APPENDIX IV

Excerpted from Harry Frumerman, "Job Vacancy Statistics" chap. in National Commission on Employment and Unemployment Statistics, Concepts and Data Needs. Counting the Labor Force Appendix Volume 1 (Washington, D.C.: U.S. Government Printing Office, 1979).

EXECUTIVE SUMMARY

Job vacancies are the stock of unfilled job openings for which firms are actively trying to recruit new workers. The job vacancy rate is computed by dividing the number of job vacancies by the sum of employment and vacancies and multiplying the quotient by 100. Detailed information on the supply of labor has been available for more than three decades. There has not been, except for the period 1969-73, a continuous and comprehensive statistical series on the unmet demand for labor, or job vacancies, despite the value that such data would have for economic research and for operational purposes. Job vacancy data can contribute to business cycle analysis, they can also help to determine whether the prescription for high unemployment should be more spending (insufficient aggregate demand) or job training and employment counseling (structural unemployment). Job vacancy data can also be used to improve the way in which labor markets function by pinpointing emerging labor shortages.

Job vacancy data were collected in major metropolitan areas during World War II and the Korean conflict; thereafter most of these programs were discontinued. Since then, the strongest impetus for undertaking a national job vacancy data collection program came from the report of the President's Committee to Appraise Employment and Unemployment Statistics (Gordon Committee) in September 1962. Thereupon BLS undertook a research program on the conceptual and definitional problems of collecting job vacancy data and initiated pilot feasibility studies. Further developmental work was conducted in state employment security agencies, by universities and by private research organizations. The 1968 amendments to the Manpower Development and Training Act mandated a regular job vacancy data collection program on a national scale. BLS initiated a job vacancy program in 1969 as an addition to its program of collecting labor turnover data from employers by mail questionnaire, rather than as a separate and independent survey, in order to launch the program more quickly and at the lowest possible cost. This new "Job Opening-Labor Turnover" (JOLT) program was conducted through December 1973, when the job vacancy part of JOLT was discontinued. Since then there has been no national program for the collection of job vacancy data.

Before any new program is undertaken, both conceptual and technical problems remain to be solved. The BLS definition of a job vacancy was designed to be symmetrical with the definition of unemployment in the Census Bureau's Current Population Survey. This concept of job vacancy, however, has caused some disagreements, and has been subjected to varying interpretations. There are also difficult problems of design in devising a sample for job vacancy data collection. In the JOLT program, sample design was never wholly satisfactory, particularly for nonmanufacturing industries, and for this reason BLS did not validate most of the data for this sector.

The job vacancy program under JOLT was criticized by some in the academic community on the grounds that there had been insufficient preparatory analysis and testing to give the program a firm foundation. The most severe and the most persistent criticism of job vacancy programs, however, has come from the labor movement, which opposed the JOLT program and is currently against any new job vacancy program. This opposition is based on the arguments that the data could be misused to minimize the seriousness of unemployment and that trying to use job vacancy data simultaneously for job placement and for economic analysis would serve neither well, and that accurate job vacancy data could not be collected.

A BLS study in the 1960's of foreign job vacancy statistical programs, together with some supplementary data, showed that no country surveyed could provide insight for a data collection program based on surveying a sample of establishments, because administrative statistics which gave only an incomplete count were used almost exclusively for this purpose. The recent Canadian experience with collection of job vacancy statistics is more directly applicable to the development of a future job vacancy program in the United States. The Canadian Job Vacancy Survey program apparently has been more successful in handling some technical problems of making job vacancy surveys than was JOLT, but the administrative arrangements in Canada which have contributed to this better performance are not readily transferable to the United States. Furthermore, pressures to reduce survey costs apparently are moving Canada to combine job vacancy data with labor turnover data collection. Another possibility being considered in Canada, using administrative statistics for this purpose, may be indicative of the course a future job vacancy program could take in this country. Canadian experience so far does not show extensive analytical or operational uses of job vacancy data.

Possible alternatives for job vacancy statistics, such existing government programs as labor turnover statistics, occupational employment information or Job Banks; or data collected from newspaper want-ads or from private employment agencies do have some of the same operational uses as would job vacancy data, but are only a partial substitute. New statistical programs now being developed by state Employment Service agencies, however, such as Employment Service Potential, and the "New Hires" program, are promising for operational applications.

For the development of a future job vacancy program, preliminary basic research needs would include reexamination of basic concepts; rethinking of the job vacancy definition; reexamination of collection methodologies, particularly of administrative statistics as alternatives for field surveys; and reevaluation both of the JOLT period and of foreign experience, particularly that of Canada.

Costs of a future job vacancy program cannot be estimated in the absence of the specifics, but current "ball-park" figures range from $15 to $25 million annually.

The following conclusions come from the analysis of the theoretical and operational aspects of job vacancy statistics and from the evaluation of the experience with job vacancy collection programs:

  1. Job vacancy statistics would serve important analytic and policy needs. Job vacancy data have unique characteristics. There are no other data series that are a fully satisfactory alternative for analytic purposes.

  2. An exploratory job vacancy data collection program should be undertaken which, if successful, should become a full-scale and continuing national job vacancy data collection program.

  3. A full-scale and continuing job vacancy data collection program, if undertaken, essentially should serve analytical and policy needs, and should have the design characteristics that would ensure the most workable organizational framework and operating plan.

Such a job vacancy program should be a wholly federal program; the Bureau of Labor Statistics should have exclusive and full authority and full responsibility for the conduct of this program, and should be funded directly for this purpose; the program should be conducted independently of other data collection programs; the program should have the fullest possible industrial coverage; the program should be designed to obtain national data only, so that the data have the greatest degree of reliability, within cost and time constraints; and the program should be designed, however, to provide the necessary detail so that the job vacancy data can be matched with data from the Current Population Survey.

