Crain, R. L., Allen, A., Thaler, R., Sullivan, D., Zellman, G., Little, J. W., and Quigley, D. D. (1999). The Effects of Academic Career Magnet Education on High Schools and Their Graduates (MDS-779). Berkeley: National Center for Research in Vocational Education, University of California.

Career Magnet Schools:
Effects on Student Behavior And Perceived Parental Support:
Part Two

      The Zellman and Quigley finding about parental support for college is one of the most important findings in recent research on high schools. Considering it, we decided for this report to take the next step and ask if the career magnet graduates were more successful in college. There is no difference in the rate at which students from comprehensive schools or career magnet programs begin college. We constructed a measure of college success by combining the number of credits earned by the time of the interview and whether they had declared a major.[13] Table 5.4 shows the result.

      When the two variables are combined by standardizing each and summing, the correlation with career magnet assignment is .269, (p = .01). The effects are large and would be even larger if there were not "assignment errors" in our "experiment." In this sample, we selected only those students who had applied to the career magnets and either won the lottery and graduated from the career magnet, or lost the lottery and graduated from a comprehensive school. Unfortunately for our research, 11 of the students who entered comprehensive schools were later placed in career magnet programs, and since no record of program placement is kept, we did not learn this until we interviewed them. These students also have a high college success rate and make the relationship shown in Table 5.4 appear weaker than it would be if we had been able to omit the career magnet students from the control group.

      As is often the case with good research, the questions that are answered raise new questions that are more interesting and often more difficult to understand. The questions asked about parental support seem straightforward--they seem to measure the family's financial resources and the parents' attitudes. Career magnets cannot possibly affect a family's finances, and it is hard to imagine how they could affect the attitudes of the parents more strongly than they affect the students' answers to any of the hundreds of questions asked about the student's attitudes and experiences. Zellman and Quigley pursue this analysis and find evidence that student commitment to a career, their sobriety, and the extent to which they press their parents for financial help[14] affect their parent's willingness to pay for college.

      Zellman and Quigley also point out that the graduates, at the same time they were earning more college credits, were also working at higher paying jobs. We had an opportunity to observe the way in which career magnet students had been prepared for work. We "hired" the 30 students who participated in the life history interviews to come in as teams of career magnet graduates and comprehensive graduates in order to perform a group work exercise. The career magnet graduates came in on time, dressed for the office, and worked together much more effectively.

      Career commitment and sobriety are affected by attending a career magnet. Zellman and Quigley's findings suggest that attending a career magnet, by changing a student in these ways and perhaps motivating them to press their parents for financial support for college, may cause their parents to change their behavior as well.

      What can we make of this? There are several plausible hypotheses, but no easy way to test them.

• Hypothesis 1: The School as Advertisement
  Zellman and Quigley stress that even when student behavioral factors are taken into consideration (and every other question in the questionnaire has been tested), attending a career magnet school affects the student's perception of their parents. Most of this effect is not the result of any of the intervening variables that were measured in the two-hour questionnaire. This raises the possibility that there is a cognitive or ideological process here--that attending a career magnet is a sign to parents that their student should go to college. The conclusion seems to be that part of what a school does is to serve as an advertisement, a persuasive argument, for college, independent of anything else the school may do for the student. According to this analysis, these data answer the question "What is in a name?" by saying "a great deal." Obviously, the name must be communicated over and over again in the words and actions of the adults and other students in the school; however, the goal of the program seems to have a value of its own, independent of how the program actually operates.
  
  The power of the idea of the academic career-focused high school is transmitted through the student to the parents. The student of a career magnet who is drinking less, and who is pressing his or her parents for college tuition, has been affected by the idea of the career magnet and, in turn, is passing that excitement on to the parents.
  
  
• Hypothesis 2: The School as Culture
  This is the hypothesis that Zellman and Quigley lean toward, although they are appropriately cautious about the data, that perhaps the career magnet program creates a culture whose diverse elements combine formally and informally to promote career discussion, career and college planning, and realism about the future.
  
  
• Hypothesis 3: The School as Career Preparation
  Finally, we should consider the possibility that the academic career focus is not a catalyst for improving school quality but simply more important than school quality. Perhaps these high school programs are not better in any way, but they are a place where one is promised a career. Of all eighth graders, 82% apply to a magnet. More than half of these magnet programs are academic career focused. Perhaps students react differently because they are aware that they are being prepared for a career. Perhaps they are more supportive of the school, and more committed to their studies, as part of giving their school the benefit of the doubt because they think the school is trying to help them.

