Employment-Related Health Insurance: Federal Agencies' Roles in Meeting Data Needs

Abstract

Employer-sponsored health insurance accounts for almost one-third of all health care spending. As health care cost growth accelerates affecting the availability of employer-sponsored insurance and depth of coverage, the importance of timely and accurate information for measuring and monitoring these changes and formulating policy options increases. Identifying a growing gap between the need for and availability of data to inform policy on employment-related health insurance issues, the Office of Management and Budget (OMB) established a committee of Federal agency representatives to evaluate and advise data collection efforts. This article reports on the committee's current efforts, focusing on evaluation of results from the Medical Expenditure Panel Survey-Insurance Component (MEPS-IC) and the National Compensation Survey (NCS).

Introduction¹

The United States Federal statistical system is highly decentralized. Many of the Nation's statistical programs are carried out by agencies such as the Bureau of Labor Statistics (BLS) and the National Center for Health Statistics who have as their principal mission the collection, analysis, and dissemination of official statistics. In other cases, agencies such as CMS carry out statistical activities in conjunction with primary missions to provide services or enforce regulations. Coordination across these diverse agencies and their statistical programs is the responsibility of OMB's Statistical Policy Office, which sponsors a number of interagency groups to focus on specific areas of activity that cut across agency lines.

A case in point involves surveys on employment-based health coverage. The results from these surveys are used in many ways, most notably in measuring the growth and structure of the economy, assessing changes in the compensation of employees, and addressing public health concerns. While these statistics provide a wide variety of information about health insurance availability, options, usage, benefits, costs, funding methods, impacts, and participating entities, it has become increasingly clear that substantially improved coordination of these data collections is essential.

In spring 1998, OMB created the Interagency Committee on Employment-Related Health Insurance Surveys (hereafter referred to as the committee) to coordinate statistical activities across the agencies involved in this area. The work of this committee serves as an excellent model for effective interagency cooperation and coordination and provides a useful illustration of how agencies can work together both to fulfill their own data needs as well as address the needs of the broader data user and policy communities. The improvements already resulting from the committee's work will align survey data elements, concepts, and definitions to facilitate analyses of employer-provided health benefits and other forms of non-wage compensation across series. Coordinating surveys also has the potential to reduce respondent burden and conserve funds by eliminating redundant requests for information. Producers and users of data stand to benefit from reviewing the activities the committee has undertaken to improve data on employer-related health insurance, the substantial accomplishments already achieved, and the challenges remaining.

Background

Employer-provided health insurance covered approximately 62 percent of Americans in 1998 (U.S. Bureau of the Census, 2001). Employers and employees spent $357 billion in premiums for this coverage (Agency for Healthcare Research and Quality, 2001; U.S. Office of Personnel Management, 2001), which in turn paid for almost one-third of all health care.² Overall health care expenditures are expected to more than double by 2008 (Heffler et al., 2001). Escalating health care spending potentially places future employer-sponsored health insurance coverage at risk for the Nation's working and retired population as some employers eliminate coverage (especially for retirees), introduce more restrictive plans, or transfer more costs—which may not be affordable—to workers and retirees.

These risks put employer-sponsored health insurance at the center of health policy debates. In turn, high priority is placed on Federal Government efforts to insure that information to support employer-provided health insurance policy decisions is available when needed. As a result, several Federal agencies currently sponsor or conduct surveys that collect data on employment-based health coverage; at the same time, during the past decade policymakers and researchers increasingly have called for additional data to inform understanding and decisions.

The committee was formed to address these gaps and to better coordinate existing survey efforts. This group represents Federal agencies from the Departments of Labor, Health and Human Services, Commerce and Treasury, Small Business Administration, and OMB that provide and/or use health care data. These agencies were to review existing statistical needs, programs, data products, and uses and identify opportunities for improvement in these statistics. Members were to assess barriers (legal, cost, and feasibility, among others) to implementing identified options for improvement and make practical recommendations for developing and enhancing available statistics. In that sense, the committee provides a forum for Federal agencies to consider and recommend collaborative efforts that will improve employment-related health insurance statistics and related data collection programs carried out by Federal agencies.

Beginning with its inception in 1998, the committee investigated and reported on a variety of employer-sponsored insurance data issues. These include an inventory of private health insurance data sources and documentation of their characteristics; development of uniform insurance definitions and concepts that the Federal Government will use consistently when reporting major health care information; refinement/coordination of the abstraction of information from employer insurance documents (including plan brochures) to reduce future duplication of effort; comparisons of different agency's employer survey response rates; and efforts to share committee results with the research community.

In this article, we summarize the results of the committee's efforts to determine employment-related health insurance data needs and gaps. We also evaluate recently released results from the Agency for Healthcare Reseach and Quality's (AHRQ's) MEPS-IC and assess the ability of MEPS-IC and BLS' NCS to meet identified data needs. We will conclude with recent reactions by members of the Federal Economic Statistics Advisory Committee (FESAC) to the work of the committee.

