Analysis of School District Fiscal Response

1993-94 to 1998-99

Technical Supplement

The State Aid Work Group

November, 2000

Analysis of School District Fiscal Response –Technical Supplement

Introduction

The question of whether school districts respond to increases in State Aid by maintaining local tax effort or by lowering their tax rates has been a traditional concern of public finance experts. In light of the Regents emphasis upon targeting State Aid to high need school districts and ensuring adequate resources are available to enable all students to achieve high academic standards, the issue of fiscal response has remained an area of concern to New York State policy makers as well. As the State focuses additional resources on high need schools, any diminution of local tax effort, particularly if local tax effort is "inadequate" to begin with, poses a significant policy concern. For this reason, particular attention was paid to understanding patterns of fiscal response in districts of varying need/resource capacity.

This paper will focus on two major areas of concern. First, it will examine the extent to which changes in State Aid are reflected in district expenditure. In other words, if districts are using their State Aid to fund educational programs while maintaining local effort, increases in State Aid should be accompanied by increases in expenditure on a per pupil basis. This paper will describe the strength of this association and describe those variables that have been found to be associated with changes in per pupil expenditure. Secondly, the strength of the relationship between changes in State Aid and changes in local effort will be examined, and specific categories of district fiscal response will be identified.

Background

According to Goertz and Natriello (1999), past research on intergovernmental grants designated for education has shown that, when districts received increases in funding for education, only a portion of that increase was used for education while another portion was used for local tax relief. The authors suggest that the portion used for tax relief is largely dependent on the degree of fiscal burden experienced by the district. In other words, districts that were high spending and high taxing to begin with tended to take more education aid for tax relief in subsequent years, whereas districts that were low spending, tended to invest more of their aid towards education. This finding suggests that high need districts already burdened with high tax rates will respond to aid increases by taking more tax relief than their low taxing counterparts.

In a study of New York State school districts, Adams (1980) found that the relationship between changes in educational expenditures and changes in State Aid was greater when multiple years were examined instead of limiting the analysis to single-year changes. For this reason, this study focused on five-year changes in expenditures rather than on single-year changes. It is also important to note that in the simulation model employed by Adams, a number of individual district characteristics affected individual district fiscal responses. For example, Adams found that low wealth, low income districts tended to use more of their State Aid to increase expenditure than did districts of high property wealth and high income.

As reported by Tsang and Levin (1983), from an extensive review of studies from different states, on average, school districts used approximately one-half of their aid increases for funding education and the other half for local property tax relief. In contrast, Goertz (1979) found that those New Jersey districts that received large increases in state aid as a result of a comprehensive education finance reform movement, used most (85 percent) of their additional funding for education instead of reducing local property taxes.

Overview

This study builds on a previous analysis of local tax effort and attempts to explain the extent to which New York State school districts have responded fiscally to changes in State Aid and the implications of this response.

Data

The fiscal and descriptive data for this study were obtained from the State Aid database and the School District Fiscal Profile database. Fiscal patterns were examined for the five-year period from 1993-94 to 1998-999. Eighteen districts with less than 225 pupils were excluded from this study. These small districts were found to have large fluctuations in expenditure and State Aid over time, largely attributable to their small size, and were therefore eliminated. A detailed list of eliminated districts can be found in Appendix A.

Throughout the study, the State Aid measure reported was total State Aid on the database minus Building Aid, Transportation Aid and Reorganization Incentive Building Aid. The expenditure measure used was total expense minus transportation, debt service, and district intra-fund transfers. The expenditure and revenue funds removed from the analysis were associated with capital-spending activities. Because these activities typically include periodic one-time investments in assets like buildings or school buses – assets that will be used over an extended period of time -- they distort accurate year-to-year comparisons.

Part I. Changes in State Aid and Changes in Total Expense

Dollar Changes

This section of the analysis examines the extent to which State Aid and expenditures are related. For example, if a district were to receive an increase in State Aid per pupil, one might expect this increase to be reflected in part in corresponding increases in expenditure per pupil over the same time period. Some districts will use more of their State Aid to increase spending while others may tend to direct a smaller portion of an aid increase towards actual expenditures. The relationship between changes in State Aid and changes in expenditure will be demonstrated for each need/resource category and the Big Five City school districts.

Chart 1a below depicts the weighted average CPI-adjusted dollar change in State Aid per pupil and the weighted average dollar change in total expenditure per pupil by need/resource category. The Big Five City school districts are reported separately in Chart 1b.

As demonstrated in Chart 1a, the largest increases in State Aid per pupil and expenditure per pupil occurred in the high need districts. The largest average increase in State Aid per pupil ($462/pupil) was observed in the Urban/Suburban High Need districts. These districts also experienced an increase in expenditure of $350 per pupil. In effect, 75.8 percent of the State Aid increase was reflected in a simultaneous increase in expenditure for Urban/Suburban High Need districts. In other words, for every one dollar in additional State Aid, districts spent 76 cents.

