The accurate measurement of employer-provided health and pension plan coverage is critically important for state policy. However, acquiring such data from employer surveys is not only expensive, but also imposes a substantial response burden on them. We propose to investigate the potential for using an administrative file - the 5500 file - to avoid such burden and cost.

The 5500 file, which contains data from annual returns filed annually with the Internal Revenue Service by all employers covered by ERISA, provides information on employee benefit plans, including data on different types of pension and health insurance plans. State partners receive a detailed report on the uses of this file, demographic characteristics of workers covered by these benefits, and if the data support it, summary data on the kinds of firms that offer different types of health benefit and pension plans.

Summary

Our match between the Form 5500 data and the Census Bureau's Business Register - which results in a match rate of about 95% - provides a reasonably good coverage of pension plans.

We use this combined data-base to address three questions:

1. What are the differences between firms that provide benefits and firms that don't - and what are the differences in their workforces?
2. What is the relationship between wages and benefits?
3. What is the relationship between the provision of benefits and subsequent firm outcomes?

Our first set of results confirmed findings from earlier work - firms that provide benefits tend to be larger and are more likely to be in manufacturing and wholesale trade. We used new measures to confirm other evidence that firms that offer benefits are better able to attract higher skilled, prime-age workers and have lower turnover (after controlling for size and industry). However, we also found that firms that offered benefits paid their employees more than those same employees would earn with the average non-benefit-offering firm-so workers appeared to earn both higher wages and better benefits than did observationally equivalent workers who worked for non-benefit offering firms.

Our last set of results was particularly interesting. We find that firms that offer benefits are less likely to fail - even after controlling for all other observable characteristics - than are firms that do not offer benefits. Many interpretations could be put on this. One is that of endogeneity - firms that are more likely to die (either due to current financial problems, or perhaps because they are an inherently more risky business) are less likely to offer benefits. This could either be as a way to cut down on current costs, or because workers value benefits less when the risk of future default is higher. Another possibility is that not enough firm-level controls were included.

These results are preliminary, so their robustness is one area for further work.

More Information

Documents