Employer Health Insurance Surveys

Federal experience in collecting data on employment-related health insurance has spanned many decades and several departments. The Department of Health and Human Services (DHHS) (formerly the Department of Health, Education, and Welfare) first began collecting such data in the Social Security Administration's survey of “Voluntary Insurance Against Sickness” in the early 1940s. Eventually, agencies within DHHS collaborated to merge several data collection efforts into the MEPS-IC.

Paralleling DHHS efforts, the Department of Labor, through the BLS Employee Benefits Survey, has provided information on the availability and characteristics of employer-sponsored health benefits since 1979. In 1986, BLS began producing estimates of employer costs for health insurance benefits.³ These efforts have since been combined into the BLS's NCS, which provides information on all employee compensation, including the cost, availability, and characteristics of employer-sponsored health benefits.

The two major surveys of employment-related health insurance—the DHHS' MEPS-IC (first fielded to collect data for 1996) and BLS NCS—have been the major focus of the committee's work over the past 2 years. While many other Federal household health care provider and insurance surveys collect limited information on employer-sponsored health insurance, these two focus heavily on this topic and are the sources of premium cost, information in CMS' National Health Accounts (NHA) and the Bureau of Economic Analysis' (BEA) National Income and Product Accounts (NIPA).

The MEPS-IC is an annual survey of employers that collects detailed insurance information by mail; 41,000 sample establishments are selected to represent the entire Nation and to support sub-national estimates. At present, estimates can be made for the 40 largest States and most large metropolitan areas. Data are collected for all employees in surveyed establishments. Data elements include the number of employees eligible for health insurance coverage, the number currently enrolled, and the total cost of providing health insurance benefits. The number and types of plans offered by each employer in the sample are also available. For each plan with enrollees, questions are asked about the number of enrollees, the employer and employee premiums, and details of plan provisions (such as deductibles and covered services).⁴ The results from MEPS-IC become available approximately 24 months after the close of the calendar year.

The NCS is an umbrella survey that captures data used to produce several quarterly and annual outputs. These include ECI—measuring the change in employer costs for wages and benefits; Employer Costs for Employee Compensation—measuring employer costs for wages and benefits per hour worked; and the Employee Benefit Survey—measuring the availability and detailed provisions of benefit plans. While statistics from these NCS component surveys are currently available as separate products from BLS, the goal of the NCS is to link these outputs. (For example, separate data are currently published on employer and employee premiums for health insurance. However, these data are not linked in such a way to allow tabulation of the proportion of total premiums paid by the employer and employee.) Such data will be available over the next few years as BLS continues to develop the NCS (Wiatrowski, 2000).

Sample establishments in the NCS are selected from within a sample of metropolitan and non-metropolitan areas to represent the entire Nation, as well as regions of the country. The initial collection of data is conducted by personal visits to establishments by BLS field economists; subsequently, telephone and mail is used to obtain updated data. Within an establishment, data are captured for a sample of occupations designed to represent all occupations in the establishment. There are currently about 18,000 establishments surveyed, with estimates produced for the Nation and for broad geographic regions. Data elements include the number of employees currently enrolled in a health insurance plan. For each plan with enrollees, questions are asked about the number of enrollees, the employer and employee premiums, and details of plan provisions. Similar data are also captured for a number of other benefits; wage data are captured as well. Some results are available as early as 1 month or as late as 24 months after the close of the quarter.

A key difference between the surveys is the concentration on health insurance in MEPS-IC in contrast with the broader view of compensation captured by NCS. Because MEPS-IC focuses solely on health insurance, more detailed questions about each health insurance plan can be asked. Conversely, NCS asks fewer questions about health insurance while attempting to capture a broader, more complete picture of all employee compensation (wages and salaries, overtime pay, vacation benefits, workers' compensation, life insurance, etc.). This survey trades off indepth information for breadth of benefits and more timely release of certain information.

The committee recognized these fundamental differences early in its work. While maintaining a limited number of similar variables, each survey provides important dimensions of information on the provision of health insurance—dimensions that are invaluable for analyzing trends and evaluating policy issues. With this recognition, the focus of the committee work has been to consider whether data needed from the two surveys have been captured, and to compare similar data elements in order to understand differences that may be found in survey results.

A major issue with both the NCS and MEPS-IC surveys involves the availability of information to researchers.⁵ Because information from specific establishments is protected by confidentiality restrictions, no public use files at the establishment level are directly available to researchers. MEPS-IC information is available at a data use center in Rockville, Maryland, where non-identifying information for approved research projects may be accessed. Similarly, BLS confidential data files, including those from NCS, are available for statistical research projects only at the BLS National Office in Washington, DC.

