Subjective health measures and state dependent reporting errors

TLDR

Subjective health measures in labour supply and retirement studies are often criticized for bias, as respondents may underreport health problems to gain financial incentives or conform to social norms, leading to state‑dependent reporting errors that make self‑reported health endogenous. The paper aims to evaluate the significance of state‑dependent reporting errors in survey responses and to develop a model that accounts for such systematic mis‑reporting. The authors propose and estimate a model that incorporates state‑dependent reporting errors to correct for bias in subjective health measures. Estimates show that respondents receiving Disability Allowance exhibit large, systematic reporting errors, and using subjective health measures in retirement models can severely bias parameter estimates and conclusions.

Abstract

Abstract The use of subjective health measures in empirical models of labour supply and retirement decisions has frequently been criticized. Responses to questions concerning health may be biased due to financial incentives and the willingness to conform to social rules. The eligibility conditions for some social security allowances, notably Disability Insurance benefits, are contingent upon bad health. Even if the decision to apply for a disability allowance is to some extent motivated by financial considerations or a relatively strong preference for leisure, respondents will be inclined to play down these motives and emphasize the importance of their health condition. As a consequence, reporting errors may depend on the labour market status of the respondent and self‐reported health variables will be endogenous in labour supply and retirement models. The objective of this paper is to assess the importance of state dependent reporting errors in survey responses and to propose and estimate a model that can be used to account for this kind of systematic mis‐reporting. The estimation results indicate that among respondents receiving Disability Allowance, reporting errors are large and systematic. Using such subjective health measures in retirement models may therefore seriously bias the parameter estimates and the conclusions drawn from these.