INTRODUCTION

The unutilized supply of labor is measured by the unemployment count and the unemployment rate. For the unmet demand for labor the equivalents are the count of job vacancies and the job vacancy rate. The U.S. Department of Labor has defined these terms as follows: "Job vacancies are the stock of unfilled jobs openings for which firms are actively trying to recruit new workers as of [a given date]. The job vacancy rate is computed by dividing the number of job vacancies by the sum of employment and vacancies and multiplying the quotient by 100."(1)

Detailed information on the supply of labor has been available for more than three decades. There has not been, except for 1969-73, a continuous and comprehensive statistical series on the unmet demand for labor, or job vacancies, despite the interest in such information expressed by the Congress, many government agencies, and by analysts in academic circles and research institutions. (The U.S. Employment Service (ES) has shown lack of interest in job vacancy data because the Bureau of Labor Statistics (BLS) has refused to allow such data to be used for referrals and job placement, in order to honor its pledge that the data would be kept confidential.) A closer examination is warranted therefore of the need for a continuing, nationwide program of gathering, analyzing and disseminating data on job vacancies, and its feasibility, applications, and potential costs.

In developing a comprehensive program of measuring unemployment and employment, in analyzing their trends and interrelationships, and in formulating a general theory of the operations of the labor market, an indispensable element would be a continuous and detailed statistical series which would measure job vacancies. In recent years of high unemployment, analysts and policy makers have asked how much of it is due to frictional and structural factors, as opposed to insufficient aggregate demand, because stimulating the economy will not necessarily reduce unemployment due to frictional or structural factors. Satisfactory answers to these questions require the collection and analysis of data on job vacancies. Job vacancy data could also be applied to improve the efficiency of labor markets. Frictional unemployment -- temporary unemployment of workers in between jobs -- might be reduced by providing current information on job openings to job seekers. By pinpointing where job opportunities are best, structural unemployment -- the mismatch of worker skills and demand for labor -- could be lessened by helping to create job training and employment counseling programs.

BACKGROUND

During World War II, and again during the Korean conflict, job vacancy data classified by occupation and industry were collected in the major metropolitan areas by the state employment security agencies to identify and alleviate manpower shortages. Although most of these programs were discontinued when hostilities ended, some 15 state agencies did collect limited data on job vacancies during the 1950's and 1960's. Some economists continued to stress the need for more complete information on labor demand. Arthur F. Burns, Chairman of the Council of Economic Advisers in the mid-1950's, was a strong supporter of the collection of job vacancy data, to help economic analysts make a more precise diagnosis of the nature of unemployment. He declared that the Employment Act of 1946 required continuous, carefully compiled and comprehensive statistics on job vacancies.

In 1956, at Burns' request, BLS studied the feasibility of collecting job vacancy data and concluded that employers did not keep records of vacancies comparable to their payroll records, and therefore "it would be impractical to initiate a mail collection of vacancy data."(2) Interest in job vacancies continued, nevertheless. At hearings before the Joint Economic Committee's Subcommittee on Economic Statistics in late 1961, expert witnesses put a high priority on the need for obtaining job vacancy data. The strongest impetus, however, for initiating national collection of job vacancy data was provided by the Gordon Committee in September 1962, which declared: "It is doubtful that any suggestion for the improvement of knowledge about the Nation's labor markets was more frequently voiced to this Committee than that calling for job vacancy statistics."(3) The Gordon Committee recommended that the government undertake a program of research in conceptual and definitional problems and mechanisms for collection and survey design, looking toward the possible development of a national system of job vacancy data.

In 1963 and 1964, an investigation of the problems involved in the collection of job vacancy data, sponsored by the University of Illinois, found that acceptable data on current job vacancies, as well as those on anticipated job vacancies, could be obtained by interviews, and that "such information possesses a degree of accuracy high enough to be useful for labor market and economic analysis."(4) Late in 1964, the National Industrial Conference Board (now The Conference Board), a private research organization, undertook an exploratory job vacancy measurement program in the Rochester, N.Y., labor market. This study found that employers had, and would provide, occupational data on job vacancies with a level of accuracy that would make the figures meaningful.(5)

Following through on the Gordon Committee's recommendation, in 1964 the Department of Labor began a "modest research program which included research on conceptual and definitional problems, pilot feasibility studies in the Chicago and Buffalo areas, and a survey of the nature and uses of job vacancy statistics being collected in foreign countries."(6) As part of this program, the Illinois Bureau of Employment Security conducted two job vacancy data collection surveys in 1964. From the findings of these studies, the investigators concluded that periodic surveys of job vacancies should be conducted.

At the request of the Secretary of Labor, in early 1965, the National Bureau of Economic Research convened a conference of academic experts, federal and state officials, labor union representatives, and private research organization specialists to review and analyze the theoretical and operational problems of measuring and interpreting job vacancy statistics, and to evaluate the experience with such measurements here and abroad. In his opening statement to this conference, Arthur F. Burns reiterated the importance of job vacancy measurements by calling the unavailability of these data "the most serious gap in our entire scheme of economic intelligence."(7) The publication the next year of the book, The Measurement and Interpretation of Job Vacancies, incorporating the research findings and discussions presented at this conference, helped to disseminate more widely in academic circles and government agencies the then current assessments of the analytical and operational potentialities of job vacancy measurements, and to stimulate further investigations and analyses.

Meanwhile the U.S. Department of Labor, by April 1965, had launched a more comprehensive experimental job vacancy collection program in 16 Standard Metropolitan Statistical Areas (SMSA's) with what was reported as favorable employer response.(8)

BLS research efforts during 1966 were directed at fashioning a new approach to the collection of job vacancy data -- the integration of the collection and analysis of job vacancy information with the existing federal-state cooperative mail collection labor-turnover program. Findings of pilot surveys conducted in Maryland and Connecticut showed that this appeared to be feasible. In 1967 and 1968 BLS undertook additional pilot studies to test this collection technique and again reported favorable results. After the 1968 amendments to the Manpower Development and Training Act (MDTA) instructed the Secretary of Labor to collect job opportunities data, BLS undertook a national program for regular collection of job vacancy data, as an addition to the existing labor turnover program. It was believed that joint collection would speed up launching the program and minimize costs. Monthly data on the number of job vacancies initially were to be obtained from a total of 50 Standard Metropolitan Statistical Areas (SMSA's), limited in 24 of these areas to manufacturing and mining industries, but covering all nonagricultural industries in the other 26 areas, with quarterly data on vacancies by occupation and by wage rate in 17 of these areas. Subsequently the monthly survey was expanded to include all nonagricultural job vacancies in 29 SMSA'S, with data by occupation and wage rate collected quarterly in 20 of these and annually in the remaining 9 SMSA'S.