Data Needs and Gaps

The next section of this article discusses the committee's efforts to identify employer-sponsored health insurance data needs and gaps. For purposes of analyzing data needs and gaps, employer-sponsored health insurance information that is required to inform policymaking was divided into three broad groupings:

• National Accounts—Measures used in formal accounting structures (such as the NIPA and the NHA) to create aggregate statistics about the overall economy or the health care sector.

• Monitoring and Trends—Information used to track changes that identify potential problems and to evaluate the effect of policy changes and changes in the labor and health care services markets.

• Research and Policy Analysis—Data, often at the micro-analytic level, needed to understand determinants of behavior and to estimate the potential impact of policy changes.

National Accounts

Two formal Federal accounting structures that supply important information on economic activity overall, and health care specifically, require information on employer-sponsored health insurance. These are the NIPA compiled by BEA and the NHA compiled in CMS. Both agencies have similar needs that include the measurement of aggregate premiums paid for health insurance coverage of workers; the amounts paid by employers and employees separately toward those premiums; and the net cost of insurance—the difference between premiums and benefits. (Similar information is also required for the NHA and NIPA on premiums, benefits, and net cost for individually purchased plans not sponsored by employers.) In addition, the NHA uses information on benefits paid under these plans by service type and requires similar State level information for its State health accounts. Each agency requires national accounts information within 6 months of the close of the calendar year.

Prior to the release of MEPS-IC information, a mutually exclusive and exhaustive set of estimates on private health insurance spending was difficult to create with any confidence. Data to meet these national accounting needs came from Federal Government surveys of employers, providers, and households, as well as from private data collections on insurance industry premiums and benefits. Data from these sources were often not compatible, resulting in gaps and overlaps when combined.

Trends and Policy Analysis

Federal agencies need timely data that allow them to describe, monitor, and understand trends in the availability, enrollment, cost, and characteristics of employer-sponsored health insurance coverage. Agencies involved in research and policy analysis need data to understand the dynamics of employer-sponsored health insurance coverage, measure the impact of policy proposals on costs and coverage, and evaluate the effects of major policy changes. For both types of needs, timing is critically important to enable policy analysts to confidently understand and address evolving issues in the provision of private health insurance through employers.

To address key issues in monitoring and trends and in research and policy analysis, the most important employer-supplied data are premiums for health insurance coverage (total monthly or annual employer, and employee share); employee offers; eligibility and enrollment; and information on the scope and content of the benefits provided.

Results from MEPS-IC and NCS

The availability of MEPS-IC results in early 2000 along with evolving plans for improvement in NCS led to much discussion on ways to evaluate the results from these surveys and whether data needs would be met through these two surveys. The committee has concentrated on evaluating new and revised data from these sources—specifically aggregate premiums, employer-paid premiums, employer-paid share of premiums, and employee enrollment and participation rates. In general, MEPS and the fully implemented NCS should be able to provide data to meet accounting data needs on a regular basis. In addition, the consistency found among similar items in the NCS and the MEPS is encouraging, although some areas of inconsistency do exist, as described in the following sections.

Aggregate Annual Premiums

Measures of aggregate insurance premiums required by NIPA and NHA include premiums for employment-related health insurance as well as for plans purchased separately by individuals. At this level of aggregation, private health insurance premiums estimated from a variety of provider, employer, household, and insurer surveys can be compared (Table 1). For 1996 and 1997, similar estimates of total health insurance premiums are generated from MEPS-IC, NCS, insurance industry data, and NHA, although MEPS data are somewhat higher than the other sources in 1998.

Table 1. Aggregate Annual Employer-Sponsored and Individually Purchased Premiums, 1996-1998.

Estimate Source	1996	1997	1998
	In Billions
MEPS1	$346.4	$345.8	$392.8
Aged BLS 1992 Survey Results 2	341.1	347.7	373.3
Insurance Industry Data 3	356.9	352.4	378.0
NHA Provider-Based Estimate 4	334.9	346.7	375.0

¹Includes private- and government-sector employer- and employee-paid premiums for hospital and physician plans and for single service plans and individually paid premiums from MEPS household survey.

²BLS 1992 survey aged using growth in employer costs from BLS Employer Costs for Employee Compensation and Consumer Expenditure survey results.

³Includes information from InterStudy, Blue Cross Blue Shield, and A.M. Best.

⁴Estimated using data from the 1996-1998 Census Service Annual Survey and the insurance industry.

NOTES: MEPS is Medical Expenditure Panel Survey. BLS is Bureau of Labor Statistics. NHA is National Health Accounts. MEPS and Aged BLS 1992 Survey results include U.S. Office of Personnel Management estimates of Federal agency expenditures for employee and annuitant health care coverage.

