Expenditures and Achievement Running header: EXPENDITURES AND ACHIEVEMENT Expenditures and District Characteristics on Student Achievement in North Carolina Jamila Jones Kennedy George Mason University, Fairfax, Virginia 1 Expenditures and Achievement 2 Background and Purpose of the Study There has been much debate in the literature on the impact of geographic location on student achievement. Many educators, state board of education members, legislators, and the general public believe that students from smaller and rural schools receive an education that is inferior to that of students from larger urban or suburban schools. Several studies have not found any significant differences between the two groups. In research completed in the state of New York, Monk and Haller (1986) found that students from smaller (often rural) schools achieved as well as students from larger schools. Moreover, in one New Mexico study, which looked at factors affecting performance of selected high school students, those attending schools in rural areas performed as well as those in urban locales (Ward and Murray, 1985). Other scholars have found, however, that rural-urban differences do exist. One study in Kansas found that the ACT scores of rural students were two points lower than scores of urban students in each of the categories on the ACT (Downey, 1980). Another examination of student performance in Hawaii public schools found sub-standard achievement to be a pattern in rural areas (McCleery, 1979). Education spending has also been a hot topic of debate for many years. Taxpayers often believe that schools receive too much funding and therefore, do not want more of their tax dollars going towards education spending. Teachers and schools, on the other hand, often claim that current funds are insufficient to finance necessary school programs. Public education is a public good financed primarily by state and local governments. Economic theory views education as an important input to the production function. In fact, many empirical studies have shown that education provides positive Expenditures and Achievement 3 returns to society as more education leads to higher productivity and wages (Angrist & Krueger, 1991; Ashenfelter & Krueger, 1994; Card, 1995). Thus, because of these gains, the government invests billions of dollars each year in education. Through the years, it has become a common belief that increasing school funding will lead to increased student academic achievement. Studies that do look at the impact of education spending on student academic achievement have found varying results. After reviewing the current literature, Picus and Robillard (2000) concluded that a clear link between school spending and student academic achievement fails to exist. In a study of schools in Austin, Texas, authors Murnane and Levy (1996) also concluded that the availability of extra resources does not equal greater student achievement. However, after reviewing 35 years of production function research, Verstegen and King (1998) concluded that “resource inputs can and do make a difference in students’ educational outcomes.” Nyhan and Alkadry (1999) also attempted to answer the question of whether school funding affects student achievement. The study averaged math, reading, and writing standardized test scores to create one dependent variable, which was then regressed on expenditures per student, the percent of students in poverty, and the percent of minority students. No conclusive results were made. But does funding lead to higher student performance or is this a misconception so easily believed by the public? Are there disadvantages for urban or rural students? The purpose of this paper is to explore the relationships between instructional expenditures, student achievement, and demographics of school districts in North Carolina. Specifically, this paper addresses the following research questions: (1) Do total Expenditures and Achievement 4 expenditures (instruction, student services, per-pupil allocations, administrative, operations, transportation, business, and other) predict average math scores on the Scholastic Aptitude Test (SAT) for school districts in North Carolina?, (2) Do North Carolina school districts' expenditures (current expenditures on instruction and teacher salaries) and demographics (enrollment, size, poverty, fte, ratio) predict their status on meeting Adequate Yearly Progress (AYP) proficiency growth in mathematics?, (3) Are there differences in current expenditure on instruction by type of district (local or charter) and geographic location (urban, rural, city, town, suburb)?, and (4) Are there differences among school districts located in various geographic locations (urban, rural, city, town, suburb) on current expenditures on instruction, student services, teacher salaries, and other general expenses? Method This paper analyzes the effects of certain school district-related expenditures and various demographics on student achievement in mathematics in school districts in North Carolina. This paper will focus specifically on the use of the following statistical procedures—multiple regression, logistic regression, two-factor ANOVA and MANOVA. Sample/ Data Description The sample used in this paper consists of data on school districts in North Carolina. These data were accessed via the internet and downloaded into spreadsheets for Expenditures and Achievement 5 analysis. For the purposes of this paper, all data were obtained at the district level for the 2007 fiscal year (or the 2007-08 academic year). Specifically, district expenditure data were obtained from the U.S. Department of Education, National Center for Education Statistics, Common Core of Data (CCD), Local Education Agency (School District) Finance Survey", 2007-08. These data have also been released as the F-33 survey, which is conducted by the U.S. Census Bureau on behalf of the National Center for Education Statistics. The primary purpose of the Local Education Agency (School District) Finance Survey (F-33) is to provide revenue and expenditure data for all school districts in the United States. Demographic and school characteristics data were obtained from the U.S. Department of Education, National Center for Education