Implementing the Levitan Commission's recommendations to improve labor data.
Five years have proven a very short period for making changes in the statistics. The Bureau of Labor Statistics has implemented a number of the major recommendations that were approved by the two Secretaries of Labor who have served during this period, and continues testing and developing programs leading to the implementation of others. Some recommendations were found to be either impractical or too costly. Still others may be reexamined when the next review panel is convened. This article summarizes the accomplishments of the BLS in implementing several of the recommendations and in conducting activities preparatory to the adoption of others. It does not cover all the recommendations directed at the BLS or those relating to programs of other agencies.
The National Commission on Employment and Unemployment Statistics was not the first outside review panel to examine the Nation's labor force data system. Earlier examinations took place in 1948, 1954-55, and 1961-62. The most well-known of these was the 1961-62 review panel, officially named the President's Committee to Appraise Employment and Unemployment Statistics, but commonly known as the Gordon Committee after its chair, Professor Robert A. Gordon. The Gordon Committee made a number of recommendations for experimentation, sharpening of concepts, and expansion of data from the BLS major surveys.
One of the basic issues confronting the Gordon Committee was the "accuracy and meaningfulness of the figures themselves." Some critics argued, for example, that the official jobless measure was overstated because it included people who searched for work but whose willingness or availability to accept a job was marginal; others believed that the official measure understated the extent of labor market underutilization because it excluded persons who would habe searched for work if they thought jobs were available. The Gordon Committee recommendations led to the 1967 implementation of several important conceptual changes in the Current Population Survey (CPS)--the national household sample survey on which the unemployment statistics are based--which "tightened" the official definition of unemployment; years later, the Levitan Commission considered, but rejected, other possible changes in the definition. Despite these review, the debate over the relevancy and objectivity of the jobless measure continues.
The National Commission on Employment and Unemployment Statistics was charged with "responsibility for examining the procedures, concepts, and methodology involved in employment and unemployment statistics and suggesting ways and means of improving them." Major areas of investigation concerned the accuracy of the BLS data program, including the household-based CPS, as well as the establishment-based Current Employment Statistics survey. Did the surveys measure what they purported to measure? More importantly, did they measure what they should be measuring, given the social and economic changes that had occurred since the last review? One of the main questions examined by the Commission, for example, was the desirability of linking information on labor force status with economic hardship. It also completed an extensive review of the establishment survey to determine the representativeness of the sample and the accuracy and reliability of the survey estimates, particularly in the fast-growing services sector.
In its report, the Commission made nearly 100 specific recommendations for improving the Nation's overall labor statistics system, most of which were concerned with major programs of the BLS. Former Secretary of Labor Ray Marshall and current Secretary Raymond Donovan examined the Commission's recommendations and, as required by the public law establishing the Commission, submitted reports to the Congress evaluating each recommendation in terms of desirability, feasibility, and cost.
Current Population Survey Issues
The BLS has made several Commission-recommended changes in the CPS program relating to the development of new or expanded labor market information, data presentation, as well as to the CPS estimation process. However, it has encountered several unresolved problems in adopting a number of recommendations for changes in Labor force measurement which had been endorsed by the Secretary of Labor.
Expanded data. One of the key recommendations of the National Commission on Employment and Umemployment Statistics was that the labor force be redefined to include the Armed Forces stationed in the United States. (It recommended against including the military in the State and local area statistics because Armed Forces installations are not part of the local labor market.) The Commission noted that because joining the military was (and still is) voluntary, it represents a viable alternative to civilian employment.