SOURCES: Bureau of Labor Statistics: Data from the Employer Costs for Employee Compensation Survey; Agency for Healthcare Research and Quality: Data from MEPS; and the Centers for Medicare & Medicaid Services: Data from the National Health Accounts.

Employer-Paid Premiums

The premium cost for employer-sponsored insurance is usually shared between the employer and employee. Estimates of the aggregate portion of premiums paid by the employer (employer-paid premiums) in 1997 are less comparable among survey sources and by industry than are total premium (employer-paid plus employee-paid portions) measures. While most industries report similar levels of employer-paid premiums in both surveys, there are two notable exceptions: construction and retail trade (Table 2).

Table 2. Total Private-Sector and State and Local Government Employer-Paid Premiums for Active Workers, 1997.

Industry	NCS	MEPS	(MEPS-NCS)/NCS
	In Billions		Percent Difference
Total Private	$160.4	$175.5	9
Mining	1.6	1.7	5
Construction	12.5	6.1	-52
Manufacturing	48.2	46.5	-4
Wholesale Trade	14.6	14.5	-1
Retail Ttrade	13.4	19.9	48
Transportation, Communication, and Utilities	13.0	17.2	32
Finance, Insurance, and Real Estate	14.2	17.1	21
Services	42.8	52.5	23
State and Local Government	49.1	35.6	-27

NOTES: NCS is National Compensation Survey. MEPS is Medical Expenditure Panel Survey. MEPS data exclude single-service plans, such as prescription drug or dental plans. NCS represents previously unpublished estimates derived from the Bureau of Labor Statistics, Employer Costs for Employee Compensation Survey.

SOURCES: Bureau of Labor Statistics: Data from the Employer Costs for Employee Compensation Survey and the Agency for Healthcare Research and Quality: Data from the MEPS-Insurance Component.

Some of the underlying differences in the MEPS and NCS cost aggregates can be attributed to different estimates of employment resulting from the use of different sample frames of establishments for each survey (Table 3). (When reweighted to account for differences in industries covered by each survey, aggregate private industry employer-paid premiums are very similar between the two surveys in 1996 and 1997, although industry-specific differences remain.) The NCS establishment frame excludes agriculture, membership organizations, and some other small sectors of the economy that are covered in the MEPS sample frame.

Table 3. Civilian Employment by Industry—Active Workers, 1997.

Industry	NCS	MEPS	(MEPS-NCS)/NCS
	In Millions		Percent Difference
Total	116.9	123.4	6
State and Local Government	17.1	16.7	2
Private Industry	99.8	106.7	7
Agriculture	NA	1.9	NA
Mining	0.6	0.6	0
Construction	5.2	5.4	4
Manufacturing	18.4	18.8	2
Wholesale Trade	6.6	7.0	6
Retail Trade	21.5	21.8	1
Transportation, Communication, and Utilities	6.4	6.4	0
Finance, Insurance, and Real Estate	7.0	8.1	16
Services	34.2	36.6	7

NOTES: NCS is National Compensation Survey. MEPS is Medical Expenditure Panel Survey. NCS data are previously unpublished estimates derived from the Bureau of Labor Statistics, NCS data relate to March 1997; MEPS data are for a typical pay period in 1997. NA is not available.

SOURCES: Bureau of Labor Statistics: Data from the Employer Costs for Employee Compensation Survey and the Agency for Healthcare Research and Quality: Data from the MEPS-Insurance Component.

In addition to the employment differences, the period for which the premium data were gathered can also account for a portion of the discrepancy between the MEPS and the NCS. The estimates from NCS use current (March) premiums and enrollment data and annual work schedules to estimate aggregate costs when available; when current rates or enrollment are not available, prior period expenditures are used as a proxy for current costs. Therefore, NCS estimates will not reflect increases to plan costs that occur after the current cost is reported in March.

MEPS estimates use current premium and enrollment data for a typical employee for a typical payroll period in the year. (MEPS uses the following procedures to capture data for the typical employee: In private sector and non-certainty [randomly selected] government establishments, data are requested for the typical premium for both single coverage and family-of-four coverage for up to four plans in the establishment. Typical is generally the value for the largest group or the average value. If there are more than four plans, the same data are captured for the three largest plans and for one additional plan selected by probability sampling. For certainty governments—about one-half the government enrollment whose large size ensures their inclusion in the sample—these data are collected for all plans. Data are also requested regarding any variations—for example, by age, sex, and/or length of service— in premiums and the details of those premiums. About 3 percent of establishments indicate that premiums vary.)

Other differences may be related to dissimilarities in sample frame draw and replenishment procedures, sample size or sample design. For example, NCS excludes those costs for owners, retirees, and recipients of continuing health benefits under the Consolidated Omnibus Budget Reconciliation Act of 1985 (COBRA) that are collected in MEPS.

