Skewness Dispersion and Stock Market Returns

Imagine the stock market as a giant, chaotic dance floor where thousands of individual dancers (stocks) are moving to the music.

Most financial experts have spent decades trying to predict the next move of the entire dance floor by looking at the average speed or direction of the dancers. They ask, "Is the crowd generally moving up or down?"

This paper introduces a new, clever way to predict the dance floor's future by looking at how much the dancers disagree with each other.

Here is the simple breakdown of what the authors, Mykola Babiak, Jozef Baruník, and Josef Kurka, discovered:

1. The Core Idea: "The Skewness Dispersion"

In finance, "skewness" is a fancy word for asymmetry.

Positive Skew: A stock is like a lottery ticket. It usually stays flat or drops a little, but occasionally, it explodes upward with a huge gain.
Negative Skew: A stock is like a ticking time bomb. It usually climbs slowly, but occasionally, it crashes hard.

The authors didn't just look at the average skewness of all stocks. Instead, they measured the spread (or dispersion) between the most optimistic stocks and the most pessimistic ones.

The Analogy:
Imagine a classroom of students taking a test.

Low Dispersion: Everyone got a B. The class is uniform.
High Dispersion: Some students got A+ (lottery tickets), and some got F (time bombs), while the rest are in the middle. The class is chaotic and divided.

The paper finds that when the "class" is highly divided (High Dispersion), the entire stock market is likely to drop in the near future.

2. The Prediction: "The Chaos Signal"

The researchers analyzed 6,770 US stocks over 22 years using high-speed data. They found a clear rule:

When the gap between the "best-case" and "worst-case" stocks widens significantly, the stock market tends to fall in the coming months.

It's like a weather vane. When the wind starts blowing in wildly different directions (some stocks soaring, others crashing), it signals a storm is coming for the whole market.

3. Why Does This Happen? (The Two Reasons)

The authors dug deep to understand why this happens. They found two main drivers, which they describe as a mix of "Fear" and "Confusion."

The "Risk" Explanation: When investors are scared, they demand a higher reward to hold risky stocks. High dispersion signals that the market is uncertain about the future, so investors pull back, driving prices down.
The "Behavioral" Explanation: This is the more interesting part. The authors found that this "chaos" is often driven by smart investors (short sellers) betting against overpriced stocks while regular investors are still optimistic.
- The FOMC Connection: The magic happens around Federal Reserve meetings (when interest rates are announced).
- The Metaphor: Imagine a rumor spreading in a town. Before the Mayor makes an official announcement, the "smart kids" in town start betting on what he will say. They bet against the houses they think are overpriced. The "regular kids" are still cheering.
- When the Mayor (the Fed) finally speaks, the smart kids were right, and the prices adjust quickly. The "skewness dispersion" captures this moment of conflicting beliefs right before the news breaks.

4. Does It Make Money? (The Wallet Test)

The authors didn't just stop at theory; they tested if this could actually make an investor richer.

The Strategy: They built a portfolio that used this "Chaos Signal" to decide when to buy stocks and when to hold cash.
The Result: It worked incredibly well.
- Compared to a "set it and forget it" strategy (just buying the market and holding), using this signal added massive value.
- The Numbers: An investor using this signal could have earned an extra 7% to 8% per year (in "certainty equivalent" terms) compared to a passive investor.
- The Sharpe Ratio: This is a score for "risk-adjusted return." Their strategy scored between 0.82 and 0.91, which is very high compared to the standard market score of roughly 0.57.

5. The "Kurtosis" Side Note

The authors also checked if the heaviness of the tails (Kurtosis) mattered. They found it did, but it was like trying to hear a whisper in a hurricane—too much noise. The "asymmetry" (Skewness) was the clear, loud signal.

Summary for the Everyday Investor

Think of the stock market as a crowd.

If everyone is moving in the same direction, the market is stable.
If the crowd is split—some people are jumping for joy (lottery stocks) and others are running for the exits (crash stocks)—get ready.

