This Article Extract Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to Citation Manager Reprints & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Acs, G. Articles by Blumberg, L. J. Search for Related Content PubMed PubMed Citation Articles by Acs, G. Articles by Blumberg, L. J. Related Collections Insurance - Employer-Based System

Health Tracking

TRENDS

How A Changing Workforce Affects Employer-Sponsored Health Insurance

Many factors affect the probability that a worker has employer coverage. For example, workers in small firms, young workers, and lower-paid workers are all less likely to have it than are older, higher-paid workers in larger firms.³ Lower employer coverage rates result from declines in offer and/or take-up rates.⁴ While research shows that changes in the workforce cannot explain all of the changes in employer-sponsored coverage, we can anticipate how certain changes in the distribution of workers across jobs and in the demographic composition of the workforce may affect this coverage in the future.⁵

Other factors—most notably, changes in the level of health insurance premiums relative to income and the share of premiums paid directly by workers—will also affect the probability that employer coverage will be offered and taken up. Although we do not have direct measures of either premiums or workers’ share of premiums, we do include variables for firm size and industry in our models; firm size and industry are highly correlated with variations in workers’ premium shares.⁶ Additionally, future expansions or contractions of eligibility for public programs may also interact with the probability of having employer coverage; these types of unforeseen changes are not captured here.

Data And Definitions

 
For the bulk of our analyses, we use the February 1997 Contingent Worker Supplement (CWS) to the Census Bureau’s Current Population Survey (CPS) merged with data from the 1997 March CPS. The March survey adds information on workers’ earnings and family incomes. Our sample, therefore, consists of those workers who are included in both data sets.

The February 1997 CPS/CWS contains data on a nationally representative sample of workers and reports information on their employers’ as well as on their own characteristics. Unlike the March files, these supplements ask whether the respondent’s employer offers health insurance and whether the worker is eligible for coverage in addition to the type of coverage held. The supplement is the source for the variables used in these analyses to identify each worker’s coverage. The February questions related to workers’ coverage refer to current coverage.

We draw a sample of workers (excluding the self-employed) ages twenty-one to sixty-four and examine their access to employer coverage and whether they take up benefits when offered. Note that our analysis excludes the self-employed, all nonworking dependents, and younger workers who may have coverage through their parents’ policies.

Forecasting Coverage Rates

Top Data And Definitions Forecasting Coverage Rates Conclusions And Policy... NOTES

 
Recent trends. In 1993, 68.6 percent of workers had employer coverage in their own name; by 1997 coverage fell to 67.8 percent.⁷ The share of workers covered through spouses fell from 12 percent to 11 percent. Thus, on net, coverage fell from 80.6 percent to 78.7 percent over the four-year period.⁸

Forecast method. We use a multivariate regression framework to estimate the probability of insurance coverage for all adult workers; we then use results from a series of regressions to forecast coverage rates under a set of assumptions about the future composition of the workforce. This approach allows us to control for changes in multiple characteristics simultaneously. We take into account the probability of offer and take-up of coverage from one’s own employer and through a spouse’s employer (if any).

Estimating the multivariate model. We estimate the likelihood of offer and take-up on using linear probability models. Offer and take-up models are estimated separately for workers who are married to other workers (dual earners) and for workers who are single or have nonworking spouses (single earners). We take this approach because the decision to take a job in a firm is strongly related to the offer status of spousal coverage. Likewise, the propensity to take up insurance may be affected by the options available to the worker through the spouse’s employer.

Our models consist of six estimated equations, which are described in detail elsewhere. ⁹ Briefly, we estimate (1) the probability of a worker’s being offered coverage by his or her own employer (estimated separately for single earners and dual earners); (2) the probability of a worker’s taking up his or her own employer’s coverage given the existence of an offer (estimated separately for single earners and dual earners); (3) the probability that a worker’s spouse is offered coverage, given that the worker does not have his or her own coverage and is a dual earner; and (4) the probability that a worker is covered by his or her spouse’s employer coverage, given that the worker does not have coverage through his or her own employer and has a working spouse with a coverage offer.

Total employer coverage for workers, then, is a weighted average of the coverage rates of workers in dual-earner and single-earner families. The rate of own-employer coverage is simply a product of offer and take-up, while spousal coverage must take the spousal offer and take-up into account as well as the probability that the worker does not have an offer of coverage from his or her own employer.

Our equations include both firm- and worker-specific information. This includes the worker’s occupation, sex, race/ethnicity, age, marital status, education, wage rate, income relative to the poverty line (in the case of take-up), region of residence, job tenure (in the case of offer), and part-time status, as well as employer’s industry and firm size.

To test the predictive accuracy of our model, we used our estimated equations to generate a "backcast" of coverage for workers in 1993 (using the April 1993 CPS). This yielded a predicted coverage rate within 0.2 percentage points of the actual 1993 rate.¹⁰

Projecting workforce characteristics. To project the distribution of workers across occupations and industries as well as the sex, race/ethnicity, and age composition of the workforce, we use two approaches. The first uses information from the U.S. Bureau of Labor Statistics (BLS). Every two years the BLS issues its employment outlook for the coming decade, providing forecasts of the occupational, industrial, and demographic structure of the workforce. The most recent BLS projections are for the year 2008.¹¹ We use the average annual rate of change in characteristics implied by the BLS projections to compute changes for our sample of workers.

