Estimating a Cardinal Attribute from Ordered Categorical Responses Subject to Noise

Estimating a Cardinal Attribute from Ordered Categorical Responses Subject to Noise This paper presents an estimate of the distribution of relative risk tolerance---a preference parameter quantifying individuals' behavior toward risk. The estimate is based on a new survey question about gambles over lifetime income. The preference parameter is treated as a latent variable subject to error; responses to the survey are modeled as depending on whether this latent variable falls within certain bounds. In contrast to the standard discrete choice model, the latent variable and the bounds have a cardinal interpretation. Moreover, for a subset of survey respondents, the survey questions are answered twice. This structure allows identification of the mean and variance of the preference parameter and the variance of the noise.

Methodology/Latent variables/Cardinal probit/Risk aversion and risk tolerance/Preference parameters/Survey methods

8102