BLS compensation programs: what will users need?

An academic analyst speculates on the future needs of labor-management practitioners, academic researchers, and government policymakers for BLS compensation data

To help mark the Monthly Labor Review's 75th year, the editors asked both data users and data producers to speculate about programs and data needs in 2015, when the Review will mark its centennial. This article, and the one beginning on page 38, deal with the Bureau's compensation programs.

Data on wages and compensation often have been less visible than data on price inflation and unemployment.(1) During the 1990's and beyond, however, changes in compensation systems may well play a critical role in reconciling conflicting pressures in the American labor market. This will make it crucial for the Bureau of Labor Statistics to monitor and disseminate compensation data.

Although I am a frequent user Of BLS data on compensation, I am not in a good position to know about the cost/benefit side of data collection. Judgments about that are made by the Bureau and the political process. My role here is to put forward user preferences, against a backdrop of current BLS compensation programs and likely changes in the labor market.

Who needs what?

BLS currently operates nine primary programs which gather compensation information as shown in exhibit 1. The data produced by these programs are of varying degrees of interest to three constituency groups. (1) practitioners who set pay (managers, unions, and sometimes neutrals), (2) academics (researchers), and (3) government macro policymakers (along with private forecasters). Although their data needs overlap, their demands for further compensation information differ in many ways. Thus, BLS, faced with resource constraints, must make decisions concerning competing needs.

Practitioners' views are officially presented tO BLS by union and management advisory committees. Federal Government policymakers are obviously in position to make their needs known and to influence resource allocation.(2) Academics lack formal channels of input and, of course, have no direct control over resources. They have, nonetheless, been active supporters Of BLS programs, particularly when budget cuts have been threatened.(3) Development of formal communications between BLS and academic users Of BLS compensation data would assist in balancing competing demands from users.

Here are ways in which each program might be improved from the viewpoint of their primary constituencies, especially in light of changing compensation practices and changing information technology.

Practitioners. In stylized terms, practitioners primarily want information on who is being paid what. Surveys of compensation-setting behavior suggest that a first step is commonly to find out what other employers (within similar industries or locations) currently are paying related groups of workers.(4) Information so obtained is not necessarily slavishly applied; the ultimate pay decision might be to pay more or less than some perceived going rate or market average. But knowledge of the outside market is a starting point in the decision process.

To be most helpful to the practitioner, surveyed compensation information must first be detailed. This involves disaggregation by occupation, location (or labor market), and by the type of pay practice under which payment is made. Apart from detail, there is the issue of frequency of data collection and the speed of publication. Information on wage rates paid a year or two ago, even if provided on a detailed basis, will be of limited interest. Finally, of relevance to practitioners is the intent of other firms regarding future compensation decisions. Unfortunately, BLS does not survey salary intentions, a significant gap in its compensation data base.(5)

Of the three user categories, compensation-setting practitioners are most likely to want data on the traditional printed page; they do not feel a need to subject the data to further processing. However, practitioners will want to know of emerging trends for competitive reasons. Which of the nine programs in exhibit I are likely to be of most concern to practitioners? The Current Employment Statistics survey provides information on pay levels and trends by detailed industry, but does not include benefits and lacks straight-time hourly earnings estimates outside of manufacturing. This is troublesome, given the growth of benefits relative to wages since World War H and the growth of the service sector industries, which are less likely than manufacturing industries to have nationwide paysetting practices. On the plus side, earnings data from the establishment survey have a fast turnaround.

Current Population Survey data on usual weekly earnings potentially are available on a detailed occupational basis. (Most practitioners are probably unaware of this source of pay information which could be valuable, especially to those who need data on occupations with national labor markets.) The key issue is speed of publication. Much annual labor market information from the cps is available immediately after each year closes, and is published in the January issue of Employment and Earnings. Inclusion of detailed annual occupational earnings data on the same schedule would be valuable to practitioners (although practitioners' understandings of who is in a given occupation may not always be in accord with cps methodology). For practitioners in the union sector, the existing system of tracking major collective bargaining settlements (those involving 1,000 or more workers) is helpful. It provides relatively frequent data with quick turnaround. And it permits tracking of such items as the frequency of use of cost-of-living adjustments and lump-sum payments. The listings of contract settlements in Current Wage Developments provides the ability to track individual bargaining situations. Generally, it has been assumed that the major agreements set patterns for smaller units. However, significant divergences between major agreements within manufacturing were found when BLS kept track of the smaller settlements (late 1950's through late 1970,S).(6) Just as firm and establishment size seemed to shrink in the 1980's, so too did the number of workers covered by major agreements relative to overall union representation.(7) Major union settlements may thus have less importance, even within the union sector, than was once the case. Industry wage surveys provide substantial occupational detail and sometimes indicate generally what type of pay system is involved, time or incentive. Obviously, these surveys are of greatest potential use to paysetters within the covered industries. Unfortunately, the long intervals between surveys and the lag between collection and publication limit the usefulness of industry wage surveys to practitioners. Area wage surveys provide data only on certain widely used occupations. The surveys are taken regularly and have quicker turnaround time than do industry wage surveys. Both area and industry wage surveys provide information on the dispersion of pay, information of potential value to the firm in considering its pay policy relative to others in the labor market.

