We examine the relationship between consumption growth, agents' self-reported layoff probabilities and effective layoff rates for evidence of bias in behavior. Agents' reports about the probability of a job loss differ substantially from the true layoff chance and this paper studies to what extent those reports can explain variation in consumption growth for individuals with similar objective probabilities. If agents with comparable layoff rates behave similarly no matter what they report about a future job loss, it would provide evidence in favor of what we call the unbiased hypothesis in behavior. On the contrary, if individuals whose layoff rates are similar but who differ regarding their reported probabilities behave differently, it would suggest the existence of bias in behavior. 

The econometric model uses reported layoff probabilities elicited in the Health and Retirement Study (HRS) as an indicator variable of beliefs and rational expectation probabilities. The estimation analysis performs 2SLS and control function procedures to consider the high concentration of focal answers observed in the distribution of reported probabilities. The estimates are consistent with evidence against the rational expectations hypothesis in behavior. Specifically, it is found that an increase in 1pp in the subjective probability generates an increase in reports of about 0.3pp.