The Hansen-Jagannathan Bound: A Speed Limit Every Asset Pricing Model Must Obey

Here is a genuinely clever kind of scientific move. Instead of building a model and testing it, build a test that every possible model must pass, then hand it to the modellers and let them see how badly they do.

That is what Lars Peter Hansen and Ravi Jagannathan did in 1991. They derived a simple inequality, now universally called the Hansen-Jagannathan bound, that any economic model of asset prices must satisfy if it wants to be consistent with the returns we actually observe in markets. It is a diagnostic tool, not a theory. And it turned out to be brutal.

The problem: too many models, no cheap way to reject them

By 1991 the consumption-based asset pricing framework was well established. The core equation says that today's price of any asset equals the expected value of its future payoff, multiplied by a special random variable, the stochastic discount factor, which measures how much investors value a dollar in each possible future state. A dollar in a recession is worth more than a dollar in a boom, and the discount factor encodes exactly that.

Different economic theories are, in the end, just different guesses about what the discount factor is. A CAPM guess. A consumption-growth guess. A habit-formation guess. And testing each one meant building it out in full, calibrating it, simulating it, and comparing it to data. Expensive, slow, and easy to argue with, because a failure could always be blamed on some auxiliary assumption rather than the core idea.

Hansen and Jagannathan wanted to skip all of that.

The key idea, via analogy

Suppose someone claims to have designed an engine, and you have not seen it. You do know one thing: their car went from zero to sixty in three seconds. Even without inspecting the engine, physics lets you say it must produce at least this much power. Any proposed engine that produces less can be rejected on the spot, without ever opening the hood.

The Hansen-Jagannathan bound is exactly this kind of argument for economics. Their reasoning runs backwards from what markets do:

• We observe that some portfolios have achieved high Sharpe ratios, meaning a lot of excess return per unit of volatility. The stock market's own long-run Sharpe ratio is respectable, and clever combinations of assets have done much better.
• The pricing equation says the discount factor has to explain all these returns simultaneously.
• Getting a large excess return to be consistent with the equation requires the discount factor to be highly volatile. Intuitively, the only way an asset can be forced to offer a big premium is if the value of a dollar swings wildly between the states where the asset wins and the states where it loses. A calm discount factor cannot generate a big risk premium, in the same way a weak engine cannot generate fast acceleration.

The result they prove is a clean inequality: the ratio of the discount factor's standard deviation to its mean must be at least as large as the highest Sharpe ratio achievable in the market. Give them a set of returns, and they will hand you back a region on a chart of mean versus standard deviation. Any valid model's discount factor must land inside that region. If yours lands outside, it is dead, and you never had to specify the rest of the model.

Now plug in the standard consumption-based model. Its discount factor is driven by consumption growth, and aggregate consumption growth is famously calm: it wobbles by a couple of percent a year. To get such a placid variable to produce a discount factor volatile enough to clear the bound, you have to crank risk aversion up to values that are, frankly, ridiculous, implying people who would pay almost anything to avoid a small gamble.

The standard model does not just miss the bound. It misses it by an order of magnitude.

Why it mattered

• It made the equity premium puzzle unavoidable. Mehra and Prescott had shown the puzzle inside one specific model. Hansen and Jagannathan showed it as a constraint that binds on entire classes of models at once. You cannot wriggle out by tweaking functional forms; you have to produce a discount factor that is genuinely volatile.
• It became the standard first check. Ask any macro-finance economist how they screen a new model and the answer is: does it clear the HJ bound? It is cheap, model-free, and decisive.
• It gave a beautifully intuitive equivalence. The bound essentially says: the most volatile thing in your economy has to be at least as volatile as the best Sharpe ratio in your market. Sharpe ratios, the bread and butter of practitioners, turn out to be a direct measurement of how wildly the economy values a marginal dollar.
• It steered a generation of research. Habit formation, long-run risk, rare disasters, and heterogeneous agents are all, at bottom, attempts to build a discount factor volatile enough to clear this bound without absurd risk aversion. Hansen went on to share the 2013 Nobel Prize.

The honest limitations

• It rejects, it does not build. The bound tells you your model is wrong. It offers no guidance about what would be right. A great deal of subsequent literature is people flailing at the bound with increasingly exotic preferences.
• Passing the bound is necessary, not sufficient. A model can clear it and still be nonsense in every other respect. It is a screen, not a certificate.
• It is sensitive to what returns you feed it. The bound is computed from a chosen set of assets. Feed it a portfolio with an implausibly high in-sample Sharpe ratio, perhaps one that is really just a data-mined artifact, and you get an implausibly demanding bound. Sampling error in estimated Sharpe ratios flows straight through into the constraint.
• It inherits the data problems of consumption. When you check a consumption-based model against the bound, you are relying on consumption data that is noisy, revised, and aggregated. Some of the failure may be measurement rather than economics, although few believe measurement can explain a gap this large.

The one-line takeaway

Hansen and Jagannathan turned market Sharpe ratios into a minimum volatility requirement that any economic model of asset prices must meet, and the standard consumption-based model fails it so spectacularly that clearing the bound has become the central challenge of macro-finance.