Paper: SSRN 1315328

Authors: Tim Bollerslev, Hao Zhou

Abstract

We find that the difference between implied and realized variances, or the variance risk premium, is able to explain more than fifteen percent of the ex-post time series variation in quarterly excess returns on the market portfolio over the 1990 to 2005 sample period, with high (low) premia predicting high (low) future returns. The magnitude of the return predictability of the variance risk premium easily dominates that afforded by standard predictor variables like the P/E ratio, the dividend yield, the default spread, and the consumption-wealth ratio (CAY). Moreover, combining the variance risk premium with the P/E ratio results in an R2 for the quarterly returns of more than twenty-five percent. The results depend crucially on the use of “model-free”, as opposed to standard Black-Scholes, implied variances, and realized variances constructed from high-frequency intraday, as opposed to daily, data. Our findings suggest that temporal variation in risk and risk-aversion both play an important role in determining stock market returns.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor
  • Basis: provisional. Scored from the abstract alone because the full text was not available, so both scores reflect what the abstract claims, not a reading of the paper.

Why this score: The paper demonstrates strong empirical rigor through its use of ‘model-free’ implied variances and high-frequency data, yielding significant predictive power. Its novelty lies in highlighting the variance risk premium’s superior predictive ability compared to traditional variables. The mathematical complexity is moderate, involving advanced statistical concepts for variance construction and analysis.

Research Flowchart

  flowchart TD
    A["Research Goal: Explain time series variation in market excess returns"] --> B{"Key Methodology: Analyze Variance Risk Premium (VRP)"};
    B --> C{"Data/Inputs:"};
    C --> C1["Model-free Implied Variances (from options)"];
    C --> C2["Realized Variances (from high-frequency intraday data)"];
    C --> C3["Standard Predictors (P/E, Div Yield, etc.)"];
    B --> D["Computational Process: Regress excess returns on VRP and other predictors"];
    D --> E["Key Finding 1: VRP explains >15% of quarterly excess returns (1990-2005)"];
    E --> F["Key Finding 2: VRP predictability > standard predictors (P/E, Div Yield, etc.)"];
    F --> G["Key Finding 3: VRP + P/E R^2 > 25% for quarterly returns"];
    G --> H["Crucial Dependence: Model-free implied and high-frequency realized variances"];
    H --> I["Conclusion: Temporal variation in risk & risk-aversion determine stock returns"];

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes that the variance risk premium, defined as the difference between implied and realized variances, can predict quarterly excess market returns and explains a substantial portion of their variation.

Source sentence
We find that the difference between implied and realized variances, or the variance risk premium, is able to explain more than fifteen percent of the ex-post time series variation in quarterly excess returns on the market portfolio over the 1990 to 2005 sample period, with high (low) premia predicting high (low) future returns.

Assumes 2

  • Use of model‑free implied variances rather than Black‑Scholes implied variances. model
    Source sentence
    The results depend crucially on the use of "model-free", as opposed to standard Black-Scholes, implied variances, and realized variances constructed from high-frequency intraday, as opposed to daily, data.
  • Realized variances are constructed from high‑frequency intraday data instead of daily data. data
    Source sentence
    The results depend crucially on the use of "model-free", as opposed to standard Black-Scholes, implied variances, and realized variances constructed from high-frequency intraday, as opposed to daily, data.

Data

  • Model‑free implied variances (options) options unknown
    1990-2005 · market portfolio
    Source sentence
    We find that the difference between implied and realized variances, or the variance risk premium, is able to explain more than fifteen percent of the ex-post time series variation in quarterly excess returns on the market portfolio over the 1990 to 2005 sample period, with high (low) premia predicting high (low) future returns. The results depend crucially on the use of "model-free", as opposed to standard
  • Realized variances (high‑frequency intraday) market-prices unknown
    1990-2005 · market portfolio
    Source sentence
    We find that the difference between implied and realized variances, or the variance risk premium, is able to explain more than fifteen percent of the ex-post time series variation in quarterly excess returns on the market portfolio over the 1990 to 2005 sample period, with high (low) premia predicting high (low) future returns. The results depend crucially on the use of "model-free", as opposed to standard

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The variance risk premium explains more than fifteen percent of the ex‑post time‑series variation in quarterly excess market returns, with high premia predicting high future returns. predictive-accuracy
    Source sentence
    We find that the difference between implied and realized variances, or the variance risk premium, is able to explain more than fifteen percent of the ex-post time series variation in quarterly excess returns on the market portfolio over the 1990 to 2005 sample period, with high (low) premia predicting high (low) future returns.
  • The return predictability from the variance risk premium dominates that of standard predictors such as the P/E ratio, dividend yield, default spread, and consumption‑wealth ratio. predictive-accuracy
    Source sentence
    The magnitude of the return predictability of the variance risk premium easily dominates that afforded by standard predictor variables like the P/E ratio, the dividend yield, the default spread, and the consumption-wealth ratio (CAY).
  • Combining the variance risk premium with the P/E ratio yields an R² of more than twenty‑five percent for quarterly market returns. predictive-accuracy
    Source sentence
    Moreover, combining the variance risk premium with the P/E ratio results in an R2 for the quarterly returns of more than twenty-five percent.

Checked by others

Contradicted by

  • 0DTE Asset Pricing (2024)
    On the question: Does the variance risk premium predict future market returns?
    Compare the findings

    This paper: The variance risk premium explains more than fifteen percent of the ex‑post time‑series variation in quarterly excess market returns, with high premia predicting high future returns.

    That paper: The high variance risk premium is mainly driven by compensation for upside risk and negatively predicts market returns.

    A finds high premia predict higher returns explaining >15% of variation; B finds high premia mainly from upside risk and predict lower returns.

Qualified by

  • 0DTE Trading Rules (2023)
    On the question: Does the variance risk premium predict short‑term market returns?
    Compare the findings

    This paper: Combining the variance risk premium with the P/E ratio yields an R² of more than twenty‑five percent for quarterly market returns.

    That paper: A positive 0DTE variance risk premium exists, but at same‑day horizons its economic magnitude is small and difficult to monetize after realistic frictions.

    A shows VRP+P/E gives R²>25% for quarterly returns; B finds a positive 0‑day VRP but its economic magnitude is small and hard to monetize.

Code

No repository by the authors found in the paper.

Extracted 2026-11-15 · same card as JSON · A language model chose these items: check the source sentence before relying on one.