To project most other workforce characteristics, we use data from the CPS. Specifically, we examine how the distribution of workers across firm-size groups, wage-rate categories, family income categories, marital status, and single-versus dual-earner households changed between 1993 and 1997. We then extrapolate these trends to 2008.¹² Finally, to project the share of workers who will be working part time in 2008, we examine the trend observed in the Survey of Income and Program Participation (SIPP) between 1990 and 1996 and assume that this trend continues between 1997 and 2008.¹³

One concern with this approach is that we use the same forecast rates of change for workforce characteristics in all six of our models. Consequently, we developed alternative predictions that allow us to generate independent projections for single earners, dual earners, and dual earners without coverage from their own employer. This model must rely solely on CPS data because, not surprisingly, the BLS does not provide projections for the industries and occupations of each unique set of workers. We assume that the CPS trends between 1993 and 1997 continue through 2008.

Results. Our first model, using the combination of BLS, CPS, and SIPP trends, indicates that anticipated changes in the composition of the workforce, on net, will drive down workers’ employer-sponsored coverage from 78.7 percent in 1997 to 77.6 percent in 2008. This translates into a coverage decline of 1.3 million workers relative to coverage rates’ staying constant.¹⁴ Exhibit 1 reveals the patterns underlying this modest forecast decline. The model predicts that coverage through a worker’s own employer will fall for workers in both single-and dual-earner households and that coverage through a spouse’s plan will also fall. We forecast that both the offer rate and the take-up rate will decline among workers in single-and dual-earner households; however, the offer rate effect is considerably stronger. Coverage through a spouse is predicted to fall by the smallest percentage, with all of the effect coming through a decline in the offer rate.

Exhibit 2 allows us to see which sociodemographic and workforce characteristics contribute the most to the predicted changes in coverage under each model.¹⁵ For the first model, the largest absolute effects come from work status and wages, both tending to decrease coverage over time. The number of part-time workers is predicted to increase between 1997 and 2008, bringing with it lower coverage rates. Changes in the wage distribution are predicted to lead to more higher-and lower-wage earners (below $7.50 and above $12 per hour in 1997 dollars), with a smaller share of the workforce falling in the middle of the distribution. The change with the largest positive effect is an increase in education. Also, a positive coverage effect comes from the predicted increase in dual-earning couples.

Conclusions And Policy Implications

Top Data And Definitions Forecasting Coverage Rates Conclusions And Policy... NOTES

 
We find that the aging of the workforce, increases in the marriage rate, changes in the occupational distribution, and continued increases in educational attainment are all likely to increase rates of employer-sponsored coverage. Changes in the racial/ethnic makeup of the workforce, increases in the share of part-time workers, changes in the wage distribution, and the increasing proportion of workers employed in smaller firms are likely to depress rates of coverage. On net, our forecasts suggest that changes in the composition of the workforce are likely to lead to lower employer coverage rates between 1997 and 2008. We anticipate that the overall decline in coverage resulting from changes in the composition of the workforce will be relatively small.

Our models indicate that both offer and take-up rates of employer coverage will tend to decline in the coming years, with the percentage- point declines in offer rates tending to be the larger factor. The single most important detrimental trend affecting offer and take-up is the rising share of part-time workers in the labor force. Part-time workers are far less likely than full-time workers are to be offered employer coverage, and their often modest incomes make the employee share of any offered coverage less affordable.

Taking these conclusions into account, number of public policy approaches may be considered. Financial incentives could be provided to encourage more employers to offer coverage. However, the effects of such an approach are likely to be uneven, leaving many without a viable source of coverage. Tightening Employee Retirement Income Security Act (ERISA) regulations regarding which workers must be offered benefits may well bolster the situation of part-time workers in the employer coverage system.

States and the federal government also might consider developing alternative stable and comprehensive sources of coverage for workers and their dependents. Permitting buy-ins to Medicaid and/or the State Children’s Health Insurance Program (SCHIP) one option. Another possibility is the development of organized purchasing entities for those in the nongroup private insurance market, coupled with reform of the insurance regulations in this market. Such efforts would be directed toward making the nongroup market easier to navigate and better able to serve those with higher-than-average health risk, and providing individuals with the benefits of administrative economies of scale and purchasing power.

Public policies focusing on the affordability affordability of the employee share of employer coverage premiums may be able to counteract the predicted decline in take-up. This might require subsidies to low-and moderate-income workers for this purpose. The political difficulty with such an approach is the trade-off between policies that provide subsidy dollars to those who probably would have taken up coverage without the subsidy and policies that exclude those with prior coverage while subsidizing others with the same incomes. Public costs will be lower under the latter approach, while the new coverage effect will, in all likelihood, be significantly larger under the former. Finally, a much more limited and low-cost approach to expanding employer coverage involves public information campaigns. These campaigns could target growing subpopulations, such as Hispanics, who tend to have lower take-up rates.