Statistics, Common Core of Data (CCD), Local Education Agency (School District) Universe Survey", 2007-08. Its purpose is to provide basic information about all education agencies and the students for whose education the agencies are responsible. These data were accessed via the Elementary/Secondary Information System (ELSi), an NCES web application that allows users to view public and private school data and create custom tables and charts using data from the Common Core of Data (CCD). Additionally, data on poverty level were obtained from the US Census Bureau's Small Area Income and Poverty Estimates (SAIPE), 2007. SAIPE produces estimates of median household income for states and counties, and poverty for states, counties, and school districts. These estimates are based on statistical models that use decennial census data, household surveys, administrative data, and population estimates. Expenditures and Achievement 6 Student achievement data were accessed from the North Carolina Department of Public Instruction's (DPI) website. DPI is the agency charged with implementing the State's public school laws and the State Board of Education's policies and procedures governing pre-kindergarten through 12th grade public education. Specifically, the analyses in this paper utilize average mathematics scores on the SAT Reasoning Test and data on whether school districts met its Adequate Yearly Progress (AYP) proficiency growth goal. Variables We used a number of variables related to district expenditures in the analyses. The expenditure amounts are derived from current spending totals and the school district’s fall membership data. Expenditures do not include spending for non-elementary and -secondary programs (community service, adult education), or spending by a school district for students not included in its fall membership counts. Total Current Spending for Instructional Activities (TCURINST). This variable is operationalized as the money spent per district for the 2007-2008 school year on instructional related activities. Instructional expenditures include costs such as textbooks, classroom supplies, and technological resources. Total Current Spending for Support Services (TCURSSVC). This variable includes school district expenditures on academic enrichment programs, after-school programs, transportation, instructional staff support (e.g., teacher aides), and the operation and maintenance of the school facility. Within this category, and also used in the analyses, are INST (spending on instructional services), PUPL (per pupil allocation, Expenditures and Achievement GADMN (general administrative expenditures), SADMN (school administrative expenditures), OPMAN (spending on operations and maintenance), TRANS (spending on student transportation), and BUSOTH (other business expenses). Total Current Spending for Other Elementary and Secondary Services (TCUROTH). This variable includes all other expenditures related to elementary and secondary districts not accounted for by the above variables. Teacher Salaries (TSAL). This variable only includes expenditures for base salaries. Benefits and other compensation were excluded from this analysis. . We also used additional variables to describe certain characteristics and achievement information for school districts in North Carolina. Geographic Location (GEOG). These data were taken from the Common Core of Data Local Education Agency Universe Survey. This data is a measure of a district’s location relative to populous areas, and is a composite of the school locale codes, weighted by school population, associated with the schools in the district’s jurisdiction. There are five categories (city, urban, town, rural, and suburb).1 Type of District (TYPE). This variable describes whether or not a school district is a local district (comprised of public elementary and secondary schools) or a charter district (comprised only of public elementary and secondary charter schools). Private schools were excluded from this analysis. Poverty (PVTY). This variable includes the number of students aged 5 - 17 whose households are in poverty. Enrollment (ENRL). This variable describes the number of students enrolled in each district for the fall portion of the 2007 school year. Size of District (SIZE). We transformed the enrollment variable, ENRL, to create this 1 None of the school districts in the data were classified as urban. therefore, the analysis in this paper only analyzes four categories - rural, city, town, and suburb. 7 Expenditures and Achievement 8 variable. For purposes of this paper, all school districts with total student enrollment of 20,000 or more, were classified as "large" districts. All other districts were classified as "other." Mathematics Achievement (SAT, AYP). In this paper, we use two measures of student achievement. The SAT Reasoning Test (formerly the Scholastic Aptitude Test or Scholastic Assessment Test) is a standardized test for college admissions in the United States. The test is intended to assess a student's readiness for college. Adequate Yearly Progress, or AYP, is a measurement defined by the United States federal No Child Left Behind Act that allows the U.S. Department of Education to determine how every public school and district in the country is performing academically according to results on standardized assessments. These assessments allow State Education Agencies to develop target starting goals for AYP. After those are developed, states must increase student achievement in gradual increments each school year. Analysis To explore whether school district predict mathematics achievement, we employ multiple regression with dependent variable, average math scores on the Scholastic Aptitude Test (SAT), and predictors, total expenditures on instructional activities (TCURINST), expenditures on student services (TCURSVC), per-pupil expenditures (PUPIL), and expenditures on instructional services (INST), general administration (GADMN), school-related administration (SADMN), operations and maintenance (OPMAN), business and other services (BUSOTH), and all other expenses (CUROTH). To determine whether school districts' expenditures and demographics predict their status on meeting Adequate Yearly Progress (AYP) proficiency growth in