Measuring the 'Dark Matter' in Asset Pricing Models
Massachusetts Institute of Technology; National Bureau of Economic Research (NBER)
Winston Wei Dou
Massachusetts Institute of Technology (MIT) - Sloan School of Management
Massachusetts Institute of Technology (MIT) - Sloan School of Management; National Bureau of Economic Research (NBER)
September 17, 2013
Models of rational expectations endow agents with precise knowledge of the probability laws inside the models. This assumption becomes more tenuous when a model's performance is highly sensitive to the parameters that are difficult to estimate directly, i.e., when a model relies on "dark matter." We propose new measures of model fragility by quantifying the informational burden that a rational expectations model places on the agents. By measuring the informativeness of the cross-equation restrictions implied by a model, our measures can systematically detect the direction in the parameter space in which the model's performance is the most fragile. Our methodology provides new ways to conduct sensitivity analysis on quantitative models. It helps identify situations where parameter or model uncertainty cannot be ignored. It also helps with evaluating competing classes of models that try to explain the same set of empirical phenomena from the perspective of the robustness of their implications.
Number of Pages in PDF File: 64
Keywords: model fragility, robustness, rational expectation, cross-equation restriction, information theory
JEL Classification: C1, D83, E44, G12working papers series
Date posted: September 19, 2013
© 2014 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo4 in 0.266 seconds