At first glance, stock prices seem to move with the fortunes of companies. If prices rise, it’s tempting to assume that fundamentals have improved: companies will earn more, pay more dividends, or grow faster. But a powerful body of academic work suggests something more subtle—and perhaps more unsettling—is at play.

John Cochrane and John Y. Campbell, two giants of asset pricing research, show that the majority of movements in stock prices are not driven by changes in expected cash flows, but by changes in expected returns. In other words, prices rise not because the future looks brighter in a fundamental sense, but because investors are willing to accept lower returns going forward.

This post explores what that means, why it matters, and how it challenges some of the core assumptions of traditional investing.

Either Dividends or Returns Must Be Predictable

Cochrane (1992, 1994, 2011) starts from a fundamental identity in asset pricing: the price of a stock reflects the present value of expected future cash flows, discounted at the expected return.

So if prices fluctuate a lot over time, one of two things must be happening:

Investors expect future dividends (or earnings) to grow significantly.
Investors are changing the discount rate that they apply to those future cash flows—effectively, they expect a different return for holding risky assets.

Empirically, it turns out that dividend (or earnings) growth is barely predictable (Cochrane 1992, 2011), whilst returns are moderately predictable—especially over long time horizons. This leads to a powerful conclusion: most of the variation in stock prices is due to changes in the discount rate—the expected return, not expected future cash flows.

Log-Linear Decomposition

To formalise this, and building on work by Campbell and Shiller (1988), Cochrane uses a log-linear approximation of the present value identity. The standard pricing formula says:

P_t = E_t [ D_{t+1} / (1 + R_{t+1}) + D_{t+2} / (1 + R_{t+2})² + D_{t+3} / (1 + R_{t+3})³ + ... ]

Where:

P_t is the stock price today
D_{t+j} is the dividend (or future cash flow) in year t+j
R_{t+j} is the expected return (discount rate)
E_t indicates expectations formed at time t

This says today’s price is just the discounted sum of future expected dividends.

By taking logs and applying a Taylor approximation, we get:

p_t - d_t ≈ ∑ (ρ^j) [ Δd_{t+j} - r_{t+j} ]

Where:

p_t is the log of the price
d_t is the log of the dividend
Δd_{t+j} is the expected dividend growth rate at horizon j
r_{t+j} is the expected return at horizon j
ρ is a constant between 0 and 1 (typically around 0.96–0.97) used to discount future periods

What This Means

The left-hand side of the equation, p_t - d_t, is the log price-dividend ratio, a common measure of market valuation.

The right-hand side shows that this ratio reflects:

Expected future dividend growth (i.e. a cash-flow story)
Expected future returns (i.e. a discount rate story)

So when stock prices go up relative to dividends, either:

Investors expect stronger future fundamentals (higher cash flows), or
Investors require a lower return going forward (they’re willing to pay more now for the same cash flows)

Cochrane’s research shows that it’s mostly the second case that drives price changes: the discount rate moves, not expectations for future dividends.

What Does ‘Dividend’ Really Mean?

It’s worth noting that when Cochrane says ‘dividends’, he often means something broader—like all future cash flows to shareholders, including earnings, free cash flow, and share buybacks.

Many companies retain earnings to reinvest in growth, so literal dividends understate the true value returned to investors. But the valuation framework still applies: whatever form cash flows take, the logic is the same. If future cash flows don’t explain price volatility, then expected returns (discount rates) must.

Dividends vs Price Movements: Understanding the Valuation Identity

A key puzzle in asset pricing is this: dividends and earnings are relatively stable over time, but stock prices are highly volatile. If fundamentals don’t swing wildly, why do prices?

This question brings us back to the valuation identity, which says:

The value of an asset equals the present value of its expected future cash flows, discounted at the required rate of return.

Formally:

Price = Present Value of Expected Cash Flows / Discount Rate

So if stock prices increase, something on the right-hand side must have changed. If expected cash flows haven’t moved much (as the data suggest), then it must be that the discount rate has fallen. Investors are willing to accept lower returns for holding stocks, pushing prices up.

This explains why metrics like the price-dividend ratio or price-earnings ratio tend to forecast future returns, not dividend growth. When valuations are high, subsequent returns are generally low—because investors have already paid for the future cash flows upfront.

As Cochrane (2011) puts it:

‘If prices move a lot and dividends don’t, then expected returns must be moving. The price-dividend ratio forecasts returns, not dividend growth.’

Long-Term Return Predictability

Campbell and Shiller (1988) were among the first to show that the dividend-price ratio can forecast long-term returns. Cochrane extended and deepened this work, both theoretically and empirically.

This predictability doesn’t mean that you can forecast returns precisely year to year. Rather, it means that valuation ratios contain information about average returns over long periods—like 5 or 10 years.