The inclusion of resident Armed Forces data in the national labor force statistics was initiated in January 1983. The new series reflect the inclusion of approximately 1.7 million resident Armed Forces members and include separate data on the noninstitutional population, labor force, participation rate, total employed, employment-population ratio, and unemployment rate. (The number of unemployed persons, of course, is not affected by the addition of the military data because Armed Forces members are, by definition, classified as employed.) Separate data are published for men and women 16 years and over, and monthly data back to 1950 are available. The new series augments, rather than replaces, the traditional civilian-based labor force series. The most conspicuous estimate from the new series, the total unemployment rate, is consistently one-tenth to two-tenths of a percentage point below the civilian-based jobless rate. Because the resident Armed Forces level has varied little in recent years, both the civilian labor force and the total labor force series show identical trends.
The BLS also adopted the Commission's recommendation that monthly data be collected on whether youth ages 16 to 24 are attending school full or part time. The Commission explained that such data are needed "to understand work and education choices, to design appropriate employment policies and training programs, and to help appraise the labor market attachment of students. A short series of questions designed to determine whether youth are currently in school, whether those in school are high or college students, and whether they are enrolled on a full- or parttime basis, was initiated in the regular CPS questionnaire in November 1983. The BLS is evaluating the responses to these new questionnaire items and expects to begin publishing the results with the January 1985 data.
The Commission did not address the controversial issue of whether youth who are attending school full time should continue to be included in the official labor force figures when they are working at or seeking part-time jobs. At present, neither school status nor other activities are considered in determining labor force status. Excluding full-time students from the official labor force count would reduce the overall jobless rate by about one-half percentage point.
Another key recommendation implemented by the BLS dealt with the preparation of an annual report that linked economic hardship resulting from low wages, unemployment, and low labor force participation with earnings and family and household income. The BLS issued its first report in January 1982, based on data for calendar year 1979. Subsequent reports have been published covering data for 1980, 1981, and 1982.
The Commission was aware that introducing the 1980 Census-based occupational classification system into the CPS (which was subsequently put into operation beginning with data for January 1983) would create discontinuity with prior CPS occupational groupings. Therefore, it recommended that a format be developed to provide for comparable historical data based on skill levels. Time and cost constraints preclude a complete recording of the individual monthly records necessary to accomplish this goal. However, BLS and the Census Bureau developed a limited set of estimates for the 1972-82 period, based on the new classification system, which can be compared with data for 1983 forward. Annual average estimates are available for all civilian, men, and women by major occupational group. The procedure was not sufficiently accurate to develop detailed occupational estimates below the major categories.
The Commission recommended more frequent collection of occupational mobility data through special supplements to the CPS in order to measure movements of workers among occupations over a single year.
In January of both 1981 and 1983, the BLS included supplements to the CPS which asked questions on the extent and nature of job changes during the prior year and length of employment on current job. The 1983 supplement also included questions on the type of job training persons received in order to obtain their current job or to improve skills in their present job.
Conceptual changes. The National Commission on Employment and Unemployment Statistics made no recommendations for changing the basic labor force concepts and definitions. It did, however, recommend changes related to the identification and measurement of the relatively small number of persons outside the labor force commonly known as "discouraged workers." As currently defined, these are persons who want a job "now" but are not looking because they believe no job is available in their line of work or community. The Commission concluded that present CPS procedures for identifying this group were too arbitrary and subjective. The procedures were considered too arbitrary because they exclude students and persons who cite home or family responsibilities as their reason for not searching for work even if such persons also indicate they believe no job is available. They were also considered too subjective because they depend on a person's stated desire for work, regardless of whether the person had, in fact, tested the job market recently.
The Commission recommended an alternative approach, one that is modeled after the Canadian Labor Force Survey. The new criteria would determine whether persons, who were neither working nor looking for work (during the most recent 4-week period), had, in fact, sought work in the previous 6 months. If so, they would be asked the reasons they were not presently looking for work, whether they were current available for work, and whether they wanted a job. After much debate, the Commission also recommended continuation of the present practice of classifying discouraged workers as outside the labor force rather than making them part of the unemployment count. (Many critics believe the jobless figures are understated and that discouraged workers should be reflected in the unemployment figures.) The Secretary of Labor accepted these recommendations, agreeing that the new methodology would provide better estimates of the number of persons whose discouragement over the prospect of finding a job has prompted them to give up their job search. However, adoption of the new procedure has been delayed indefinitely because tests of the feasibility of introducing the necessary series of questions (as well as other potential changes) into the CPS questionnaire have been inconclusive. (See more detailed discussion later in this section.)
