By almost any metric imaginable, U.S. equities are “overvalued.”

But by how much is a topic of constant debate, as it is dependent on a) the valuation metric used (ex: P/E, P/B, P/FCF, etc.), b) the length of historical data over which you are comparing to (ex: average P/E over last 10 years, 20 years, 100 years, etc.), and whether or not you are “smoothing” fundamentals or simply taking the most recent data point (ex: today’s earnings or earnings averaged over the past 10 years, etc.).

One of the more popular valuation metrics used in recent years is the CAPE (Cyclically Adjusted Price to Earnings) ratio, also known as the Shiller PE or PE 10. Unlike a standard P/E Ratio which uses earnings over the prior year, the CAPE smooths earnings over a 10-year period and adjusts for inflation. This is said to mitigate some of the inherent volatility in earnings, giving a more accurate depiction of value.

At over 31 today, the CAPE ratio using 10 years of earnings is higher than any period in history with the exception of the late 1990s to early 2000s (aka the “tech bubble” or “dot-com bubble”).

Data Sources for all charts/tables herein: Robert Shiller, Bloomberg, Pension Partners.

A common critique of the CAPE ratio today is that it still includes earnings from the financial crisis in 2008-2009, making the denominator (average earnings) lower than it otherwise should be. Excluding those earnings, it is often said, would give a more accurate reading and make stocks look relatively cheap. How accurate is that assertion? Let’s take a look…

Going back to 1900, we can calculate the average CAPE ratio for the S&P 500 using various time periods for smoothing earnings (1 year to 10 years).

As you’ll notice in the table below, the 10-year CAPE (hereinafter “CAPE10”) is 86% above its historical average. By comparison, the CAPE1 based on the past year of earnings is 59% above its historical average.

Which is more a more accurate depiction of value? The answer depends on your bias. If you’re leaning bearish, you’ll argue the CAPE10 and if you’re leaning bullish the CAPE1 (or more likely a different ratio entirely, one using high forward earnings estimates to drive the P/E ratio lower).

Irrespective of the CAPE time period used, though, it is difficult to argue that stocks are cheap. You can observe that in the chart below, with the CAPE1 through CAPE10 in the 93-96 percentiles of historical valuations.

What have top decile valuations meant for stocks historically?

Below average forward returns with a lower probability of a positive return over the subsequent 1-10 years.

This is true irrespective of the CAPE time period used, though the longer CAPE ratios seem to have a more inverse correlation with subsequent S&P 500 returns.

Why is that the case? One reason is that during the earnings collapse of 2008, P/E ratios using 1-year of earnings skyrocketed, making stocks appear expensive. In the chart below, you’ll immediately notice the sharp spike in the CAPE1 during this period, to its highest level ever. Meanwhile, the CAPE10 was relatively unaffected and still showed a cheap market as it was averaging earnings over a 10-year period.

Following 2008, equities rallied, skewing the data points for CAPE1 to show strong returns from their most overvalued levels in history. In contrast, the strong returns for CAPE10 came from relatively undervalued starting levels. The 2008-09 example exposed one of the critical weaknesses of using only one year of earnings data: that it can make stocks appear to be extremely expensive when they are actually cheap. The same can be said in the opposite scenario, where high earnings and high profit margins at the end of a cycle can portray stocks to be cheaper than they actually are.

Does that mean anything other than 10 years of earnings is insufficient? Not at all. There is no holy grail valuation indicator and mining the data for a time period that worked best in the past is not going to help you predict the future.

The weight of the evidence today suggests that stocks are overvalued irrespective of the number of years over which you are smoothing earnings. Historically, that has meant below average forward returns with lower odds of a positive return.

But averages are just that, hiding the wide range of possibilities that is a defining feature of markets.

From top decile valuations in the past, we’ve seen stocks return as high as 8.4% annualized over the subsequent 10 years and as low as -8.3% (average return of 2.8%). What will the next 10 years bring? Today’s high valuation levels may tilt the odds, but the wide range of possible outcomes remains.