REFINEMENTS OF AVAILABLE EMPLOYMENT AND
EARNINGS MEASURES

Notes: The reference quarter is 1991:4.
"Percent of Total" is the percentage of the total population.
"Percent of Employed" is the percentage of the former students who were reported as employed in 1991:4.
"Percent of Subtotal" is the percentage within each category defined by the relatedness determination methods.
The "Based on Industry Only" rows include only those whose job titles were not available.
The "relatedness" approach used to prepare this table was provided by Florida's Employment and Training Placement Information Program.
The approach is considered to be in a pilot use phase and is subject to future refinement.

The omission of the somewhat related category from the industry only assignment method reflects the FETPIP staff's unwillingness to adopt such an aggressive classification approach.

Three columns of results are presented in Table 1:

1. Percent of Total--Distributes all of the 1990-1991 reference group community college vocational program completers among the five determination method categories.
   
   
2. Percent of Employed--Distributes only those reference group members who had been reported to the state employment security agency as employed during the reference quarter.
   
   
3. Percent of Subtotal--Distributes, within subgroup, only those who have been assigned a training-relatedness code on the basis of either the job title or industry only method.
   
   

For many reporting and management decision purposes, the middle and right-hand columns are more useful than the "Percent of Total" column. However, it is always important to be able to account for all members of an original reference population, so detractors are not allowed to falsely assert that there must be a devious reason why some have been omitted from a tabulation. The upbeat news in the middle column is that 77% of the former students who had been reported as employed during 1991:4 were successfully assigned to a training-relatedness category using the job title approach, and another 11% were assigned using the fall-back industry only method.

The only number that many interested parties will look for and remember is the top figure in the middle column--the percent of those who were reported as employed in 1991:4 and were determined to be working in a directly related job. Many of those who seek this single number are likely to depart without recognizing the importance of the number at the top of the right-hand column: Two-thirds of those who could be assigned a training-relatedness status using the job title method were assigned to the directly related category.

Table 1 typifies one reason why vocational educators are distributed along a continuum of enthusiasm about the use of state employment security agency employment data, and associated attempts to assign training-relatedness designations to these jobs. There is something in this tabulation for everyone. Detractors can point to the not employed, unable to determine, and not related cells. Advocates can focus on the directly related and related cells. This is why either of two alternative approaches is encouraged--do nothing or do more.

Table 2 illustrates one way in which FETPIP has done more. This tabulation replicates the Percent of Total column from Table 1 for each of six postsecondary vocational programs. This choice of presentation format reflects the anticipated interests of likely readers of this volume. It does not represent the recommended selection for a press conference.

Notes: The "relatedness" approach used to prepare this table was provided by Florida's Employment and Training Placement Information Program.
The approach is considered to be in a pilot use phase and is subject to future refinement.

The comparative data presented in Table 2 exemplifies the theme of the entire volume: Available data can be organized in ways that support and promote local and state management diagnostics--formats that may be unrelated to required federal reporting or to responding to constituent inquiries. Here, the word diagnostics is meant to convey a motive of behind-the-scenes troubleshooting or curiosity, rather than triggering a public argument about the accuracy and relevance of particular numbers. Note, for example, that the range of combined not employed/unable to determine cases extends from a high of 50% to a low of 18%. Awareness of this pattern alone is cause for management action to learn more about why, and whether, anything should be done to try to reduce this disparity. Note that the terminology used here is whether anything should be done, not what should be done. The disparity may be traced to origins that do not warrant corrective action.

Table 3 probes to a deeper level of understanding of the training-relatedness issue. Four post-schooling measures of employment status and earnings are reported by each of the relatedness categories:

1. The percentage of each subgroup of former students who were reported as employed by a different employer at some time during 1992 than had been used as their reference affiliation in 1991:4
   
   
2. The average length of first job held, in quarters
   
   
3. The average level of reported earnings during the reference quarter, 1991:4
   
   
4. The average level of reported earnings one year after the reference quarter, or 1992:4

Notes: A "different employer" is defined as any employer identification code other than that of the reference quarter, which determined the training-relatedness code.
The "relatedness" approach used to prepare this table was provided by Florida's Employment and Training Placement Information Program.
The approach is considered to be in a pilot use phase and is subject to future refinement.

Table 3 reveals as low a turnover rate, as long an average length of first job held, and a higher average earnings level during the 1991:4 reference quarter, for the directly related subgroup relative to each of the other five designations. A statistically significant difference in turnover rate and average length of first job, favoring the directly related subgroup, is found in a comparison with the not related subgroups alone.

