Risk and time preference and health insurance decisions

We investigate individual preferences for risk and time using a population of low-income residents of Richmond, VA. We elicit risk and time preference using two different methods. We compare revealed preferences for risk and time using incentive compatible tasks with self-reported preferences elicited with a survey questionnaire. Understanding the stability of preferences for risk and time across multiple measures also has important policy implications. Stability would indicate that preferences can be measured with either method, and at the same time would suggest that we can use appropriately designed experiments to test economic policies.  We then assess the correlation of each method with measures of health status, health behaviors and health care utilization history. Finally, we examine the associations among methods of time and risk preference elicitation and preferences for insurance coverage, the consistency between stated and revealed preferences for insurance plans (choice consistency), and the extent to which participants choose a cost minimizing plan given expected health care needs (choice quality).

This study uses an unusual population of particular interest for the implementation of the Affordable Care Act; a culturally and ethnically diverse low-income, uninsured, urban group of individuals. We first measure the preference for risk and time in a group lab-setting, albeit in the field, with individuals completing two tasks. We then have the subjects answer a questionnaire with multiple questions about personal information, health status and utilization and personal attitude toward risk and time. The subjects finally are presented several health insurance choices and asked to choose the most preferred plan.

Our preliminary sample is comprised of 103 uninsured individuals primarily residing in low-income urban areas (85%) who are predominantly African American (83%).  Initial results using Pearson’s and Spearman’s correlation indicate incentive compatible and survey methods used to assess risk preferences were not significantly correlated.  However, a significant correlation was found among the stated and revealed time preference measures employed (rho =0.34 for both correlation measures; p<0.01).  For survey methods of eliciting risk preference, less risk averse participants were more likely to visit the ER and be admitted to the hospital (p<0.01).  For incentive compatible preference measures, the propensity to be risk-loving was also positively correlated with having an inpatient admission in the past year (p<0.10).  For survey methods of eliciting time preference, we found less patient participants were more likely to have a chronic disease (p<0.10) and smoke (p<0.05), while impatience elicited via incentive compatible methods was positively correlated with smoking (p<0.05), visiting the ER (p<0.05), having an inpatient admission in the past year (p<0.10), and negatively correlated with comprehension of health insurance terms (p<0.10). Other than this latter result, we find no evidence that either incentive or stated preference methods are related to health insurance coverage decisions, although both methods agree on multiple measures of health and health care choices, which must be considered when purchasing insurance.