High need rural districts experienced an increase in State Aid per pupil of $418. Additionally, rural high need districts had an average increase in total expenditure per pupil of $574, suggesting, not only that State Aid increases were reflected in expenditures, but that local contributions were increased as well.

Average need districts had an increase in State Aid per pupil of $53 and an increase in expenditure of $42 per pupil. After adjusting for inflation, low need districts experienced a decrease in State Aid per pupil of eight dollars, and a decrease in expenditure per pupil of $264. This suggests that decreases in State Aid were accompanied by larger decreases in expenditures, or simultaneous decreases in local effort.

Chart 1b describes the same State Aid-Expenditure relationship in the Big Five districts. As shown in the chart, each of the Big Five districts experienced increases in State Aid per pupil. The largest increase was $732 per pupil and occurred in the Buffalo school district. New York City, Buffalo, and Yonkers experienced increases in expenditure per pupil during the same time period, while Rochester and Syracuse experienced decreases in per pupil expenditure. Although Chart 1b does not directly address the local effort issue, it is apparent that Rochester and Syracuse did not respond to State Aid increases with corresponding spending increases. In sharp contrast, increases in expenditure per pupil exceeded State Aid per pupil increases in New York City, Buffalo, and Yonkers.

It is important to remind the reader that any analyses based on changes in per pupil measures are also affected by changes in enrollments that occur during the time period being examined. Chart 2 below shows the percent change in pupils for New York City, the Big Four and each need/resource category.

With the exception of rural high need districts, all other need/resource categories experienced increases in pupils from 1993-94 to 1998-99. New York City experienced an increase of 8.70 percent while the Big Four City school districts experienced an increase in pupils of 6.16 percent. The greatest percent increase in pupils occurred in the low need category, which experienced an increase of 11.66 percent.

While these descriptive data illustrate how various types of districts have tended to respond, their value is limited for prediction of districts’ behavior under different circumstances. For example, the foregoing analysis suggests that a decrease in State Aid to low need districts will result in a decrease in local effort as well. It does not, however, provide any indication of how those districts may respond to an increase in State Aid. For that purpose, we must rely on a more sophisticated, but somewhat more abstract statistical analysis.

Scatterplots provide a very useful depiction of the bivariate relationship between changes in State Aid and changes in expenditure. As shown in Chart 3, the relationship is linear and positive (r=0.54), meaning that increases in State Aid per pupil were found to be associated with increases in expenditure per pupil. Conversely, decreases in State Aid per pupil were found to be associated with decreases in expenditure per pupil. An R2 value of 0.288 indicates that 28.8 percent of the variability in changes in expenditure per pupil can be accounted for by changes in State Aid per pupil, for all districts.

When we examine similar scatterplots within each major need/resource category, it becomes clear that the strength of the relationship between changes in State Aid per pupil and changes in expenditure per pupil varies. For example, Chart 4   displays the relationship between dollar change in State Aid per pupil and dollar change in expenditure per pupil for urban/suburban high need districts. In this instance, the relationship between changes in State Aid and changes in expenditure is weaker (r=0.44) and the expenditure per pupil response pattern is far more disperse than when compared to all districts statewide (r =0.54).

While the scatterplots depicted in these charts are useful in "seeing" the whole picture, you will also note that the "line of best fit" for this pattern is also described. The b-coefficient for this line indicates the dollar amount of change in expenditure per pupil associated with a one-dollar change in State Aid per pupil. Thus, the b-coefficient (b=0.679) that defines the linear relationship shown in Chart 4 can be easily interpreted. In urban/suburban high need districts, a one-dollar increase in State Aid per pupil was associated with a 68-cent increase in per pupil expenditure. Although the relationship

between changes in State Aid per pupil and changes in expenditure per pupil was weaker for urban/suburban high need districts when compared with all districts statewide, it is evident that most of the districts in this category fall to the right of the y-axis, meaning that they have experienced State Aid increases. Only five of the 37 urban/suburban high heed districts experienced State Aid per pupil decreases from 1993-94 to 1998-99.

When we examine high need rural districts (Chart 5), we find that the relationship between changes in State Aid per pupil and changes in expenditure per pupil is much "tighter" and less dispersed around the line of best fit. (r=0.69) than the corresponding relationship in urban/suburban high need districts (r=0.44).