The Commission saw a need for procedures which would provide a more comprehensive picture of work time lost due to economic conditions in order to more accurately measure the extent of slack labor market resources. Thus, it recommended that all CPS respondents, whether working full or part time, be asked if they usually worked more hours than they had during the survey reference week. Those on reduced hours would then be asked why they worked fewer hours than usual, and, if the response was an economic factor affecting the job, such persons would be counted as involuntary short-hour workers and tabulated by hours worked. This recommendation was accepted by the Secretary of Labor and is expected to be implemented during the mid-1980's by revising the CPS questionnaire.
The Commission recommended replacing the current CPS question which asks unemployed persons whether they are seeking full- or part-time jobs with one that asks whether they want to work more or less than 35 hours per week. The reply would determine their classification as full- or part-time jobseekers. This recommendation needs to be tested further to determine whether the change would have any effect on the distribution of the full- or part-time job-search categories.
In fact, the Commission recognized that testing was necessary to determine the feasibility and possible impact of its recommendations on existing labor force measures prior to any revision of the basic, questionnaire. To this end, it recommended that the Census Bureau's Methods Development Survey (then called the Methods Test Panel) be doubled in size--from 1,600 to 3,200 households per month--and be used to test questionnaire revisions. The Methods Development Survey was expanded, but even a doubling of its size was insufficient for its intended purpose. It soon became apparent that although none of the questions being tested was intended to change the concepts of employment and unemployment, they did have a sizable impact on the employment and unemployment levels resulting from the test questionnaires. Throughout the testing period, estimates of such key labor force measures as the unemployment rate and the employment-population ratio were somewhat lower when based on the test document than those obtained from the control document (the present CPS questionnaire). In other words, the inclusion of the new questions resulted in lower estimates of labor force activity. Reasons for this effect were difficult to understand. The variations may have resulted from the content differences being tested, from procedural problems in conducting the test survey which had developed over time, or from the fact that the survey was quite limited geographically and the results were not representative of the Nation. The Methods Development Survey's monthly sample size reverted back to 1,600 households in mid-1981 because of funding cutbacks, thus becoming inadequate to support the objectives of the test.
The Methods Development Survey continued in that format until it was discontinued in September 1983. A scaled-down questionnaire (limited essentially to the new questions for identifying discouraged workers) was tested in connection with the Census Bureau's Random Digit Dialing project from October 1983 to September 1984. Those results also showed an unexplained effect on the basic labor force measures. Currently, there are no plans for futher testing.
Where does this leave the proposed implementation of a revised CPS questionnaire? It is apparent that the Methods Development Survey test questions produced significant differences in the level and rate of unemployment, even though no conceptual changes were made. Both Census Bureau and BLS technicians agree that, prior to introduction in the ongoing survey, revised questions would need to be tested under tightly controlled conditions in a major national overlap sample survey of a minimum of 10,000 households monthly for at least 1 year (in addition to a sufficient break-in period). A separate panel of that size would enable analysts to assess any difference in the levels and trends of the major labor force measures over the course of a year. For example, any potential breaks in series--such as in the unemployment rate--could be identified and quantified through the overlap sample testing, and the extent of the effect could then be clearly delineated at the time the new questions were adopted. Because it is impossible to measure the impact of the new questions on key labor force estimates in the absence of an overlap survey, it is not feasible to introduce major revisions into the CPS at this time.