This paper tells us that extreme disagreement among stocks is a warning sign. It suggests that the market is digesting big news (like interest rate changes), and once the dust settles, the overall market is likely to take a dip. By watching for this "disagreement," investors can potentially protect their money or position themselves to profit.

The Bottom Line: When the stock market gets too chaotic with some stocks looking like gold mines and others like mineshafts, it's a signal that the whole market might be about to stumble.

1. Problem Statement and Motivation

The equity risk premium literature has long struggled to identify robust predictors for aggregate stock market returns, particularly those that perform well out-of-sample. While higher-order moments (like skewness) are well-established predictors for individual stock returns, their ability to predict aggregate market returns has been limited. Previous work (e.g., Jondeau et al., 2019) showed that the average firm-level skewness has predictive power, but the authors argue that focusing solely on the mean masks critical information contained in the heterogeneity of skewness across firms.

The central research question is: Does the cross-sectional dispersion (variation) of firm-level realized skewness contain incremental information for predicting future aggregate stock market returns?

2. Methodology and Data Construction

Data

Sample: 6,770 U.S. stocks listed on NYSE, AMEX, and NASDAQ.
Period: December 2000 to December 2022.
Frequency: High-frequency intraday data (5-minute intervals) from 9:30 to 16:00 EST.
Source: Kibot and CRSP.

Construction of the Predictor: Skewness Dispersion ( $SD_{a-b}$ )

The authors construct a novel predictor based on the inter-percentile range of the cross-sectional distribution of daily realized skewness.

Realized Variance ( $RV_t^s$ ): Calculated as the sum of squared intraday log-returns for stock $s$ on day $t$ .
Realized Skewness ( $RS_t^s$ ): Calculated as the standardized sum of cubed intraday returns:
$RS_t^s = \frac{\sqrt{N} \sum (r_{t,k}^s)^3}{(RV_t^s)^{3/2}}$
Note: High-frequency realized skewness primarily captures jump components rather than the leverage effect.
Dispersion Measure: The daily skewness dispersion is defined as the difference between the $a$ -th and $b$ -th percentiles of the cross-sectional distribution of $RS_t^s$ . The authors test various ranges (e.g., 75th–25th, 95th–5th).
Monthly Aggregation: The monthly predictor is the average (or median) of the daily dispersion measures over the last five trading days of the month.

Econometric Framework

Predictive Regressions: Standard univariate and bivariate regressions of future market excess returns ( $r_{t,t+h}$ ) on the lagged skewness dispersion ( $SD_t$ ).
Horizons: 1, 3, 6, and 12 months.
Inference Corrections: To address the Stambaugh (1999) bias and overlapping observations, the authors employ:
- Newey-West (1986) standard errors.
- Non-overlapping regression estimations (Martin, 2017).
- IVX-Wald statistics (Kostakis et al., 2015) to account for regressor persistence.
Out-of-Sample (OOS) Tests: Expanding window estimation (2005–2022) comparing the Mean Squared Forecast Error (MSFE) against a historical average benchmark using the Clark and West (2007) test.
Asset Allocation: Evaluation of economic value via Certainty Equivalent Return (CER) gains and Sharpe ratios for a mean-variance investor (risk aversion $\gamma=3$ ).

3. Key Contributions

Novel Predictor: Introduces "Skewness Dispersion" as a distinct predictor derived from high-frequency data, capturing the heterogeneity of return asymmetry across the cross-section of stocks.
Robust Predictive Power: Demonstrates that skewness dispersion is a strong, negative predictor of future market returns, robust to various estimation techniques and not driven by recessionary periods.
Incremental Information: Shows that skewness dispersion provides information distinct from 50 established predictors (including sentiment, variance, macro variables, and other skewness measures).
Economic Significance: Quantifies substantial economic gains for investors using these forecasts, outperforming passive strategies and many alternative predictors.
Mechanism Identification: Provides evidence that the predictability is driven by an information-based mechanism, specifically concentrated around Federal Open Market Committee (FOMC) announcements, linking it to the gradual incorporation of macro news and heterogeneity in investor beliefs.

