 |
This paper proposes an empirical procedure to identify investors' subjective beliefs from observed asset prices, i.e. deviations from rational expectations in the conditional distribution of macroeconomic and financial variables. Our methodology relies on the smoothed empirical likelihood technique, which is non-parametric, allowing us to be agnostic on the nature of behavioral biases (if any). Conditional Euler equation restrictions for a chosen cross-section of assets and a parametric pricing kernel enable us to infer the subjective conditional distributions, given the investors' conditioning set. When using inflation and consumption growth as conditioning variablea, we show that deviations from the objective distribution can be quite large and that belief distortions seem to affect both the conditional mean and higher order moments of consumption growth. We show that the estimated beliefs distortion are remarkably similar for many of the popular consumption-based asset pricing models.