Two other Commission recommendations dealt with the labor force classification of participatns of government classroom training and work experience programs. The Commission recommended that participants in programs that provide only classroom training be classified as not in the labor force and that participants in work experience programs be classified as employed. Formerly, if such information were volunteered, each group was classified as unemployed. Both recommendations were implemented in January 1984 through modifications to the CPS interviewers' instruction manual. However, there is no direct inquiry in the CPS questionnaire about participation in these programs, and we know that previously such information had been volunteered by only a few respondents. As expected, this definitional change has not had any measurable effect on the labor force estimates.
Data presentation. The National Commission on Employment and Unemployment Statistics recommendations related not only to the type of data to be collected through the CPS and how it should be measured, but also to the methods by which data and related information should be presented to the public. The Commission deemed the explanations of labor force concepts, definitions, and data reliability contained in the BLS monthly news release, "The Employment Situation," too technical and detailed for the intended audience. In 1981, the BLS rewrote these explanations in plainer language, particularly the explanation of seasonal adjustment. Another change in the news release related to the placement of the table showing alternative measures of unemployment (U-1 through U-7). In keeping with the Commission's suggestion, the table has been given greater prominence in the release. Also, the BLS had intended to simplify and revise the components of these measures, but encountered problems with the testing of new questions (described earlier) which prevented it from doing so.
The BLS adopted one of the two recommendations for changing the method of seasonally adjusting labor force statistics. In January 1980, it shifted to the X-11 ARIMA method, as suggested by the Commission. This method continues as the official method for seasonally adjusting employment and unemployment data. The Secretary of Labor withheld judgment on the Commission's recommendation that the BLS adopt the concurrent method of seasonal adjustment (that is, calculating new factors every month by including the current month's data in the seasonal adjustment process) primarily because such a procedure would necessarily preclude the present practice of advance announcements and publication of seasonal factors for future months. The BLS does make available each month the civilian worker unemployment rate based on a concurrent adjustment, as well as four other unofficial alternative approaches. Both BLS and the Census Bureau are continuing to conduct research in this area and could adopt the concurrent method in the future if the perceived advantages outweigh the disadvantages.
The Commission urged that greater use be made of CPS gross flow data in order to provide insights into the dynamics of labor market behavior. Gross flow data provide monthly estimates of the total number of persons entering the labor force, those leaving the labor force, as well as shifts between employment and unemployment. Despite their potential value, until recently these data had not been published in many years because of serious deficiencies. The Secretary of Labor agreed with the Commission that publication of these data should be resumed on a regular basis, together with an explanation of the discrepancies between the gross flow data and the official estimates. The first two such reports were published in March 1982 and November 1983, and regular publication of the data is expected to continue.
In mid-1984, the BLS and the Census Bureau sponsored a conference on gross flow statistics, bringing together experts from both inside and outside of government. Discussion focused on the identification of the specific reporting errors and biases which affect the accuracy of gross flow data, methods of adjusting the data to overcome these deficiencies, alternatives to gross flow data, and recommentations for future research directions. Several promising techniques for adjusting the gross flow data to make them more useful were discussed. These techniques will be tested by the BLS, and the adjusted data will be published if they are acceptable.
Estimation process. The accuracy of CPS estimates depends largely on the accuracy of the population data that are used in the sample design and selection. For this reason, after every decennial Census since the inception of the CPS in 1940, the sample has been redesigned to reflect changes in the size, location, and characteristics of the population. Throughout most of its history, the CPS has been viewed as a national sample survey which produced national estimates. However, growing pressures for more accurate and consistent State and local area estimates (in some cases mandated by law for the implementation of Federal revenue sharing programs) brought three State and subState expansions to the survey during the 1970's. This made the CPS a sort of hybrid, that is, a national sample modified on an ad hoc basis to provide data for States and large labor market areas. The pressures for more reliable area statistics placed a strain on the CPS design. So that the post-1980 Census redesign of the CPS would provide a more efficient sample design, the National Commission on Employment and Unemployment Statistics supported the plan of the BLS and the Census Bureau to design 51 individual samples--each State plus the District of Columbia--as the basis for the overall national frame. The new sample is being phased in between April 1984 and July 1985, and is expected to improve the reliability of the State and subState estimates and the overall cost effectiveness of the sample.