mathematics, we Expenditures and Achievement 9 employ logistic regression with dependent variable, AYP = 1 (met AYP) or AYP = 0 (did not meet AYP) and independent variables, expenditures on instructional services (INST), teacher salaries (TSAL), enrollment (ENRL), large vs. other districts (SIZE), number of students in poverty (PVTY), number of full time teachers (FTE), and student: teacher ratio (RATIO). To determine if there are differences in current expenditure on instruction by type of district (local or charter) and geographic location (urban, rural, city, town, suburb), a two-factor ANOVA is employed with dependent variable, total expenditures on instructional activities (TCURINST), factor A, geographic location (GEOG - rural, city, town, suburb) and factor B, type of district (local or charter). We will also compute effect sizes for the data in order to estimate the practical significance of any effects. To determine if there are differences among school districts located in various geographic locations on current expenditures on instruction, student services, teacher salaries, and other general expenses, we use MANOVA with dependent variable, geographic location (GEOG - rural, city, town, suburb) and independent variables, total expenditures on instructional activities (TCURINST), student services (TCURSVC), teacher salaries (TSAL), and all other expenditures (TCUROTH). Results The results of the analyses are discussed below. Multiple Regression The results from the multiple regression show that the prediction of SAT mathematics test scores from current expenditures on instruction, current expenditures on student Expenditures and Achievement 10 services, per-pupil expenditures, INST, GADMN, SADMN, OPMAN, BUSOTH, and CUROTH is statistically significant, F(9, 108) = 3.34, p = .001. Further, R2 = .218 indicates that 21.8 percent of the differences in average SAT mathematics test scores are explained by the differences in the various types of school district expenditures. There is a statistically significant unique contribution to the prediction of average SAT mathematics test scores for GADMN (p = .002), and TRANS (p = .010). The multiple regression equation for predicting average SAT mathematics test scores is SAT = - 0.000016 (GADMN) – 0.000009 (TRANS) + 494.61.The interpretation of the regression coefficients indicates that the predicted average SAT mathematics score decreases by 0.000016 when general administrative expenditures increase by one dollar, holding other variables constant. Similarly, the predicted average SAT mathematics score decreases by 0.000009 when expenditures on student transportation increase by one dollar, holding other variables constant. As noted previously, each predictor GADMN and TRANS, are statistically significant at the .05 level. This indicates that each predictor has its own unique contribution to the explanation of the variance in average SAT mathematics test scores. The magnitude of the unique explanatory contribution for each predictor is indicated by the squared value of its part correlation with average SAT mathematics test scores. Specifically, 6.97 percent of the variance in average SAT mathematics scores is uniquely accounted for by the variance in GADMN. Likewise, 4.97 percent of the variance in average SAT mathematics scores is uniquely accounted for by the variance in TRANS. For these data, GADMN is relatively more important than TRANS for the prediction of average SAT mathematics test scores. Expenditures and Achievement 11 In this analyses, we defined a restricted model by removing the 7 predictors that were not statistically significant, thus obtaining a restricted model with two predictors: GADMN and TRANS. The results show that the two predictors in the restricted model (model 1: TRANS, GADMN) do not account for a statistically significant amount of the variance in SAT math scores, R2 = .024, F(2, 115) = 1.43, p = .243. The change in R2 from Model 1 to Model 2 is statistically significant, R2change = .193, F(7, 108) = 3.81, p = .001. This indicates that CUROTH, BUSOTH, OPMAN, INST, PUPIL, and CURINST account for a statistically significant proportion of the variance in SAT math scores over and above the proportion accounted for by the independent variables in the restricted model. Since the restricted model and the full model differ in how much variance in SAT math scores they account for, it was better to use the full model. See enclosures I and II for the results of the multiple regression analysis. Logistic Regression The results in the Omnibus Tests of Model Coefficients indicate that the prediction of a school district's status on meeting its AYP proficiency goal from current expenditures on instruction and teacher salaries, as well as district characteristics such as enrollment, size, poverty level, number of full time teachers and pupil: teacher ratio, is statistically significant, χ2(7) = 29.94, p < .001. The results of the Hosmer and Lemeshow goodness-of-fit test show that the chisquare statistic is non-significant, x2(8) = 7.70, p = .46, thus indicating a good fit for the logistic regression model. The Nagelkerke R2 value is .311, indicating a relatively high explanatory effect in the prediction of meeting AYP from all 7 predictors together. Expenditures and Achievement 12 The descriptive information provided in the Classification Table indicates a relatively good hit rate (74.8%). The sensitivity in this prediction is also relatively high, 60/(60+19)=75.9%, whereas the specificity is somewhat lower, 26/(26+10)=72.2. Further, the false positive rate is 19/(26+19)=.42 or 42% and the false negative rate is 10/(10+60)=.14 or 14%. Results from the Wald test indicate that there is one stat sig regression coefficient: B5 = -0.002 (for students in poverty). The value of the odds ratio for poverty, Exp(B) = 0.998 indicates that the odds of a school district meeting AYP proficiency status decreases by a factor of .998 (or about 1 time) when the number of students in poverty increases by one unit, when controlling for all other predictors. Table 1 provides data for the analysis of school districts' status on meeting AYP. See enclosure III for results of the logistic regression analysis. Two-Factor ANOVA The means and standard deviations