The Limits of the CAPE: What Siegel’s Critique Changes

Robert Shiller’s CAPE divides the real level of the S&P 500 by the ten-year average of its real earnings. Jeremy Siegel argues that the accounting earnings behind it have become less comparable over time, pushing the ratio up.

The critique can be tested by swapping index earnings for the corporate profits of the national accounts. The CAPE then falls, without returning to normal.

What the CAPE measures, and what Siegel objects to

The cyclically adjusted price-to-earnings ratio relates the index’s real price to the average of real earnings over the previous ten years. Smoothing over a decade neutralizes recessions and short booms. In the dataset Robert Shiller maintains, the ratio stood at 40.2 in June 2026, the latest month with published earnings, against a 1881-2025 median of 16.6 (Robert Shiller’s data).

In 2016, Jeremy Siegel published “The Shiller CAPE Ratio: A New Look” in the Financial Analysts Journal (vol. 72, no. 3). His argument fits in one sentence: return forecasts drawn from the CAPE may be too pessimistic, because the GAAP earnings it uses have changed in how they are computed, notably with mark-to-market accounting. Substituting the after-tax corporate profits of the US national accounts (NIPA), which are more consistent over time, improves the model’s forecasts and raises expected returns, according to the paper’s abstract.

The critique is about earnings, not prices.

Two crises, two measures of earnings

The difference shows in downturns. Between June 2007 and March 2009, trailing twelve-month earnings per share of the S&P 500 collapsed in Shiller’s data. That is a 92% collapse. US corporate profits after tax measured by the Bureau of Economic Analysis (series CP) fell 48% from their peak in the third quarter of 2006 to their trough in the fourth quarter of 2008. The same gap appeared in 2000-2001: −54% for earnings per share from September 2000 to December 2001, −13% for NIPA profits from late 1999 to late 2001. In real terms and with 1990 set at 100, December earnings per share fell to 87.6 in 2001 and 44.4 in 2008, while national profits held at 142 and 253.

Those collapses owe much to asset write-downs that accounting standards require companies to book through earnings. The US standard FAS 142, issued in June 2001 and applied to fiscal years beginning after December 15, 2001, replaced goodwill amortization with impairment tests: when markets fall, large losses hit the accounts at once. The national accounts measure profits from current production and ignore such losses in value.

A ten-year average keeps the mark of those holes. The 2009 trough weighed on the CAPE’s denominator until 2019, inflating the ratio throughout. That is the mechanism that gives Siegel’s critique its empirical weight.

The CAPE recomputed on national profits

The test replaces earnings per share with NIPA after-tax profits, deflated by the same price index. The two ratios are not comparable in level, since one divides a price by earnings per share and the other by a total of profits. Each is read as a multiple of its own 1957-2025 mean, the period the two series share.

In June 2026, Shiller’s CAPE stood at 1.88 times its mean and the NIPA version at 1.56 times (calculation by Eco3min on Shiller and BEA data). The gap between the two readings widened mostly after the crisis: in 2015, Shiller’s CAPE was 22% above its mean, the NIPA version 19% below. In 2000, at the top of the technology bubble, the two measures agreed, at 1.75 and 1.69 times their means.

The critique softens the warning without removing it. On national profits, US equities remain valued more than half above their historical norm.

What buybacks actually change

Share repurchases became a major payout channel in the United States after 1982, when SEC Rule 10b-18 gave them a framework. Grullon and Michaely showed in “Dividends, Share Repurchases, and the Substitution Hypothesis” (Journal of Finance, 2002) that companies financed them with funds that would otherwise have gone to dividends, and that before 1983 regulatory constraints held them back.

That shift matters for the CAPE through one channel. A company that buys back shares instead of paying dividends grows its earnings per share faster. The faster that growth, the further the ten-year average lags current earnings, and the higher the CAPE looks against the twelve-month P/E. In June 2026, current earnings per share were 59% above their ten-year real average, against 11% on average over 1881-2025: the trailing P/E stood at 25.2 while the CAPE reached 40.2.

Part of the gap between the CAPE and its median therefore reflects faster earnings growth than in the past. Whether that growth lasts is something the ratio cannot say.

What the CAPE still tells

Neither correction turns the CAPE into a timing tool. It describes a valuation level against history, and its link with returns over the following decade is a statistical regularity. The NIPA version has limits of its own: national profits cover all US corporations, listed or not, and include foreign earnings under rules that differ from consolidated accounts.

Two measures are enough to follow the debate: Shiller’s CAPE, updated monthly, and its version recomputed on BEA profits, published quarterly with a lag of about two months. As long as the gap between them stays close to its June 2026 size, Siegel’s critique explains about a third of the CAPE’s excess over its mean. The main episodes of the series are traced in the history of the CAPE since 1881, its place among valuation measures in the comparison between CAPE and P/E, and its foundations in the presentation of Shiller’s ratio.

Last updated — 8 October 2026