The relationship between changes in State Aid per pupil and expenditure per pupil in high need rural districts depicted in Chart 5 can be characterized as fairly strong. The dollar change in State Aid per pupil accounts for 48 percent of the variability in dollar change in expenditure per pupil. Additionally, a dollar of State Aid "buys more" in terms of expenditure in rural high need districts when compared to urban/suburban high need districts. A one-dollar increase in State Aid per pupil was found to be associated with a 92-cent increase in expenditure per pupil in high need rural districts.

Chart 6 displays the relationship between change in State Aid per pupil and change in expenditure per pupil for the 332 average need districts in the study.

Twenty-three percent of the variability in change in expenditure per pupil is accounted for by change in State Aid per pupil in average need districts. A one-dollar increase (or decrease) in State Aid per pupil is associated with an 89-cent increase (or decrease) in expenditure per pupil in average need districts.

The relationship between changes in State Aid per pupil and changes in spending per pupil is even weaker (r=0.22) when considering only low need districts (Chart 7). As shown in this chart, a change in State Aid per pupil accounts for only five percent of the variability in dollar change in expense per pupil. Additionally, it is interesting to note that the low need districts are the only category to have a negative Y-intercept value (-$341.15). This indicates that if a low need district were to receive no per pupil change in State Aid (after accounting for inflation), it would still be predicted to have a decrease in expenditure per pupil of $341.15.

Additionally, a one-dollar increase (or decrease) in State Aid per pupil was found to be associated with an 89-cent increase (or decrease) in expenditure per pupil. However, it is important to note that the low R2 is also an indication that there is considerable error to be expected with such a prediction.

Table 1 provides a detailed summary of the individual relationships displayed in the previous charts. As shown in the second column of the table, the strength of the linear relationship between dollar change in State Aid per pupil and dollar change in expenditure per pupil is greatest in rural high need districts (r=0.69) and weakest in low need districts (r=0.22). Additionally, this relationship varies depending on the type of high need districts being examined, with urban/suburban high need districts displaying a much weaker linear relationship (r=0.44). In other words, their fiscal response is not as tightly linked to State Aid changes as is the case in rural high need districts.

The magnitude of the effect of a dollar change in State Aid per pupil was found to be greatest in rural high need districts, where a one-dollar change in State Aid per pupil was associated with a 92-cent change in expenditure per pupil (+/-$0.15). Urban/suburban high need districts as well as low need districts have margins of error that are too large to accurately predict the magnitude of the State Aid per pupil effect.

Percent Changes

Table 2 highlights the relationship between percent change in State Aid per pupil and percent change in expenditure per pupil for districts in each need/resource category. A similar trend is found when examining percent changes.

For example, the relationship is strongest for high need rural districts and weakest for low need districts. Additionally, a one percent change in State Aid per pupil is associated with a 0.64 percent change in expenditure per pupil in high need rural districts but only a 0.12 percent change in expenditure in low need districts.

Finally, while there is clearly a positive linear relationship between changes in State Aid per pupil and changes in expenditure per pupil, the strength of the relationship, as well as the magnitude of the State Aid effect, varies according to district need.

State Aid Increases and Decreases in Total Expenditure

Of particular importance in the scatter diagrams displayed in Charts 3 through 7, is the bottom right quadrant. Districts that fall into this quadrant represent cases for which there has been an increase in State Aid per pupil and a decrease in expenditure per pupil during the same time period.

Statewide, there were 430 districts that received dollar increases in State Aid per pupil, after controlling for inflation. Of these 430 districts, 84 also experienced decreases in per pupil expenditures. Therefore, statewide 19.5 percent of the districts that received aid increases also experienced decreases in per pupil expense. Chart 8 displays the proportion of those districts that experienced increases in State Aid per pupil that also experienced decreases in per pupil expense, by need/resource category.

As shown in Chart 8, the percentage of State-Aid-increase districts that experienced a simultaneous decrease in total expenditure per pupil varies dramatically across need/resource categories. Of those 32 urban/suburban high need districts receiving State Aid increases, 21.9 percent experienced decreases in per pupil expense. However, in rural high need districts, only 5.8 percent of the 138 districts receiving increases in State Aid per pupil experienced decreases in per pupil expense. Of the 202 average need districts that experienced increases in State Aid per pupil, 18.8 percent also experienced decreases in per pupil expense. There were 53 low need districts that received State Aid increases from 1993-94 to 1998-99. Of these, 54.7 percent experienced decreases in per pupil spending over the same five-year period.

As shown in Chart 8 above, average need and urban/suburban high need districts were found to be roughly three times more likely than high need rural districts to experience decreases in per pupil expenditure in light of State Aid per pupil increases. Additionally, low need districts were found to be three times more likely than average need districts and nine times more likely than high need rural districts to decrease per pupil expenditures while receiving State Aid per pupil increases over the same five-year period.