The Commission also made several recommendations for research on and improvement of the reliability of sample estimates. These included the development of information on sources of bias, characteristics of refusals and non-response, the differential effects of rotation group bias, and possible biases arising from the use of proxy respondents. Results of research in some of these areas have already been useful to the Census Bureau in the CPS redesign process. Research continues on some of the more difficult methodological aspects of the survey.
Improved Establishment survey data
The National Commission on Employment and Unemployment Statistics made considerably fewer specific recommendations for improving the Current Employment Statistics survey. The Commission's report noted that the lack of many detailed recommendations was due to several basic shortcomings of the survey, particularly its nonprobability sample design, relatively poor survey documentation, and inadequate quality measures.
The Commission recognized that the extensive industry and geographic detail on employment, hours, and earnings produced from the survey were essential for current economic analysis as well as for providing basic building blocks for construction of estimates of industrial production, personal income, gross national product, and productivity. Therefore, while questioning the sample design and other basic statistical underpinnings of the survey, it urged caution in making major changes that might disrupt major economic series. The Commission stressed the importance of first documenting all survey and operational aspects of the existing program before proceeding to the conduct of basic research on the technical aspects of the survey. The BLS has addressed these recommendations in just that fashion. The process of documenting the current program operations has been completed, and a number of specific concerns noted by the Commission have been answered. These improvements have taken place during a period of great change in the basic structure of the program, as funding and administrative responsibility for the surveys has been transferred from the Department of Labor's Employment and Training Administration to the BLS. As part of a major modernization effort, the BLS is now in the process of conducting the research and planning needed to place the establishment survey on a firmer statistical base.
Documentation of current program. Documentation has long been a problem because of the way the Current Employment Statistics program operated. The survey is conducted as a Federal/State program, with cooperating State agencies responsible for soliciting sample members and collecting data each month on a mail shuttle collection document (BLS-790 form). The microdata (individual employers' reports) are then transmitted to BLS in Washington for use in preparing national estimates. The State agencies also use these data to prepare State and area (Metropolitan Statistical Area) estimates of employment, hours, and earnings by industry. While obviously avoiding a duplicate reporting burden on employers to provide data separately to local, State, and national levels of government, the decentralized conduct of the survey by 51 separate agencies makes it difficult for BLS to ensure adherence to standards in survey operations. At the time of the Commission's review of the Current Employment Statistics program, the BLS did not have in place sufficient standards or the appropriate tools to monitor State performance.
After the Commission's report, BLS undertook a complete review and rewriting of the State Operations Manual which was used as the basis for a full round of training sessions with BLS regional and State personnel working on the survey. Annual updates to this manual reflect program changes and new developments. The manual is currently being revamped and is scheduled for reissue in 1985. Over the last several years, a training program has been developed for new employees in the State agencies and separate training has been provided to State employees involved in preparing estimates.
The first task in monitoring State agencies' performance was to review all aspects of each State's survey operations. BLS national and regional staff conducted the initial review in 1980. Based on the results of this review, improvement plans were developed for individual States to bring them into compliance with standards contained in the State Operations Manual. After the initial review of State operations, an annual review procedure was developed by BLS regional office staff to use in monitoring the implementations of improvement plans. Information on critical aspects of survey operations such as sample selection, solicitation, data collection, editing, data processing, estimation, and publication is now being maintained in a computer data base and is updated each year based on these reviews.
A new regional office operations manual was developed to systematically document the rapidly changing role of the BLS technical staff in the eight regional offices. This documentation has proven particularly important in keeping abreast of the application of data processing and telecommunications technology to survey operations. In addition, annual meetings are held with BLS Washington and regional technicians to review and discuss new developments.
