What Risk Are Investors Actually Being Compensated to Bear? The Answer Isn’t Volatility

Every serious framework for investing in equities begins with the same foundational question: What risk are investors being compensated to bear? The standard answer—that equity investors are paid for taking on exposure to a broadly diversified portfolio of risky assets—is intuitive but incomplete. It raises a deeper question: Investors are being compensated for which feature of that risk? Variance? The possibility of catastrophic loss? Something else entirely?

C. Christopher Hyland and Niklas Schmitz address this question in their April 2026 study “Pricing the Left Tail: Consumption Skewness and Expected Returns.” Their answer: Investors are compensated for the asymmetry between downside and upside risk in aggregate consumption growth—specifically, the degree to which the distribution of future consumption outcomes is tilted to the left. Symmetric risk measures like variance explain almost nothing, while left-tail asymmetry explains a great deal.

What the Researchers Examined

The core challenge in testing consumption-based asset-pricing theory is measurement. Traditional finance theory predicts that expected risks to future consumption growth should forecast equity returns. But direct measures of conditional consumption risk—what investors expect the distribution of future consumption to look like—have never existed. Survey data on this topic doesn’t exist. Using realized consumption growth conflates expected and unexpected components. Extracting risk from asset prices requires assumptions about preferences that remain actively contested.

Hyland and Schmitz solve this problem by building a model-free estimate of the conditional distribution of US aggregate consumption growth at each point in time. Their methodology combines three elements:

• Quantile regressions targeting the fifth, 25th, 50th, 75th, and 95th percentiles of year-ahead consumption growth
• A large macroeconomic predictor set drawn from the FRED-MD panel of 128 monthly series—covering output, labor markets, housing, credit, interest rates, prices, and equities—as opposed to the single-predictor approach used in earlier work
• Machine-learning variable selection (lasso quantile regression, designed to be less cluttered, more stable, and better at selecting important variables) combined with partial quantile regression, which weights predictors by their actual forecasting power for each specific quantile rather than just their common variation

The entire exercise was conducted out of sample: At every monthly step from January 1970 through September 2025, they forecasted the distribution using only information available at that date—the resulting risk measures genuinely reflect what investors could have known in real time, not what we know with hindsight.

From the predicted quantiles, they constructed multiple candidate risk measures: the conditional mean, variance, skewness, kurtosis, dispersion, expected shortfall, and Kelley’s measure of skewness. The last of these is their main variable of interest. Kelley skewness compares the distance from the median to the 90th percentile against the distance from the median to the 10th percentile. When the distribution is left-skewed—when the downside is wider than the upside—Kelley skewness is negative.

Having constructed these measures, they tested three things: whether any of them predict aggregate equity returns in the time series; whether they are priced in the cross-section of returns; and which macroeconomic shocks are most closely linked to movements in the priced measure.

Key Findings

1. Left-tail asymmetry—not variance—predicts equity returns.

Kelley skewness is the only consumption-risk measure with robust, statistically significant return-forecasting power across horizons. A one-unit decline in Kelley skewness (say, from 0.5 to negative 0.5, representing a shift toward greater left-tail risk) is associated with a 16.07-percentage-point increase in excess market returns over the next 12 months. The in-sample R-squared at that horizon was 9.5%.

Variance and dispersion—the symmetric risk measures that dominate the uncertainty literature—were small and insignificant across all horizons. The conditional mean, while significant in some windows, was less stable than Kelley skewness. The result is not about the level of consumption expectations or the width of the distribution. It is specifically about its asymmetry.

Out of sample, Kelley skewness delivers a 7.8% R-squared, outperforming standard valuation ratios (which often produce negative out-of-sample R-squared) and the Brian Kelly and Hao Jiang “Tail Risk and Asset Prices” index (which reports 4.5% at the same horizon). The statistical evidence is not decisive—the null of equal forecast performance against a constant-only benchmark is not formally rejected—but a 7.8% out-of-sample R-squared is economically meaningful in the context of equity return forecasting.

Critically, none of this holds for risk measures constructed using the simpler single-predictor approach from earlier literature. When the researchers used the National Financial Conditions Index as the sole predictor, the resulting Kelley skewness measure had the wrong sign and essentially no predictive power. The return-predictability result depends on the richer, machine-learning-based forecast distribution. You need a better measurement to find the better result.

2. Consumption skewness is priced in the cross-section.

Using Fama-MacBeth regressions on 25 size/book-to-market portfolios and 12 industry portfolios, the researchers found that consumption-skewness betas strongly predict average returns. The cross-sectional price of risk is negative 0.51 (Shanken-corrected standard error of 0.08), with the negative sign indicating that assets with greater exposure to left-skewed consumption states earn higher average returns—exactly the direction theory predicts.

The cross-sectional R-squared was 90.6%. A single factor (consumption skewness beta) explains more than nine-tenths of the return variation across size, value, and industry portfolios. The slope is negative in every nonoverlapping monthly subsample tested and statistically significant in most of them.

3. Financial uncertainty and bad technology news are the primary drivers of the priced state.

Having established that Kelley skewness is priced, the researchers ask what moves it. Using local projections (horizon-by-horizon regressions) over the 1991–2025 sample, they estimate how four broad macro shocks affect the conditional distribution of consumption growth.

The ranking by average Kelley skewness response over horizons two through six quarters is: financial uncertainty innovations (negative 0.029), bad investment-specific technology news (negative 0.022), credit-condition tightening (negative 0.014), and adverse oil-supply news (negative 0.006). Financial uncertainty and bad IST news both produce asymmetric lower-tail widening: the 10th percentile of expected consumption growth falls substantially more than the median, while the 90th percentile barely moves. The lower tail stretches; the center and upper tail remain relatively stable.