The turnover rate and average length of first job, as these appear in Table 3, are not independent events. Also, care must be exercised in defining length of first job. Here, it means the number of quarters after a former student leaves school during which the reported employer affiliation is the same. Even this seemingly straightforward definition can become complicated, when it is realized that the appearance of more than one wage record in a quarter may indicate that an employee has moved between employers during the quarter. Also, for some reporting purposes, the pre-school-leaving quarters of employer affiliation may be considered relevant.

One year after the reference quarter of 1991:4 the initial earnings advantage enjoyed by the members of the directly related subgroup (determined using the job title method) has been closed by the industry-related subgroup. This is the type of finding that should trigger further inquiry: What explanation comes to mind for this pattern? Is information conveyed by an industry affiliation that has potential value as a predictor of a former student's long-term earnings prospects? If so, this readily available data element, which is contained in state employment security agency records (but not in the wage record database in most cases) can be used for selected analytical and reporting purposes.

Table 4 represents a preliminary attempt to investigate the potential value of the readily available Standard Industrial Classification (SIC) code as a complement to, and perhaps even as a substitute for, the costly and often challenged documentation of training- relatedness. Each column of numbers in Table 4 presents regression coefficients for variables that might reasonably be thought to be correlates of differences in reported earnings during the reference quarter of 1991:4. The basic purpose for this specification is to explore how the coefficients for training-relatedness and SIC code behave together and separately. The results indicate that the industrial classification variable absorbs enough
of the training-relatedness variable's explanatory power that some signs change and the statistical significance of some correlations is lost.

Notes: Italic indicates a 0-1 dummy variable.
Local-Avg-Earn is the average 1992:1 earnings of all 1990-1991 of the state's high school and community college students by groups of counties chosen by the authors.
Pre-Job refers to the job held by individuals in 1991:2.
Industry-Specific Avg. Earnings are based on two-digit SIC code. The 1991:4 earnings of all workers (in a different state for which the necessary data was available) are used to calculate the average earnings and standard error of the average earnings.

This brief excursion through the topic of training-related employment is intended to increase reader awareness of the complexity of the issue, but also to indicate some promising paths for future inquiry. The priority that is given to this issue ranks right behind the placement measure based on a management support criterion. Since the two concepts are traditionally paired in management use, together they warrant the highest available designation for serious attention.

Alignment of Enrollment and Employment Affiliations

A typical vocational education follow-up design identifies a reference population of former students, usually a school-year of leavers, and carries out a single snapshot of their employment status during some interval soon after this. Adoption of this approach in high school settings is of minor concern because most seniors who graduate do so during May or June; but there are many reasons why an investigator might want to know the year/month of last school attendance for those who leave without receiving a diploma.

Figure 8 shows why such complacency is not justified in the case of community college inquiries. Each of the four panels in Figure 8 represents one subgroup disaggregated from a total of 6,491 vocational program completers in school year 1989-1990. Those in the top panel, 10.4% of the total, left school in the summer quarter (July-September 1989). Those in the next panel, 21% of the total, left in the fall quarter (October-December 1989). The third panel from the top represents 14.3% of the full year's class who left in the winter quarter (January-March 1990). And the bottom panel covers the 54.3% who left in the traditional spring quarter (April-June 1990).

The asterisk in each of the four panels represents the July-September 1990 quarter, which can be thought of as a snapshot of the percent of the reference subgroup of former students who were reported as employed in this quarter. Note that for the 1989:3 graduates, this is the fourth full quarter following school-leaving; while for the other three subgroups of the class of 1989-1990, this is the third, second, and first full quarter after leaving school respectively.

Those represented in the top panel had a full year to hold, and perhaps to leave, one or more jobs. Those included in the bottom panel had a maximum of three months, if they left school in June 1990. Nothing more can be said, based on Figure 8 alone, about the sequences of employment affiliations that have occurred.

Figure 9 reveals why the limitations associated with Figure 8 should be of concern from a management support perspective. Here, a seven-quarter reference period is identified for each member of the 1989-1990 community college vocational program completer population. Keep in mind that the particular seven-quarters differ among the four subgroups of summer, fall, winter, and spring school leavers. For instance, for the summer 1989 completers, the seven-quarter reference period begins in January 1989 and ends in September 1990; while for the spring 1990 completers, the reference period begins in October 1989 and ends in June 1991. This means that ten quarters of data were required to cover the thirty reference months. This compares with the typical one-quarter coverage of most follow-up designs.