Employer Premium Shares

Until recently, the main source of employer/employee share data was a special 1992 BLS survey of employer-provided health care expenditures that was used as a basis for estimates in several NHA reports. The release of the 1996-1998 MEPS adds an additional, more recent source of information. Estimates of employer shares from the 1992 BLS survey are noticeably higher than those measured by MEPS in 1996 and 1997 (Table 4). Some of the difference is likely due to changes in the premium share arrangements over time. Other private surveys that measure the employee/employer shares using methods similar to those in the MEPS support the 1996 and 1997 MEPS estimates of shares.⁶ They also generally confirm the declining portion of cost borne by the employer between 1992 and 1996, although the amount of difference over time is smaller in the private surveys than is indicated by the share difference in the 1992 BLS and 1996-1997 MEPS-IC survey results.

Table 4. Employer-Paid Share of Employer-Sponsored Health Insurance Premiums, by Selected Years.

Industry	1992	1996	1997	1998
	BLS	MEPS	MEPS	MEPS
	Percent
Total	86	77	78	77
State and Local Governments	85	81	83	82
Private Industry	86	76	76	76
Construction	89	76	75	75
Manufacturing	89	82	79	83
Wholesale Trade	86	78	73	75
Retail Trade	77	71	71	69
Transportation, Communication, and Utilities	93	83	85	84
Finance, Insurance, and Real Estate	81	71	74	74
Services	84	75	71	73

NOTES: BLS is Bureau of Labor Statistics. MEPS is Medical Expenditure Panel Survey. MEPS information is for active workers only.

SOURCES: Bureau of Labor Statistics: Data from the Expenditures for Health Care Plans by Employers and Employees, 1992, USDL 93-560, 1993 and the Agency for Healthcare Research and Quality: Data from MEPS-Insurance Component.

Different survey methods and potential survey error are other reasons cited for dissimilar results between the 1992 BLS survey and the later MEPS-IC. The 1992 BLS survey was designed specifically to collect separate annual employer and employee expenditures for health insurance for the purpose of estimating employer shares. Collecting expenditure data required more contacts with employers than are usually made in BLS survey collections, thereby raising costs and lowering response rates.

On the other hand, MEPS estimated employer shares using employer and employee premium and enrollment data collected for each plan for typical employees in a typical payroll period.⁷ The consistency of the MEPS data for 1996 and 1997 seems to rule out problem response patterns or reporting errors in the plan premium rate and enrollment data.

NCS will provide yet another source for measuring premium shares and tracking trends in the future, as its survey questions and databases are being modified to allow future tabulation of the share of health insurance costs borne by employers and employees.

Employee Participation Rates

NCS and MEPS estimates of employee participation rates are similar, except for government and the private sector construction and retail trade industries (Table 5). Also similar are participation rates estimated at the national level using NCS and MEPS for small private establishments (those with fewer than 100 employees) in 1996 (49 and 51 percent, respectively) and for medium and large private establishments (those with 100 or more employees) in 1997 (67 and 69 percent, respectively).

Table 5. Percent of Active Employees Participating in Health Care Benefits, 1996-1998.

Industry	MEPS			NCS
	1996	1997	1998	Total	19961	19972
State and Local Government3	61	66	64	79	NA	NA
Private Industry	60	57	57	57	49	67
Agriculture	40	31	35	NA	NA	NA
Mining	84	75	84	NA	NA	NA
Construction	47	44	44	61	56	78
Manufacturing	78	79	81	79	70	84
Wholesale Trade	72	72	72	73	74	73
Retail Trade	41	37	41	32	26	44
Transportation, Communications, and Utilities	75	77	74	73	63	80
Finance, Insurance, and Real Estate	75	70	68	75	73	77
Services	55	52	51	51	45	57

¹Survey of small establishment with fewer than 100 employees.

²Survey of medium and large establishment with 100 employees or more.

³NCS State and local government estimate total is for 1996.

NOTES: MEPS is Medical Expenditure Panel Survey. NCS is National Compensation Survey. Estimates by industry for NCS are previously unpublished and, due to the small sample sizes, may be subject to higher than normal variance. NA is not available.

SOURCES: Bureau of Labor Statistics: Data from the Employee Benefit Survey and the Agency for Healthcare Research and Quality: Data from the MEPS-Insurance Component.

MEPS-IC employee participation rates in employer-provided health plans exhibit some instability over time at the industry level. Among the largest changes are those in the divisions with the smallest samples—agriculture and mining—indicating that this volatility may be due to sample noise rather than changes that are occurring in participation rates. Despite this instability within individual sectors, the private sector participation estimate remains stable at 57 percent in 1997 and 1998.