The Commission questioned the sample design for the establishment survey (specifically the methods used for selecting the sample and its representativeness). It recommended that BLS study possible new sample designs, including a pilot version of a probability sample. While BLS has begun such a study, it recognizes that solicitation procedures would have to be grealty improved before any new design could be implemented. Various tests have been conducted to determine the most effective solicitation approach to assure high response rates from new sample members. These tests have shown that response rates in the 75- to 80-percent range are possible when solicitation is conducted by trained staff, with extensive followup for nonresponse. Based on this experience, a solicitation procedure has been developed for use in all States, with a requirement for stringent control and recordkeeping of solicitation efforts consistent with the current sample design. This procedure will be incorporated in the revised State Operations Manual, to be issued in 1985.
The use of telecommunications by State agencies to transmit microdata to Washington has grealty speeded the flow of data and enhanced the quality of the national estimates by providing more complete samples. The effect of this technology is most apparent in preliminary estimates included in "The Employment situation" news release. Prior to the use of telecommunications, each State agency had to stop collecting and editing the reports of sample respondents by Wednesday or Thursday of the week prior to the news release. They had to mail the reports to Washington in time for them to be processed Monday evening of the release week. The advent of telecommunications has meant that the States can continue to collect and edit data from respondents through Monday afternoon, and then transmit the data electronically to Washington for processing that same evening. This capability, coupled with better controls and increased followup, has resulted in a dramatic increase in the sample response included in the preliminary estimates (from about 65,000 to more than 100,000 reports).
The Commission was concerned about the accuracy of the establishment survey data. This concern was intensified by the fact that the BLS had been unable to benchmark the survey results in 4 of the 7 years prior to the Commission's review. Furthermore, at the time of the review, the ES-202 (Quarterly Report on Employment and Wages of Employees Covered by Unemployment Insurance), the Federal-State cooperative program which is the cornerstone for the benchmark, was not being compiled on a timely basis by many of the State Employment Security Agencies. In addressing this concern, BLS devoted more resources to the annual benchmark adjustment, and the required adjustment has been made each year for the last 5 years. The benchmarks were not complete the first several years, as several State agencies continued to miss deadlines for submitting their ES-202 reports. However, in the last 2 years this situation has grealty improved, and all State agencies are now submitting the reports on time. As a result, the most recent benchmark adjustment was complete and the BLS has reduced by 3 months the time required for preparation and release of the benchmark revision. The latest establishment survey revisions were published in the May--June period rather than the August--October period as was customary prior to the Commission's review.
The Commission recommended that the number of cells and the degree of sample stratification for the establishment survey be reevaluated annually at the time of the benchmark revision. This recommendation was immediately implemented, and a review has been performed in conjunction with the annual benchmark revision each year. Nearly 1,100 strata are now used in the benchmark estimation process, an increase of more than 20 percent from 1978. The increased stratification has improved the accuracy of the hours and earnings estimates in addition to those for employment, and has permitted the publication of additional industry detail.
The Commission recognized that the industry detail produced from the establishment survey, while extensive, was inadequate for the large and growing service-producing sector of the economy. However, it stopped short of recommending a major expansion in industry detail because the sample appeared to be particularly weak in this sector. Many industries in the service-producing sector are characterized by small firms, which makes it very difficut to develop and maintain an adequate sample of employers. Despite the difficulties, the cooperating State agencies have increased sample coverage in service-producing industries by 26,000 additional employer reports. As a direct result of this marked buildup of sample, the BLS expanded its publication of industry detail in the service-producing sector by 82 additional industries in 1984.
