Time-varying risk aversion and asset prices

George LiJournal of Banking & Finance 2006 · DOI

TL;DR

This paper investigates how time-varying risk aversion influences asset prices using a variant of the Campbell and Cochrane (1999) model. Surprisingly, it finds that countercyclical risk aversion, contrary to intuition, leads to a procyclical risk premium and return volatility. The study also suggests that while high risk aversion is often cited as a solution to asset pricing puzzles, its effectiveness depends critically on other model parameters, particularly the volatility of the surplus consumption ratio.

What Problem Does This Paper Address?

A long-standing puzzle in finance is the "equity premium puzzle" and the "risk-free rate puzzle," which refer to the difficulty of explaining the historically high average stock returns and low average risk-free rates using standard economic models. One prominent explanation for these puzzles is that investors' risk aversion is not constant but varies over time, specifically increasing during economic downturns (countercyclical risk aversion). This paper aims to rigorously test this hypothesis by dissecting the mechanisms through which countercyclical risk aversion affects asset prices within a well-established consumption-based asset pricing model. It seeks to answer: (1) Does countercyclical risk aversion indeed drive a countercyclical risk premium? (2) Can high risk aversion, even when countercyclical, resolve the equity premium and risk-free rate puzzles? (3) Does countercyclical risk aversion explain other observed asset market phenomena like return predictability, price mean-reversion, and the leverage effect?

Methods

The paper employs a variant of the Campbell and Cochrane (1999) model, which features a representative agent with a habit-formation utility function. This setup allows for time-varying risk aversion that is countercyclical with respect to the business cycle, proxied by the surplus consumption ratio. The authors use a continuous-time version of the model to derive closed-form solutions for key asset pricing variables like the risk-free rate and the market price of risk. A crucial aspect of their methodology is to isolate the effect of risk aversion by setting the volatility of the log surplus consumption ratio (the "second source" of a countercyclical risk premium in the original CC model) to a constant. This allows them to specifically examine the impact of countercyclical risk aversion (the "first source"). They then calibrate the model to empirical data and generate artificial data to test whether the model can replicate observed asset market dynamics.

Datasets

The paper uses historical data from the postwar period (1947-1995) and a century-long dataset (1871-1993) for calibration and comparison. These datasets include value-weighted NYSE stock index returns, three-month Treasury bill rates, and per capita nondurables and services consumption. The authors do not introduce new datasets or benchmarks but rather use existing, well-established data to evaluate their model's performance.

Key Insights and Ideas

  • Counterintuitive Procyclicality: The most striking finding is that countercyclical risk aversion, contrary to the intuition that it should lead to higher risk premiums during recessions, actually results in a procyclical risk premium and return volatility. This is because while investors become more risk-averse in recessions, stock returns also become less volatile, and this latter effect dominates, leading to a lower risk premium.
  • The Dominance of the Market Price of Risk: The paper suggests that in the original Campbell and Cochrane model, it is the time-varying market price of risk (driven by the volatility of the surplus consumption ratio) rather than countercyclical risk aversion itself that is the primary driver of countercyclical risk premiums and return volatility.
  • Puzzles Depend on Parameters: The ability of high risk aversion to resolve the equity premium and risk-free rate puzzles is shown to be highly sensitive to other model parameters, particularly the volatility of the log surplus consumption ratio. A small value for this volatility can render high risk aversion ineffective in solving these puzzles simultaneously.
  • Limited Explanatory Power for Other Phenomena: The calibrated model with countercyclical risk aversion fails to replicate several other empirically observed asset market phenomena, including the predictability of long-horizon stock returns, the mean-reversion of stock prices, and the leverage effect in return volatility.

Key Results

  • Countercyclical risk aversion leads to a procyclical risk premium and return volatility.
  • When the volatility of the log surplus consumption ratio is small, high risk aversion may not resolve the equity premium and risk-free rate puzzles.
  • The model, with countercyclical risk aversion as the sole driver, does not generate the observed predictability of long-horizon stock returns, stock price mean-reversion, or the leverage effect in return volatility.
  • The paper's calibration suggests that a large value for the volatility of the log surplus consumption ratio is crucial for resolving the equity premium and risk-free rate puzzles.

Limitations and Future Work

The authors acknowledge that their numerical exercise is a "step stone" and that the results regarding the inability of countercyclical risk aversion to explain certain phenomena may not be robust. They suggest that future research should focus on developing consumption-based representative agent models that incorporate a constant value of risk aversion alongside a time-varying countercyclical market price of risk to better explain aggregate stock market behavior. This implies a shift in focus from the source of risk aversion to the dynamics of the market price of risk as the key driver of empirical regularities.

Summary updated

BibTeX
@article{li2006time,
  author = {George Li},
  title = {{Time-varying risk aversion and asset prices}},
  journal = {Journal of Banking \& Finance},
  year = {2006},
  doi = {10.1016/j.jbankfin.2006.02.005},
}
Figure