This decision to undertake a full-scale program of collecting job vacancy data was questioned at the time by some government and academic spokesmen, and organized labor strongly opposed it. The opponents argued that insufficient research and analysis had gone into the conceptual, definitional and operational problems involved, and that expansion beyond the pilot stage could not be justified. These dissenters urged instead that the government merely continue its exploration of the theoretical and operational aspects of job vacancy measurement.

This new "Job Opening-Labor Turnover" (JOLT) program went ahead, nevertheless, with the first national monthly job vacancy figures being published for April 1969. Job vacancy data from the JOLT program were published monthly thereafter through December 1973, when the job vacancy portion of the JOLT program was discontinued.

An evaluation of the JOLT program is of prime importance in arriving at a decision on whether another job vacancy collection program should be undertaken now. In announcing the termination of JOLT, the Commissioner of Labor Statistics stated:

The job vacancy program, as presently constituted, is not meeting the needs of the Manpower Administration or the state employment security agencies, our cooperating partners in the program, since the data cannot be used for direct placement. In addition, the current program does not appear to meet the needs of economic analysts because the data cannot be matched with the components of the employment and unemployment surveys.(9)

Local Employment Service offices wanted specific job vacancy data for direct job placement activities by company name, plant location, and occupation. BLS and state employment security research offices, however, had given a pledge of confidentiality to the employers who supplied the data, and therefore provided only summary information which was not very useful for operating purposes. The lack of detail and the uneven quality of the data that were collected gave them only limited value for research purposes. The basic questions about end-uses of job vacancy data therefore remain unanswered. Whether job vacancy data of an acceptable quality can be collected has not been fully determined either. Answers to these questions are central to a decision on whether another job vacancy statistics program should be undertaken now.

Since the discontinuation of JOLT there has been no national program for the collection of job vacancy data, although comparable data are still collected and published by the Wisconsin and Minnesota state employment security agencies. Currently, other sources of job vacancy data are the local Job Banks, operated by almost all of the state employment security agencies, which provide daily listings of job openings for many labor market areas; The Conference Board, which publishes a monthly index of help-wanted advertising; and private employment agencies which have some current data on job vacancies.

MEASUREMENT OF JOB VACANCIES

Problems of Concept and Definition
The BLS Definition. The official definition of job vacancies by the U.S. Department of Labor (as given above) lays down the three basic stipulations for an employment opportunity to be reported as a job vacancy: (1) It must be unoccupied; (2) it must be immediately available; and (3) it must be the object of an active search for a new worker from outside the firm. These criteria in the definition of job vacancies were designed to provide maximum comparability with the definition of unemployment, consistent with other uses of the data.(10)

The definition of unemployment, to which that of a job vacancy is to be symmetrical, is that used in the monthly Current Population Survey (CPS):

Unemployed persons comprise all persons who did not work during the survey week, who made specific efforts to find a job within the past 4 weeks, and who were available for work during the survey week (except for temporary illness). Also included as unemployed are those who did not work at all, were available for work, and (a) were waiting to be called back to a job from which they had been laid off; or (b) were waiting to report to a new wage or salary job within 30 days.

This concept of a job vacancy is subject to varying interpretations and has caused some disagreements. The major points at issue include the following:

  1. The condition that the job opening must be unoccupied eliminates from the count those in which the search by employers for workers needed to fill jobs on a normal turnover basis is undertaken before those jobs are actually vacated -- a large part of total accessions. When employers are recruiting in anticipation of turnover and expansion, it is often difficult to determine whether the recruiting activity is in connection with a filled or an unfilled job. It has been noted that "the importance of hiring in anticipation of turnover, on a forecasting basis, is that such recruiting may increase when labor markets become tight...."(11) Such increased activity is not reflected in job vacancy data collected on the basis of currently unoccupied openings.

  2. The condition that the job vacancy count be limited to those immediately available provides symmetry with the definition of unemployment, which requires that an unemployed person be available for work during the survey week. This requirement, however, eliminated from the count any job vacancy with a future starting date, that is, "one for which an employer is willing to hire a worker as of the survey reference date, even though the worker will not normally begin work until a later time."(12) Job vacancies with future starting dates were found in the Rochester, N.Y., job vacancies study to be typical of education, where teachers are hired in advance of the September starting date of the new school year. Vacancies with future starting dates were found also in this study in manufacturing, construction, public utilities and transportation, and a high proportion of these were considered to be due probably to seasonal influences. Inclusion of future job vacancies in the count would require, for proper interpretation, identification of seasonal patterns in each situation.

    Data on future vacancies are useful for business forecasting, for labor market analysis, and for the planning of training programs. Should future job vacancies as well as current job vacancies therefore be collected? Would future job vacancies data be accurate? Would employers exaggerate or underestimate future job vacancies? These questions remain unanswered.

  3. The condition that the job vacancy count include only those jobs which are the object of active search by the employer provides symmetry with the unemployment count which includes only those persons who make specific efforts to find a job during the preceding 4 weeks. It eliminates such vague responses by employers as that they are not recruiting, but would, nevertheless, be amenable to additional hirings. At the same time, this requirement also eliminates job vacancies for which there has been no active searching because employers are discouraged by the absence of qualified applicants.

    The further condition that the employer's active search must be for a new worker from outside the firm bars job openings that are filled by promotion or transfer within a company. This restriction has resulted in questions being asked about both the meaning and uses of job vacancies as so defined by those who believe that the operations of internal labor markets are at least as significant as those of the external markets for making job openings available and filling them. John T. Dunlop, for one, has asserted that relatively few job openings, so-called "port of entry" jobs, are filled from the outside, with the rest filled internally. If this interpretation is correct, a count of job vacancies based on outside recruitment would be of questionable value.

  4. Because of the emphasis on the immediate job opportunity, the definitions of job vacancies and unemployment are not fully symmetrical. Openings to be filled by recalling laid off workers would not be included as job vacancies, but laid off persons waiting to be recalled (and not working elsewhere) would be classified as unemployed. Similarly, a job for which a person had been hired with a future starting date would not be defined as vacant, but the person waiting to report to that job would be classified as unemployed.

