CAPE and the next ten years

Robert Shiller's cyclically adjusted P/E compares the S&P 500 with ten years of inflation-adjusted earnings. I first met it in his Financial Markets course, so I took his data back to 1881 and checked how well it lined up with the real returns of the decade that followed.

Built with: Python, pandas, statsmodels
Data: Shiller's monthly S&P 500 data, 1871 to 2023
Method: Total-return index, OLS with Newey–West errors

Shiller CAPE at the start of each month against the next ten years of real returns. — Each dot is a month from 1881 to 2013, against the real total return of the following ten years, dividends reinvested.

Cheap markets did better

When CAPE was in its cheapest fifth (below 11), the next ten years returned 10.9% a year after inflation on average. In the dearest fifth (above 21) it was 3.0%.

The fitted line explains about 29% of the variation, which is a lot for one number, but the spread around it is wide. The dearest fifth still ranged from −5.9% to +13.1% a year.

Shiller CAPE from 1881 to 2023. — CAPE since 1881. The dashed line is the median, 16.5.

How I'd use it

At the end of the data, September 2023, CAPE was 30.8, higher than 95% of past months. The fit pointed to about 1.7% a year after inflation for 2023 to 2033, with a range of roughly −7% to +10%.

Three years in, the market has run well ahead of that: the S&P 500 went from about 4,516 to about 7,669 by September 2026. CAPE is a guide to long-run expectations, and it's no use for timing. The overlapping ten-year windows also mean there are only about 13 truly independent decades in the sample.