The NCS sample contains fewer establishments than the MEPS sample (Table 6) (8,000 for 1996 small establishment survey and 1997 medium and large establishment survey in NCS, and 35,000-38,000 for 1996-1998 MEPS). In addition, the longitudinal sample design of the NCS will not represent establishment births as efficiently as do the annual samples drawn for MEPS. Thus, more mature establishments, which are thought to offer health insurance more often and have increased employee participation rates, will be more prevalent in the NCS sample than in the MEPS-IC. In industries with more frequent business startups and failures, this effect will be exacerbated. Though this outcome might explain in part the much higher NCS participation rate for construction, it does not account for retail trade differences.

Table 6. Sample Size and Response Rates for the Medical Expenditure Panel Survey-Insurance Component, 1996-1998.

Category and Industry	1996	1997	19981
Number of Establishments
State and Local Government	1,946	1,913	NA
Private Industry	34,945	35,839	38,093
Agriculture	1,568	1,232	1,345
Mining	204	136	198
Construction	3,128	3,038	3,164
Manufacturing	2,965	3,012	3,733
Wholesale Trade	2,240	2,412	2,306
Retail Trade	8,217	8,197	8987
Transportation, Communications, and Utilities	1,608	1,606	1,731
Finance, Insurance, and Real Estate	2,889	3,021	2,948
Services	12,126	13,185	13,681
Number Responding
State and Local Governments	1,735	1,636	NA
Private Industry	23,452	25,635	27,657
Agriculture	995	868	864
Mining	137	91	141
Construction	2,069	2,187	2,190
Manufacturing	2,026	2,229	2,849
Wholesale Trade	1,350	1,580	1,570
Retail Trade	5,509	5,797	6,614
Transportation, Communications, and Utilities	1,051	1,119	1,236
Finance, Insurance, and Real Estate	1,900	2,095	2,137
Services	8,415	9,669	10,056
Response Rate
State and Local Governments	0.89	0.86	NA
Private Industry	0.67	0.72	0.73
Agriculture	0.63	0.70	0.64
Mining	0.67	0.67	0.71
Construction	0.66	0.72	0.69
Manufacturing	0.68	0.74	0.76
Wholesale Trade	0.60	0.66	0.68
Retail Trade	0.67	0.71	0.74
Transportation, Communications, and Utilities	0.65	0.70	0.71
Finance, Insurance, and Real Estate	0.66	0.69	0.72
Services	0.69	0.73	0.74

¹1998 data not yet available for State and local governments.

NOTE: NA is not available.

SOURCE: U.S. Bureau of the Census tabulations from the Medical Expenditure Panel Survey.

NCS will produce enrollment estimates for the entire private sector (regardless of employment size) with reference years beginning in 1999. The first such estimates were released in late 2001. Some detail by major industry division will be provided; additional detail may be available over time.

Are Data Needs Met?

MEPS-IC and NCS have supplied information that will address most national accounts needs. For monitoring and trends and research and policy, gaps in needed data remain.

National Accounts

Federal Government (largely through the new MEPS-IC and NCS) and private surveys or administrative records that are available from employers, households, health insurers, and/or providers currently meet national accounts needs for employer-sponsored private health insurance. NCS and MEPS-IC are used most effectively in combination: MEPS premiums in the most recent year available can be moved forward to more current periods using data from NCS. MEPS will also fill previous State health accounts data gaps on employer- and employee-paid expenditures for employer-sponsored private health insurance information by State,⁸ along with counts of enrolled workers.

While BEA has adopted the aggregate MEPS estimates in the NIPAs, CMS has used information from MEPS, along with other sources, in preparing the NHA. MEPS data have improved the estimates of the aggregate level of expenditures—particularly the expenditures for the self-administered, self-insured component of the market where data has not been available for many years—although the year-to-year growth rates in aggregate expenditures remain subject to large variation. This may result from refinements in the questionnaire over time and is expected to become more reasonable in the future. More frequent than annual estimates from NCS will provide an additional source of information on the annual growth rate estimates. NCS's estimates of the employer and employee share of premium costs should also be helpful. Finally, NCS will soon provide the more timely (available one-quarter following the survey date) estimates of costs that cannot be produced by MEPS. In summary, we expect NCS to result in substantial improvement in the aggregate national account data available on employment-related health insurance, just as MEPS has already.

Trends and Policy Analysis

Even with the availability of MEPS-IC, several key types of descriptive information are missing for monitoring and trends purposes (in priority order):

• Purchasing mechanisms (for example, multiple employer welfare arrangements and health insurance purchasing cooperatives) and the reasons for their use or non-use by small employers.

• Greater detail on cafeteria plan provisions and other fringe benefits that would allow for calculation of the value of these alternatives to employees.

• Coverage by firm size for the Health Insurance Portability and Accountability Act monitoring purposes.