for total current expenditures on instruction by type of district and geographic location are given in enclosure IV. The results from the Levene’s test indicate that the homogeneity of variance assumption is not met, F(7, 199) = 40.686, p = .000. Nonetheless, we present the results here. The ANOVA F-test shows that there is a statistically significant main effect for TYPE, F(1, 199) = 98.10, p = .000, pη2 = .33, a statistically significant main effect for GEOG, F(3, 199) = 30.71, p = .000, pη2 = .32, and a statistically significant interaction between TYPE and GEOG, F(3, 199) = 30.81, p = .000, pη2 = .32, at the .05 level of significance. The partial eta squared (pη2) measure of effect size for TYPE (.33) indicates that 33 percent of school districts' Expenditures and Achievement 13 expenditures on instruction is accounted for by gender differences, controlling for the effects of GEOG and the interaction between TYPE and GEOG. Likewise, the value of pη2 for GEOG (.32) shows that 32 percent of the differences in school districts' expenditures on instruction are accounted for by differences among the geographic locations, controlling for the effects of TYPE and the interaction between TYPE and GEOG. Also, pη2 = .32 for the interaction effect size shows that the interaction between TYPE and GEOG accounts for 32 percent of the differences in school districts' expenditures on instruction, controlling for the effects of TYPE and GEOG. These results are presented in table 2. Given the statistically significant main effects for TYPE and GEOG, the Tukey post hoc method of multiple comparisons was used to determine which groups differ in average SAT mathematics test scores. The results indicate that, at the .05 level, there is a statistically significant difference between school districts located in cities and school districts located in towns (p = .013), but not between rural and city districts (p = .084), rural and town districts (p = .514), rural and suburb districts (p = .967), city and suburb districts (p = .703), and town and suburb groups (p = .545). Specifically, the results provided by the 95 percent confidence interval for the difference between the means of the two groups indicate that spending on instruction for school districts located in cities was higher than spending on instruction for school districts located in towns, by a difference of at least $5,851,745 but no more than $69,383,711 (see table 3). See enclosure IV for the results of the two-factor ANOVA and Tukey post hoc tests. Expenditures and Achievement 14 MANOVA The results from the MANOVA analysis are as follows. Wilk's lambda (Λ = .92) is statistically significant. Therefore, we can conclude that there is a statistically significant difference between at least two groups on some linear combination of the dependent variable. The results from the Eigenvalues table show that that the first linear discriminant function (LDF1) accounts for 85.5 percent of the total variance for the set of dependent variables across the groups. The remaining 14.5 percent is accounted for by the second linear discriminant function (LDF2). The results from the Wilk's Lambda table indicate that LDF1 is statistically significant, Λ = .92, χ2(6) = 16.37, p = .012. The second linear discriminate functions, LDF2, is not statistically significant, Λ = .99, χ2(2) = 2.44, p = .295. The Structure Matrix shows that the dependent variables, TCUROTH, TCURINST, TCURSVC, correlate with LDF1. The table Standardized Canonical Discriminant Function Coefficients shows that INST, TCURSVC and TCUROTH define LDF1 but comparison of their standardized coefficients shows that the meaning of LDF1 is defined primarily by TCURINST and TCURSVC because their standardized coefficients , (-5.040 and 7.353, respectively) are much higher than that for TCUROTH (-1.886). Given that TCURINST and TCURSVC include expenditures for instruction and student support services, we label LDF1 as student learning. Therefore, one dimension--student learning--emerged as a linear combination of the dependent variables that provided the best separation of districts by geographic location. See enclosure V for the results of the MANOVA analysis. Expenditures and Achievement 15 Discussion The issues surrounding efforts to assess the achievement of students are by no means simple. To really assess school districts' impact on students, comparisons must be made among students who are matched by origin, background, and access to information before any meaningful conclusions about achievement can be rendered. Further, an examination of the extent to which certain school districts have less total access to educational information would be important to understand the differences in student achievement in school districts nationwide. The findings of the analyses may suggest that the district aid formula is successful in creating equity across the state in how funding impacts achievement. Additionally, this suggests the school finance funding formula constructs parity in the amount a district is able to receive in funding per year and thus spend on student instruction. Further, the results also imply that in North Carolina there is more to student achievement than the amounts a district spends per pupil. Clearly, poverty plays a role in student achievement, according to these findings. Perhaps it is not the dollar amounts that make a difference but how the funds are used. Possible future research on this issue could examine whether an association exists between how a district spends it money and its impact on achievement levels. Further, other kinds of funding, such as family and school resources, could also be examined to determine if there is a relationship with student outcomes. Expenditures and Achievement 16 References Angrist, J. & Krueger, A. (1991). Does compulsory school attendance affect schooling & earnings? Quarterly Journal of Economics, 106(4), 979-1014. Ashenfelter, O. & Krueger, A. (1994). Estimates of the economic return to schooling from a new sample of twins. American Economic Review, 84(5), 1157-1173. Card, D. (1995). Earnings, schooling & ability revisited. Research in Labor Economics, 14, 23-48. Dimitrov, D. (2008). Quantitative research in education. Oceanside, NY: Whittier