Other Problems of Concept, Definition and Interpretation. Other problems of concept, definition and interpretation of job vacancies include:

  1. The usefulness of job vacancy data for analytical and operational purposes is greatly enhanced if detailed occupational characteristics are obtained. Present systems of job classification, however, are not wholly suitable for converting shop job titles into standard nomenclature. The Dictionary of Occupational Titles (DOT) and the Census of Population's occupational classifications are not fully compatible; the conversion tables do not bridge the gap entirely. Whether the Standard Occupational Classification (SOC) system will be more satisfactory remains to be determined.

  2. Should part-time as well as full-time job openings be included in the total job vacancy count? Should temporary as well as permanent job vacancies be counted? Where do you draw the line to exclude occasional jobs (for example, baby-sitting) or jobs of brief duration (for example, yard cleaning)?

  3. Should the counting of an employment opportunity as a job vacancy be conditional on its carrying a prevailing wage rate, as the labor unions have argued? Or, should all job vacancies be counted, but separated into those with going wage rates and those with wage rates below prevailing norms? Or, should the count cover only vacancies that equal or exceed a stated minimum entry wage? Should job vacancies be differentiated by other terms of employment such as hours of work, and working conditions? Should job vacancies be differentiated by educational or experience requirements?

  4. Should job vacancies originating from strikes or lockouts be counted?

  5. Should we try to obtain data on job vacancies by duration? If so, how long should a job have to be vacant before it is classified as "hard-to-fill"? For operational convenience the Department of Labor surveys defined any vacancy existing 30 days or longer as "hard-to-fill." Because the average time needed to fill a vacancy varies from one occupation to another, should the definition of "hard-to-fill" vary accordingly?

    Data on the duration of job vacancies would be useful in evaluating labor shortages. A long average duration of job vacancies may indicate either real shortages or inefficient operation of the labor market, but it may also reflect such factors as low wages, poor working conditions, undesirable plant locations, and unrealistic hiring specifications. Should all "hard-to-fill" job vacancies therefore be counted? If not, which ones?

  6. What is the minimum time that should elapse before a job opening is counted as a "job vacancy"? In the JOLT program, a job opening must have been vacant for at least 1 full day before being counted as a job vacancy. This restriction was intended to eliminate job openings, particularly for unskilled work, which can be filled almost instantaneously, and to avoid an artificial inflation of the job vacancy count.

    Labor union representatives and others have maintained that the 1-day rule for counting job vacancies is not symmetrical with the 1-week out-of-work rule for counting unemployment. These critics argue that this rule would overstate job vacancies and hence could be used to minimize the seriousness of unemployment at a given time. An analysis of the concept by BLS in 1966 showed, however, that 85 to 90 percent of the vacancies listed in the reports examined had been unoccupied for 1 week or longer.

  7. Should job vacancy data relate to the same dates as the unemployment statistics? Under the JOLT program, employers were asked to list the number of job vacancies immediately available as of the last business day of the month, because the data were reported on the same form used to collect labor turnover data, which were obtained as of the end of the month. The CPS, however, counts unemployment as of the week containing the 12th day of the month. It would be feasible, of course, to count job vacancies as of any given day if the data were collected independently.

  8. Should a survey count the stock of job vacancies as of a given time, or should it measure the flow of vacancies over a given time period? The JOLT program used the stock concept in order to be symmetrical with the unemployment count. The stock concept is useful if job vacancy information is to be used for job placement, but it is less satisfactory than the flow concept for economic research. Furthermore, using the flow concept would enhance the value of job vacancy data for the planning of training and counseling programs. Collecting job vacancies as a flow would require estimates of future vacancies and future quits and layoffs. The undertaking of such a program would be conditional on the solution of new conceptual and definitional problems, and the ability of respondents to make reasonable forecasts.

Sample Design. Because a canvass of the universe of employers is impractical, it is necessary to use a sampling procedure, which entails the solution of difficult design problems. The JOLT experience is instructive: The original intent was to develop a probability sample based on the labor turnover sample of approximately 40,000 manufacturing establishments, but to get valid job vacancy data, the sample for manufacturing industries had to be expanded and a new sample for nonmanufacturing industries had to be devised. At the end of 1969, an effort was made to solicit reports from the two-thirds of the some 100,000 establishments in the Current Employment Survey (CES) sample which were in the scope of the JOLT program, but the response was insufficient. Starting in 1970, many of the state employment security agencies began soliciting firms which were not in the CES sample, but with very mixed results. By January 1973, however, the JOLT sample included 90,000 establishments. Even after this expansion, however, many states did not have adequate samples for several nonmanufacturing categories, especially construction, trade and government. Satisfactory sample size was never reached for all industries in the scope of the JOLT program in any state.

Difficulties of sample maintenance necessitated that the state employment security agencies continually attempt to expand their JOLT sample in an effort to achieve representatives. By January 1972 many states had exhausted their list of potential respondents and, instead of having probability samples, the samples were essentially those firms within the universe which were willing to cooperate. As a result, traditional statistical techniques could not be used to evaluate the quality of the JOLT sample; the determination whether a sample was valid often rested primarily on judgmental factors.

Job vacancy data for manufacturing industries in most of the 50 SMSA's in the JOLT program were validated by BLS based upon criteria that were not strictly statistical. The regional BLS offices evaluated the behavior of the job vacancy statistics themselves to determine whether the sample data could be weighted up to the universe. The state agencies, in cooperation with BLS, evaluated the quality of the sample and, based on their judgment, the sample was or was not validated. In nonmanufacturing industries, where experience had been extremely spotty, BLS generally did not accept the job vacancy sample as being sufficiently large or detailed enough to be considered for validation. By December 1973, at the end of the JOLT program, there was not one SMSA in which all series of job vacancies for nonmanufacturing industries had been validated. Many state agencies, however, had their samples for such nonmanufacturing industries as finance, insurance, real estate and banking validated by BLS. But BLS rejected other requests for validation of samples in nonmanufacturing industries because the samples were not of the probability type and benchmarked data were not available. Such difficulties with validation of basic job vacancy data ruled out the possibility of obtaining comprehensive occupational job vacancy information.

Although the JOLT experience is not encouraging, nevertheless any future job vacancy data collection program should attempt to obtain industrial, occupational and geographical coverage which will permit matching job vacancies with the corresponding characteristics of the unemployed. If these objectives are to be achieved, the following considerations and problems will have to be taken into account:

Occupational Detail. Here, the problems are determining the level of refinement in obtaining job vacancy data by job titles; matching respondents' lists with standardized job titles and standardized job descriptions; and training staff to code job titles accurately. During the JOLT period, occupational data were collected in only 29 areas. This limited program did not meet either analytical or operational needs. Obtaining comprehensive, detailed and statistically reliable occupational data would require a large sample and involve high costs.