The darker bars in Figure 9 represent the summation of seven sequential snapshots of reported employment status from the state employment security agency's wage records database. The pattern is what might reasonably have been expected, with a stable employment rate during the two quarters prior to completion, which increases somewhat during the school-leaving quarter, and then increases again to a higher stable plateau over the next year. This contrasts with the lightly shaded bars, which reveal a continuous decline in the percentage of former students who are still affiliated with the same employer who
had reported them as an employee during their last full quarter prior to the quarter of vocational program completion.

One important implication of Figures 8 and 9 together is that when a single quarterly snapshot of employment status is recorded for an entire school year's reference population of former students, the probability of capturing the first job held will differ for subgroups within this population. Based on the data that underlie these two figures, the range of sustained employer affiliation would be expected to range between a low of 23% (four quarters following completion) and a high of 38% (one quarter after completion).

The diagnostics reflected in Figures 8 and 9 require more data than is needed for the typical one-quarter snapshot of post-schooling employment status. In this case, ten quarters of data were used. The following issues arise in attempting to conduct this type of investigation:

1. Most state employment security agencies maintain only the most recent available five quarters of wage records on-line because these are used in the routine conduct of the state's unemployment compensation program. Currently, this statement does not apply to Michigan and New York, which are referred to as wage request states because they request verification of earnings when a former employee files a claim to receive unemployment compensation benefits. Michigan does require employers to submit a quarterly report of employee earnings, but these are not used in the same way that other states use them. Massachusetts is the most recent state to enact wage reporting legislation, so the potential availability of historical coverage will be extremely limited there.
   
   
2. Employers are required by law to submit their report on a particular reference quarter within thirty days of the end of that quarter. The state employment security agency then has at least sixty days to process the data so they are available for administrative use by the beginning of the next quarter. Late reporting does occur, which should be considered in deciding when a request for wage records will be submitted to a state employment security agency.
   
   
3. State employment security agencies have uneven capacities to respond to external requests for information. The authors' ability to perform these diagnostics through the auspices of a small federal grant can be traced to the pioneering, and still unique, agreement between the University of Baltimore and Maryland's Department of Labor, Licensing, and Regulation. This interagency agreement between two branches of state government provides for the archiving and research use of the universe of Maryland's wage records, which currently cover the period from April 1985 through March 1995. The data is maintained in encrypted form in a secure facility managed by The Jacob France Center in the University's Merrick School of Business. The Center's database manager and her research colleagues have each signed an oath acknowledging their awareness of and intention to honor the confidentiality stipulations that apply to these records. Each research use of the data must be approved by an authorized person in the originating Department of Labor, Licensing, and Regulation. The Center's executive director is responsible for assuring that any public release of information conforms to the rule that no individual or reporting entity can be identified, either directly or inferentially, on the basis of available data elements such as sex, race, school attended, year graduated, and program completed. This interagency agreement frees the researchers from the queuing issues discussed earlier. The quid pro quo is that the Department of Labor, Licensing, and Regulation has expanded the timeliness and scope of its own ability to answer questions using this database. Based on this unique agreement, the four-state database that underlies the present discussion was assembled by entering into fourteen agreements with multiple state agencies in the four states for the specific research purpose stated in each agreement. While there is no known precedent for this approach, federal government contractors have engaged in project-specific data acquisition of a similar type. Recently, the Unemployment Insurance Service in the U.S. Department of Labor has taken advantage of the archived data at the University of Baltimore to conduct inexpensive inquiries in a timely manner; investigations that could not have been accomplished as quickly at low cost, and may not have been feasible at all through other sources. Legal issues aside, each agency has multiple internal data processing and reporting obligations. Both predictable seasonal fluctuations, and unpredictable pressures associated with recessions that increase the number of unemployment compensation claimants, will affect the agency's ability to respond. Even established interagency agreements that specify a particular schedule of response times may be ignored when unexpected legislative actions impose new demands on the agency's data processing unit.
   
   
4. The states vary widely in the pricing of interagency requests for record linkages. In many cases, informal understandings of the past have given way to formal agreements. There does not appear to be any commonly followed standard for pricing. Some agency's charge on a per-record-matched basis. Others charge on a per-quarter batch run basis, or on the number of records provided by the external party. Some charge on a marginal direct cost basis, while others include standard overhead rates. Many have different rates for particular categories of request.