      There are no doubt other hypotheses one could offer. Whatever the explanation, there is an effect worth explaining here. The effect of graduating from a career magnet is perhaps most striking in terms of what Zellman and Quigley call "risk behavior." Table 5.5 shows the impact of attending a career magnet on alcohol consumption.

      Zellman and Quigley are correct in assuming that the randomized experiment eliminates the need for controls on background factors. To be certain of this, Table 5.6 tests for the effects of bias. There is a bias introduced, not in the randomization, but because the two types of schools differ in the type of students who can graduate. Graduates of career magnet schools had higher seventh-grade math scores than comprehensive graduates; almost none of the career magnet graduates in our sample have very low math scores. Chapter 2 shows that the career magnet programs are academically more demanding; however, Table 5.6 also shows that the presence of students with very low math scores among the comprehensive school graduates does not explain any of the difference in alcohol use. The regression shows that impact of the career magnets is only significant in eleventh and twelfth grades and in the postgraduate interview; in these grades, controlling on middle-school test scores, ethnicity, and parents' educ`tion actually increases the effect of career magnets. In Model II, in eleventh grade through postgraduate, graduates who have high math scores actually drink more, not less, but the results are not significant. In Model III, the multicolinearity among the four test scores makes it impossible to interpret the coefficients, including the supposedly statistically significant ones (in opposite directions!) for seventh- and eighth-grade math scores for postgraduates, so we cannot conclude that middle class graduates drank more than poorer graduates, but it does seem safe to conclude that they do not drink less.

      Career magnet attendance not only reduces alcnhol use; it also lowers the total risk scale scores (drinking, drugs, smoking, pregnancy, and fighting). Table 5.7 shows that controlling on background variables does not alter the impact of career magnet programs in reducing risk behaviors. (Combined risj behaviors are not measured separately for each year, so this table is in a different format.)

      We were concerned that the lower levels of smoking, drinking, and pregnancy among career magnet graduates might be the result of the presence of a number of students from a Health Careers magnet; however, Table 5.8 shows that graduates of the three non-health career magnet schools also have lower rates than do graduates from comprehensive high schools. Tables 5.9 and 5.10 show similar results for respondents' perceptions of parental support for their college education and number of college credits earned. Again, controlling on background factors makes the apparent effect of career magnets slightly stronger. Table 5.10 shows that students with higher test scores have slightly more college credits, but the difference is not significant.

      It is important to bear in mind that the experimental design eliminates all the unknown factors that might distinguish the lottery winners from the lottery losers; the only significant factors are those that occur after randomization has occurred, namely the differential acceptance of students into the comprehensive schools or career magnet programs, and the differential dropout rates of the two kinds of schools.

      The argument presented below, that the difference in graduation rates cannot explain the difference in parental support, uses what is sometimes referred to as a "contradiction proof." The demonstration will show that if it were possible to eliminate all the relationships between career magnet attendance and parental support by controlling on a factor biased by the difference in graduation rates, then the impact of this biasing factor on the graduation rate would be so large as to defy credibility.

      Let us assume that there is such a factor--call it "f"--perhaps built by combining a large number of factors into a single variable. Since the career magnet graduation rates cannot directly affect parental behavior, or even the students' perceptions of parent behavior, we must postulate a two-step process:

      Graduation rate "g" -> unknown factor "f" -> perceived parent support, "p"

      In educational research, correlations between student-level attitudes or behaviors are rarely higher than .3; correcting for measurement error might push the correlations "rgf" and "rfp" as high as .5. If so, the correlation between career magnet graduation rates and parental support, "rgp," would be no higher than .25. To be on the safe side, let us assume it is .95. The assumption we are testing in this "proof" is that the correlation between attending and graduating from a career magnet compared to a comprehensive school, "c," and parental support, "p," will be reduced to zero if we control on the graduation rates of the two types of schools, "g." Since the randomization prevents the career magnet attendees from differing from the students attending the comprehensive schools, any difference in the value of "f" for the career magnet and comprehensive graduates must be because of the lower graduation rate in the career magnets. (This assumes that there is no difference in the kind of students who graduate--both kinds of schools have the same kind of standard for graduation in terms of test scores and grades, and the same kind of work is required to obtain passing scores in classes and on tests--except that the standard is higher for the career magnets. In other words, we assume no second-order interaction effect between the "f," the graduation rate, and whether the program is career magnet or comprehensive. We drop that assumption in the more complicated version of this proof, below.)