4. Empirical Results

In-Sample Predictability

Significant Negative Relationship: Higher skewness dispersion predicts lower future market returns.
Coefficients: For a 1-month horizon, the slope coefficient is approximately -0.10 (statistically significant at the 1% level).
Robustness: Results hold when excluding NBER recessions, using non-overlapping data, and applying IVX corrections.
Bivariate Analysis: The predictive power of skewness dispersion remains significant even when controlling for 50 other variables (e.g., market skewness, sentiment indices, variance risk premiums). Conversely, most other variables lose significance when skewness dispersion is included.

Out-of-Sample Performance

$R^2_{OOS}$ : Skewness dispersion yields positive and significant out-of-sample $R^2$ values (ranging from ~2.3% to 8.9% depending on the horizon and specification), significantly outperforming the historical average benchmark.
Forecast Encompassing: In convex combination tests, skewness dispersion captures the majority of the predictive information (optimal weight $\lambda > 0.5$ for most competitors), indicating it subsumes the information in other predictors.

Economic Value (Asset Allocation)

CER Gains: Incorporating skewness dispersion forecasts yields Certainty Equivalent Return gains of 709 to 875 basis points (annualized) for a 1-month horizon compared to a historical average benchmark.
Sharpe Ratios: Strategies based on skewness dispersion achieve Sharpe ratios between 0.82 and 0.91 (1-month horizon), significantly outperforming the buy-and-hold strategy (Sharpe ~0.57) and most other predictors.
Cumulative Wealth: Portfolios rebalanced using these forecasts generate substantially higher cumulative wealth over the sample period.

Mechanism Analysis

Risk vs. Behavior:
- Risk: Skewness dispersion is negatively correlated with aggregate risk measures (option-implied variance, realized variance), suggesting it captures variation in perceived economic conditions.
- Behavior: It is positively correlated with investor attention and AAII survey sentiment (bullish/bearish spread), suggesting it reflects heterogeneity in investor expectations.
FOMC Concentration: The predictive power is sharply concentrated in months surrounding FOMC announcements.
- Pre-FOMC and FOMC months: Strong, significant negative prediction.
- Non-FOMC months: Predictability vanishes.
- Interpretation: This supports a mechanism where skewness dispersion reflects the heterogeneity of beliefs among sophisticated investors (e.g., short sellers) regarding macro news, which is gradually priced in before the announcement and resolved immediately after.

Kurtosis Dispersion

The authors also test Kurtosis Dispersion (dispersion of tail heaviness). It also predicts returns negatively but is generally weaker than skewness dispersion, especially at longer horizons. This suggests that asymmetry (skewness) is the more critical driver of predictability than general tail risk (kurtosis).

5. Significance and Conclusion

This paper makes a significant contribution to asset pricing by identifying a robust, economically meaningful predictor for aggregate stock returns that relies solely on high-frequency return data.

Theoretical Implication: The findings challenge the view that higher-order moments are irrelevant for aggregate returns. Instead, they suggest that the dispersion of these moments across firms is a proxy for the heterogeneity of investor beliefs and the pace of information diffusion.
Practical Implication: Skewness dispersion offers a practical tool for portfolio management, delivering superior risk-adjusted returns and substantial economic value compared to traditional benchmarks.
Mechanism: The concentration of predictability around monetary policy announcements highlights that the stock market's reaction to macro news is not instantaneous but involves a period of belief heterogeneity and gradual price adjustment, which skewness dispersion effectively captures.

In summary, the paper establishes that the cross-sectional variation in the asymmetry of stock returns is a powerful, robust, and economically valuable signal for forecasting the equity risk premium.