• Prevalence of consumer-protection provisions within health plans offered to employees.

• Number of contract and contingent workers and their eligibility for health insurance and other benefits.

• Annual revenue as an additional establishment-size measure for smaller firms.

• Premiums and plan enrollment by worker type (including COBRA beneficiaries and retirees).

• Net cost (the difference between the premium and the benefits paid in return for that premium) included as part of health insurance premiums.

Information on these items would greatly improve our ability to describe and track change in employer-provided health benefits at a level of detail useful to Federal agencies for informing policymakers.

For research and policy analysis, several key pieces of information are missing, cannot be linked to other necessary data sources, or are inaccessible in micro data form to the people who need it. In priority order, these gaps include:

• Information about the workforce (including wages, age and health status) from the same establishment survey providing premium cost and employee participation rates.

• Identification of both who in the organization is responsible for determining which health insurance plans are offered and where in the organization (at the establishment or at the firm level) this decisionmaking takes place.

• Factors related to decisionmaking on self-insurance (e.g., understanding of workforce health risk and its effect on premium rates).

• Detailed information about other employee benefits.

• Linkability to information on State-specific policies and regulations.

• Linkability to information on local labor, health insurance, and medical market conditions.

• Risk status of employees in individual establishments.

• Insurer requirements, particularly for employers who sought/investigated coverage but turned down insurer quotes.

In addition to the data gaps previously specified, the current system of employer surveys lacks a quick response mechanism to obtain information relating to an emerging policy issue or to establish a baseline in advance of the implementation of a new policy.

Priorities for Filling Gaps

The main data gap that continues to exist for national accounting is the measurement of aggregate benefits paid.

The highest priority is given to items that would greatly enhance our ability to measure, describe, and monitor changes in employer-sponsored insurance. Since respondent burden is an issue when new items are added to an existing survey, consideration is generally given to whether old survey items should be dropped in order to accommodate new items. We recommend evaluations of MEPS and NCS to determine which survey could most easily incorporate the elements needed to fill our high priority gaps. In addition, a review and evaluation should be undertaken periodically to identify those items whose importance and/or quality has diminished and whose usefulness is less than a question to be added.

Among the gaps identified for research and policy analysis, the highest priority items are workforce characteristics, more detailed information on cafeteria plan provisions, and locus of health insurance decisionmaking.

Workforce characteristics (for example, age and health status of the group to be insured) constitute the biggest gap in data needed to understand the determinants of the cost of health insurance and firms' decisions about offering insurance. Since employer-sponsored health insurance is offered to the group, information about key characteristics of that group will likely affect the cost and desirability of health insurance. Such information would greatly enhance our ability to understand employers' decisions about whether to offer health insurance, what types of health plans to offer, and whether to self-insure.

Research Issues Related to Data Gaps

Some data gaps exist because it is not clear how to collect the needed data. Additional investigation is required to answer such questions as how best to gather information on workforce characteristics and on benefits paid out under employer-sponsored plans; and how to make better use of the full range of public and private surveys and of administrative data such as the Internal Revenue Service Form 5500 (Annual Return/Report of Employee Benefit Plan) filed by most employers providing self-insured health benefit plans.

Workforce Characteristics

Data on workforce characteristics would assist researchers in answering questions about the effect of group characteristics on the availability, cost, and design of fringe benefits from different employers. Extracting information from employers that they may not be able or willing to give presents great challenges for workforce surveys.⁹ However, work in progress by the U.S. Bureau of the Census matching employer and household data across large surveys shows promise for meeting some health data needs and should be explored.

Data on Benefits Paid

While employers generally know the cost or premium of plans offered to employees, they generally do not have access to information about payout for benefits under those plans, unless the employer is self-insured.¹⁰ Such data may sometimes be available at the firm level, but not at the specific surveyed local establishment. An additional complication exists for benefits paid in some types of health maintenance organizations in which billing transactions for individual services may not occur. As a result, benefits, and therefore net cost for specific plans, are usually not known by the employer. This gap hampers the analysis of insurance companies' micro-level responses to experiences of individual employers; it also hampers the measurement of aggregate benefits paid for NIPA and NHA purposes.

Comparisons with Non-Federal Survey Data

Questions repeatedly arise about the comparability of private¹¹ and public sector surveys of employer-sponsored health insurance. To enhance the understanding of data users and to provide a common platform from which to work, it would be advisable periodically to prepare a comprehensive research document to educate the user community about attributes and comparability of employer-based public and private health surveys. Comparisons would be helpful because data from non-Federal sources often provide detailed information quicker than do current Federal surveys.