Publications, Inc. Downey, R. (1980). Higher education and rural youth. Paper presented at the Kansas State University Rural and Small Schools Conference, Auburn, AL. McCleery, M. (1979). Stranger in paradise: Process and product in a district office. Washington, D.C.: National Institute of Education. Monk, D. & Haller, E. (1986). Organizational alternatives for small rural schools. Cornell University: New York State College of Agriculture and Life Sciences, New York. Murnane, R. J., & Levy, F. (1996). Evidence from fifteen schools in Austin, Texas. In G. Burtless (Ed.), Does money matter? The effect of school resources on student achievement and adult success (pp. 93-96). Washington, D.C.: Brookings Institution Press. Nyhan, R. C., & Alkadry, M. G. (1999). The impact of school resources on student achievement test scores. Journal of Education Finance, 25(2), 211-227. Expenditures and Achievement 17 Picus, L. & Robillard, E. (2000). The collection and use of student level data: implications for school finance research. Educational Considerations, 28(1), 2631. Verstegen, D. A., & King, R. A. (1998). The relationship between school spending and student achievement: A review and analysis of 35 years of production function research. Journal of Education Finance, 24(2), 243-262. Ward, A. & Murray, L. (1985). Factors affecting performance of New Mexico high school students. Paper presented at the Meeting of the Rocky Mountain Educational Research Association, Las Cruces, NM. Expenditures and Achievement 18 Tables Table 1: Logistic Regression Analysis of School Districts' Status on Meeting AYP as a Function of Expenditures and District Characteristics ________________________________________________________________________ 95% Confidence Interval for Odds Ratio ______________________ Variables B Wald χ2 Odds Ratio Lower Upper ________________________________________________________________________ INST .00 .065 1.00 1.00 1.00 TSAL .00 .258 1.00 1.00 1.00 ENRL .00 1.770 1.00 1.00 1.00 SIZE -1.286 1.080 .276 .024 3.125 PVTY -.002 12.56* .998 .996 .999 FTE -.014 1.818 .986 .967 1.006 RATIO -.389 1.661 .678 .375 1.224 (Constant) 6.049 2.142 _______________________________________________________________________ Note: Wald (df = 1). *p < .01. Table 2: Analysis of Variance for Total Expenditures on Instruction _______________________________________________________________________ Source df F pη2 p _______________________________________________________________________ District type (TYPE) 1 98.10 0.33 .000 Geographic location (GEOG) 3 30.71 0.32 .000 TYPE X GEOG 3 30.81 0.32 .000 199 (3.016E15) S within group error _______________________________________________________________________ Note: The value enclosed in parentheses is the mean square error (MSw). S = subjects. Expenditures and Achievement 19 Table 3: Multiple Comparisons for Total Expenditures on Instruction Among Geographic Locations Geographic Location ΔM 95% CI for ΔM SEΔM Rural - City -22873379.00 9609749.423 -47771001.02 2024243.01 Rural - Town 14744349.27 1.069E7 -12960550.80 42449249.33 Rural - Suburb -6445558.67 1.399E7 -42691984.80 29800867.46 City - Town 37617728.27* 1.226E7 5851745.07 69383711.46 City - Suburb 16427820.33 1.522E7 -23009720.26 55865360.92 Town - Suburb -21189907.94 1.593E7 -62457201.45 20077385.57 _____________________________________________________________________________________ Note: ΔM = Mean difference. SEΔM = Standard error of ΔM. Expenditures and Achievement Enclosures Enclosure I: SPSS Output for Multiple Regression - Full Model Variables Entered/Removedb Model d 1 Variables Variables Entered Removed CUROTH, Method . Enter i GADMN, m BUSOTH, e OPMAN, n TRANS, INST, s PUPIL, i CURINST, o SADMNa n 0 a. Tolerance = .000 limits reached. b. Dependent Variable: SAT Model Summary Model R d 1 .466a R Square Adjusted R Std. Error of the Square Estimate .218 .152 32.479 i m e n s i o n 0 a. Predictors: (Constant), CUROTH, GADMN, BUSOTH, OPMAN, TRANS, INST, PUPIL, CURINST, SADMN 20 Expenditures and Achievement ANOVAb Model 1 Sum of Squares Regression df Mean Square 31689.122 9 3521.014 Residual 113929.971 108 1054.907 Total 145619.093 117 F Sig. 3.338 .001a a. Predictors: (Constant), CUROTH, GADMN, BUSOTH, OPMAN, TRANS, INST, PUPIL, CURINST, SADMN b. Dependent Variable: SAT Coefficientsa Model Standardized Unstandardized Coefficients B 1 Std. Error (Constant) 494.611 5.713 CURINST 6.490E-7 .000 PUPIL 3.003E-6 INST Coefficients Beta t Sig. 86.569 .000 1.711 1.427 .156 .000 .783 .805 .422 -6.809E-6 .000 -1.074 -1.710 .090 GADMN -1.605E-5 .000 -.676 -3.107 .002 SADMN 5.962E-6 .000 1.722 1.065 .289 OPMAN 1.251E-6 .000 .392 .500 .618 TRANS -7.862E-6 .000 -1.706 -2.622 .010 BUSOTH -2.333E-6 .000 -.354 -.640 .523 CUROTH -3.913E-6 .000 -.787 -1.515 .133 a. Dependent Variable: SAT 21 Expenditures and Achievement Coefficientsa Model Correlations Zero-order 1 Partial Part (Constant) CURINST .195 .136 .121 PUPIL .190 .077 .069 INST .157 -.162 -.146 GADMN .089 -.286 -.264 SADMN .185 .102 .091 OPMAN .203 .048 .043 TRANS .148 -.245 -.223 BUSOTH .155 -.061 -.054 CUROTH .162 -.144 -.129 a. Dependent Variable: SAT Excluded Variablesb Model Collinearity Beta In 1 CURSVC t .a Sig. . . Partial Statistics Correlation Tolerance . .000 a. Predictors in the Model: (Constant), CUROTH, GADMN, BUSOTH, OPMAN, TRANS, INST, PUPIL, CURINST, SADMN b. Dependent Variable: SAT 22 Expenditures and Achievement 23 Enclosure II: SPSS Output for Multiple Regression - Comparison with Restricted Model Variables Entered/Removedc Model d 1 Variables Entered Removed TRANS, Method . Enter GADMNa i m Variables 2 e n s i o CUROTH, . Enter BUSOTH, OPMAN, INST, PUPIL, CURINST, SADMNb n 0 a. All requested variables entered. b. Tolerance = .000 limits reached. c. Dependent Variable: SAT Model Summary Model Change Statistics R d i R Square Adjusted R Std. Error of the R Square Square Estimate Change F Change df1 df2 Sig. F Change 1 .156a .024 .007 35.150 .024 1.431 2 115 .243 2 .466b .218 .152 32.479 .193 3.812 7 108 .001 m e n s i o n 0 a. Predictors: (Constant), TRANS, GADMN b. Predictors: (Constant), TRANS, GADMN, CUROTH, BUSOTH, OPMAN, INST, PUPIL, CURINST, SADMN Expenditures and Achievement 24 ANOVAc Model 1 2 Sum of Squares Regression df Mean Square 3536.696 