The Employee Benefits Surveys provide substantial information on the degree to which particular types of benefit programs cover the work force. Emerging benefit programs, such as profit-sharing and employee stock ownership plans are reported in the surveys, but more detail is provided about traditional pension and health care programs.

BLS does not survey employers concerning their intent to establish new pay and benefit plans or terminate old ones. Information on such plans, to the extent that it exists, has been provided by the private data collection firms, often as part of surveys with questionable sampling techniques. Sometimes such surveys are undertaken by organizations which advocate use of particular kinds of compensation programs. A well sampled, periodic survey of this type from BLS would be of great interest to paysetting practitioners.

In the future, practitioners will need more timely compensation data on occupation and labor market area. At present, the surveys which provide the most rapid turnaround tend to be those which give general compensation trends in the labor market, but which are not sufficiently detailed for many compensation-setting purposes. Those with great detail, such as the industry wage surveys, tend to be published with a considerable lag.

Academics. If practitioners need to know who and what, academics need who, what, and why. They are also interested in the effects of particular pay policies. Timeliness is less of an issue for academics, who are more likely to be concerned with completeness of information and the ability to link compensation data sets with other information on the firms or establishments supplying them. Alternatively, they are likely to need a combination of compensation data sets and information on the characteristics of the firms, establishments, or work forces from which they are gathered.(8)

Modem computer technology makes possible the linking of data sets, provided issues of confidentiality do not prove to be insurmountable hurdles.(9) Because of their interest in probing the reasons for and effects of compensation outcomes, academics are likely to want information do can be accessed by computers. Such dissemination facilitates statistical analysis of data sets. Although academics often have made use of earnings data from the establishment survey, questions of "why" (as opposed to "what" and "who") are not easily addressed from this source. Information is not available by establishment characteristics, such as size or union or nonunion status. Other than the categories of nonsupervisory employees and others, there is no occupational information.

The Current Population Survey has been used by academics for exploring such issues as racial or sex discrimination in pay, or union/nonunion pay differentials. An issue that will need to be addressed in the future, however, is the relationship between CPS-reported earnings and data from other earnings series. For example, the heavy volume of concession bargaining suggests that union/nonunion wage differentials fell during the 1980's. Although other data sources reflect the impact of concession bargaining, it has been, at best, weakly and unevenly reflected in cps data.(10) Because of the growing use of cps earnings data to research such policy issues as comparable worth, it would be helpful if BLS itself undertook research on the reliability of these data as trend indicators.

The Employment Cost Index also could be a more valuable source for academics. A limited sample of establishments is repeatedly surveyed, potentially allowing for longitudinal analysis. From the academic viewpoint, an ability to link compensation outcomes with establishment characteristics would be a boon for research. Similar issues can be raised with regard to area and industry wage surveys and the PATC survey.

Academics have devoted substantial research to union wage determination, even during the era of declining unionization. Among the reasons for this interest is the fact that the field of labor economics was very heavily focused on the union sector until the 1960's. Also, there is the advantage that the union sector continues to provide researchers with wage information because of its relative openness. One can still track union settlements through Current Wage Developments and use the employer's name to link to other data on the firm. The discontinued series of "wage chronologies" was helpful in pulling such information together for certain major firms. Access to a data base containing the history of settlements reported in Current Wage Developments could enable users to generate their own wage chronologies.

There are significant differences between union and nonunion pay practices, for example, regarding cost-of-living adjustments.(11) Academics would thus greatly benefit from a data set tracking nonunion pay adjustments comparable to those in the union sector.