      In order for "f" to explain away the relationship between career magnet attendance and parental support, we must assume that the higher parental support of the career magnet graduates shown by a correlation between parental support and being a career magnet graduate, rcp = .306, is entirely or largely explained by the correlation between being a career magnet graduate and being in a school with a lower graduation rate, "rcg." This means the partial correlation of career magnet attendance and parental support, controlling on the student's program's graduation rate, must be near zero.

      The correlation between being a career magnet student and the student's program's graduation rate is based on the difference in graduation rates of the two kinds of schools and is .052. The correlation between being a career magnet graduate and possessing high parental support is .306. Substituting those values,

      A value of rgp = .95 would produce a value of rcp.g = .14, still considerably greater than zero, but it is impossible to imagine a particular student's program's graduation rate being correlated .95 with any aspect of a student's life, or a student's perception of her or his parental support being correlated.95 with any other variable. Worse yet, eliminating the correlation between career magnet graduation and parental support would require a correlation above.95. Our initial assumption, that career magnets do not raise parental support, has led to a contradiction, and is therefore false.

      We can demonstrate the same point in another way. We found that 55% of career magnet graduates reported having parental support for college, while only 25% of those from comprehensive schools said this. Imagine that the student's perception of parental support for college is perfectly correlated with the child having the ability to graduate from high school, and that only 19% of the students who enter these high schools, whether career magnet or comprehensive, have this parental propensity. Suppose that in career magnets, all 19% graduate, and an additional 17% of the student body, who do not have parental support, also graduate, giving an overall graduation rate of 36%. Now suppose that all 19% in comprehensive schools also graduate, but an additional 22% of the students graduate as well, so that the graduation rate rises to 41%. The relationship between graduating and perceiving parental support could not possibly be stronger than this (still assuming no interaction effect). With these assumptions, the percentage of graduates from career magnets who perceive that their parents will support them is 53% (.19/.36), which is slightly below our actual results. In the comprehensive schools, the percentage of graduates with parental support is 46% (.19/.41). The difference between career magnets and comprehensive schools is 7% (53% - 46% = 7%). The actual difference is 30%, over four times as large. At a minimum, over three-fourths of the effect of career magnet attendance on the graduates' perceived parental support cannot be explained by the lower graduation rate of career magnets under the most extreme assumptions.

      Since the effects of attending a career magnet on the risk behavior of students and their college achievement are about two-thirds as strong as their effect on perceived parental support, these effects are also strong enough so that they cannot be explained away by a difference of graduation rates.

      We noted earlier that we had assumed that the correlation between parental support and graduation is the same in career magnets and comprehensive schools--that is, that graduates had more parental support than did dropouts, and that this was equally true for career magnets and comprehensive students. The more complicated of our two contradiction proofs drops this assumption, and allows for the possibility that career magnet dropouts might have higher or lower levels of parental support than do dropouts from comprehensive schools. Differences in who drops out can explain the higher support rate of career magnet graduates, but only if we assume that the graduates of comprehensive schools believe they have less parental support than do the comprehensive students who did not graduate, which seems not credible. There are two proofs needed. The first is for a purely theoretical model that assumes that graduating from high school does not affect level of parental support for college. The proof is arithmetical as follows:

      Solutions for this linear relationship show that for reasonable values of parental support for career magnet dropouts, "Y," we get

      In all cases, the amount of parental support perceived by dropouts from comprehensive schools exceeds that perceived by dropouts from career magnets, and in the last two cases, noticeably exceeds the amount of support perceived by comprehensive school graduates, none of which seems possible. The original assumption, that the positive effect of career magnets is due entirely to the fact that career magnets have a lower graduation rate, is contradicted.

.019 + 1.088Y = Z

This equation once again shows the amount of support perceived by comprehensive school nongraduates to be greater than that perceived by career magnet school nongraduates, so the contradiction remains.

[13] The exact formula is to add the standardized versions of the number of credits, which ranged from 0 to 77, and the dichotomous statement that they did or did not have a major. When correlated separately with type of school, their correlations are .194 and .212, respectively.

[14] The actual wording of the question is, "A. Did you do any of the following things while you were in high school? . . . Talk with your parents about financial help to go to college?" Since the question is about what the student did ("Did you do?), it is safe to assume that this question refers to students approaching their parents rather than parents bringing the subject up.

Crain, R. L., Allen, A., Thaler, R., Sullivan, D., Zellman, G., Little, J. W., and Quigley, D. D. (1999). The Effects of Academic Career Magnet Education on High Schools and Their Graduates (MDS-779). Berkeley: National Center for Research in Vocational Education, University of California.