2 1768.348 Residual 142082.397 115 1235.499 Total 145619.093 117 31689.122 9 3521.014 Residual 113929.971 108 1054.907 Total 145619.093 117 Regression F Sig. 1.431 .243a 3.338 .001b a. Predictors: (Constant), TRANS, GADMN b. Predictors: (Constant), TRANS, GADMN, CUROTH, BUSOTH, OPMAN, INST, PUPIL, CURINST, SADMN c. Dependent Variable: SAT Coefficientsa Model Standardized Unstandardized Coefficients B 1 2 (Constant) Std. Error 491.231 4.976 GADMN -1.860E-6 .000 TRANS 9.705E-7 .000 494.611 5.713 GADMN -1.605E-5 .000 TRANS -7.862E-6 CURINST PUPIL Coefficients Beta Correlations t Sig. Zero-order Partial Part 98.719 .000 -.078 -.516 .607 .089 -.048 -.048 .211 1.389 .168 .148 .128 .128 86.569 .000 -.676 -3.107 .002 .089 -.286 -.264 .000 -1.706 -2.622 .010 .148 -.245 -.223 6.490E-7 .000 1.711 1.427 .156 .195 .136 .121 3.003E-6 .000 .783 .805 .422 .190 .077 .069 -6.809E-6 .000 -1.074 -1.710 .090 .157 -.162 -.146 SADMN 5.962E-6 .000 1.722 1.065 .289 .185 .102 .091 OPMAN 1.251E-6 .000 .392 .500 .618 .203 .048 .043 BUSOTH -2.333E-6 .000 -.354 -.640 .523 .155 -.061 -.054 CUROTH -3.913E-6 .000 -.787 -1.515 .133 .162 -.144 -.129 (Constant) INST a. Dependent Variable: SAT Expenditures and Achievement Excluded Variablesc Model Collinearity Beta In 1 Sig. Statistics Correlation Tolerance CURINST 1.680a 3.442 .001 .307 .033 CURSVC 1.971a 2.979 .004 .269 .018 PUPIL 1.585a 3.201 .002 .287 .032 .282a .789 .432 .074 .067 SADMN 2.052a 3.431 .001 .306 .022 OPMAN 1.332a 3.318 .001 .297 .048 BUSOTH .222a .461 .646 .043 .037 CUROTH .219a .786 .433 .073 .109 CURSVC .b . . . .000 INST 2 t Partial a. Predictors in the Model: (Constant), TRANS, GADMN b. Predictors in the Model: (Constant), TRANS, GADMN, CUROTH, BUSOTH, OPMAN, INST, PUPIL, CURINST, SADMN c. Dependent Variable: SAT 25 Expenditures and Achievement Enclosure III: SPSS Output for Logistic Regression Logistic Regression Case Processing Summary Unweighted Casesa Selected Cases N Included in Analysis Missing Cases Total Unselected Cases Total Percent 115 100.0 0 .0 115 100.0 0 .0 115 100.0 a. If weight is in effect, see classification table for the total number of cases. Dependent Variable Encoding Original Value dim ens Internal Value Not Met AYP 0 Met AYP 1 ion 0 Block 0: Beginning Block Classification Tablea,b Observed Predicted AYP Not Met AYP Step 0 AYP Percentage Met AYP Correct Not Met AYP 0 45 .0 Met AYP 0 70 100.0 Overall Percentage a. Constant is included in the model. b. The cut value is .500 60.9 26 Expenditures and Achievement Variables in the Equation B Step 0 Constant S.E. .442 Wald .191 df 5.347 1 Variables not in the Equationa Score Step 0 Variables df Sig. INST .051 1 .822 TSAL .067 1 .795 ENRL .080 1 .778 SIZE .085 1 .771 PVTY 1.146 1 .284 FTE .081 1 .776 RATIO .965 1 .326 a. Residual Chi-Squares are not computed because of redundancies. Block 1: Method = Enter Omnibus Tests of Model Coefficients Chi-square Step 1 df Sig. Step 29.943 7 .000 Block 29.943 7 .000 Model 29.943 7 .000 Model Summary Step -2 Log likelihood 1 124.003a Cox & Snell R Nagelkerke R Square Square .229 a. Estimation terminated at iteration number 6 because parameter estimates changed by less than .001. .311 Sig. .021 Exp(B) 1.556 27 Expenditures and Achievement Hosmer and Lemeshow Test Step Chi-square 1 df 7.701 Sig. 8 .463 Contingency Table for Hosmer and Lemeshow Test AYP = Not Met AYP Observed Step 1 AYP = Met AYP Expected Observed Expected Total 1 8 10.143 4 1.857 12 2 10 7.762 2 4.238 12 3 8 6.403 4 5.597 12 4 4 5.254 8 6.746 12 5 4 4.678 8 7.322 12 6 5 3.797 7 8.203 12 7 2 3.025 10 8.975 12 8 2 2.332 10 9.668 12 9 2 1.489 10 10.511 12 10 0 .118 7 6.882 7 Classification Tablea Observed Predicted AYP Not Met AYP Step 1 AYP Percentage Met AYP Correct Not Met AYP 26 19 57.8 Met AYP 10 60 85.7 Overall Percentage a. The cut value is .500 74.8 28 Expenditures and Achievement Variables in the Equation B Step 1a S.E. Wald df Sig. INST .000 .000 .065 1 .798 1.000 TSAL .000 .000 .258 1 .611 1.000 ENRL .001 .000 1.770 1 .183 1.001 SIZE -1.286 1.238 1.080 1 .299 .276 PVTY -.002 .001 12.559 1 .000 .998 FTE -.014 .010 1.818 1 .178 .986 RATIO -.389 .302 1.661 1 .198 .678 Constant 6.049 4.133 2.142 1 .143 423.836 a. Variable(s) entered on step 1: INST, TSAL, ENRL, SIZE, PVTY, FTE, RATIO. Variables in the Equation 95% C.I.for EXP(B) Lower Step 1a Exp(B) Upper INST 1.000 1.000 TSAL 1.000 1.000 ENRL 1.000 1.002 SIZE .024 3.125 PVTY .996 .999 FTE .967 1.006 RATIO .375 1.224 Constant a. Variable(s) entered on step 1: INST, TSAL, ENRL, SIZE, PVTY, FTE, RATIO. 29 Expenditures and Achievement Enclosure IV: SPSS Output for ANOVA Between-Subjects Factors Value Label TYPE GEOG N 0 charter 92 1 local 115 2 rural 107 3 city 47 4 town 35 5 suburb 18 Descriptive Statistics Dependent Variable:TCURINST TYPE GEOG charter rural 1366200.00 847591.818 30 city 1179270.27 653408.952 37 town 1340687.50 920167.138 16 suburb 1649777.78 1357105.447 9 Total 1314326.09 847866.594 92 rural 42656337.66 3.552E7 77 2.49E8 2.312E8 10 town 28962368.42 1.949E7 19 suburb 73400666.67 4.509E7 9 Total 60761643.48 9.371E7 115 rural 31079663.55 3.538E7 107 city 53953042.55 1.449E8 47 town 16335314.29 1.991E7 35 suburb 37525222.22 4.817E7 18 Total 34340613.53 7.574E7 207 local city Total Mean Std. Deviation N 30 Expenditures and Achievement 31 Levene's Test of Equality of Error Variancesa Dependent Variable:TCURINST F df1 40.686 df2 7 Sig. 199 .000 Tests the null hypothesis that the error variance of the dependent variable is equal across groups. a. Design: Intercept + TYPE + GEOG + TYPE * GEOG Tests of Between-Subjects Effects Dependent Variable:TCURINST Source Type III Sum of Squares Partial Eta df Mean Square F Sig. Squared Corrected Model 5.817E17 7 8.310E16 27.555 .000 .492 Intercept 3.129E17 1 3.129E17 103.770 .000 .343 TYPE 2.958E17 1 2.958E17 98.102 .000 .330 GEOG 2.778E17 3 9.260E16 30.705 .000 .316 TYPE * GEOG 2.788E17 3 9.292E16 30.813 .000 .317 Error 6.001E17 199 3.016E15 Total 1.426E18 207 Corrected Total 1.182E18 206 a. R Squared = .492 (Adjusted R Squared = .474) Expenditures and Achievement 32 Post Hoc Tests GEOG Multiple Comparisons TCURINST Tukey HSD (I) GEOG (J) GEOG Difference (I-J) rural dimensio 95% Confidence Interval Mean Std. Error Sig. Lower Bound Upper Bound city -22873379.00 9609749.423 .084 -47771001.02 2024243.01 town 14744349.27 1.069E7 .514 -12960550.80 42449249.33 suburb -6445558.67 1.399E7 .967 -42691984.80 29800867.46 rural 22873379.00 9609749.423 .084 -2024243.01 47771001.02 town 37617728.27* 1.226E7 .013 5851745.07 69383711.46 suburb 16427820.33 1.522E7 .703 -23009720.26 55865360.92 -14744349.27 1.069E7 .514 -42449249.33 12960550.80 -3.76E7 1.226E7 .013 -69383711.46 -5851745.07 -21189907.94 1.593E7 .545 -62457201.45 20077385.57 6445558.67 1.399E7 .967 -29800867.46 42691984.80 city -16427820.33 1.522E7 .703 -55865360.92 23009720.26 town 21189907.94 1.593E7 .545 -20077385.57 62457201.45 n3 di city m dimensio e n3 n si town rural dimensio o city n3 suburb n 2 suburb rural dimensio n3 Based on observed means. The error term is Mean Square(Error) = 3015678513561469.000. *. The mean difference is significant at the .05 level. Homogeneous Subsets TCURINST Tukey HSDa,b,c GEOG Subset N 1 2 town 35 16335314.29 rural 107 31079663.55 31079663.55 suburb 18 37525222.22 37525222.22 city 47 Sig. 53953042.55 .375 .306 Expenditures and Achievement Means for groups in homogeneous subsets are displayed. Based on observed means. The error term is Mean Square(Error) = 3015678513561469.000. a. Uses Harmonic Mean Sample Size = 34.859. b. The group sizes are unequal. The harmonic mean of the group sizes is used. Type I error levels are not guaranteed. c. Alpha = .05. Profile Plots 33 Expenditures and Achievement Enclosure V: SPSS Output for MANOVA Analysis 1 Stepwise Statistics Variables Entered/Removeda,b,c,d Step Wilks' Lambda Entered Statistic df1 df2 df3 1 TCURSVC .969 1 2 186.000 2 TCURINST .927 2 2 186.000 3 TCUROTH .915 3 2 186.000 At each step, the variable that minimizes the overall Wilks' Lambda is entered. a. Maximum number of steps is 8. b. Minimum partial F to enter is 0. c. Maximum partial F to remove is 0. d. F level, tolerance, or VIN insufficient for further computation. Variables Entered/Removeda,b,c,d Step Wilks' Lambda Exact F Statistic df1 df2 Sig. 1 3.006 2 186.000 .052 2 3.555 4 370.000 .007 3 2.774 6 368.000 .012 At each step, the variable that minimizes the overall Wilks' Lambda is entered. a. Maximum number of steps is 8. b. Minimum partial F to enter is 0. c. Maximum partial F to remove is 0. d. F level, tolerance, or VIN insufficient for further computation. 34 Expenditures and Achievement Variables in the Analysis Step Tolerance F to Remove Wilks' Lambda 1 TCURSVC 1.000 3.006 2 TCURSVC .009 4.580 .973 TCURINST .009 4.124 .969 TCURSVC .009 3.574 .951 TCURINST .007 1.261 .928 TCUROTH .047 1.208 .927 3 Variables Not in the Analysis Step 0 1 2 3 Tolerance Min. Tolerance F to Enter Wilks' Lambda TCURINST 1.000 1.000 2.556 .973 TCURSVC 1.000 1.000 3.006 .969 TCUROTH 1.000 1.000 1.911 .980 TSAL 1.000 1.000 2.602 .973 TCURINST .009 .009 4.124 .927 TCUROTH .064 .064 4.068 .928 TSAL .009 .009 3.400 .934 TCUROTH .047 .007 1.208 .915 TSAL .000 .000 3.400 .934 TSAL .000 .000 3.400 .934 Wilks' Lambda Step Number of Variables Lambda df1 df2 df3 1 1 .969 1 2 186 2 2 .927 2 2 186 3 3 .915 3 2 186 35 Expenditures and Achievement Wilks' Lambda Step Exact F Statistic df1 df2 Sig. 1 3.006 2 186.000 .052 2 3.555 4 370.000 .007 3 2.774 6 368.000 .012 Pairwise Group Comparisonsa,b,c Step GEOG 1 rural rural F Sig. city town 2 rural F rural town .050 .393 .050 .023 F .732 5.264 Sig. .393 .023 F F F 1.232 .003 .294 3.196 .003 .043 1.232 3.196 .294 .043 F F 5.966 5.966 Sig. city .732 Sig. Sig. 3 3.901 5.264 Sig. town town 3.901 Sig. city city 4.778 .993 .003 .397 4.778 2.238 Sig. .003 .085 F .993 2.238 Sig. .397 .085 a. 1, 186 degrees of freedom for step 1. b. 2, 185 degrees of freedom for step 2. c. 3, 184 degrees of freedom for step 3. 36 Expenditures and Achievement Summary of Canonical Discriminant Functions Eigenvalues Function Canonical Eigenvalue % of Variance Cumulative % Correlation 1 .078a 85.5 85.5 .269 2 .013a 14.5 100.0 .115 dimension0 a. First 2 canonical discriminant functions were used in the analysis. Wilks' Lambda Test of Function(s) Wilks' Lambda Chi-square df Sig. 1 through 2 .915 16.368 6 .012 2 .987 2.443 2 .295 dimension0 Standardized Canonical Discriminant Function Coefficients Function 1 2 TCURINST -5.040 3.264 TCURSVC 7.383 -3.578 TCUROTH -1.886 1.209 Structure Matrix Function 1 2 TCUROTH .339 .932* TCURINST .468 .883* TSALa .477 .877* TCURSVC .542 .841* 37 Expenditures and Achievement Pooled within-groups correlations between discriminating variables and standardized canonical discriminant functions Variables ordered by absolute size of correlation within function. *. Largest absolute correlation between each variable and any discriminant function a. This variable not used in the analysis. Functions at Group Centroids GEOG Function 1 d i m 2 rural -.195 .060 city .471 .042 town -.038 -.239 e n s i o n 0 Unstandardized canonical discriminant functions evaluated at group means 38 Expenditures and Achievement Classification Statistics Classification Processing Summary Processed Excluded 207 Missing or out-of-range 0 group codes At least one missing 0 discriminating variable Used in Output 207 Prior Probabilities for Groups GEOG Cases Used in Analysis Prior d i m e n s i o n 0 Unweighted Weighted rural .333 107 107.000 city .333 47 47.000 town .333 35 35.000 Total 1.000 189 189.000 39 Expenditures and Achievement T-TEST GROUPS = GEOG(1, 2) /VARIABLES = Y1 TO Y4. 40 Expenditures and Achievement T-Test Group Statistics GEOG TCURINST d i 1 rural N Mean Std. Deviation Std. Error Mean 0a . . . 107 31079663.55 3.538E7 3420789.931 0a . . . 107 15341205.61 1.611E7 1557844.675 0a . . . 