Campbell and Shiller: Why a High Price Must Mean Bad News Somewhere

Some results in finance are empirical: you find them by digging in data, and they might not hold tomorrow. Others are accounting identities, true by construction, and those are the ones with real teeth, because you cannot argue with them. Campbell and Shiller's 1988 paper is the great example of the second kind.

They took the standard present-value formula, which is nonlinear and awkward, and produced an approximate version that is linear, which means you can actually do statistics with it. That sounds like a technical footnote. It is in fact one of the most consequential moves in modern asset pricing, because it turned a vague debate into a closed logical box with no exits.

The problem: the present value formula was useless for empirical work

Everyone agreed a stock's price should equal the discounted value of its future dividends. The trouble is that this relationship is multiplicative and messy: prices, dividends, and discount rates all interact in a nonlinear way. You could not simply run a regression on it. So the theory sat there, universally accepted and empirically inert, while people fought about market efficiency using far weaker tools.

The key idea, via analogy

Think of a household budget identity: income minus spending equals savings. That is not a theory, it is arithmetic. And precisely because it is arithmetic, it forces a conclusion: if I tell you someone's income went up and their savings did not, then you know their spending went up. There is nowhere else for the money to go. You have learned something real without any theory of human behavior at all.

Campbell and Shiller built the equivalent identity for the stock market. Working with logs and a first-order approximation, they showed that the dividend-price ratio (dividends divided by price, essentially the inverse of how expensive the market is) can be written as a sum of two things:

Today's dividend-price ratio is approximately equal to the expected future returns you will earn, minus the expected future growth in dividends.

Rearranged, it is a trap. Suppose the market is trading at a very high price relative to dividends, meaning a low dividend-price ratio, which is exactly the situation in most bubbles and most bull markets. The identity says at least one of two things must be true:

1. Investors expect dividends to grow unusually fast. The high price is justified by a great future.
2. Investors are willing to accept unusually low future returns. The high price is not about growth at all; it just means you are paying a lot for the same cash flows, so you will earn less on them.

There is no third possibility. If prices are high, the market is either forecasting fast growth or forecasting weak returns. This is not an empirical claim you can argue with. It follows from the definitions.

So Campbell and Shiller went to the data and asked the obvious question: which is it?

The answer, tested across a long span of US history, is uncomfortable for the classic efficient markets story. High prices relative to dividends have historically not been followed by unusually strong dividend growth. Dividend growth is stubbornly hard to predict from valuations. What high prices have been followed by is low subsequent returns.

In other words, when the market gets expensive, it is not because the future got brighter. It is because investors got willing to accept less. And they get repaid accordingly.

Why it mattered

• It made valuation ratios respectable predictors. "Buy when the market is cheap" went from folk wisdom to a claim with a rigorous logical backbone. The predictive power of the dividend yield, and later of Shiller's CAPE, sits on this framework.
• It killed a comfortable illusion. Many investors implicitly believe both that stocks are fairly priced and that they will earn the historical average return regardless of the starting price. Campbell and Shiller prove those beliefs are inconsistent. Your expected return depends on what you paid.
• It relocated the entire debate. Because almost all price variation turns out to be about expected returns rather than about expected cash flows, finance spent the next thirty years asking why expected returns move around so much. Cochrane later called this the central question of the field. That reframing starts here.
• It gave us the log-linear toolkit. The Campbell-Shiller decomposition is now standard equipment, used to split return variation into cash-flow news and discount-rate news, and it appears in an enormous share of modern asset pricing papers.

The honest limitations

• It cannot tell you why expected returns move. The identity proves that a high price implies low future returns. It is completely silent on whether that is because investors rationally require less compensation in good times, or because they are irrationally exuberant. The rational and behavioral camps both cite this paper, happily, for opposite conclusions.
• The predictability is real but weak, slow, and treacherous. Valuation ratios predict returns over horizons of many years, with wide error bands. A famously expensive market can get much more expensive for a decade. As a market-timing tool, this is nearly unusable.
• The regressions have nasty statistical properties. Predictive regressions using persistent variables like the dividend yield suffer from small-sample bias and overlapping-observation problems, and Stambaugh in particular showed the standard t-statistics overstate the evidence. The finding survives, but with less confidence than the raw numbers suggest.
• Dividends have become a poor measure of payouts. With the rise of share buybacks, the dividend-price ratio has drifted structurally lower, which muddies any comparison to historical levels.
• It is an approximation. The linearity comes from a Taylor expansion around an average valuation level. It is very accurate in normal conditions, less so at extremes, which is precisely when you most want it.

The one-line takeaway

Campbell and Shiller showed by pure arithmetic that a high market price must mean either fast dividend growth ahead or low returns ahead, and the historical record says it is almost always the second, so what you pay determines what you get.