Credit tightening behaves differently. It shifts the whole distribution downward rather than widening the lower tail selectively—more like a level shock to expected consumption than an asymmetric tail-widening event. This explains why credit conditions rank third despite being economically important: Kelley skewness measures asymmetry, and credit tightening produces relatively symmetric effects.

4. Monetary policy is a conditional stabilizer, not an unconditional driver.

The unconditional effect of monetary policy surprises on Kelley skewness was negligible—less than 5% of the index’s in-sample standard deviation. But the average masks significant state dependence. When recent Kelley skewness has been low (indicating elevated downside risk), monetary easing is associated with a 0.05-point improvement in Kelley skewness, achieved almost entirely through lifting the 10th percentile of expected consumption growth back toward the median. In high-skewness states, the same easing produces roughly zero effect or slightly negative effects.

Monetary easing is, therefore, most effective at reducing the priced component of consumption risk precisely when the equity premium is highest. This pattern is consistent across 11 alternative high-frequency monetary surprise series and multiple sample windows.

Their findings led the authors to conclude: “The contribution of this paper is to show that what investors are compensated for in equity markets is not consumption uncertainty in general, but a specific feature of the conditional consumption-growth distribution: the length of its lower tail relative to its upper tail.”

They added that the “Kelley skewness, a simple summary of left-tail asymmetry, predicts the equity premium and prices the cross section of returns, while the conditional mean, variance, and quantile dispersion do not.”

Key Investor Takeaways

1. Investors are compensated for left-tail risk, not uncertainty in general.

The finding that variance and dispersion predict nothing while left-tail asymmetry predicts equity returns has real implications. The large literature on uncertainty—VIX, economic policy uncertainty, macroeconomic uncertainty indexes—typically focuses on symmetric measures of dispersion. This paper suggests that is the wrong object. What markets price is specifically the risk that the bad outcomes are much worse than the good outcomes are good: Not just that the future is uncertain, but that the distribution is tilted in the wrong direction.

2. Consumption risk matters—but measuring it correctly is essential.

Consumption-based asset pricing has struggled empirically for decades, in part because the available proxies for conditional consumption risk have been poor. This paper’s contribution is to show that with a better measurement apparatus—one that draws on a rich macro panel rather than a single predictor—the theory’s predictions hold up quite well. The equity premium is high when the conditional distribution of consumption growth is left-skewed. The issue was not the theory; it was the measurement.

3. The equity premium is the highest when the macroeconomy is most fragile.

Kelley skewness falls—implying higher expected returns—during recessions, financial crises, and periods of high financial uncertainty or bad technology news. This is exactly the environment in which most investors want to reduce risk, not add it. The equity risk premium is a compensation for holding on when conditions look bleakest. Investors who systematically reduce equity exposure during these states may be doing so at precisely the worst time from a forward return perspective.

4. Be cautious of factor strategies anchored to symmetric risk.

If the priced object is left-tail asymmetry rather than variance, then factor strategies built around exposure to realized volatility or uncertainty may be capturing the right general concept but the wrong specific moment. Strategies that explicitly target exposure to consumption skewness—or that use it as a state variable for timing or risk management—would be more directly aligned with what the evidence suggests is actually being priced.

5. Financial uncertainty and technology shocks are the leading drivers of the equity premium.

The finding that financial uncertainty and bad IST news are the primary movers of the priced consumption-risk state variable has practical implications for thinking about economic regimes. Episodes of high financial uncertainty and contracting productive capacity are the states in which the equity premium is highest—and, by the theory, the states in which long-term investors should most want to maintain or increase their exposure. The discomfort of those environments is precisely the source of the compensation.

6. Monetary policy works when it matters most—but don’t confuse correlation with comfort.

The state-dependent result on monetary easing is reassuring in one sense: Accommodative policy does appear to reduce priced downside risk precisely when that risk is elevated. But the mechanism is improvement in the lower tail of expected consumption growth, not a broad risk-on signal. Investors should not read monetary easing as automatically reducing the equity premium—it depends on whether the easing is occurring in a state where the lower tail of the consumption distribution has room to improve.

A Note of Caution

The authors are transparent about the limitations. The consumption-risk estimates use final revised data rather than real-time vintages, meaning the out-of-sample R-squared likely overstates genuine real-time forecastability. The calibration tests show that the predicted quantiles are not exactly calibrated—particularly in the middle and upper parts of the distribution—so the fitted distribution is best read as a summary of relative tail shape, rather than a precise probability statement. And the cross-sectional evidence, while striking, is based on 37 test portfolios over a sample that includes the same period used to estimate the consumption betas—the usual concerns about in-sample circularity apply.

These are real limitations. The paper is a substantial advance, but it is not the final word. Replication using real-time data vintages and alternative cross-sectional test assets would be the natural next step.

Closing Thoughts

We began with the question of what risk investors are actually being compensated to bear. The standard answer has always been vague, something like “the risk that markets go down when times are bad.” This paper sharpens that answer considerably. Investors are compensated for bearing exposure to states in which the lower tail of future consumption is unusually wide relative to the upper tail—states where the bad outcomes are disproportionately bad. Not for volatility in general. Not for the level of expected consumption growth. For the asymmetry.

That precision matters. It tells us when the equity premium is the highest (when the distribution is most left-skewed), what drives it there (financial uncertainty and adverse technology news), and how policy affects it (by lifting the lower tail in fragile states). It also tells us what has not been working in the empirical literature on consumption-based asset pricing: The right theory was being tested with the wrong measurement.

For investors, the lesson is familiar but worth restating with this new precision: The equity premium is compensation for bearing risk that is concentrated in bad times. Those who stay invested through periods of elevated left-tail consumption risk are the ones who earn the premium. Those who cannot tolerate that discomfort—and exit precisely when skewness is most negative—are the ones who pay for it.