Even if acceptable data on job vacancies by occupation were obtained, their value would depend on a satisfactory system of coding occupational titles and job descriptions. The U.S. Department of Labor's Dictionary of Occupational Titles (DOT) is designed largely for Employment Service operations, and is not wholly suitable for analytical uses. The Census Bureau's occupational classification system can be more readily applied to analysis, but it cannot easily be linked with the DOT system. Conversion tables prepared to bridge the two classification systems have not solved all of the problems. The BLS Occupational Employment Surveys (OES) follow another occupational classification system with some 400 occupations listed. A Standard Occupational Classification System has recently been developed for use by all government agencies. Already there are reports that it is not fully suitable for the analysis of occupational job vacancies and for coordination of these data with the occupational classification of the unemployed.

Geographic Detail. During the JOLT period, details on job vacancy data were provided for some of the Standard Metropolitan Statistical Areas (SMSA's). At the end of the JOLT program about 50 SMSA's were being surveyed but it proved difficult to obtain job vacancy data of publishable quality for small areas with low number of job vacancies. Theoretically, publishable data could be obtained for all 243 SMSA'S, and even for the non-SMSA areas, but the costs of expanded geographical coverage would have to be weighed against the value of the additional detail. The level of industry detail desired would also be a factor in this determination.

Industrial Detail. Establishments covered by state unemployment insurance (Ul) laws are already classified according to the Standard Industrial Classification. Classifying job vacancy data by industry, therefore, would not seem to present any special problem. There are, however, practical problems: the slow pace of the state Ul agencies in updating the industrial classification of covered firms and in classifying new firms by SIC.

Job vacancy data collection programs, either of the experimental types of the 1960's or of the JOLT period, were most successful in manufacturing. 'Surveying nonmanufacturing industries was more difficult and more costly; these firms are smaller and more widely dispersed geographically. Although there may not be any conceptual, definitional, operational, or analytical obstacles that would make impossible broad industrial coverage, nevertheless a number of industries are "problem areas."

The concepts and definitions of job vacancies that apply to the private sector are not readily transferred to the public sector. For example, the existence of job vacancies in a government agency may simply reflect the difference between actual employment levels and the employment ceiling imposed by a legislative body or a budgetary authority. Agencies may have a large number of such "vacancies" which they have no intentions of filling; at the same time, they might fill other "vacant" jobs if funds were to become available. Even if funds are obtained, agencies might or might not be actively recruiting, and even if they were, the jobs might or might not be available immediately. Weeks or months might elapse between hiring and actual placement. Assuming that these conceptual and definitional problems could be surmounted, if the JOLT experience is a guide, it would probably be difficult, sometimes impossible, to obtain data on job vacancies in governmental agencies because of the lack of interest and cooperation by government officials.

The small size of the average construction company, the geographic dispersion of construction, the widespread use of subcontractors, the lack or incompleteness of employment records by most builders, the high business mortality rate, the short time that each craft worker generally spends on a construction job, and the intermittency of construction work because of weather, and delays due to material shortages, equipment breakdowns and other such factors, all make for difficulties in collecting job vacancy data in the construction industry. Union hiring halls could be an alternate source of job vacancy data for some crafts, but a job vacancy data collection program geared to reports from employers might not be compatible with such a procedure.

A study of the feasibility of collecting agricultural job vacancy data made by the Kansas State Employment Security Agency during the JOLT period showed that while it was possible to collect job vacancy data for farms, caution had to be used in evaluating the results obtained from those with less than 200 acres. Presently, too few data are available for judging the general feasibility of collecting such data. Another possible way of collecting data on seasonal job vacancies in agriculture would be to use Employment Service data on such job placements.

To round out the job vacancy count, it has been suggested that such other kinds of job vacancies be included as those in domestic service, those for commission salesmen, stevedoring, temporary and part-time work, and casual work. Whether these counts should be attempted would depend on finding satisfactory answers to such questions as whether such inclusions would result in any significant addition to the total count of job vacancies; whether such data could be counted accurately (avoiding overcounting of job vacancies for commission salesmen, for example); and whether the value of the additional detail would justify the high costs.

A feasibility study of collecting job vacancy data from private household employers by the New York State Department of Labor in 1973 showed that valid data could be collected on a continuing basis from private household employers. There was, however, a problem of nonresponse from "zero vacancy" employers, and the problems of collecting and editing such occupational employment data seemed to outweigh the benefits of having such information.(13)

SURVEY METHODS

Data Collection. In the JOLT program, job vacancy data were collected monthly on a shuttle-type mail schedule -- a modification of the labor-turnover form used by BLS for a number of years. The labor turnover program was selected as the vehicle for collecting job vacancy data in part because of the obvious relationship between the two concepts, but also because of the feasibility studies in Maryland and Connecticut in 1966 which showed that over 90 percent of the labor turnover respondents could and would voluntarily cooperate in reporting job vacancies.

Job vacancy data can also be collected by field investigators, and by telephone. Field investigation may yield more complete and more accurate data, but its slowness and cost generally precludes its use in such survey work.

On the relative merits of different data collection methods Myers and Creamer, in their job vacancy study of the Rochester area, concluded:

In our judgment, the collection of data by interview and on-the-spot transcription is much more effective than collection through the mail. This is particularly true during the introductory stage of a new survey. Once the value of the survey is recognized by the respondent and there is a full understanding of the data requested, the balance of advantage probably shifts to the use of collection by mail, since at this point relative costs must enter into the decision.(14)

They found that collection of job vacancy data by personal interview initially had several advantages. A fixed appointment with an interviewer had a "disciplinary effect" on a respondent who, despite good intentions, might procrastinate. A knowledgeable interviewer could answer questions concerning interpretation of definitions and their application in special cases. They also found that a personal interview was more likely to obtain a response from an uninterested employer than a mail questionnaire. Furthermore, an uninterested employer, in their view, was less likely to give a casual, off-the-cuff response to an interviewer than in answer to a mail questionnaire. A personal interview might elicit a bona fide response, they believed, where an employer was unwilling to commit himself to paper.

