Expected Spending: The variable mis is the expected level of spending by each individual for each category of service. Estimating expected spending requires assumptions about the information available to individuals. The literature reflects a wide range of conceptions of what consumers might know about their health risks. Newhouse et al. (1989) suggest that individuals know all the information contained in measurable aspects of health status plus the time invariant-person specific component of the unobserved factors contributing to variation in health care spending.
Welch (1985) makes a similar assumption, referring to a “permanent” component of health spending that is individual-specific. Welch speculates that individuals might know more than this and be able to forecast use of some acute services such as births and some other illnesses. Some empirical work on plan choice confirms the presence of considerable individual knowledge. Ellis (1985) and Pemeger (1995) show that an individual’s historical pattern of spending affects health plan choice. Other research points to the fact that individuals appear to select plans on the basis of information not contained in risk adjustment systems (Cutler, 1994; Ettner et al., 1998).
We consider the implications of several informational assumptions. Recall that if individuals can predict nothing, there is no selection problem, so no simulation needs to be done for this case. We start with the assumption that individuals can predict based on age and sex. That is, we assume each individual predicts they will use the average of a person of their age and sex for each service category. Alternatively, we assume individuals can use the information contained in prior use. As will be seen shortly, if individuals know all the information contained in prior use, existing risk adjusters cannot cope with the selection-induced inefficiencies, and some services would have very high or very low q’s in profit maximization.
In the simulations, we therefore equip individuals with some of the information in prior use, 10%, 20%, 30% and 40%, to show the impact of more information. In order to construct these estimates under different information conditions, we estimate a series of two-part models. Each two-part model uses right hand side variables (e.g. age and sex) at their 1991 values to explain service specific spending in 1992. Variables included in the model correspond to information individuals are assumed to be able to use to predict spending. We estimate two sets of regressions, one with age and sex as right-hand variables and one with age, sex, and prior spending. The estimated coefficients from each pair of service specific regressions are then applied to 1992 values of the right hand side variables to generate estimates of expected spending for each individual.
Following Duan et al. (1983) and Manning et al. (1981), each two-part model is specified as:
where i indexes the individual enrol lee, X is a vector of individual characteristics (either age, sex, or age, sex, and prior use), p is a vector of coefficients to be estimated and e is a random error term. payday loan lenders online