EMPIRICAL METHODOLOGY
By comparing the standardized test scores between students, the better school type can be determined. The test scores used in this study are from the 2001 version of the Iowa Test of Educational Development (ITED). Iowa tests are administered to students in public schools and cover about 92.7 percent of all enrolled ninth graders in Washington State (Office of Financial Management 2001). The enormity of the sample, which includes 72,161 subjects, is very important as the variation in it approaches the variation of a standard normal sample. Of these, about 21 percent (17,191) are in junior highs and about 73 percent (58,777) in high schools. With both junior high and high school students present in the sample, the variation between them is inherently included in the data. Since some students have undergone the transition to high school and some have not, this test will provide a good way to measure which student performs better. Also, the cross-sectional nature of the data allows for the impact of adolescence to be held constant. Since the high school students are not on average older than the junior high students, the impact of the transition cannot be a result of being in a different stage of adolescence.
One of the reasons why the ITED tests are so valuable is that they provide a rich body of information accompanying student test scores. Surveys completed by students, teachers, and school administrators collect information such as race, language spoken at home, student computer access and parental involvement in school. Students give their perceptions of things such as school safety and plans after school. Teachers and administrators fill out portions of the survey regarding, among other things, student GPA, school attendance, and student behavioral problems.
For ease of interpretation, this paper will use a standardized version (mean zero, standard deviation equal to one) of the core developmental standard score (CoreSS). CoreSS is a number that describes a student’s location on an achievement continuum. The continuum is a number line that describes the lowest level of knowledge or skill on one end (lowest numbers) and the highest level of development on the other end (highest numbers). The table below shows the national CoreSSs assigned to typical performance in grade groups on each test at grades 4-12 in the spring of the year.
The scale used with the ITED was established by assigning a score of 260 to the median performance of students in the spring of grade nine. Using multiple choice and essay questions, the different sections in the ITED cover topics such as vocabulary, reading comprehension, revising written materials, spelling, concepts and problem solving, computation, social studies, and science (Iowa Testing Programs 2003).
Grouping the data by school type allows the ability to analyze the different characteristics between junior high and high school students. A summary of selected demographic and academic information is given in Table 1. Looking at the standardized CoreSS between the junior high and high schools, junior high students on average perform about .143 standard deviations above the mean score while high school students on average perform about .019 standard deviations below the mean. Performing a 1 tailed paired t-test on the two means for standardized scores, junior high scores are statistically higher than high school scores at the 99 percent level of significance. In order to ensure that this difference is in fact a product of being in junior high and not because of some hidden variable, such as junior highs being located in larger communities, a linear regression should be run on the data given by surveys and test scores.
Ordinary Least Squares will be employed to regress student CoreSSs against independent variables that potentially influence them. Eighty different variables measure the relative impacts of numerous potential score-influencing factors. The majority of these variables are qualitative measures; the student, teacher or administrator effectively answered either a yes or a no on the survey. With qualitative independent variables, it only makes sense to use dummy variables to account for the respective impacts of each independent variable on a student’s standardized score. The regression model will use the following mathematical form where is the dependent variable, is the regression constant and is the variable accounting for high school enrollment. is a vector of coefficients; X is a matrix of dummy variables and is the stochastic error term.
In addition to , there are six other variables in matrix X that account for school type. The omitted variable in this case is the junior high variable. Hence, the coefficient on each of these is interpreted as the respective impact of not being in junior high for ninth grade. is the variable of interest, as the coefficient of such will predict the impact of being in high school relative to junior high. The significance and direction of the coefficient on will indicate whether students in high school perform relatively better or worse than do students in junior high. In addition to what type of school a student is in, this paper employs a large body of other data relevant to student test scores. So matrix X also includes an array of variables accounting for characteristics ranging from computer use to parental help with homework and ethnicity. District enrollment is also included to control for the possibility that junior highs occur in large communities.
But as has been suggested by numerous researchers, what is true for the majority is not necessarily true for all. Along with many others, Kathryn Schiller and David Kinney showed that the effects of such a transition are not universal among students (Schiller 1999, Kinney 1993). Even if Ninth graders in general perform better in junior highs, there could still be students who perform better in high school. Determining a direction on the coefficient for high school, the next step is to sort out which groups of students are affected more or less by such a transition. In total there are two regressions, the first aimed at whether junior high or high school is better for ninth graders in general, the second at what type of student does better in each setting. Interactive dummy variables combine student characteristics with school type and allow the ability to discern which groups of students, if any, respond more or less to being in a high school versus junior high environment.
New variables, constructed by interacting racial and family characteristics with the high school variable are included with the original variables in regression 2, the form of which is below.
represents a vector of coefficients and Y is a matrix of interactive dummy variables. The rest is the same as described for regression 1.