Multivariate Analysis of Expenditure Changes

Thus far, this analysis has been confined to the impact of State Aid changes on local spending decisions. In actuality, however, the decision-making process involves a host of other considerations. In the next stage of the analysis, we will expand the focus of the paper to consider other fiscal factors that may help to explain districts’ decisions.

As described above, the connection between percent change in State Aid per pupil and percent change in expenditure per pupil was much stronger in high need districts when compared to average and low need districts. However, it is reasonable to expect that the relationship is weaker in low need districts simply because they are much less dependent on State Aid to achieve their spending levels. In other words, in low need districts, State Aid represents a much smaller share of total expenditure. The fact that the observed linear relationship was found to change depending on need/resource category suggests that, not only is the change in State Aid per pupil an important independent predictor of expenditure changes, but that there is likely to be an interaction effect between both a district’s need/fiscal capacity status and percent change in State Aid per pupil. In order to test this assumption statistically, the need/fiscal capacity index, percent change in State Aid per pupil and the cross-product of the two were used to predict percent change in expenditure per pupil in a multivariate regression model.

Consistent with the findings of Goertz and Natriello that districts will use more aid for tax relief if they are already in a high spending or high taxing situation, expenditure per pupil at time one and tax rate at time one were added to the model. The multi-variate model was run for all districts using ordinary least squares regression. This technique permits us to identify the net independent effects of variables on percent change in expenditure while exerting statistical control over common variance in the model. The descriptive statistics for the variables in the model are shown in Table 3 below.

The average percent change in expenditure per pupil (column two of the table) was 3.71. The average un-weighted tax rate in 1993-94 was $15.19 per $1,000 actual value. Two-thirds of the districts in the study fell within $3.08 of the average tax rate. The un-weighted average percent change in State Aid per pupil, using 1993-94 constant dollars, was 5.9 percent and in 1993-94, the average per pupil expenditure was $8,062.

Consistent with Table 2, the correlation between percent change in State Aid and percent change in expenditure per pupil was 0.48. There was a strong negative correlation between time one expenditure per pupil and percent change in expenditure per pupil (r=-0.49). This finding clearly indicates that districts that are already in a high spending situation are more likely to experience a decrease in expenditure per pupil five years later. A similar although moderate relationship exists for tax rate at the start of the reporting period and percent change in expenditure per pupil (r=-0.17). Districts that are already in a high taxing situation will be more likely to experience a decrease in expenditure per pupil. The summary statistics for the five-variable expenditure model are displayed in Table 4 below.

The multivariate expenditure model displayed above accounts for 46 percent of the total variability in percent change in expenditure per pupil. This represents an increase in explanatory power of 60.4 percent when compared to the simple bivariate linear model using percent change in State Aid per pupil.

Since each of the predictor variables in this statistical model are measured differently, it is desirable to determine the relative importance of their impacts upon expenditure per pupil – using a common "yardstick." The column titled Beta indicates the standardized amount of change in expenditure per pupil associated with one (standard deviation) unit of change in each of the independent variables. Because Betas are standardized, they can be compared directly in order to gauge the relative importance of each variable’s effect upon expenditure per pupil after controlling for all other variables. Chart 9 displays graphically the magnitude of the independent effects for the variables in the expenditure model.

Chart 9 provides compelling evidence that districts already in a high spending situation are more likely to decrease expenditure per pupil, holding constant changes in State Aid, tax rates, and district need/fiscal capacity. That is, the negative (Beta = -0.367) effect indicates that in districts whose 1993-94 spending was already one standard deviation above the statewide average, spending per pupil five years later would drop by 0.367 standard deviation units. More importantly, this model indicates that the initial expenditure level of districts is approximately 1.5 times as important as percent change in State Aid per pupil in predicting percent change in expenditure per pupil.

The second most important predictor of changes in expenditure is the interaction between district need relative to fiscal capacity and changes in State Aid per pupil. This finding suggests that State Aid changes affect subsequent expenditures differently, depending on a district’s need relative to fiscal capacity. As shown in the previous set of scatterplots, State Aid increases are more likely to result in increases in expenditure per pupil in high need districts, than in low need districts. This interaction effect is 1.37 times as important as percent change in State Aid per pupil. However, percent change in State Aid per pupil is an important variable in the model. A one standard deviation unit change in State Aid per pupil is associated with a 0.238 standard deviation unit change in expenditure per pupil. This represents a sizable effect.

Finally, a district’s tax rate at the start of the period (1993-94) also exerted a moderate negative independent effect on percent change in expenditure per pupil. As expected, districts that were already in high taxing situations were more likely to decrease expenditures per pupil, holding constant change in aid, initial spending level, and district need/fiscal capacity.

Conclusions Part 1:

proceed to Part 2

FARU home page