returns but are by far the most important factor determining long term returns. As the time horizon rises to around 4-5 years some of the more reliable valuation metrics are able to explain almost half of the S&P500’s total real subsequent returns. At the 10-year horizon the r-squared rises to around 0.9.
The second thing to note is that there is a large degree of variability in the predictive power of different valuation metrics. Although the price/earnings ratio is the most widely used, it is among the least reliable indicators in terms of its ability to predict returns. The reason is that earnings are highly sensitive to the business cycle. The market has looked artificially cheap on a PE basis in the past due to unsustainably high profits during boom times and looked artificially expensive on a PE basis when profits have been temporarily depressed during recessions. The price to forward earnings ratio does a much better job of predicting returns but the cyclically adjusted PE ratio has a closer correlation as it smooths out these earnings over 10 years.
The table below shows the correlation between valuations and 10-year subsequent total real returns for the main valuation metrics, along with 10-year implied future returns.
|Implied Forward Annual Real Returns
|R-Squared Log Scale
|Price / Earnings
|Price / Dividend
|Price / Forward Earnings
|EV / EBITDA
|Price / Book Value
|Price / Sales
Real return estimates range from 3.1% in the case of the PE ratio to -3.2% in the case of the PS ratio. The PE ratio’s weak historical reliability suggests caution is warranted, and when we smooth out the impact of profit margins using the cyclically adjusted PE ratio expected returns fall to 1.0% and the correlation significantly improves. The price to book value ratio has the closest correlation on a 10-year horizon, with an r-squared of 0.9, and the current figure implies annual real total returns of -2.8%.
The price to book value estimate seems like the most reliable indictor to use not only because it has the strongest correlation but also because it is the most stable indictor. The 3.1% annual real return figure implied by the PE ratio implicitly assumes that profits continue to grow faster than book values over the next decade even with returns on equity at record highs, which is extremely unlikely.
Mean Reversion In Required Rate Of Return: -6.4%
The second methodology seeks to calculate the returns that would result over the next decade if valuations were to mean revert to the levels that implied historical average subsequent returns of 6.5%. For this, we need to take the dividend yield and estimate its likely future growth rate in dividends per share. Over the long term dividends tend to grow at the pace of earnings, which grow in line with sales, book values, and GDP. This methodology therefore factors in the ongoing slowdown in trend real GDP growth into returns.
My core view is that real GDP growth will average around 1% over the next decade in line with the trend growth rates seen in the rest of the developed world over the past two decades. To see why, we can break down real GDP growth into three main components; productivity growth which is captured by output per worker, growth in the labour force reflecting population growth and net immigration, and changes in the unemployment rate. The chart below shows the 10-year moving average of the three components. The 2.3% annual growth rate enjoyed over the past decade can be broken down roughly into 0.8% from growth in the labour force, 0.5% from a decline in the unemployment rate, and 1% growth in output per worker.
Going forward, the CBO projects the US total population to grow by around 0.4% annually over the next decade while the Conference Board projects the working age population to grow by around 0.2% annually, driven entirely by net immigration. This means that even if the downtrend in productivity growth is halted and remains at 1%, it would only take a move in the unemployment rate back to around 5% for real GDP growth to average below 1%.
Assuming 1.0% real GDP and dividend growth, the dividend yield would have to rise to 5.5% for returns to meet the required rate of 6.5%. With the dividend yield currently at just 1.4%, this means that if this rise were to take place gradually over the 10 year forecast period it would result in real total returns of -8.5% annually as shown below.
(1+g)(Original Yield/Terminal Yield)1/N – 1 + (Original + Terminal)/2
(1.01)(.014/.055)1/10 – 1 + (0.014 + 0.055)/2 = -0.085 = -8.5%
The -8.5% return figure above assumes that investors will ultimately value stocks to return 6.5% annually in the future even as growth is assumed to be slower. Slower growth in the future means that investors may be willing to accept lower future returns, which may prevent a full mean reversion in valuations. Assuming the required rate of return falls to 5.0% due to 1.5pp slower real GDP growth relative to the past, the return figures rises slightly to -6.4%.
(1.01)(.014/.04)1/10 – 1 + (0.014 + 0.04)/2 = -0.064 = -6.4%
Taking an average of the two methods gives an expected annual total real return forecast of -4.6%. This is broadly in line with the returns observed from 1999 to 2009 and 1964 to 1982.
Rate Cuts Have Never Prevented Valuation Mean Reversion
There is a widespread belief that if interest rates fall then the required rate of return on stocks would also fall, keeping valuations from mean reverting lower. However, even if real bond yields were to fall back to zero it would not necessarily prevent a large decline in stock valuations in the event of a reversal in the current optimism surrounding market returns. The equity risk premium has moved independently of real bond yields throughout history. Even amid deeply negative real yields stocks have often traded at levels implying strong long-term returns. As we are seeing right now, and as saw at the 2000 peak and the 2007 peak, high equity valuations can exist alongside high yields for a long period of time, but eventually economic reality sets in and investors begin to require higher equity returns. This usually occurs in a self fulfilling manner as such extreme optimism means even small losses cause investors to seek the safety of cash.
Over the long term a key factor driving swings in equity valuations has been economic uncertainty related to inflation. Falling valuations have tended to occur when inflation has deviated significantly from its long term average, either to the downside like during the Great Depression or to the upside like in the 1970s, valuations have fallen significantly. This makes current market valuations particularly troubling. As we saw in 1929 and the late-1990s, valuations are not only extreme but are no longer justified by a stable macro backdrop.
The combination of extreme valuations and slowing trend real GDP growth suggests real S&P500 returns are likely to be deeply negative over the next decade, with my baseline assumption sitting at -4.6%. This is broadly in line with the returns observed from 1999 to 2009 and 1964 to 1982 which occurred despite falling real bond yields. Macroeconomic uncertainty as reflected in inflation volatility is at levels that have triggered downside moves in valuations in the past.