In an earlier paper (Haber, 1989), logit regression was used to analyze response rates to a special survey, namely, the Survey of Income and Program Participation (SIPP) Fringe Benefits Survey. The goal of this survey was to determine the feasibility of collecting information from employers about their contributions for health insurance, life insurance, and private pension plans on behalf of specified individuals in their employ. Interest in this objective stems from the continuing effort by the Census Bureau to improve its estimates of income by including in income the value of government and private in-kind transfer payments (see U.S. Bureau of the Census, 1988).

For reasons given in the earlier paper, only the health insurance information in the Fringe Benefits Survey (denoted below as the FBS to distinguish it from the SIPP survey itself were examined in detail. The major conclusion of that study was that, with minor changes in questionnaire design, it is feasible to collect employer health insurance cost data. This conclusion, however, was based only on considerations of response rates and on internal consistency with other information collected in SIPP but not found in the special survey.

Still to be examined is the question of whether or not the health insurance cost figures provided by employers are consistent with other empirical data and intuition, insofar as intuition helps in distinguishing between reasonable and unreasonable cost estimates. To answer this question, an empirical regression model is estimated from the usable data in the FBS. Based on the regression model, employer health insurance contributions are imputed for individuals with similar characteristics in the much larger parent SIPP sample for which health insurance cost data are absent. These estimates are then summed over all individuals and the total compared with the national income estimate of employer contributions for group health insurance. The major finding of this research is that although the regression sample size is small, the estimated total of employer contributions is fairly close to that reported in the national income accounts. The remaining findings pertain to variations in employer contributions among different groups of workers and the impact of rising health care costs on employer contributions for family plan coverage.

In Section 1, additional background information is provided about the FBS. A description of the regression model for estimating employer contributions for health insurance is found in Section 2. The results of applying the model are given in Section 3. In Section 4, the estimate of the amount employers contribute for health insurance is benchmarked against the national income figure. Section 5 contains estimates of employer health insurance contributions by worker and other characteristics based on the model. In Section 6 we discuss the question of which workers will be most likely affected by rising medical care costs. A short summary of the paper is given in the last section.