Long-Run Risk: The Faint Signal Buried in Economic Growth That Terrifies Investors

The equity premium puzzle rests on one observation: aggregate consumption is smooth. Year to year, people's spending barely wobbles, so stocks cannot be very frightening, so the premium should be small. It is not.

Campbell and Cochrane attacked this by changing how people feel about consumption swings. Bansal and Yaron, in 2004, attacked it from an entirely different direction. They accepted that consumption looks smooth, and argued that we are looking at the wrong feature of it.

The problem: we measure the wobble and ignore the drift

When economists look at consumption data, they see a series that grows at roughly two percent a year with small fluctuations around that trend. The fluctuations are what everyone measures, and they are small. Hence the puzzle.

Bansal and Yaron asked: what if the trend itself is not constant? What if the long-run growth rate of the economy quietly drifts up and down over time, by tiny amounts, so tiny that they are essentially undetectable in the data?

Such a component would be nearly impossible to see. A shift in the long-run growth rate from 2.0% to 1.8% would be buried under the year-to-year noise. Statistically, you could stare at decades of data and never confidently reject the idea that growth is constant.

But investors are not statisticians looking backwards. They are forward-looking, and here is the crucial asymmetry: a tiny change in the long-run growth rate has an enormous effect on the value of everything, because it compounds forever.

The key idea, via analogy

Compare two pieces of bad news about your career.

News A: you will lose ten thousand dollars this year. Unpleasant. A one-off.

News B: your annual raises will be 0.2 percentage points smaller, forever. In the first year, that costs you a few hundred dollars. It is barely detectable in your bank statement. You might not even notice.

But over a forty-year career, compounded, News B costs vastly more than News A. It reduces your lifetime wealth, your retirement, your children's education. It is a catastrophe wearing the disguise of a rounding error.

The small permanent shock is far scarier than the large temporary one. And crucially, the small permanent shock is nearly invisible in short-term data, while the large temporary one is glaring.

That is the entire logic of long-run risk. Bansal and Yaron model consumption growth as containing a small, highly persistent predictable component: a faint, slowly-varying signal about the economy's long-term growth prospects. News about that component barely moves this year's consumption, but it moves the entire discounted stream of future consumption, and therefore it moves stock prices violently.

They add a second ingredient: time-varying economic uncertainty. The volatility of the economy is itself not constant. Sometimes the future looks murky, sometimes clear. Rising uncertainty is itself bad news that investors will pay to avoid.

And they need a third ingredient to make the machine run, which is the technical heart of the paper: Epstein-Zin preferences. The standard utility function has a defect, it forces a single parameter to control two completely different things: how much you dislike risk, and how much you dislike having your consumption vary across time rather than across states. Those are not the same psychological quantity. Epstein-Zin preferences separate them.

Why does this matter so much? Because it is what makes investors care about news rather than only about outcomes. Under Epstein-Zin preferences, an investor who wants to resolve uncertainty early and who dislikes risk is genuinely hurt by learning that the long-run growth rate has fallen, even before any of that lost consumption actually materializes. The bad news itself is the blow. Under standard preferences, news about the distant future carries almost no weight. Under Epstein-Zin, it is the main event.

Put the three together and the model can generate a large equity premium, a low and stable risk-free rate, volatile stock prices, and returns that are predictable from valuation ratios, all with a level of risk aversion that is high but not absurd.

Why it mattered

• It is one of the two leading resolutions of the equity premium puzzle. Alongside Campbell and Cochrane's habit model, it defines the modern macro-finance landscape. Most subsequent work builds on one or the other.
• It made Epstein-Zin preferences standard equipment. Separating risk aversion from the willingness to substitute over time is now routine, and this paper is a large part of why.
• It explains why markets react so violently to information about the distant future. A modest downgrade to long-term growth expectations can move markets far more than a bad quarter. Under long-run risk, that is not irrational overreaction; it is exactly right.
• It gives a useful mental model for real investing. The risks worth worrying about are not the ones that make headlines this quarter. They are the slow, permanent, barely visible shifts in the trajectory: demographics, productivity growth, institutional decay. Those are the ones that actually compound.

The honest limitations

• The key variable is essentially unobservable, and that is a problem. The persistent growth component is calibrated to be so small and so persistent that you cannot reliably detect it in the data. Critics, most forcefully in work examining whether long-run consumption predictability actually exists, have pointed out that a model whose central mechanism is by construction almost impossible to falsify is on uncomfortable ground.
• The results are sensitive to the calibration. The persistence of the growth component and the size of the preference parameters have to be chosen carefully. Modest changes can substantially alter the model's implications, which invites the accusation of tuning.
• It requires a preference for early resolution of uncertainty. The model needs investors who genuinely want to know bad news sooner rather than later. This is a real psychological posture, but the experimental evidence that people actually have it is not overwhelming.
• Risk aversion is still fairly high. The model reduces the required risk aversion compared with the original puzzle, but not to a level everyone finds comfortable.
• It competes with habit and disaster models, and cannot easily be distinguished from them. All three explain broadly the same set of stylized facts. Designing an empirical test that cleanly separates them has proven very difficult, which is itself a mild indictment of the whole enterprise.

The one-line takeaway

Bansal and Yaron argued that investors are frightened not by this year's economic wobble but by tiny, permanent, nearly invisible shifts in the long-run growth rate, because a small change that compounds forever is far more damaging than a large one that reverses.