The response rate is crucial in deciding which collection instrument should be used. In BLS experimental mail surveys of job vacancy data in the 1960's, there were large variations in response rates. In one such survey, the response rate ranged from 97.8 percent in Portland, 93.9 percent in Milwaukee, 93.8 percent in Los Angeles, and 92.7 percent in Birmingham down to 64.5 percent in Chicago and 60.2 percent in Providence; the median was 86.4 percent.(15) In a 1969 survey, during the JOLT period, BLS reported that nonrespondent firms amounted to 10 percent of the sample in manufacturing.

It has been suggested that one reason for nonresponse or underreporting of job vacancies might be some employers' unwillingness to accept referrals from the Employment Service. Hence, federal laws requiring the public posting of job openings by companies holding federal contracts or otherwise subject to federal regulations could well determine employer cooperation in a job vacancy program. If such employers are not listing all of their job openings with the Employment Service, they might fear being exposed by revealing a larger number of openings in a job vacancy collection program, even if the data were furnished on a confidential basis.

In analyzing the extent and variations of nonresponse, consideration must be given to the care with which a particular survey and survey community are prepared, the technical competence of the survey staff, the reaction of respondents to the survey organization and its objectives, and the survey method.

Establishing the best point of contact in the firm was a source of continuing difficulty during the JOLT period. How can the most knowledgeable official be identified? Should there be a single point of contact, on the assumption that large employers either have central hiring points where complete records are kept or prefer to assemble these data themselves to eliminate duplication? Or, is hiring, in fact, decentralized in large companies, especially multiplant firms, or in retailing? For these reasons, collecting job vacancy data from multiunit firms was a trouble spot during the JOLT period.

One special problem of data collection, especially if done by mail, is possible lack of interest by respondents in reporting "zero" job vacancies month after month. Substantial nonresponse for this reason could endanger the validity of the sample. Collecting job vacancy data on the labor-turnover shuttle form is believed to have reduced the number of such nonresponses in the JOLT program, because the job vacancy question was one of a number of questions, so employers did not have to make only a single "zero" entry month after month on a questionnaire limited to job vacancy data.

"Zero" vacancy reporting was nevertheless a continuing problem under the JOLT program. In the effort to minimize employer nonresponse because of "zero vacancy discouragement," followups were made by telephone contacts and personal visits. Still, a high proportion of firms continued to report "zero" vacancies, and this complicated the problem of maintaining the sample because of the lack of month-to-month continuity.

Data Processing. Job vacancy data obtained from a sample of employers must be edited, coded, and classified. Editing involves reviewing the schedules for "reasonableness" and for internal consistency, and is aimed at eliminating omissions, errors of transcription, and inconsistencies with earlier reports. The accuracy and completeness of editing depends upon proper instructions and procedures, thorough staff training, and proper supervision and review. A problem in the processing of job vacancy schedules is that of ensuring the accuracy of both industrial and occupational coding. A study by the Wisconsin agency during the JOLT period found that 85 percent of the occupationally coded data were sufficiently valid for planning purposes.

Evaluation of Data. To insure the accuracy of job vacancy data, it is necessary to evaluate the accuracy of reporting by employers and the quality and consistency of internal procedures-sample design, editing, coding, and analysis. During the JOLT period the regional offices of BLS reviewed the internal procedures of the state agencies for sample adequacy, follow-up procedures, and estimating techniques. Fairly soon after the JOLT program had gotten under way, BLS found that in 29 areas the manufacturing sample seemed adequate, the data seemed to be well edited, and the estimates were properly prepared. In the nonmanufacturing industries, however, not one area was found to meet these tests -- weaknesses which continued throughout the JOLT experience.

Job vacancy data are subject to several kinds of errors on the part of respondents. Response errors can be accidental or willful. Accidental response errors can arise from misunderstanding of the definition or the specific questions, or simply inattention or distraction. Such willful errors as understatement (for reasons cited above), or the response that the employer has no vacancies (the "zero vacancy" situation discussed above), or overstatement of the number of job vacancies, have been cited by those who are critical of the value of job vacancy information. While not ruling out with certainty that willful errors were important, Myers and Creamer did not think that they actually were so in their studies.(16) They pointed out that purposeful response errors can be minimized by skillful and tactful personal interviews.

Another source of response error, lack of information or faulty recall on the part of the respondent, very often results from poor or no recordkeeping. In fact, a 1956 job vacancy feasibility study concluded that job vacancy data could not be collected by mail because a high proportion of employers did not keep record of job vacancies.(17) Pilot studies by BLS in the 1960's, however, showed that mail collection of job vacancy data was feasible, because of better record keeping by employers. The current job vacancy collection programs in Wisconsin and Minnesota do not appear to be having any difficulty with employer records. Those programs, however, are limited to manufacturing establishments, and the favorable experience might not necessarily be duplicated in a job vacancy data collection program that extended to all nonmanufacturing industries as well.

Accuracy of response can be tested by such methods as conducting response analysis surveys and checking the consistency of current data with previous data or data from other sources. A 1966 response analysis survey by BLS showed that, although most employers understood job vacancy concepts and reported accurately, the total number of vacancies was understated by 4 percent. Another 1966 survey found a 12 percent understatement.

In February 1970, response analysis surveys of manufacturing industries were conducted by the Arizona, Illinois, Oklahoma, Oregon and Utah agencies. In these studies, interviewers visited a randomly selected sample to check on the quality of the submitted reports. The findings showed that employers understood the questions, were willing to provide the information, and reported accurately in 9 out of 10 cases. Underreporting far exceeded over-reporting, but amounted to only about 2 percent of all job vacancies reported. This low rate was attributed both to collection of job vacancy data in conjunction with labor-turnover data and to limiting the surveys to manufacturing establishments. Based on these findings, the job vacancy data for manufacturing industries were found to be reasonably accurate. When BLS compared job vacancy estimated for several months with data from the Baltimore Job Bank program, it found that Job Bank openings averaged a little over 40 percent of the job vacancy figure, ranging from about 35 percent to 50 percent. This was considered to be consistent with expectations based on earlier experimental programs.