The Commission recommended that the BLS intensify research for the implementation of cyclically sensitive bias adjustment procedures in the establishment survey. This technique was used in the construction industry in 1981. It was more difficult, however, to develop procedures for other industries that actually improved the accuracy of current employment estimates. A major breakthrough in the research occurred in 1983, and BLS introduced cyclically-sensitive bias adjustment procedures for all industries coincident with the introduction of the latest benchmark revision in 1984. The new model used to produce the bias adjustment factors represents a refinement of the old method that utilized a 3- ear average of the differences between benchmarks and estimates. The new method uses the 3-year average of differences, but also relates the bias to the employment change in the sample in the most recent quarter. Thus, bias factors are now recomputed quarterly, instead of annually, and are much more sensitive to the economic cycle as measured by the month-to-month employment changes in the establishment sample.
Modernizing the establishment survey. The Bureau is currently involved in a full-scale modernization of the establishment survey that encompasses many of the broader recommendations made by the National Commission on Employment and Unemployment Statistics. This long-range project involves the review of all aspects of the program at the national, State, and area levels. The modernization effort stems from the fact that the nature of the establishment survey, as well as the basic cooperative Federal/State relationship under which it is conducted, has been relatively unchanged for the last 30 years. During this period, significant advances have occurred in the areas of survey design techniques, data processing, telecommunications, and employer recordkeeping practices. As discussed earlier, the ongoing survey has benefited from some of these developments, but major changes must await a systematic redesign.
The objectives of the modernization are:
* to provide data relevant to current public policy uses and needs;
* to develop a firm statistical foundation for the establishment survey within which objective measures of current sampling and nonsampling errors can be made for all survey estimates; and
* to modernize survey operatons and procedures, and telecommunications and software systems.
Current research testing is focused on collection methodology and the control of response/nonresponse errors for the national preliminary estimates which appear in the monthly news release. In addition, the collection of new data items is also being tested, including total payroll--and total hours for all workers, and part-time employment, hours, and earings for workers in service and trade establishments.
Change in program responsibility. The Ntional Commission on Employment and Unemployment Statistics found that the division of technical and administrative responsibilities between the BLS and the ETA for supervising the conduct of the establishment survey in the State agencies led to duplication of effort and conflicting demands. Funds for the survey were not earmarked or itemized in the State agencies, making it difficult to asses the adequacy of funding levels. The Commission recommended, therefore, that BLS responsibility for the establishment survey (and several other Federal/State cooperative statistical programs) should include sole funding authority, and that allocations to State agencies for the survey should be earmarked.
After lengthy negotiations within the Department of Labor, the BLS was granted administrative responsibility for the survey in State agencies. In fiscal 1984, for the first time, cooperative agreements were signed between the BLS and each State agency specifying survey deliverables. The deliverables were in the form of specific State performance relative to program requirements and specified State compliance with specifications and standards outlined in the State Operations Manual. Variances were allowed only when improvement plans were agreed to that would bring the State(s) into compliance. This process also resulted in a funding reallocation based on each State's required sample size, the number of Metropolitan Statistical Areas, and other factors.
Although these new administrative procedures have been in effect for a comparatively short time, their positive impact is being felt in pratically all aspects of the survey. Coupled with the annual State operations review, the cooperative agreements offer a degree of control for standardizing procedures and for implementing improvements in States that was not possible before.
The ES-202 Report
As mentioned earlier, the National Commisson on Employment and Unemployment
Statistics had expressed concern with the timeliness of another of the BLS Federal-State cooperative programs--the ES-202 Report--which is the primary input to the annual industry employment benchmarks for the establishment survey. The Commission had also expressed concern with two other aspects of the ES-202 program--the absence of effective quality control and the failure to clearly identify funding and committed resources.
With regard to quality control, the BLS insituted a formal State operations review procedure in 1980 to analyze all was to identify problem areas (notably failure to meet standards contained in the State Operations Manual) and develop a plan with the State agencies to remove these deficiencies over a reasonable period. Initially, the reviews were conducted annually but are now being done every 2 years with an annual followup to assess progress made in implementing the previous year's program improvement recommendations.