Form 5500 Series Report

Most health plans that provide employer-sponsored insurance must file an annual Form 5500 series report with the Internal Revenue Service. Despite certain limitations, including the exemption of plans with fewer than 100 employees and data that are not edited for accuracy or internal consistency, it would be useful to explore tabulation of selected information from the Form 5500. At the U.S. Bureau of the Census, work is ongoing to link these data to household surveys as a crosscheck on benefit coverage. At the same time, it would be worthwhile to see what, if any, changes might be made to the Form 5500 that would either reduce processing costs or provide more relevant information.

The provision of health insurance in the United States is constantly evolving. As new methods for insuring health care are tried in the public or private sector, new questions arise, and new types of data are needed to understand these changes. Nevertheless, we need to retain the capability to analyze trends over time that in turn depends on stable survey questions and methods. Thus, the committee should explore methods for developing a quick-response mechanism to add to the current portfolio of surveys. BLS is exploring such a mechanism, by examining methods to collect answers to simple questions and disseminate establishment data within a year of collection as part of the activities of the NCS¹² or of a subsample of the Consumer Expenditure Survey. At the same time, the committee is exploring other options to achieve quick responses to important and emerging issues.

Federal Economic Statistics Advisory Committee

A report to the Federal Economic Statistics Advisory Committee (FESAC) summarizing the major work of the committee was presented in June 2001. FESAC is a technical committee composed of private sector economists, statisticians, and behavioral scientists who are recognized for their attainments and objectivity in their respective fields. Committee members analyze issues involved in producing Federal economic statistics and recommend practices that will lead to optimum efficiency, effectiveness, and cooperation between BLS, the Bureau of the Census, and BEA. Members of FESAC commented on the Committee's efforts to date, noting that its work sends a very important signal of the desire to improve information, to resolve discrepancies in results of different surveys, and to share data collection knowledge among agencies.

FESAC members noted that many critical health care policy decisions are being driven by best guess estimates because data are either unavailable or not available in a timely fashion. While improvements in collecting information on private health insurance have been made in the last 10 years, decisionmakers are often making decisions based on outdated information.

Simple, real-time data are needed from quick response surveys. FESAC members cited several scenarios where this is critical. For example, quick response surveys could provide more timely information on current insurance experiments such as the shift to three-tier prescription drug benefits. In addition, some States, in the absence of major changes in Federal policy, are initiating prescription drug programs to fill an important gap in Medicare coverage for the elderly and disabled. As a result of these policy changes, to what extent are employers dropping post-retirement benefits? Further, would these actions affect the employer-related health plan provisions of the near elderly who are anticipating retirement?

Other questions center on the growing role of insurance, and the amount of money spent on prescription drugs, which has doubled since 1994-1995. How much of insurance growth is due to a tight labor market in which employers use desirable prescription drug benefits to attract and retain workers? Would employers shed benefits if labor markets were to loosen?

Timely data are not currently available to address these issues. The committee can help to identify such critical data gaps and improve the timeliness of future data. They also need to evaluate how useful old data are in assessing current issues and, at the same time, ascertain the ability of existing surveys to bridge new data gaps as they develop.

FESAC members suggested that future data collection efforts should follow a broad strategy of diversification. The portfolio should include large integrated (linkable) data sets as well as quick response surveys that can provide information on emerging issues. FESAC suggested that opportunities to integrate surveys be more fully explored. For example, matching data from employers and households is one way to link workforce characteristics to employer provided information about health insurance. Whether this is accomplished by using employers as the base to select households, or households as the base to select employers will be determined by the specific questions to be answered.

The committee was cautioned against viewing micro and aggregate data needs as competing interests. Rather, these are overlapping concerns. It should also consider accessing State resources, in addition to consulting with representatives from the employer community. Broader perspective from non-Federal Government sources would allow the committee to better understand issues and incorporate responses to these issues in Federal surveys.

In summary, FESAC recommended that the committee continue to allow some surveys to perform specific tasks, while seeking a broader portfolio of survey instruments. They urged the creation of large integrated data sets to respond to issues that single surveys alone cannot address, while also encouraging the creation of quick response surveys to produce timely information. Failure to modify surveys to keep abreast of the changing insurance environment will generate new gaps to replace those just filled, causing the state of insurance information to revert to the condition it was in during the early 1990s.

Next Steps

The committee is charged with aligning current and future data to meet national accounts, monitoring, and research needs. Through efforts to date, those needs have been more clearly focused and articulated, and the available data have been scrutinized to ensure conformance. Data gaps have been identified, and changes have been made or planned that can help to fill these gaps. However, more work is required to fill existing research and monitoring data gaps. Identifying the best methods for capturing quick response information on emerging issues for policymakers is currently being explored and will present additional future challenges.

Incorporating views of other health care data constituents—private researchers, academics, insurers, States, employers, and more—into the committee's deliberations will enhance its ability to meet future data needs of a wider range of users. The continually evolving health insurance sector and policy initiatives responding to those changes guarantee sustained work for the committee.