While our forecast declines over an eleven-year period are not alarming in and of themselves, it is important to keep in mind that they assume that the probability of coverage for any given subgroup remains at its 1997 level. However, if health insurance premiums begin to rise dramatically, as they did during the early 1990s, the probabilities that a specific worker will be offered coverage and will take up benefits are both likely to decline. Similarly, if unemployment rises, firms may not feel a need to offer health insurance to attract workers. If employer coverage among workers falls substantially, coverage of dependents is also likely to fall, and current public health insurance programs will be unable to fully compensate.

Editor's Notes

 
Gregory Acs and Linda Blumberg are senior research associates at the Urban Institute.

This research was funded by a grant from the Commonwealth Fund. The authors thank John Holahan, Len Nichols, and anonymous reviewers for their helpful comments, and Daniel McKenzie and Tracy Roberts for research assistance. The opinions expressed herein are those of the authors and do not necessarily reflect those of the Commonwealth Fund, the Urban Institute, or its funders.

NOTES

Top Data And Definitions Forecasting Coverage Rates Conclusions And Policy... NOTES

 

1. J. Holahan and J. Kim, "Why Does the Number of Uninsured Americans Continue to Grow?" Health Affairs (July/Aug 2000): 188–196. Indeed, the declining rate of employer coverage over the past two decades has been well documented. See, for example, G. Acs, "Explaining Trends in Health Insurance Coverage between 1988 and 1991," Inquiry (Spring 1995): 102–110; P. Fronstin, Sources of Health Insurance and Characteristics of the Uninsured, EBRI Issue Brief no. 217 (Washington: Employee Benefit Research Institute, 2000); R. Kronick, "Health Insurance 1979–1989: The Frayed Connection between Employment and Insurance,." Inquiry (Spring 1991): 318–332; and E. O’Brien and J. Feder, Employment-Based Health Insurance Coverage and Its Decline: The Growing Plight of Low-Wage Workers (Washington: Kaiser Commission on Medicaid and the Uninsured, 1999).
2. Fronstin, Sources of Health Insurance.
3. L. Nichols et al., Small Employers, Their Diversity, and Health Insurance, Urban Institute Monograph (Washington: Urban Institute, 1997).
4. Recent studies attempt to decompose the decline in employer coverage into its components of offer and take-up of coverage. See P.F. Cooper and B.S. Schone, "More Offers, Fewer Takers for Employment-Based Health Insurance: 1987 and 1996," Health Affairs (Nov/Dec 1997): 142–148; and H. Farber and H. Levy, "Recent Trends in Employer-Sponsored Health Insurance Coverage: Are Bad Jobs Getting Worse?" Journal of Health Economics 19, no. 1 (2000): 93–119.
5. Acs, "Explaining Trends in Health Insurance Coverage."
6. L.M. Nichols and L.J. Blumberg, "Modeling Decisions to Buy Private Health Insurance," Final Report to the U.S. Department of Labor, Pension, and Welfare Benefits Administration (Washington: Urban Institute, forthcoming).
7. Our data on workers from 1993 come from a file merging information on workers in the March 1993 CPS and the April 1993 Survey of Employee Benefits supplement to the CPS. Data on 1997 workers are from the February and March 1997 CPS files.
8. In 1997 another 1.4 percent of workers were covered as dependents under employer-sponsored plans by someone other than a spouse.
9. G. Acs and L.J. Blumberg, "Changes in the Composition of the Workforce and Their Effects on the Future of Employer-Sponsored Health Insurance," Urban Institute Report prepared for the Commonwealth Fund (Washington: Urban Institute, 2000).
10. The adjusted R² values ranged from .29 to .31 for the offer equations and from .08 to .12 for the take-up equations.
11. U.S. Bureau of Labor Statistics, BLS Looks Ahead to 2008, vol. 122, no. 11 (Washington: BLS, November 1999).
12. While we control for job tenure in our model estimation, there were no data on tenure in the 1993 survey. Consequently, our projections implicitly assume that tenure stays constant.
13. We use SIPP data rather than CPS data because projections of the share of workers working part time in 2008 based on the trend observed in CPS data between 1993 and 1997 imply that approximately 50 percent of the workforce would be working part time by 2008, a projection that seemed implausibly large. The sensitivity of the model to this assumption is made clear when this base forecast is compared with the alternative forecast. Actual workforce characteristics for 1997 and predicted characteristics for 2008 appear in Acs and Blumberg, "Changes in the Composition of the Workforce."
14. This assumes 113.6million workers ages twenty-one to sixty-four at a point in time in 2008.
15. In more basic shift-share models focusing solely on coverage, we find similar patterns of results. See Acs and Blumberg, "Changes in the Composition of the Workforce."

                      What's this?

This article has been cited by other articles:

				J. D. Reschovsky, B. C. Strunk, and P. Ginsburg Why employer-sponsored insurance coverage changed, 1997-2003. Health Aff., May 1, 2006; 25(3): 774 - 782. [Abstract] [Full Text] [PDF]