With regard to the funding and resource problems, the BLS and the ETA agreed in 1983 to jointly manage the ES-202 program. Under this agreement, BLS and ETA share fiscal and administrative management of the program, while BLS retains full technical responsibilities. Resources for the program were moved from the overall Unemployment Insurance Program budget and were included among several programs that are administered under ETA's Employment Service Reimbursable Grants process. The net result of this transfer was that funds for the ES-202 program for each State are now clearly earmarked. ETA and BLS jointly determine the resources necessary for each State to operate the program and monitor each State agency's performance in the program, and work closely to resolve the "priority" problems mentioned earlier.
Local Area Unemployment Statistics Program
The National Commission on Employment and Unemployment Statistics noted that "the inadequacy of State and local data has become especially glaring in recent years as the use of these data ... has expanded." The accuracy of statewide unemployment statistics was (and still is) being questioned. In contrast to the national figures, monthly unemployment estimates for most State and local areas are not based directly on the CPS because the sample size in those areas is not adequate to provide data which meet the minimum statistical standards on a monthly basis. Under the current "two-tiered" system, monthly estimates for 10 large States and 2 large area (Los Angeles and New York City) are based directly on CPS estimates because the sample size in those areas is large enough to support direct use of CPS data. Monthly estimates for the remaining 40 States and all other subState areas are developed using the so-called "Handbook" method. This method is based on (but not limited to) administrative statistics from the State Unemployment Insurance system, which, in turn, are subsequently adjusted in accordance with annual CPS statewide estimates. The Commission examined several options for improving State and local area unemployment estimates and concluded that, given the legislative requirements for the production of monthly estimates for thousands of areas and the prohibitive cost of expanding the use of CPS data to all areas, the continuation of a two-tiered system was warranted.
Based on the Commission's recommendations, the BLS implemented a research program aimed at improving the methodology for developing State and area estimates through the development of a system that would:
* use state-of-the-art statistical techniques for estimation and testing;
* provide minimum annual estimation errors relative to the CPS;
* reflect local seasonal patterns and business cycle movements;
* incorporate local counts and sample-based estimates;
* be susceptible to rigorous statistical testing;
* be capable of being updated at least annually; and
* be efficient to operate.
Results of ongoing research have shown that the use of regression techniques for developing subnational estimates, which the Commission also had supported, may well be more consistent with BLS objectives for an accurate, efficient, and cost-effective system than the present "Handbook" procedures. Thus, recent BLS research has focused on regression methods.
The use of regression techniques for developing estimates may have several advantages over the "Hanbook" method. First, and foremost, they result from well-established mathematical and statistical principles, which make it possible to hypothesize a casual relationship (that is, a model) between a set of economic variables, estimate the relationship among those variables, and test not only the adequacy of the model as a whole, but the individual relationships as well. Statistical advances in the "state-of-the-art" could easily be incorporated into the estimation techniques. Moreover, the model's coefficients could be updated periodically to adjust for changes in the interrelations. Results of this research to data have been promising, but further development and testing, including evaluation by cooperating State agencies, is necessary before a regression-based system could be substituted for the present methodology.
A NUMBER OF THE LEVITAN COMMISSION recommendations have either been implemented by the BLS or are in the process of being implemented. Changes in the methods by which the BLS data are collected, classified, and made available to the public have resulted in several improvements in the quality of the labor force statistics program and will lead to additional improvements in the future.
Several other recommendations were considered by the Secretary of Labor, but were deemed unacceptable because of substantive or cost considerations. Some of these, along with new issues that have surfaced in recent years, may be studied again in the next comprehensive review of the Nation's labor force data system.
|Printer friendly Cite/link Email Feedback|
|Author:||Hamel, Harvey R.; Tucker, John T.|
|Publication:||Monthly Labor Review|
|Date:||Feb 1, 1985|
|Previous Article:||Employment and unemployment in 1984: a second year of strong growth in jobs.|
|Next Article:||Changing employment patterns of organized workers.|