Institutional features of NYS Findings

355 N.A. Alexander Economics of Education Review 19 2000 351–361 specific subjects over the time frame studied have chosen to apply for exemption. It is therefore possible that the factors that led these districts to adopt this strategy would have also led to higher student achievement in the absence of said policy. Assume, for instance, that school districts with more “effort” have higher student performance even without this curriculum initiative. Assume also that “effort” is one of the factors that drive districts to apply for a vari- ance see Fig. 1. Excluding this variable from the model thus leads to an upward bias in the estimation of the impact of curriculum standards on student achievement. Concerns regarding selectivity bias are especially rel- evant to this study because of the historical and insti- tutional processes that led to the formal institution of the new policy on standards in New York State. In summary, there are several reasons that may have pushed districts to make this appeal for a “50–64” vari- ance, and each may have an individual impact on student performance. For instance, parents may have forced higher standards into schools. Many of the early requests came from relatively wealthy suburban districts where a majority of their students already take the Regents exam. School districts may have had sufficient institutional “slack”. Being granted a variance obliged school districts to make annual reports to the NYS Department of Edu- cation and often entailed having smaller classes and improved teacher training. 4 In addition, the receipt of variances by districts may have induced neighboring dis- tricts to apply for the same. Besides the above district- specific attributes, time and the increased emphasis nationwide on student performance may have contrib- uted to changes in educational output. 1.2.4. Controlling for district-specific influences and historical trends One way of isolating the effect of curriculum stan- dards on student achievement is to adopt a one-way fixed effects model and to include a variable that represents each year of the sample. This strategy controls for dis- Fig. 1. Tracing the true impact of curriculum standards. 4 This information comes from anecdotal evidence and dis- cussions with Nick Argyros of the NYS Department of Edu- cation, who handles the awarding of the variances in question. trict-specific characteristics by creating dummy variables for districts in the analysis. Similarly, inclusion of a dummy variable for each year allows historical trends to be taken into account. These strategies allow the model to address concerns that apparent effects of curriculum policy on student achievement may be merely historical artifacts or caused by particular district attributes. Thus, average student performance in a district A i is thought to be a function of district-specific characteristics D i , time year b i , a vector of explanatory variables c i , and an error term e i : A i 5a i D i 1b i c i 1e i Note that it is assumed that implementation of the variance is immediate. As indicated, school districts had to have resources in place and to have presented a viable plan of action to the NYS Department of Education in order to gain approval for their variance request. 5

2. Institutional features of NYS

2.1. Research population New York State is the only state with a long-standing reliance on a curriculum-based examination system covering the majority of high school graduates. In the fall of 1995, the high school student population of New York State, excluding NYC, comprised 1.9 Asians, 7.8 blacks, 3.8 Latinos, 0.004 Native Americans, and 86.1 whites NYS Department of Education, 1995. The relative diversity of the population allows investigation of the associations between socioeconomic characteristics, curriculum policy, and student outcomes. The following analysis focuses on the population of public school students in grades 9–12 in New York State, excluding New York City. 6 Those grades are examined because much of the discussion on curriculum standards centers on high school students. To the extent that cur- riculum reform has some universal effects, the findings of this study may have important implications for the rest of the students and to the nation as a whole. The definitions and simple statistics weighted by high school 5 Ideally, a lag of the standards policies would be included to reflect that the full effect of these strategies is not instantaneous. Instead, the impact of these polices may be distributed over a period of years and would be more appropriately captured using distributed-lag models. Given the shortness of the panel and other data limitations, this option was not possible. 6 As indicated, there were problems surrounding data on New York City. They were not available in a consistent fashion from the NYS Department of Education, and a departmental contact indicated that direct contact with NYC would lead to no better results. 356 N.A. Alexander Economics of Education Review 19 2000 351–361 student population of the variables used in the analyses are presented in Table 3.

3. Findings

The data in Table 4 are the estimates derived for the models of generic measures of student performance as outputs in the education production process. The general measures of achievement used are 4-year college attend- ance, diploma type, and retention rates. Dummy vari- ables are used to capture the award of a variance in any subject. Interaction variables were created between the student demographic variables and the receipt of a vari- ance to explore if approved variances altered the associ- ation between the profile of schools and student out- comes. Faculty, district size, school diversity and organization are also included as controls. Most of the background variables have the expected associations with educational outcomes. Even when non-policy variables e.g. historical trends, diversity of the student body, etc. are controlled for, the association between the award of a subject variance and portion of graduates with Regents diplomas remains. The data suggest receipt of a variance causes the portion of graduates receiving Regents diplomas to increase by 0.5. This represents a 1.87 increase over pass rates before this strategy was pursued. However, the results also indicate that this curriculum policy is not as relevant for retention rates or the rates at which graduates sub- sequently attend 4-year colleges. The estimates imply that increased portions of black and Latino students are associated with lower portions of Regents diplomas granted. Poverty considerations do not play a big role in any of the generic educational out- comes. Moreover, there is no substantial difference in the link between school profile and educational achievement before and after applying a variance. That is, curriculum standards policies do not seem to affect the relationship between race, poverty, and student outcomes. By contrast, subject variances matter for pass rates i.e. 65 in Regents Global Studies, Regents English, and Regents Biology. Approved variances in those sub- jects resulted in an increase in the portion of students passing; this increase was by 1.68, 1.88, and 3.96 for global studies, English, and biology, respectively. The data in Table 5 are the estimates derived for subject- specific models of curriculum policy. Educational out- puts are defined as the percentage of enrolled students passing the Regents in a particular field. The standards are also categorized according to specific subjects, where biology variances are distinguished from English vari- ances, math variances, and so on. To facilitate presentation, Table 5 does not specify the class size nor the education and the experience of the faculty for each Regents course. Instead, a generic for- mat is adopted for the table where the average size of the Regents class culminating in the specified exam is represented by “Course class size”. Thus, when the dependent variable is the percentage of average 9–12 grade enrollment passing biology, “Course class size” is the average class size for Regents Biology. Similarly, the average size of the 10th grade class in a particular field is represented by “Average subject size”. When the dependent variable is the percentage of average 9–12 grade enrollment passing biology, “Average subject size” is the average size of a 10th grade science class in the district. The experience and education of the faculty teaching the Regents courses culminating in the specified exams are represented by “Average experience of course teacher” and “Course faculty with masters”, respectively. Thus, when the dependent variable is the percentage of average 9–12 grade enrollment passing biology, “Aver- age experience of course teacher” reflects the average number of years of teaching experience of faculty teach- ing Regents Biology. “Course faculty with masters” rep- resents the portion of the faculty teaching Regents Biology that has a master’s education or above. Similar procedures were followed for all other subjects. Given the percentage points noted above and the mean pass rates of global studies, English, and biology, the effect of awarding variances becomes more meaningful. Absence of approved variances in those subjects would likely result in 3 fewer students passing global studies, 3.4 fewer students passing English, and a striking 10.2 fewer students passing biology. These improved pass rates suggest that implementation of a standards policy can influence curricula mastery.

4. Discussion and implications for public policy