Health Overconfidence and Risky Health Behaviors: Evidence and Theory

Accurate beliefs are key to making optimal decisions and to assessing potential benefits and costs of actions accurately. Yet dozens of empirical studies have shown that people systemically overestimate their skills relative to others.

First, we provide evidence on how a simple yet methodologically sound measure of health overconfidence can be implemented in surveys. To our knowledge, this is the first attempt to measure relative health overconfidence in a representative population survey. Second, we show that the assumption of perfectly informed forward looking consumers is often violated in the domain of health. It also shows that biased individual health beliefs are systematically correlated with risky health behaviors. Third, we provide a novel theoretical framework of health overconfidence and risky behavior.

The empirical part is based on two different high-quality surveys from Germany which we used to implement our health overconfidence measures. The first database is the Berlin Aging Study II (BASE-II). BASE-II includes a range of standard survey questions in addition to objective health measures and cognitive tests among elderly residents of Berlin. The second database is the Innovation Panel of the nationally representative German Socio-Economic Panel Study (SOEP-IP). We included the same health assessment measures in SOEP-IP and in BASE-II.

The findings show that our proposed health overconfidence measure is reliable and has desirable distributional properties. It is robust to different health measures and different empirical definitions of overconfidence. Next, we show that a large share of the German population has biased beliefs about their personal position in the population health distribution. According to our definition, between 25 and 50% of all respondents are overconfident about their relative health status. Only five percent underestimate their relative health. Then, we show that individuals who are overly confident about their health are significantly more likely to eat unhealthy and to be overweight. In addition, they are also more likely to smoke, sleep short hours, and are less likely to exercise.

The last part of the part develops a novel model that links health overconfidence to risky health behavior. The model is flexible and general enough to explain the observed empirical nonlinear patterns. Moreover, it provides evidence on possible underlying mechanisms. It demonstrates how exactly a biased self-image can lead to an unhealthier lifestyle. Our theory formalizes health overconfidence as a bias in the agent's health assessment. The basic intuition is that optimal health behavior depends on marginal benefits and marginal costs of such behavior. If these marginal benefits and costs are also a function of the individual's health status, then a bias in the perceived health status can affect health behavior. For example, if smoking is more harmful for a sick person than for a healthy person, then an individual who is overconfident about her health status believes that smoking has lower marginal costs than it actually has. Hence, the person could smoke more than under accurate beliefs.