From the Employee Benefits Survey, academics would want to find out why particular benefits programs were offered. Unfortunately, as currently structured, the survey does not provide information on the employers' costs of benefits. There are admittedly substantial difficulties with the measurement of employer costs, especially for pay practices such as defined benefit pensions in which unfunded liabilities may accrue. And there is a conceptual difference between cost to the employer and value to the employee. Nonetheless, the marriage of benefit/cost figures-such as those now available from the Employment Cost Index-with benefit/incidence data will be a boon to academic research. The development of these data presently is in progress.(12)

Although academic interests in the effects of pay practices often seem abstract to practitioners, such information could have direct payoffs for those in the human resource area. Involving the human resource area in the strategic plan of the employer became a popular notion in the 1980's, at least among human resource executives. However, to achieve involvement in the future, human resource executives will need evidence on how (or if) compensation-setting policy (and other issues) affects the economic performance of the enterprise.

Policymakers. Macroeconomic policy relies heavily on aggregated information which, in the compensation area, involves measures of labor cost. Accuracy is important, because critical economic policy decisions may be made based on the data produced. Quick turnaround time between data gathering and dissemination is essential as policy is updated.

In the past, macroeconomic studies of wage determination relied heavily on data from the establishment survey, because that was the main source available. Aggregate establishment data, however, vaned because of such factors as shifts in employment across industries and variations in overtime usage, along with adjustments in actual pay scales. From the macro viewpoint, the latter type of adjustment is most critical. In addition, benefit information was omitted.

The Employment Cost Index (ECI) has offered a better alternative for macro judgments in recent years than either establishment survey data or even the more comprehensive index of compensation per hour.(13) Indeed, a case can be made for computing unit labor cost trends utilizing the ECI rather than compensation per hour.(14) The ECI could provide still more useful information to policymakers in the future if data showing the dispersion of pay change were published along with the movement of the average. For example, a widening of the dispersion of pay decisions might indicate a growing diversity of labor market conditions, with some employers experiencing tight labor markets while others still operated in soft markets.

ECI data on benefits reflect only current employer expenditures, not necessarily the value of promised future benefits. In the area of pension plans, such values can be significant where there are unfunded liabilities. Although private sector pension plans are subject to the Employee Retirement Income Security Act (ERISA) regulations concerning funding adequacy, State and local government plans are not. While it would be difficult to reflect the true value of promised benefits in the quarterly ECI, periodic reports on the unfunded liability issue could be useful supplements. The expenditure versus value dichotomy will become progressively more important as the aging American work force nears retirement age.

Because of the shrinkage in the union sector, the heavy emphasis once placed on trends in union pay settlements by macro policymakers has been diminished. However, the visibility of collective bargaining ensures that there will continue to be some macro interest in the aggregate settlement data, and in particular settlements seen as bellwethers. From the macro viewpoint, the lack of base wage information for the individual settlements reported in Current Wage Developments has long been a problem; it is difficult to compute percentage pay increases for individual settlements from the cents-per-hour increments that are often reported.

Another longstanding problem has been the exclusion of possible cost-of-living adjustments from reported new settlements data. BLS understandably is reluctant to forecast inflation rates. However, the index might be reported with an adjustment assuming that the current inflation rate will continue, or a menu of inflation assumptions might be provided. (15) Consideration should be given to parallel treatment of other forms of contingent pay, notably profit sharing. To the extent that data on union settlements are made available for computer use, options might be provided for the user to plug in alternative assumptions about inflation and profitability.

The growth of lump-sum payments in the union sector has posed a similar difficulty. Given the frequency with which such payments are now made, lump sums can no longer be considered a temporary aberration. Indeed, one survey of larger firms suggests these bonuses are used in the nonunion sector almost as frequently as in the union sector.(16) BLS plans to reflect lump-sum payments in the collective bargaining statistics, average hourly earnings (from the Current Employment Statistics survey), and occupational wage survey programs.(17)

The BLS data on foreign labor costs in manufacturing are valuable tools for policymakers and economists interested in understanding swings in international trade competitiveness. Data are presented in indexes although the absolute values are also of great use. Because of the interest in contingent pay and lump-sum bonus arrangements in the United States, it would be useful to provide more information on this component of foreign pay.(18) The existence of large bonus payments relative to total compensation has been a noteworthy feature of pay practices in Japan and certain other industrial countries.

The lump-sum issue points to a more general need on the part of macro policymakers. Macroeconomics inherently involves the use of aggregate indexes and data. But exactly what should be included in the aggregate series can be debated. As both macro theory and institutional arrangements change, the kind of data demanded will also vary. Indeed, policymakers and analysts may require different aggregations. As in the case of academic needs, access to the data base can resolve the problem of changing demands in the future. With appropriate access, users can calculate customized indexes which meet their analytical requirements.

