P/E in practice: CAPE, bond yields, and the cheap-stock trap
The P/E only looks cheap once you know which earnings it uses. The CAPE ratio smooths ten years of earnings — above 30 only in 1929 and 2000 — and a 10-year Treasury at 5.17% sets a high bar for stock's earnings yield.
The earnings yield has a rival
Flip the P/E over and you get the earnings yield: one divided by P/E. A stock with a P/E of 20 carries an earnings yield of 5% — one dollar of earnings for every $20 invested. That number becomes much more interesting when you put it next to the bond market. On November 13, the 10-year Treasury yield reached 5.17%, its highest in years, after hot producer-price data put a Federal Reserve hike on the table. A 10-year Treasury note was paying more than the earnings yield of a stock trading at 20 times earnings.
The comparison is the old 'Fed model': stocks look dear when their earnings yield falls below the bond yield, and cheap when it sits above it. It is a rough yardstick, not a rule. Bond coupons are fixed; company earnings can grow, which is why investors accept a lower starting yield from stocks. But the yardstick still asks a fair question: if a 10-year Treasury pays 5.17% without needing earnings to grow at all, how much growth must a stock deliver to justify a 5% earnings yield? The higher bond yields go, the harder that question gets.
CAPE: ten years of earnings instead of one
A second problem with the ordinary P/E is that one year of earnings can be a fluke — a boom year, a crisis year, a one-off gain. The cyclically adjusted P/E, or CAPE ratio, fixes that by dividing the price by the average of the last ten years of earnings, adjusted for inflation. The idea goes back to Benjamin Graham and David Dodd, who recommended in their 1934 book Security Analysis that valuation should use an average of earnings over seven or ten years. Yale professor Robert Shiller popularized the measure, which is why it is also called the Shiller P/E.
The history behind it is striking. The ratio stood at 28 in January 1997 — a record at the time — when Shiller and John Campbell warned the Federal Reserve that prices were rising faster than earnings. Investopedia notes the ratio previously exceeded 30 only in 1929 and 2000, the two great bubbles of the past century. Their warning proved prescient in the end: the S&P 500 went on to plunge 60% from October 2007 to March 2009, in the crash that followed the housing bubble.
That does not make the CAPE a timing tool. It is backward-looking by design, it relies on accounting earnings that have changed over the decades, and Jeremy Siegel argued in 2016 that those changes make the CAPE overstate how expensive the market is. What it does is put a single year's earnings in context: a P/E built on one boom year can look cheap for exactly the wrong reason, and a ten-year average is harder to flatter.
The value trap: when the E falls after you buy
The cheapest-looking P/E can be the most dangerous, because the E in the ratio is a snapshot of the past. Take a hypothetical stock trading at $40 with earnings of $5 a share: a P/E of 8, the kind of number that looks like a bargain. Now suppose earnings fall by half, to $2.50 a share, while the price does not move. The P/E is suddenly 16 — the 'cheap' stock was never cheap at all; the market had already priced in the fall that the buyer had not yet seen.
This is the classic value trap, and cyclical companies are where it strikes most often: at the top of a boom their earnings are at their peak and their P/E looks temptingly low, just before earnings fall. The market is not offering a bargain — it is pricing in the earnings decline that the ratio's backward-looking E cannot yet see. A low P/E is a claim about the future of earnings, and the future is the one thing the ratio cannot show.
- CAPE, January 199728 — a record at the time
- Above 30 in historyOnly in 1929 and 2000
- S&P 500, Oct 2007–Mar 2009Fell 60%
- Graham & Dodd's rule (1934)Average earnings over seven to ten years
Three questions before trusting a P/E
So a P/E deserves three questions before it earns any trust. First, which E is it — trailing earnings that really happened, forecast earnings that might not, or a ten-year average like the CAPE that smooths the cycle? Second, what does the earnings yield say next to the bond yield — is the stock offering meaningfully more than the 5.17% a 10-year Treasury paid on November 13? Third, would the P/E still look cheap if earnings fell 30%? In the example above, a 30% fall takes earnings from $5 to $3.50 and lifts the P/E from 8 to 11.4, before the price has even moved.
None of these questions produces a recommendation, and no single ratio ever will. What they do is stop the P/E from being read as a price tag and start it being read as what it is: a bet on earnings that have not happened yet, priced against a bond market that pays more than it has in years. Educational content, not investment advice.
