Estimates of the uninsured population range from 21 million to 40 million, depending on the definition.(1) Over 60 percent of adult nonelderly individuals who were uninsured cited the inability to get health insurance from employers as a reason for the lack of health insurance and 83 percent among those who were uninsured for twelve months or more. This highlights the importance of taking a closer look at the characteristics of firms that do and do not provide health benefits as one component of research on the uninsured population.
Existing data provide important insights on employer provided health benefits, particularly the percent of workers who have health benefits available through their employers and the percent who participate. However, we have much less detail available on the firms themselves particularly how these firms benefit offerings evolve over time, how costs of health benefits and cost-sharing with employees is changing within firms, and whether those firms employees are aware that their employers offer benefits.
This study examines a new data set constructed by the Longitudinal Household Dynamics Program (LEHD) of the Census Bureau, and it assesses its potential not only to answer questions like these but also to triangulate evidence on employer provided benefits with other data sources.(2) The new data enhance existing administrative records on benefit plans filed by firms improving the match between plans and the individual employers who offer them and tracking those employers offerings over time. The enhanced data also contain linkages to SIPP and CPS survey responses for the employees who work within these firms.
Because these data offer significant enhancements to firm-level data that were not previously possible, and because they are part of a larger data set already being explored within an ASPE-funded project, ASPE felt it was worthwhile to conduct a preliminary assessment of the datas potential value for informing research questions on employer-provided health benefits. As these data have not yet been fully analyzed, it is important to assess their quality and representativeness as a first step in determining whether they merit further exploration, and that is the primary purpose of this report. It is well-known that the stand-alone administrative data on employer filed health benefits (the 5500 data) have several limitations, but it is hoped that the enhancements resulting from the linked file explored in this study can overcome some of these problems and support informative analysis on several key questions regarding employer provision of health insurance.
We first explore what proportion of the health benefit plans on file in the 5500 data can accurately be linked to specific employers. Second, we examine how the proportion of firms offering health plans observed in these data coincides with what we observe in other data sources. Next, we look at individual workers and whether their employment by firms that have and have not filed health plans is consistent with their self-reports of access to health benefits as reported in the CPS and the SIPP. Finally, we discuss the types of questions that these data could inform, and offer some descriptive findings on changes in employer health benefits over time as an illustration of what the first steps of such a research agenda might look like.
We find that while the linkages to the Census Business Register did increase our ability to match the health plans on file in the 5500 data back to individual employers, the resulting count of employers that provide health benefits is substantially smaller than that observed in data drawn from worker and employer based surveys. Thus, the data are not well suited to measuring the prevalence of employer offerings of health plans.
Nonetheless, the patterns of health plan offerings by firm age, size and industry appear similar to those identified in other data. Offer rates are much higher among older and larger firms and within particular industries consistent with those seen in other data. In addition, we see an encouraging level of concordance between employer filings of health plans within the 5500 data and survey responses of their workers regarding access to employer provided health benefits. The instances of discordance may themselves be informative, as additional analyses indicate that some such cases may actually reveal a significant share of workers that are unaware of health benefits available on the job.(3)
While we do understand that the Form 5500 health data do not cover all existing health plans, we believe that these findings provide a preliminary yet promising indication that the 5500 data are representative of the firms that provide health plans, at least among large firms. Thus, these data could contribute much to our understanding of the types of firms that offer benefits and the dynamics behind those offers over time. Additional analyses, proposed below, could provide a more definitive assessment of the datas representativeness and appear merited based on these preliminary results.
In summary, these data provided a useful triangulation with several other data sets regarding the availability of employer provided health plans. More importantly they offer the potential for a new analysis of the way in which changes in firms and firm characteristics affect the availability of employer provided health benefits over time. The new types of questions that could be answered with these data include:
- how does economic turbulence new firms entering a sector, existing firms dying, or changes in existing firms affect the benefits offered in a sector?
- do changes in worker reports of changing levels of employer benefits accurately reflect changes in employer offerings?
- how much of increases in employers health insurance costs are passed along to workers and
- how much do such increases affect participation?(4)
In our example of descriptive analyses on the evolution of employer provided health plans between 1997 and 2001, we find some interesting results. In particular, we find that there are differences by industry, age and size one example being that more large firms ended health plans than began them over the five year span, resulting in an overall decline in health plans among large firms. Expanding these analyses to address questions like those posed above could greatly increase our understanding of the factors underlying these changes.
1. How Many People Lack Health Insurance and For How Long? CBO, May 2003
2. A similar integration of the Form 5500 pension benefits data and the Census Business Register data was explored previously through funding by the Office of Policy and Research at the Department of Labor.
3. Other reasons for discordance are misunderstanding of the question or the unwillingness to answer correctly by the respondents
4. In particular, the richness of the Census Business Register together integrated with the information of the Form 5500 data could help answer questions a) and b), and some information on the Form 5500 and the SIPP and CPS surveys could be used to answer questions c) and d).