Documentation of Continuing Education

The importance of this conclusion is illustrated in Table 5, which displays continuing education information for three populations of vocational program completers. Both high school and area school postsecondary levels are represented here. Three annual classes of completers are presented to offer a sense of the stability of the transition flows that are represented. The sex, race, and age attributes reveal the heterogeneity of the vocational education population.

Table 5 is based on the reference state's own matching of school district records (vocational completers only) and subsequent reporting of a former student's enrollment in one of the state's higher education institutions. Particular attention is drawn to the sensitivity of the enrollment rate of former African-American students to the 1990-1991 recession; a pattern that is not observed for any of the other groups. This revelation exemplifies how the data can be used to identify possible opportunities for administrative action. In this case, both instructional staff members and counselors can be alerted to the apparent vulnerability of African-American students when economic conditions weaken. Higher education authorities will be interested in the cyclical volatility of their expected enrollments, and the demographic twists that might be expected. They should also be alerted to the possibility that a transition difficulty has been shifted from one educational level to another. The relevance of this concern depends in part on the economic circumstances that arise at the time when these enrollees leave the higher education cocoon.

State governance of public education varies too much to offer detailed recommendations for the steps that should be followed to acquire higher education data. Some states, including FETPIP, are introducing coverage of private postsecondary institutions.

Multiple Employer Affiliations

Multiple Wage Records in One Quarter

The best practical rule that can be recommended for identifying sequential affiliations is to conduct quarter-to-quarter matches that reveal whether one of the affiliations disappears in the next quarter. If so, then there can be a reasonable presumption that this was a previous employer and that the one that appears in both quarters followed.

The primary job issue is more difficult to clarify because there is no reliable time-unit of employment available (e.g., weeks worked). Different combinations of low earnings and many weeks, or high earnings coupled with just a few weeks of employment, will result in the same total quarterly earnings amount. One approach is to define primary on the basis of multiple quarters of data. If a former student maintains one affiliation continuously, and another only occasionally, then the first job can reasonably be designated as the primary affiliation.

Ultimately, the decision rule that is applied should be based on the specific intended purpose that the investigator has in mind. What may be thought appropriate in one case might be rejected in another situation.

The Relevance of Multiple Employer Affiliations

The two community college vocational programs that appear in Figure 10 were chosen to highlight differences. Neither is representative, or typical, of the entire range of vocational offerings. Indeed, historically, one of the problems in discourse about vocational education has been a failure to distinguish among extraordinarily varied curriculums. Here, a three and one-half year post-graduation reference period is covered.

The first figures that virtually leap off the page are the continuing employer affiliation rates--73% for the completers of the health/medical curriculum, and 54% for the completers of the marketing and retail curriculum. These are the percentages of the respective reference groups who cannot be said to have been placed. There has been no
transition from school-to-work for these former students, at least not in the traditional sense of that term.

A second pairing of numbers represents those who were still employed in what is referred to here as the first job at the end of the three and one-half year reference period--53% of the health/medical program completers and 38% of the marketing and retail program completers. The differences in the combined rates of left- and right-truncation (i.e., continuous employment that started before leaving school and, for some, continued through the end of the observation period) account for the more than six months difference in average length of first job held between the two program completion populations.

Diagnostics of the kind displayed in Figure 10 help an investigator, and any other interested party, to understand the interplay between the school curriculum, enrollees in varied components of this curriculum, and the employment opportunity set that was included in Figure 1 earlier.

Censoring Reported Earnings

Tables 6, 7, and 8 provide a pot pourri of data elements for reader consideration. Two new concepts are introduced here. The first, full earnings, requires a former student to have reported earnings in each of the four quarters of the reference year, and to have earned more than $8,667, which is the appropriately inflated 1989 earnings level that was self-reported in the 1990 Census by those who said they had worked forty hours or more per week for forty-eight or more weeks during the year, and who fell at the 5% point in the lower tail of the distribution of earnings for this group. The 5% in the lower tail of the distribution was chosen to eliminate outliers that might be questioned as data entry errors or special circumstance cases. This concept is intended to include only those who have reported earnings in each of the four quarters and who earned at least as much as the members of this comparison group of low-earners among all respondents to the 1990 Census who can be classified as having been employed full-time year-round in 1989. This concept of full earnings is used in each of the three tables in the series. The second concept, full-time earnings, only appears in Table 8. This concept is defined as those who were reported to have worked at least 1,920 hours during the reference year of 1991. This is the equivalent of forty hours a week for forty-eight weeks a year. The data to carry out this calculation were obtained from Washington's State Employment Security Department, which is one of the few state agencies that requires employers to report hours of work associated with the earnings of each employee.