A future perspective Because of the difficulties in predicting how employers will compensate employees, the best approaches to gathering such data are those which preserve options for users. With regard to dissemination, the best approaches are open ones which provide users with the ability to tap into the data set within the limits of confidentiality. Even when resource constraints make publication of a BLS data set difficult, options should be preserved for private sector dissemination of related data. (19)

Footnotes

1 See Joseph P. Goldberg and William T. Moye, The First Hundred Years of the Bureau of Labor Statistics (Washington, Superintendent of Documents, 1985). In this official historical account, the authors devote substantially more space and attention to controversies and developments surrounding the Consumer Price Index and data relating to employment and unemployment than to compensation. A general history of BLS wage gathering and dissemination can be found in H.M. Douty, "A century of wage statistics: the BLS contribution," Monthly Labor Review, November 1984, pp. 16-28.

2 Thus, when the issue arose of publishing information on the absolute levels of wage and benefit costs as part of the Employment Cost Index, a combination of practitioner and government users were the main force in obtaining the new data. See G. Donald Wood, "A New Measure of the Cost of Compensation Components," Survey of Current Business, November 1988, p. 43. Academic researchers had long been interested in wage-versus-fringe tradeoffs, but their need for such data was frustrated when BLS Stopped producing an earlier series on wage and funge costs in the 1970's. The only private source of such data, an annual survey by the Chamber of Commerce of the United States, is not made available for academic use. Despite their concerns, academics had no formal avenues to express their continuing interest in data on wage and benefit costs.

3 Academic members of the Executive Board of the Industrial Relations Research Association (IRRA) made various efforts to have the sm take positions against proposed budget cutbacks affecting BLS. Because of the IRRA's tripartite structure, it was not possible to achieve consensus on this issue, but the organization has since maintained a statistical subcommittee to monitor budget and other developments affecting Federal statistical programs. (Records of the Board debates on this issue can be found in the association's 1981 and 1982 annual Proceedings volumes.) A related organization made up of major academic industrial relations programs, then known as the IR Center Directors, took a more active role in making contact with congressional representatives and staff and administration officials.

4 One study found that 93 percent of respondent employers reported using wage surveys as part of the paysetting process. Only 34 percent of those using such surveys reported that they used BLS data, perhaps because of some of the problems related to speed and detail discussed in this article. See Bureau of National Affairs, Wage & Salary Administration, PPF Survey, 131 (Washington, Bureau of National Affairs, 1981), p. 3.

5 There are precedents in other settings for data collection about intentions by government agencies. For example, information is collected about intended future investment outlays, and persons not in the labor force are asked about their future jobseeking plans.

6 In the 1970's, median percentage wage adjustments in manufacturing were generally higher for major union agreements than for all union agreements. Thus, the nonmajor agreements apparently provided smaller adjustments than the major agreements. See Handbook of Labor Statistics, Bulletin 2070 (Bureau of Labor Statistics, 1980), pp. 30607.

7 BLS estimates the number of workers represented under major union agreements as part of its annual bargaining calendar. An estimate of total union representation is made as part of the Current Population Survey. In 1980, the ratio of workers under major agreements to those who were union represented was 60 percent in the private nonagricultural sector. By 1988, the ratio had fallen to 52 percent.

8 Alice Nakamura and Masao Nakamura, "New Measures of Nonwage Compensation Components: Are They Needed?" Survey of Current Business, March 1989, p. 6 1.

9 Because of budgetary pressures, concern has been expressed concerning duplication of surveying efforts by different government agencies. If these difficulties can be overcome, one byproduct could be more information about the surveyed establishments which would be linked to their pay practices. See the statement of Courtenay Slater in U.S. Joint Economic Committee, The Quality of the Nation's Economic Statistics, hearings of March 17 and April 17, 1988, 99th Cong., 2d session (Washington, Superintendent of Documents, 1986), pp. 27-28, 83-84.

10 Union workers' wage and salary increases reported by the Employment Cost Index have been lower than those of nonunion workers since 1983. Yet the CPS-reported ratio of union to nonunion usual median weekly earnings for full-time employees has shown comparatively little change during the 1980's. For example, the cps ratio for the private nonagricultural sector rose from 34 percent in 1984 (the first year available) to 36 percent in 1986, and then dropped to 33 percent in 1988. During this period, the ratio as calculated from the Ea dropped steadily and showed a decline of more than 5 percent from June 1984 to June 1988. For discussion of the lack of evidence of concession bargaining in cps data, see Richard B. Freeman, "In Search of Union Wage Concessions in Standard Data Sets," Industrial Relations, Spring 1985, pp. 131-45.

