Comparison of Annual Total and Cohort Data,
1994 School Year

The results in this section have been anticipated in the discussion of the outcome measures based on the annual total data. The cohort data were collected for the first time for the 1994 school year. It will not be possible to provide trend lines for the cohort data until additional years have been collected. Consequently the charts in this section show comparisons of similar, but not identical, outcomes derived from the annual total and cohort data for the 1994 school year only.

We stressed identical in the previous sentence because it should not be expected that the annual total and cohort measures would be the same. The annual totals reflect only the 1994 school year.The cohort data include some students who enrolled in programs more than three years earlier.

Schools were instructed to provide information about students who during the 1994 school year had had one and one-half times the scheduled lengths of their programs in which to complete them. This means that students who began two year programs during the first month of the 1992 school year (July 1991) or earlier would be included in the cohorts who completed their allowed time during the 1994 school year. Programs of six months or less, in contrast, could have completed their allowed time entirely within the 1994 school year.

Given the much different time periods covered by the annual total and cohort data, the school outcome measures calculated for them are more similar than might have been expected. Figure 13 presents the comparisons.

The annual total data yielded lower figures than the cohort data for all three outcomes. In the cohort data, graduates plus withdrawals equal 100 percent. This is not true for the annual total data. In the annual total measures, the graduation rate is based on the number leaving school either through graduation or withdrawal. The withdrawal rate is based on the number enrolled. Since the bases for these two rates are not the same, they do not sum to 100.

Some of the differences between the annual total and cohort data are due to how those students who withdrew because they obtained related employment were counted. In the cohort measures, those who withdrew for related employment were included in both the percentage trained and the percentage placed in related employment. In the annual total measures, these withdrawals were not included in the graduation or TRP rates. The inclusion of those who withdrew for related employment contributes to the higher rates found for the cohort data.

school year 1994

The biggest difference between the annual total and cohort data concerns withdrawal rates. The explanation of this difference that appears most reasonable to us involves students who drop out between school years. They enroll one year but do not complete their programs. They inform their schools that they intend to return the next school year, so they are not counted as withdrawing during their first year. When they do not return for their second year, they are not counted as either enrolling or withdrawing in the annual total data for that year. The cohort data requires a more accurate tracking of such students and thus yields higher dropout rates.

Multiple Regression Comparisons

Figures 14 and 15 compare the multiple regression results obtained for the similar measures obtained from the annual total and cohort data for the 1994 school year. The figures show the school characteristics that were found to have statistically significant relationships in both sets of data.

=

With the cohort data, a less rigorous level of significance (p. = .10) was used than with the annual total data (p. 01). The less rigorous level was used so that comparisons could be made with more characteristics. Additional years of data will be needed to determine if the characteristic found significant in the cohort data are as stable as those presented for the annual total data.

Figure 14. Net effect of selected characteristics on graduation/percentage trained from annual total and cohort data in the 1994 school year

Figure 15. Net effect of selected characteristics on training-related placement from annual total and cohort data in the 1994 school year

The most dramatic difference between the results for the two data sets concerns the effect of enrollments of 300 or less on graduation/percentage trained. The effect of the estimate in the cohort data is one-third of that found in the annual total data. Part of the explanation for the difference has already been discussed the different time periods covered by the two data sets. The annual total data covers only one year. The cohort data may include three years or longer. During the longer time period, school characteristics that are related to outcomes can change: Enrollments may become larger or smaller, program length may change, the school may become more selective in the admission of ATB students. These changes could alter the relationships observed during individual school years between these characteristics and outcomes.

Having said this, it must be acknowledged that the estimates of the effect found for turnover of fulltime faculty were virtually identical in the two data sets. The magnitude of this effect in the annual total data was much larger than previous estimates. As discussed earlier, we feel these results are due to a few schools that in 1994 had unusually high rates of both turnover and withdrawals.

Figure 15 presents the three school characteristics found to have significant net effects on trainingrelated placement in both data sets. It is hard to come to any reasonable interpretation for the results of program length and default. The program length effect is negative in the annual total data and positive in the cohort data.

The explanation for the third characteristic is quite straight-forward and reasonable: it is harder to find related employment in areas with high rates of unemployment. For every one point increase in the unemployment rates for the areas where the schools and colleges are located, TRP decline by onehalf of a percentage point. The net effect of area unemployment rates was not shown as significant in Table 2 because information on these rates has only been collected for the past two school years.

It is surprising that the school characteristic found to have the strongest net effect on TRP in the annual total data (Table 2), school enrollments of 300 or less, was not significant in the cohort data. On the basis of only the 1994 school year, school size does not appear to be as powerful an influence on outcomes in the cohort data as it has proved to be in the annual total data.

The final section of the report compares the findings for the ACCSCT schools and colleges to those for similar postsecondary institutions.

Comparisons with Similar Programs

A review of available studies of postsecondary technical training (sources are listed in the Technical Notes) imply that schools and colleges accredited by ACCSCT graduate a higher percentage of their students than similar public institutions. Prior to the 1994 school year, we had less confidence in the validity of this statement. That is because the data from previous years were based on annual totals. These totals summarized the operations of schools for a given year, but they did not track students across school years.

In 1994, cohort data, similar to that used in most other studies, were also available. The cohort data traced students form initial enrollment to their status at the end of the cohort period. The results from the cohort data indicate that the outcomes derived from the annual totals underestimated both the percentage of students graduating and the percentage withdrawing.

How could the measures used with annual totals underestimate both percentages? They could because graduation and withdrawal were calculated on different bases. The graduation rate was based on the number of students graduating as a percentage of the total graduating and withdrawing during the school year. The definition of graduation used with the annual total data was developed to allow for the students who neither graduated nor withdrew within one school year. In comparison to the rates found for the cohort data, the definitions used with the annual total data underestimated graduation by about 5 percent and withdrawal by about 10 percent. We attribute these higher figures to the more detailed tracking of students required for the cohort data.

The measures of related placement used to evaluate technical training typically report the number of graduates who obtain jobs that are related to their training as a percentage of the number available for employment. By this measure, the rate at which ACCSCT schools place their graduates appears to be about the same or slightly lower than those at public institutions. If however, Commission accredited schools graduate more of their students, and about the same percentage obtain related employment, these students are being better served than are those at public institutions. This is a surprising finding given that the ACCSCT schools enroll proportionally more high school dropouts, minorities, and economically disadvantaged students.

In summary, this report shows that Commission accredited schools and colleges graduated over twothirds of their students and that over three-fourths of these graduates obtained employment that was related to their training. Other studies suggest that these outcomes compare very favorably with the results of technical training provided by comparable institutions.