Notes: "Full Earnings in 1991" is defined as "Earnings in 1991" if earnings were reported for each of the four quarters, and if this earnings amount is equal to or greater than $8,667; which is the inflated 5% quantile of 1989 full-time workers' earnings in the corresponding 1990 census group. The cut-off point for 1992 earnings is $8,887. Full-time is defined as 40 hours or more per week, 48 weeks or more per year.
The significance tests are for the difference between mean earnings levels for the vocational and academic groups, adjusted for the different distributions of "sex" in these groups. The tests are based on 5% significance level.
1992 earnings are deflated to 1991 by factor 1.025.

Table 6 presents actual earnings with no restriction on the number of quarters of reported employment, and full earnings using the censoring criteria that have been described in the previous paragraph, for the two years following the school year during which the members of the reference population received an associate's degree. This information is presented for male and female degree recipients separately, within vocational and academic groupings. This type of presentation can be replicated for any combination of curriculum and demographics, as long as the confidentiality stipulations are honored.

Notes: "Full Earnings in 1991" requires that earnings were reported for each of the four quarters, and that the combined earnings amount is equal to or greater than $9,751, which is the inflated 5% quantile of 1989 full-time workers' earnings in the corresponding 1990 census group. "Full-time" is defined as 40 hours or more per week, 48 weeks or more per year.
The significance tests are for the difference between mean earnings levels for the degree and certificate groups, adjusted for the different distributions of "sex" in these groups. The tests are based on 5% significance level.

Table 7 provides a more detailed look at the actual and censored earnings levels within the vocational curriculum. Here, three levels of formal recognition of educational accomplishment are identified--associate's degree recipients, completers of a vocational certificate program that lasted at least the equivalent of one year's credit hours, and those who received a certificate for a shorter course of study. Again, male and female students who reached these plateaus are identified separately. The diversity of average annual earnings that appears in Table 7 provides yet another bit of the accumulating evidence that aggregates and snapshots mask major differences beneath the surface. These differences are often of great importance in the decision-making process. Diagnostics of this type can improve the quality of these decisions. It is particularly important to recognize the comparative earnings levels associated with the three types of degree or certificate. There is a clear hint here that previous work experience, and perhaps other education credentials, should be considered in any attempt to treat these earnings figures as vocational education outcomes.

Notes: "Full Earnings in 1991" requires that earnings were reported for each of the four quarters, and that the combined earnings amount is equal to or greater than $9,751, which is the inflated 5% quantile of 1989 full-time workers' earnings in the corresponding 1990 census group. Full-time is defined as 40 hours or more per week, 48 weeks or more per year.
"Full-Time Earnings in 1991" is the average earnings of those who worked at least 1,920 hours (40 hours per week, 48 weeks per year) in 1991.

