Researchers assembled a target data base of 774 districts that had complete financial data and performance ratings for three years (1996-97, 1997-98, and 1998-99). As explained above, researchers separated the target data base into three groups based on district performance as measured by the Texas accountability system. Level one districts were the highest performing districts. Level two districts had somewhat lower performance overall but still performed at or above minimum performance standards. Level three districts were the lowest performing districts. Table 1 shows expenditures by function for districts within each level and compares those expenditures with the state average for the three-year period.

Table 1: Average Per-Pupil Expenditures for Target Districts by Function and Performance Level (1996-97 to 1998-99)

Table 1 fails to show strong or unambiguous patterns of expenditures associated with student performance. Level one, the group of highest performing districts, spends more than the other categories and the state average. Level two, however, is the lowest spending of the three levels.

Table 2 shows program expenditures by performance level. High-performing level one districts spend more for regular education and less for compensatory education and bilingual education. Level three districts have expenditures for compensatory education and bilingual education that exceed the state average.

The information presented so far shows that level one districts spend a greater percentage of their resources on instruction and regular education, suggesting a connection between resource allocation and performance. But these results require further analysis to determine if the findings are the result of decisions by school districts or just differences that occur by chance.

Table 2: Average Per-Pupil Program Expenditures (All Funds) for Target Districts by Performance Level (1996-97 to 1998-99)

Typically in analysis of academic performance and its relationship to resources, researchers attempt to model the effect of certain inputs (like teacher salary and instructional expenditures) on the achievement outcomes of individual students. The outcome is often specified using a single standardized test or measure of academic achievement. Resources may include financial information, school characteristics (such as size and percent of low-income students), and teacher characteristics. Researchers use linear regression to determine the contribution of resource input to increases in test performance. In this study, the goal was to understand differences in spending patterns among districts that were classified into three different performance levels. In other words, the “outcomes” or dependent variables are expenditures (expressed as functions or as program expenditures) and the independent variation is expressed in the three performance levels. Researchers used multivariate analysis of covariance (MANCOVA) to test the hypothesis that there are significant differences among the performance groups (levels one, two, and three). Statistically significant results would reveal important relationships between resource allocation and academic performance levels.

Researchers conducted two separate analyses using data for the 774 target districts. The first analysis used expenditure functions (instruction, instructional related, etc.) and the second analysis used program intent codes (regular education, gifted and talented education, etc.). The dependent variable was adjusted for differences in the percent of low-income students and the district size in terms of enrollment.
Researchers compared level one with level two. Then they compared level one with level three. In a third step, they compared level two with level three. These are referred to as pairwise comparisons, and they confirm the finding that expenditure allocations for instruction are positively related to student academic performance. Statistically significant differences occur between level one districts and level two districts as well as between level one districts and level three districts. The differences between level two districts and level three districts are not statistically significant. The table of results for pairwise comparisons appears in Appendix D.

Table 3 displays results of this analysis of expenditures by function using MANCOVA. There are statistically significant differences among the three performance levels, meaning that the differences among the levels are not simply the result of chance. Significant relationships (shown with an asterisk on the table) are revealed for instructional expenditures, instructional resources, school leadership, co-curricular activities, and general administration expenditures. In all of these areas, level one districts are spending more than their level two and level three counterparts. Significant relationships are also revealed for social work expenditures and instructional leadership expenditures. However, in these areas level one districts spend less than their level two and level three counterparts.

It is important to note that the MANCOVA analysis confirms what Table 1 suggested about a direct and positive relationship between high performance (level one districts) and higher instructional expenditures.

Table 3: Results of MANCOVA Analysis Using Per-Pupil Function Expenditures

Researchers performed the second MANCOVA analysis on the expenditures in seven program areas: regular education, gifted and talented education, career and technology education, special education, compensatory education, bilingual education, and athletic programs. Descriptive statistics for these programs appeared previously in Table 2.

The MANCOVA results appear in Table 4. There were statistically significant differences among the performance categories program expenditures only for regular education and career and technology education, but not for the other categories.

Table 4: Results of MANCOVA Using Per-Pupil Program Expenditures

Pairwise comparisons were made for the program areas. The significant results occur only between level one districts and the other two levels. There were not significant differences between level two and level three districts. Results are significant only for regular education and career and technology education. These findings reinforce the previous results suggesting that there is a relationship between resource allocation and high levels of student performance. Specifically, it suggests that resource allocation decisions have an important role in helping school districts achieve student learning goals. The table of results for pairwise comparisons appears in Appendix D.

Summary

Level one districts consistently spent more per-pupil than level two and level three districts on instruction, instructional resources, school leadership, general administration, co-curricular activities, and total operating expenditures. The expenditure differences were statistically significant, indicating a direct and positive link between resource allocation and academic performance that does not occur by chance.

Program intent areas where level one districts consistently spent significantly more per-pupil than both level two and level three districts are regular education and career and technology education. Expenditures for programs designed to serve the unique educational needs of special populations (gifted and talented education, special education, compensatory education, and bilingual education) are not positively related to overall performance levels measured by the Texas accountability system. Allocations to the regular program are the most strongly linked to performance.