EMPIRICAL RESULTS
Table 2 displays the included independent variables with their estimated coefficients and subsequent standard errors from regression 1. It is important to note that heteroskedasticity was present and the standard errors given are White’s robust standard errors. The first independent variable listed is that of the one under scrutiny, whether or not a student is in high school. The coefficient value of approximately -.06 indicates that on average, students enrolled in high school scored about .06 standard deviations below their peers enrolled in junior highs. A value of .06 is a significant impact, looking at it in relation to the impacts of some of the other variables. For example, the marginal impact of being a male is about .029 standard deviations. In this case, the impact of being in high school relative to being in junior high is over twice that of the impact of being male relative to being female. Another student characteristic that has received a lot of attention is ethnicity. Many researchers have documented the negative effects of being a minority student, (Becker 2002, Levine 1990, Jencks 1998) and comparing the impact of being in high school to the impact of being a minority can be illuminating. Table 2 provides a proxy for how well minority students do with respect to White students. The variable “African American” indicates that on average African Americans perform about .477 standard deviations below white students. So being in high school for ninth grade is equivalent to about 1/10 of the effect of being African American. Alternatively, high school is about a fourth of the impact of being Hispanic. This supports the theory that students in high schools for ninth grade in general perform worse than students in junior high do. While the move to high school may have positive effects on student performance, this suggests that the negative effects outweigh them.
Regression 1 shows that high schools are worse for students in general, but there may still be groups of students for whom high school is better. Regression 2 attempts to determine which, if any, groups perform better in high school. The variable coefficients and White’s robust standard errors are also in Table 2. African Americans and students where English was sometimes spoken at home on average did worse if they were in high school. African Americans for example, performed an additional .055 standard deviations below the mean if they were in high school. On the other hand, bilingual students performed relatively better if they were in a high school, with Bilingual averaging about .182 standard deviations above their junior high counterparts. The original high school variable coefficient rose by about .03 standard deviations in regression 2. So by adding the additional interacted variables, the new regression accounted for about .03, or about 60 percent of the variation in the high school variable.
The only statistically significant interacted variables are those of African American students in high school and bilingual students in high school. There are a couple of possible explanations for why bilingual students appear to perform better in a high school setting. Public high schools in the state of Washington have a larger concentration of bilingual students than do junior highs which would seem to imply that there would be more supportive programs designed for them. Perhaps the expanded array of potential programs for bilingual students creates a better learning atmosphere, where this group can achieve more. With the bilingual students, high schools possibly provide more programs than do junior highs to help the students adapt to English speaking teachers. It is interesting to notice that on average bilingual students do better in high school while on average students who have non-English speaking parents do relatively worse in high school. So foreign speaking students do better in high school, but students who have parents that only speak a foreign language do relatively worse. It is possible that the variable “English never spoken in the home” is not so much measuring language barriers to learning as it is measuring a family’s socioeconomic status (SES). Adults who do not speak English would be limited in their employment opportunities and as a result, likely do not hold very high paying jobs. If this variable is indeed a proxy for SES, then perhaps the students are worse off financially than students whose parents do speak English, and do not have the necessary resources for a successful transition. African Americans may also be victims of a lower SES, which would account for part of the negative correlation. Another possible reason why they do relatively worse in a high school setting can be attributed to the social pressures put on them by their peers. To do well in school is to “act white” and so many of them do not. The pressure not to succeed is not as prevalent in junior high as in high school (Levine 1990). The rest of the variables examined are not statistically different from zero.
CONCLUSION
This study is consistent with others in its findings that the instability and adjustment required of students in the high school transition are associated with educational performance. The implications of the resulting data suggest that in general, ninth grade students enrolled in high schools perform worse academically than do their counterparts enrolled in junior highs. Since the students examined in this study are all relatively the same age, the instability and changes associated with adolescence and the resulting impacts on performance are held constant. Hence it can be concluded that it is the transition and difference in school types, not the adolescence that accounts for the discrepancy between junior high and high school students. So during their ninth grade year, high schools in general are worse for ninth graders.
Why high schools are worse could potentially be due to the recent transition into the new high school environment, where ninth graders are still adjusting to new roles and responsibilities. This idea of negative effects associated with transitions is a prominent one within educational literature (Schiller 1999, Wigfield 1991, Nottelmann 1987, Lord 1994, Hirsch 1987). If this is the case, then all students will eventually experience this dip in academic performance, it is just a matter of when. The ninth graders currently in junior high will simply experience that dip in performance a year later than the ninth graders currently in high school. There is the possibility, however that the extra year in the junior high is beneficial to students in general. Indeed, if scores can be raised as a result of keeping students out of high school their freshman year by more than one twentieth of a standard deviation, it is an option to be explored. Perhaps some 13-14 year olds are not yet ready for the high school environment, and keeping them out of it would be helpful.
Continued student performance based on whether they were in a junior high for 9th grade is beyond the scope of this study, but is interesting in its policy implications. If in fact it were better for some students to stay in junior high during ninth grade, it would make sense to abolish four year high schools, instead moving seventh, eighth, and ninth graders into the junior high environment and keeping the tenth, eleventh, and twelfth grades as part of the high school experience. And with bilingual students in high school performing relatively better than their counterparts in junior high, it may be better to move some groups of students on to high school, with other groups remaining in junior high. This would be a complicated procedure though, and only justified through continued performance evaluations to the ninth graders tested in this analysis to determine if in fact the negative effects of being in high school are a result of it being a high school, or a result of the recent transition. The development of future performance models lies outside the range of this data, but would need to be devised to explain whether ninth grade school type plays any role in student performance.
Daniel Levine and Eugene Eubanks found that their African American and Hispanic groups performed about a half a standard deviation below the scores for the White groups, this is consistent with findings using the ITED (Levine 1990).
High schools have about a 2% concentration of bilingual students, while junior highs have about a 1% concentration