Dependent Variable																																																			Earnings in 1991:4																																									Earnings in 1992:4																																									First Job Length																																									Chngd Emp. in 1992
Regression Type																																																			Linear Regression																																									Linear Regression																																									Linear Regression																																									Logistic Regression
Number of Observations																																																			11,048																																									11,048																																									14,674																																									14,674
R-squared or C																																																			0.5452																																									0.4419																																									0.2549																																									0.637
Population																																																			Employed in 1991: 4																																									Employed in 1991: 4																																									Have Post- School Job																																									Have Post- School Job
																																																			Estimates     $																				P-Value																					Estimates    $																				P-Value																					Estimates  (qtrs.)																				P-Value																					Estimates    $																				P-Value
Intercept																																																			-624.243																				0.112																					-1,089.821																				0.024																					3.015																				0.000																					-1.599																				0.000
Demographic Variables
Male																																																			358.952																				0.000																					650.800																				0.000																					-0.001																				0.965																					0.053																				0.246
African American																																																			-359.124																				0.000																					-402.615																				0.000																					-0.059																				0.167																					0.123																				0.039
Hispanic																																																			124..937																				0.162																					113.325																				0.300																					0.018																				0.774																					0.041																				0.648
Vocational Program
Agriculture																																																			-184.914																				0.714																					77.211																				0.897																					-0.515																				0.159																					0.769																				0.111
Office Ocupations																																																			-67.847																				0.410																					-92.355																				0.361																					0.023																				0.687																					0.024																				0.763
Engineering Technologies																																																			-104.061																				0.310																					161.113																				0.198																					-0.024																				0.738																					-0.044																				0.675
Allied Health																																																			-4.262																				0.934																					333.042																				0.000																					-0.303																				0.000																					0.362																				0.000
Medical Science																																																			2,226.776																				0.000																					2,714.662																				0.000																					-0.005																				0.912																					0.244																				0.000
Child Care/Food Service																																																			-817.925																				0.000																					-1,016.304																				0.000																					-0.342																				0.007																					0.094																				0.590
Local Economics Conditions
Local-Avg.-Earn ($)																																																			1.017																				0.000																					1.555																				0.000																					0.000																				0.188																					0.000																				0.003
Pre-Graduation Job Info.
Pre-Job																																																			-806.899																				0.000																					-1,098.645																				0.000																					0.725																				0.000																					0.045																				0.045
Earnings in 1991:2																																																			0.614																				0.000																					0.559																				0.000																					0.000																				0.000																					0.000																				0.000
1991:4 Job Relatedness
Directly Related																																																			662.065																				0.000																					505.058																				0.000																					1.243																				0.000																					-0.393																				0.000
Somewhat Related																																																			478.926																				0.000																					404.306																				0.000																					1.214																				0.000																					-0.263																				0.002
Not Related																																																			-274.206																				0.000																					-155.603																				0.067																					0.690																				0.000																					0.167																				0.008
Industry Related																																																			347.044																				0.000																					546.003																				0.000																					1.135																				0.000																					-0.276																				0.001
Directly Not Related																																																			-538.108																				0.000																					156.325																				0.422																					0.519																				0.000																					0.358																				0.013
Changed Employers in 1992																																																																																												-660.157																				0.000

Notes: Italic indicates a 0-1 dummy variable.
Local-Avg-Earn is the average 1992:1 earnings of all 1990-1991 Florida high school and community college students in the state by groups of counties selected by the authors.
Pre-Job refers to the job held selected by the authors in 1991:2.
Industry-Specific Avg. Earnings are based on two-digit SIC code. The 1991:4 earnings of all workers (in another state because the data was available) are used to calculate the average earnings and standard error of the average earnings.

Table 8 uses the same format as Table 7, and repeats the full earnings figures from that table. The new feature here is the introduction of the full-time earnings figures. The intent is to demonstrate the sensitivity of reported earnings figures to the censoring rule that is used, and to highlight the gap between either of these censored figures and the actual reported levels that emerge when no restriction is imposed.

Specification and Estimation of Models

The clear theme of Figure 1 is that interdependence of many forces must be considered in any serious attempt to estimate the outcomes of vocational education. Three accessible examples of serious expert investigation of these forces are Grubb (1995) and Kane and Rouse (1995a, 1995b).

The authors of the present guide are now collaborating with Rouse in the creation of a new database that will include longitudinal coverage of sequential cohorts of high school students in the Baltimore City Public Schools, some of whom continued on to one or more of Maryland's public community colleges and/or to one of the eleven teaching campuses of the University of Maryland System. These files complement the decade of Maryland employment and earnings data that are already available to the authors.

Each database has unique strengths and weaknesses. Grubb (1995) and Kane and Rouse (1995a, 1995b) have used the National Longitudinal Survey of the High School Class of 1972 (NLS-72) and the National Longitudinal Survey of Youth (NLSY) to conduct sophisticated investigations of the payoff to investments in community college education. Each of these data sets contains some variables that do not appear in a simple merger of student transcript information with state employment security administrative
records. These data sets are well-suited for the type of research conducted by Grubb and Kane and Rouse, but not for management diagnostics. Management diagnostics occur at the local and state levels. These require timely information that can be used to motivate exemplary performance by subordinate program managers and teachers in these programs.

Congressional budget debates make it clear that sustained federal investment in high-quality longitudinal data sets may be in jeopardy. State vocational education leaders have rarely received actionable support from research conducted with these data sets. This is not their intended purpose. This is why complementary reliance on a state's own administrative records is encouraged. When and if a national distributed database capability becomes a reality some convergence between the two approaches might occur. Meanwhile, there is ample evidence of success in pioneering states, complemented by awareness of how to advance to a higher plateau of understanding, that should be sufficient motivation for any state leader to join the growing team of wage record users.