While there was some underreporting of job vacancies, there is no evidence of over-reporting during the JOLT period, or in the Myers and Creamer study in Rochester. Labor unions, however, have expressed strong misgivings on this score. They have objected to the inclusion of job vacancies at "substandard" wages, arguing that it would exaggerate the total count significantly. A counterargument favors counting all job vacancies, pointing to the parallel between the inclusion of "substandard" job vacancies and the inclusion in the unemployment count of workers who would work only if they got higher wages than those prevailing in available jobs for which they are qualified. A 1969 BLS survey of 12 of the 17 SMSA's in which data on wages were being collected under the JOLT program showed that the number of jobs offered at less than the prevailing wage rate ranged from 5 to 10 percent.

To attempt to exclude such job vacancies would require occupational detail for all job vacancies and the determination of the "substandard" wage cutoff for each occupation. If this proves to be impossible, one proposed solution is to assume a certain percentage of substandard job vacancies, and adjust the total figure by this factor.

Coordination. In the federal-state cooperative programs of labor market data collection, the division of responsibility (apart from financing) has been that BLS, the federal agency, is responsible for sample design, questionnaire design, evaluation, analysis and publication; and the state agencies for operations -- collecting the data, editing, coding, and assembling. In such a new and large-scale program as JOLT, a substantial coordination effort therefore was necessary to insure uniformity of data collection procedures and of data processing. Another difficult problem of coordination came from the divided responsibility for the JOLT program at the federal level between BLS and the Manpower Administration.

A future job vacancies data collection program could follow this pattern. But an attractive alternative would be a federal program centralized in one agency, such as BLS or Census, for example, that would greatly minimize, if not almost eliminate, the need for coordination with state agencies.

USES OF JOB VACANCY DATA

Analytical and policy applications of job vacancy data include analyzing and forecasting business cycle trends; evaluating the "expansionist" and "structuralist" arguments about the causes of high rates of unemployment and inflationary pressures; measuring the efficiency of labor markets; and assessing measures designed to reduce frictional and structural unemployment.

Business Cycle Analysis. Cycles in the stock of job vacancies would be expected to have a close positive relationship to business cycles. When business is slack, there is little demand for additional workers and available jobs can be filled quickly. When business is good, there is need for more workers and more difficulty in filling vacant jobs. Thus the number of job vacancies at any one time will conform positively to employment cycles and inversely to unemployment cycles, with certain exceptions. In a tight labor market job vacancies can go up without any obvious ceiling, while unemployment, once reduced to frictional levels, cannot decline much further. Conversely, in very slack labor markets, when job vacancies cannot decline any more, unemployment can still go up. These differences also affect the amplitude of cyclical fluctuations. Job vacancies and unemployment would be expected to behave similarly with respect to cyclical timing, leading turns in general business activity at business cycle peaks and lagging slightly at troughs. In the absence of a job vacancy time series, however, these relationships cannot be tested. If such time series were to become available, the data would augment key time series now used for business cycle analysis. Comparative analyses could lead to a better understanding of the components of the total of job vacancies at various stages of the cycle: What proportion of job vacancies is attributable to business cycle changes? to structural factors? to "friction" in the labor market?

Analysis of Causes of Unemployment. One of the most controversial aspects of analyzing job vacancy data concerns the relationship of job vacancies to unemployment, and the policy questions that arise from these relationships. The focus of this controversy was the position taken by Arthur F. Burns on this relationship, as shown in the following statement:

... if the number of job vacancies equaled the number of unemployed, there would then be sufficient employment opportunities to permit, in principle, a job for all who are able, willing, and seeking work. This line of reasoning leads at once to a basic criterion of full employment, namely, equality between the number of jobs seeking men and the number of men seeking jobs ...(18)

Burns considered this equality as a "... necessary but not sufficient condition of full employment. The other two conditions are: first, that the equality hold at prevailing wages; second, that the labor market is so organized that practically all of the unemployed could obtain a job after a brief search or after obtaining some special "training."(19) Thus, Burns did not claim that job vacancy and unemployment aggregates, by themselves, provided all the criteria for "full" employment.

This juxtaposition of job vacancies and unemployment as the determinant of whether "full employment" exists subjected Burns to very strong criticism from organized labor, and also from some in academic circles. Burns' critics pointed out that this model represented only one of the many standpoints from which the job vacancy-unemployment relationship could be analyzed. It did not, for example, permit explicit distinction between long-term and short-term forces in the economy. It did not concern itself with an evaluation of the behavior of wages and its interaction with the vacancy-unemployment relationship. Furthermore, in its aggregative form this model did not explicitly deal with changes in contributing factors, such as union practices, labor force characteristics, technology and industrial organization, and the relative importance of industries and regions -- all of which might affect and be affected by structural imbalances, and aggregative imbalance, or both. Critics of his position also pointed out that it did not take into account job vacancies which could never be filled, perhaps because the skills required were not available in the existing labor force, or because jobless workers lacked marketable skills, which made them unemployable.

Even if the equality of job vacancies and unemployment were accepted as an analytical measure, dissenters pointed out that it need not be accepted as a guideline for policy decisions. One academic critic, for example, stated that the equality of job vacancies and unemployment had no simple relationship to the adequacy of aggregate demand. He argued that there was no basis for assuming that job vacancies are excess demand, that they can be used systematically as a measure of market maladjustment, or that the equality of vacancies and unemployment signals the adequacy of aggregate demand.(20)

Measuring the Efficiency of Labor Markets. Trends in job vacancies, especially if classified by occupation, could throw light on the ability of our economy to adjust to changes in labor demand. Job vacancy data can be applied more directly to the measurement of the efficiency of labor markets. Are labor shortages developing? If so, are they due to geographical imbalances between job openings and job seekers? Are these imbalances on a skill level, or on an occupational basis? Together with other labor market information, job vacancy data could be used in such analysis, and could contribute to the formulation of policies that would help to minimize frictional unemployment and reduce structural joblessness. Job vacancy data on an occupational basis, and with sufficient locality detail, would also be a frame of reference for the evaluation of job matching, and for the design of job training programs. Job vacancy data could also throw additional light on demand-supply conditions in the labor market in relation to changing wage levels.

