Ten Thousand Democracies: Politics and Public Opinion in America's School Districts

Appendix b: Analysis and Supporting Tables for Chapter 4

Appendix B: Analysis and Supporting Tables for Chapter 4 Figure 4.1 in chapter 4 is based on an ordinary least squares regression model to see how spending levels vary by fiscal dependence. Per-pupil spending is the dependent variable, and we have a dummy variable coded 1 for fiscally dependent districts and 0 for independents ones. There is no variation in dependence within many states, so we cannot estimate this as a hierarchical model. Instead ,we add dummy variables for each state to account for the variety of factors that may influence the levels of spending and levels of key independent variables across the states. The slope estimates, standard errors, t-ratios and significance of each variable, along with model fit and sample size, are reported in table A4.1 (the estimates for the forty-two state dummy variables are not reported). We also control in this model for public opinion, as well as median income and the share of all education spending in the district that comes from local Table A4.1. Dependent School Districts Spend More Than Independent Districts District Characteristic B SE t p Opinion 785.37 39.62 19.82 .00 Median income –15.94 1.59 –10.00 .00 Median housing values –.98 .57 –1.73 .08 Population (logged) –150.26 10.91 –13.78 .00 Dependent district 489.94 117.00 4.19 .00 Local dependence on property tax –10.58 1.65 –6.41 .00 Property tax dependence x housing values .13 .01 15.77 .00 Local share of revenues 20.79 1.37 15.17 .00 Intercept 6456.16 106.15 60.82 .00 R2 .74 N 8,200 Note: Dependent variable is per-pupil instructional spending in fiscal 1995. 169 Table A4.2. The Impact of Public Opinion on Spending Is Higher in Dependent Districts District Characteristic B SE t p Opinion 767.28 39.91 19.23 .00 Dependent district –26.01 185.47 –.14 .89 Opinion x dependent 484.60 135.23 3.58 .00 Median income –1.16 .57 –2.05 .04 Median housing values –15.84 1.59 –9.95 .00 Population (logged) –150.69 10.90 –13.83 .00 Local dependence on property tax –10.42 1.65 –6.32 .00 Property tax dependence x housing values .13 .01 16.04 .00 Local share of revenues 20.52 1.37 14.96 .00 Intercept 6481.66 106.31 60.97 .00 R2 .74 N 8,200 Note: Dependent variable is per-pupil instructional spending in fiscal 1995. 170 Appendix B sources. To account for a conditional effect where housing values explain school spending more in communities that have the heaviest reliance on the property tax, we compute an interaction term that multiplies school spending by property tax reliance. The negative sign result on income is somewhat puzzling and appears in many models throughout the text. The reason for this is primarily multicollinearity with median housing values (r = 0.79) and local share of revenues (r = 0.62).When income is entered alone, the slope is always positive, large, and significant. To assess the conditional nature of policy responsiveness, we add to the model an interaction term between public opinion and fiscal dependence. Throughout the book, we show the interaction between opinion and some institutional characteristic to show how that characteristic mediates the effect of opinion. In this case, we are interested in how the effect of opinion on policy differs in independent and dependent districts as well as how the effect of opinion varies by the extent to which localities rely upon their own resources. The key result, illustrated in figure 4.2, is based on the model reported in table A4.2 (again, estimates for forty-two state dummy variables are not reported). With the interaction term included, the“main effect”on spending between dependent and independent districts is no longer statistically significant and changes sign. The reason dependent and independent districts systematically spend differently does not rest on this basic institutional distinction. Rather, we now know that the reason spending is higher in dependent districts is that spending is responsive to public preferences for higher spending. The positive coefficient for the interaction term (484.60) indicates the amount of policy responsiveness attributable to fiscal dependence, or looked at another way, how much responsiveness is lost by the use of independent school districts. Figures 4.3 and...