11 William M. Davis and Fehmida Sleemi, "Collective bargaining in 1989: negotiators will face diverse issues," Monthly Labor Review, January 1989, p. 14.

12 The first set of data, relating to establishments with fewer than 100 employees, is scheduled for publication in the summer of 1991. Publication of data on State and local government is scheduled for the fall of 1991, and for the entire economy (private industry and State and local government), 1992.

13 To assist in the evaluation of establishment survey hourly earnings data, BLS published an hourly earnings index through 1988 which adjusted for interindustry employment shifts and overtime in manufacturing. The ECI is intended as a replacement for the hourly earnings index and the latter is no longer published.

14 Because the currently used measure of output per hour reflects changes in the mix of industry output, it might be desirable to match the ECI with a fixed-base index of output per hour. The two indexes in combination would permit calculation of a consistent measure of unit labor costs.

15 This suggestion is made in Donald A. Nichols, "Wage Measurement Questions Raised by an Incomes Policy," in Jack E. Triplett, ed., The Measurement of Labor Cost (Chicago, University of Chicago Press, 1983), p. 461.

16 See John Thomas Delaney, David Lewin, and Casey Ichniowski, Human Resource Management Policies and Practices in American Firms, reprint series (Columbia University, Industrial Relations Research Center, Graduate School of Business, 1988), p. 22.

17 Under special agreement with the aerospace industry, information on such topics as lump sums has been provided recently on that industry in both the establishment survey and the ECI.

18 BLS does have information on this component of pay, although it has not been prominently featured when the foreign pay data have been disseminated.

19 For example, the Industrial Relations Center for Cleveland State University has revived the discontinued BLS bulletins on characteristics of major collective bargaining agreements-which contained information on a variety of paysetting and other practices-using BLS contracts collected by the center. See "Characteristics of Major Private Sector Collective Bargaining Agreements as of January 1988," Report 8801-1 (Cleveland, Cleveland State University, Contract Library and Information Services, Industrial Relations Center, May 1989).

Exhibit 1. Nine major BLS compensation programs

1) Current Employment Statistics survey. A monthly establishment survey providing information on average hourly and weekly earnings of production and nonsupervisory workers by detailed industry, with some manufacturing industries by region and metropolitan areas.

2) Current Population Survey (CPS). A monthly household survey conducted by the Bureau of the Census for BLS. It now provides quarterly data on usual weekly earnings with some occupational detail and demographic characteristics (race, age, and sex). Detailed occupational information is available on an annual basis. Annual data also are available by union status with limited industrial and occupational information for full-time workers.

3) Employment Cost Index ECI). A quarterly survey of the rate of change in compensation per hour, which includes wage and salary and benefits costs with a union/ nonunion breakdown and some occupational and industrial detail. ECI information recently has been extended to include actual (dollar) costs of wages and various broad benefit categories by union/nonunion status.

4) Compensation per hour. A quarterly series of total labor costs per hour (including benefits and employer-paid payroll taxes) linked to related productivity and unit labor cost information. As published, the series does not separate wages from other forms of compensation. Information is available only for very broad industrial sectors.

5) Major union settlements. Quarterly survey of private industry, and semiannual survey of State and local government union-management agreements involving 1,000 or more workers. Related data on individual settlements and deferred and cost-of-living adjustments are also published.

6) Area and industry wage surveys. A program of periodic collection of occupational data for selected urban areas and industries. Area wage surveys are conducted annually or semiannually and provide data on a limited number of occupations. Industry wage surveys provide substantial occupational detail, with some regional breakdown, on a 3- or 5-year cycle.

7) National Survey of Professional, Administrative, Technical, and Clerical Pay (PATC). An annual survey of white-collar salaries in private industry for selected occupations, by work level within each field. This survey is designed primarily for guidance in setting Federal civil service salaries.

8) Foreign hourly compensation costs. An annual survey of pay of manufacturing production workers in the United States and in selected industrial countries, and of related information on productivity and unit labor costs.

9) Employee Benefits Survey. An annual survey of certain benefit practices in medium and large firms. Data are presented by broad occupational categories. The private sector is surveyed in even years and State and local government, in odd years.