HISP impact assessment: comparison between registered and unregistered households

After reflecting more closely on the before-and-after comparison, the evaluation team concludes that there are still numerous factors that may explain part of the change in health expenditures over time (specifically, the ministry Finance is concerned that a recent financial crisis has affected household incomes, and may explain the observed change in health spending.)

Another consultant suggests that it would be more appropriate to estimate the counterfactual in the period after the intervention, that is, two years after the start of the program. The consultant points out that of the 4,959 households in the sample, only 2,907 enrolled in the program, so that about 41% of households still do not have HISP coverage.

The consultant maintains that all households in the 100 pilot villages

they fulfilled the conditions to register. These households share the same health clinics and are subject to the same local prices for pharmaceuticals. Furthermore, most of the members of these households work in similar economic activities. The consultant is of the opinion that, in these circumstances, the results of the non-enrolled group after the intervention could serve to estimate the counterfactual outcome of the group enrolled in the HISP. Therefore, it decides to calculate the average health expenditures in the post-intervention period, both for households that enrolled in the program and for those that did not (the results are shown in Table 3.3). Using the average health expenses of unregistered households as the counterfactual estimate, the consultant concludes that the program has reduced average health expenses by almost US $ 14.46.

Now, households that decided not to enroll in the program

Can they be systematically different from those that did? Perhaps the households that enrolled in the HISP had higher health expenditures or were people with more information about the program or more caring for their family's health. Another possibility is that perhaps the households that did enroll were poorer, on average, than those that did not enroll, since HISP targeted poor households. The consultant ensures that regression analysis can control for these potential differences between the two groups. Therefore, it performs another multivariate regression that controls for all the characteristics of the home that can be found in the database, and estimates the impact of the program.

With a simple linear regression of health expenditures on a variable indicative of whether or not a household enrolled in the program, it is possible to find an estimated impact of US $ -14.46, that is, that the program has decreased the average of health expenses in US $ 14.46. However, when controlled for all other data characteristics, the program is estimated to have reduced health expenditures by $ 9.98 a year.

The question is:

A. Does this analysis control for all the factors that determine the differences in health expenditures between the two groups?

B. Based on the results produced by the enrollment-no enrollment method, should HISP be expanded to the national level?