Job vacancy data should dovetail with related labor market indicators. A good fit would indicate the validity of the job vacancy data; conversely, if the job vacancy statistics were found to be valid, they could be used in turn to test the validity of other labor market series. Thus, the trend of employment and unemployment could be related to the trend of new hires, quits, layoffs, and job vacancies. In an analytical test in 1969, BLS did find statistically significant relationships between job vacancies and unemployment, new hires, quits and layoffs. Job vacancy data also should bear a close relationship to labor turnover. Analysis of the movement of the components of the labor turnover series and of job vacancy series would also contribute to the validation of job vacancy data. In turn, analysis of job vacancies together with data on labor turnover would suggest what proportion of the job vacancies at any time represent the typical turnover experience of an industry. The construction industry, for example, generally would be expected to have a higher rate of job vacancies than an industry characterized by more stable employment, such as banking.

Operational Uses The following operational uses of job vacancies data have been proposed from time to time:

  1. Aid the Employment Service in direct job placement by matching unemployed and under employed workers with identified job vacancies.

  2. Indicate the training and counseling needs of prime sponsors (primarily units of state or local government) under the Comprehensive Employment and Training Act (CETA) programs, the Work Incentive (WIN) program and other such government programs.

  3. Point to the direction that vocational training courses and apprenticeship training programs should follow.

  4. Contribute to improved counseling on where the best job opportunities are for workers and new entrants to the labor market.

  5. Aid the Employment Service in job development activities by identifying expanding occupations, industries and areas.

  6. Aid business and industry in formulating more effective recruiting policies. This would be especially valuable to firms which are expanding or relocating.

  7. Stimulate firms to improve their manpower planning.

  8. Facilitate worker mobility by making more effective programs designed to assist in the geographic transfer of workers.

  9. Help to develop more effective programs to deal with plant closures.

  10. Make more serviceable the labor-market attachment test by the Employment Service for unemployment insurance applicants.

  11. Help to implement the alien employment certification requirement of the Immigration and Nationality Act.

  12. Help to assure the migrant and seasonal workers equal access to available job opportunities.

  13. Contribute to the assessment of the effectiveness of the Employment Service in penetrating. occupations, industries and geographical areas.

  14. Help to implement the Resource Allocation Formula for funding of state Employment Service agencies.

  15. Help labor organizations to evaluate the demand for the services of their members and to develop their policies on training, apprenticeship and collective bargaining.

Many of these suggested operational uses of job vacancy data, however, remain mere possibilities in the absence of detailed and continuing job vacancy data.

The most obvious, but in some respects most controversial operational use of job vacancy data -- that of direct job placement -- did not materialize during the JOLT period. The job vacancy data were generally not available to Employment Service offices quickly enough for direct placement purposes, and they did not identify employers and plant locations. This problem might have been solved but for the objection of BLS to such use of job vacancy data. BLS took the position that if the job vacancies in the data gathering sample were filled by pinpointed placement activity, the sample would cease to be representative of the universe. Furthermore, BLS maintained that using the data for direct job placement would violate the pledge of confidentiality given as a condition of obtaining these data from employers, and this breach of confidence could jeopardize all other BLS data collection programs. This conflict between the BLS interest and that of the Employment Service could not be resolved during the JOLT period and was an important factor leading to the discontinuation of the job vacancy program.

Job counseling and training programs could benefit from job vacancy data and even more so if such data were available for future, as well as current job vacancies. The concept of future vacancies, however, does not fit into the design of a job vacancy program intended to be symmetrical with a monthly unemployment count.

The other possible operational uses of job vacancies are subsidiary and, in themselves, not of such overriding importance as to justify a job vacancy program. Another factor to be taken into account is the availability of other sources of information and whether they can serve these operating needs. In the last few years, for example, there has been considerable development of labor market indicators derived from operating statistics of the Employment Service and from other related BLS programs, as well as from nongovernmental sources. These include the Occupational Information Program (OIS) of the Employment and Training Administration (ETA), and the associated Occupational Employment Statistics (OES) program of BLS; as well as the "new hires" program, and the Employment Service Potential (ESP) program, the Job Bank (the computerized job matching program of the Employment Service), newspaper "help-wanted" advertisements, and job listings of private employment agencies....


ENDNOTES

  1. U.S. Department of Labor, "Job Vacancies, Hires, Quits and Layoffs in Manufacturing; July 1972" (press release, August 31, 1972), p. 2.

  2. President's Committee to Appraise Employment and Unemployment Statistics (popularly known as the Gordon Committee, after its chairman), Measuring Employment and Unemployment (1962), pp. 200, 278-279.

  3. Ibid., p. 199.

  4. National Bureau of Economic Research, The Measurement and Interpretation of Job Vacancies (a Conference Report) (1966), p. 459.

  5. John G. Myers and Daniel Creamer, Measuring Job Vacancies, A Feasibility Study in the Rochester, N.Y. Area (New York: The Conference Board, 1967).

  6. Irvin F. O. Wingeard, "Experimental Job Vacancy Survey Program of the United States Department of Labor," in National Bureau of Economic Research, op. cit., p. 350 ff.

  7. National Bureau of Economic Research, op. cit., p. 23.

  8. Job Vacancy Statistics, Hearings before the Subcommittee on Economic Statistics of the Joint Economic Committee, 89th Cong., 2d sess. (Washington: Government Printing Office, 1966), p. 11.

  9. Julius Shiskin, Commissioner of Labor Statistics, Memorandum to Paul Krueger, Office of Management and Budget, December 7, 1973.

  10. Raymond A. Konstant and Irvin F. O. Wingeard, "Analysis and Use of Job Vacancy Statistics: Part I," Monthly Labor Review, August 1968, p. 23.

  11. Myers and Creamer, op. cit., p. 123.

  12. Ibid., p. 51.

  13. New York State Department of Labor, Division of Research and Statistics, Collecting Job Vacancy Data from Private Household Employers: A Feasibility Study (Labor Research Report 1973-13, June 1973).

  14. Myers and Creamer, op. cit., p. 136.

  15. From unpublished reports on the Experimental Job Vacancy Program, 1966.

  16. Myers and Creamer, op. cit., p. 136.

  17. President's Committee, op. cit., pp. 200, 278-279.

  18. Arthur E. Burns, "Economics and Our Public Policy of Full Employment," Morgan Guaranty Survey, July 1963, p. 11.

  19. Ibid., p. 15.

  20. Myron L. Joseph, "Job Vacancy Measurement," Journal of Human Resources, Fall 1966, pp. 59-80.


RETURN TO Job Vacancies Manual

RETURN TO Employment and Training Institute Homepage

GO TO TOP