Cochrane frames this as evidence of time-varying expected returns: investors demand different returns in different environments. When risk aversion is high or uncertainty is elevated, discount rates rise—and prices fall. When sentiment is strong, discount rates fall—and prices rise.

**Figure 1. Dividend Yield vs Subsequent 10-Year Stock Returns**
This chart shows the historical relationship between dividend yields and subsequent 10-year annualised real returns for US stocks, based on data from Nobel Laureate Robert Shiller.
Each dot represents a single year between 1920 and 2013. For each year, we plot:
**X-axis**: The dividend yield at the start of the year (calculated as dividends divided by the S&P Composite Index price).
**Y-axis**: The annualised return over the *next 10 years*, adjusted for inflation.
The **dashed line** is a trend line showing the overall relationship between these two variables.
**What the Chart Tells Us**
The chart hints at a modest relationship between dividend yields and subsequent 10-year returns. The trend line slopes upwards, indicating that—on average—years with higher dividend yields have been followed by somewhat better long-term returns.
When dividend yields were high (for example, above 4%), the crosses tend to appear higher on the chart—indicating stronger real returns over the following decade.
When dividend yields were low (below 2%), future returns were often weaker, though with considerable variation.
Whilst the pattern is far from precise, it supports the general idea that valuation matters: ***when investors pay less for a dollar of dividends (higher dividend yield), their future returns have historically tended to be higher—on average.***
This fits with research by Campbell and Cochrane, which finds that valuation ratios like the dividend yield can contain some predictive power for long-term returns, even if the signal is noisy and best used in a long-horizon context.

Can Investors Use This to Time the Market?

In theory, yes. If high dividend yields imply higher future returns, an investor could shift more into equities when yields are high and reduce exposure when they’re low.

This is the basis of valuation-based asset allocation strategies—sometimes called return forecasting or tactical tilts. But Cochrane urges caution:

1. Predictability Is Statistically Real, but Economically Weak

Valuation ratios forecast returns, but only weakly. The R² values are low: they explain only a small portion of actual return variance. So whilst high valuations tend to lead to lower returns on average, there is a lot of noise amongst the signal.

2. These Strategies Are Slow-Moving and Long-Term

Return predictability is strongest at long time horizons. This is not a call to trade frequently, but to adjust asset allocations gradually based on extreme valuation signals.

3. It’s Not a Free Lunch

If these signals were reliable and easy to exploit, they would be arbitraged away. Their persistence suggests that they either:

Reflect risk (i.e. higher returns compensate for more risk), or
Persist due to market frictions, such as transaction costs, behavioural biases, or institutional constraints.

Cochrane tends to favour the risk-based explanation: markets are efficient relative to the constraints and preferences of real investors.

Campbell’s Broader Contribution

John Y. Campbell’s work with Robert Shiller helped establish that valuation ratios predict returns better than they predict dividend growth. Their research laid the groundwork for modern empirical asset pricing.

In later work, Campbell and Cochrane (1999) introduced the habit formation model of consumption. This model helps explain why expected returns vary over time: investors become more risk-averse in bad economic times and demand higher returns to hold risky assets. This, in turn, pushes prices down—offering a powerful explanation for stock market volatility.

Summary

The price of a stock equals the present value of future cash flows discounted at the expected return.
If stock prices are volatile but dividends are stable, it must be that expected returns (discount rates) are changing.
Cochrane shows that stock returns are predictable over long horizons, and this predictability comes from changes in discount rates, not in expected cash flows.
This challenges the idea that markets are always perfectly efficient in the short run.
Investors could use valuation signals to inform long-term asset allocation decisions—but the signals are noisy, and the edge is modest.

‘When the price-dividend ratio is high, it forecasts low future investment returns, not high dividend growth. Returns are predictable, not dividends.’
— Cochrane (2011, paraphrased)

References

Campbell, John Y., and Robert J. Shiller. 1988. ‘Stock Prices, Earnings, and Expected Dividends’. Journal of Finance 43(3): 661–676.

Campbell, John Y., and John H. Cochrane. 1999. ‘By Force of Habit: A Consumption-Based Explanation of Aggregate Stock Market Behavior’. Journal of Political Economy 107(2): 205–251.

Cochrane, John H. 1992. ‘Explaining the Variance of Price–Dividend Ratios’. Review of Financial Studies 5(2): 243–280.

Cochrane, John H. 1994. ‘Permanent and Transitory Components of GNP and Stock Prices’. Quarterly Journal of Economics 109(1): 241–265.

Cochrane, John H. 2011. ‘Presidential Address: Discount Rates’. Journal of Finance 66(4): 1047–1108.

Shiller, Robert J. 2024. U.S. Stock Markets 1871–Present and CAPE Ratio. Yale School of Management. Available at: http://www.econ.yale.edu/~shiller/data.htm (accessed July 8, 2025).

Can Stock Returns Be Predicted? Cochrane, Campbell, and the Case for Time-Varying Discount Rates