107 2796112.15 3154302.211 304937.905 0a . . . m e n s i o n 1 TCURSVC d i 1 rural m e n s i o n 1 TCUROTH d i 1 rural m e n s i o n 1 TSAL d 1 41 Expenditures and Achievement i rural 107 22705205.61 2.584E7 2498358.948 m e n s i o n 1 a. t cannot be computed because at least one of the groups is empty. T-TEST GROUPS = GEOG(1, 3) /VARIABLES = Y1 TO Y4. T-Test Group Statistics GEOG TCURINST d i N Mean Std. Deviation Std. Error Mean 1 0a . . . city 47 53953042.55 1.449E8 2.113E7 1 0a . . . m e n s i o n 1 TCURSVC d 42 Expenditures and Achievement i city 47 29556744.68 7.852E7 1.145E7 1 0a . . . city 47 4046234.04 1.091E7 1591415.427 1 0a . . . city 47 40070255.32 1.080E8 1.575E7 m e n s i o n 1 TCUROTH d i m e n s i o n 1 TSAL d i m e n s i o n 1 a. t cannot be computed because at least one of the groups is empty. T-TEST GROUPS = GEOG(1, 4) /VARIABLES = Y1 TO Y4. T-Test Group Statistics 43 Expenditures and Achievement GEOG TCURINST d i N Mean Std. Deviation Std. Error Mean 1 0a . . . town 35 16335314.29 1.991E7 3365340.806 1 0a . . . town 35 8487542.86 1.016E7 1718010.158 1 0a . . . town 35 1447314.29 1791190.596 302766.471 1 0a . . . m e n s i o n 1 TCURSVC d i m e n s i o n 1 TCUROTH d i m e n s i o n 1 TSAL d 44 Expenditures and Achievement i town 35 11951971.43 1.474E7 2491982.419 m e n s i o n 1 a. t cannot be computed because at least one of the groups is empty. T-TEST GROUPS = GEOG(2, 3) /VARIABLES = Y1 TO Y4. T-Test Group Statistics GEOG TCURINST d i N Mean Std. Deviation Std. Error Mean rural 107 31079663.55 3.538E7 3420789.931 city 47 53953042.55 1.449E8 2.113E7 rural 107 15341205.61 1.611E7 1557844.675 m e n s i o n 1 TCURSVC d 45 Expenditures and Achievement i city 47 29556744.68 7.852E7 1.145E7 rural 107 2796112.15 3154302.211 304937.905 city 47 4046234.04 1.091E7 1591415.427 rural 107 22705205.61 2.584E7 2498358.948 city 47 40070255.32 1.080E8 1.575E7 46 m e n s i o n 1 TCUROTH d i m e n s i o n 1 TSAL d i m e n s i o n 1 Independent Samples Test Levene's Test for Equality of Variances F TCURINST Equal variances assumed t-test for Equality of Means Sig. 27.283 t .000 Equal variances not df -1.538 152 -1.069 48.428 -1.796 152 assumed TCURSVC Equal variances assumed 32.638 .000 Expenditures and Achievement Equal variances not 47 -1.230 47.711 -1.090 152 -.772 49.411 -1.570 152 -1.089 48.330 assumed TCUROTH Equal variances assumed 21.921 .000 Equal variances not assumed TSAL Equal variances assumed 27.547 .000 Equal variances not assumed Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Sig. (2-tailed) TCURINST Mean Std. Error Difference Difference Difference Lower Equal variances assumed .126 -2.287E7 1.487E7 -5.226E7 Equal variances not .291 -2.287E7 2.141E7 -6.590E7 Equal variances assumed .075 -1.422E7 7917094.342 -2.986E7 Equal variances not .225 -1.422E7 1.156E7 -3.746E7 Equal variances assumed .277 -1250121.893 1146989.003 -3516221.102 Equal variances not .444 -1250121.893 1620367.300 -4505686.956 Equal variances assumed .119 -1.737E7 1.106E7 -3.922E7 Equal variances not .282 -1.737E7 1.595E7 -4.943E7 assumed TCURSVC assumed TCUROTH assumed TSAL assumed Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Difference Upper Expenditures and Achievement TCURINST Equal variances assumed 6512786.149 Equal variances not 2.016E7 assumed TCURSVC Equal variances assumed 1426216.143 Equal variances not 9028274.745 assumed TCUROTH Equal variances assumed 1015977.315 Equal variances not 2005443.170 assumed TSAL Equal variances assumed 4487804.840 Equal variances not 1.470E7 assumed T-TEST GROUPS = GEOG(2, 4) /VARIABLES = Y1 TO Y4. T-Test Group Statistics GEOG TCURINST d i N Mean Std. Deviation Std. Error Mean rural 107 31079663.55 3.538E7 3420789.931 town 35 16335314.29 1.991E7 3365340.806 rural 107 15341205.61 1.611E7 1557844.675 m e n s i o n 1 TCURSVC d 48 Expenditures and Achievement i town 35 8487542.86 1.016E7 1718010.158 rural 107 2796112.15 3154302.211 304937.905 town 35 1447314.29 1791190.596 302766.471 rural 107 22705205.61 2.584E7 2498358.948 town 35 11951971.43 1.474E7 2491982.419 49 m e n s i o n 1 TCUROTH d i m e n s i o n 1 TSAL d i m e n s i o n 1 Independent Samples Test Levene's Test for Equality of Variances F TCURINST Equal variances assumed t-test for Equality of Means Sig. 8.090 t .005 Equal variances not df 2.343 140 3.073 104.703 2.364 140 assumed TCURSVC Equal variances assumed 6.571 .011 Expenditures and Achievement Equal variances not 50 2.955 92.779 2.403 140 3.139 103.729 2.337 140 3.047 103.242 assumed TCUROTH Equal variances assumed 8.059 .005 Equal variances not assumed TSAL Equal variances assumed 8.065 .005 Equal variances not assumed Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Sig. (2-tailed) TCURINST Mean Std. Error Difference Difference Difference Lower Equal variances assumed .021 1.474E7 6292551.508 2303636.901 Equal variances not .003 1.474E7 4798679.244 5229140.470 Equal variances assumed .019 6853662.750 2899354.415 1121483.319 Equal variances not .004 6853662.750 2319146.165 2248153.575 Equal variances assumed .018 1348797.864 561413.737 238852.746 Equal variances not .002 1348797.864 429714.628 496631.388 Equal variances assumed .021 1.075E7 4601655.750 1655513.563 Equal variances not .003 1.075E7 3528707.101 3755070.971 assumed TCURSVC assumed TCUROTH assumed TSAL assumed Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Difference Upper Expenditures and Achievement TCURINST Equal variances assumed 2.719E7 Equal variances not 2.426E7 assumed TCURSVC Equal variances assumed 1.259E7 Equal variances not 1.146E7 assumed TCUROTH Equal variances assumed 2458742.982 Equal variances not 2200964.340 assumed TSAL Equal variances assumed 1.985E7 Equal variances not 1.775E7 assumed T-TEST GROUPS = GEOG(3, 4) /VARIABLES = Y1 TO Y4. T-Test Group Statistics GEOG TCURINST d i N Mean Std. Deviation Std. Error Mean city 47 53953042.55 1.449E8 2.113E7 town 35 16335314.29 1.991E7 3365340.806 city 47 29556744.68 7.852E7 1.145E7 m e n s i o n 1 TCURSVC d 51 Expenditures and Achievement i town 35 8487542.86 1.016E7 1718010.158 city 47 4046234.04 1.091E7 1591415.427 town 35 1447314.29 1791190.596 302766.471 city 47 40070255.32 1.080E8 1.575E7 town 35 11951971.43 1.474E7 2491982.419 52 m e n s i o n 1 TCUROTH d i m e n s i o n 1 TSAL d i m e n s i o n 1 Independent Samples Test Levene's Test for Equality of Variances F TCURINST Equal variances assumed t-test for Equality of Means Sig. 13.525 t .000 Equal variances not df 1.523 80 1.758 48.321 1.575 80 assumed TCURSVC Equal variances assumed 14.222 .000 Expenditures and Achievement Equal variances not 53 1.819 48.060 1.393 80 1.604 49.303 1.527 80 1.763 48.290 assumed TCUROTH Equal variances assumed 12.247 .001 Equal variances not assumed TSAL Equal variances assumed 13.488 .000 Equal variances not assumed Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Sig. (2-tailed) TCURINST Mean Std. Error Difference Difference Difference Lower Equal variances assumed .132 3.762E7 2.470E7 -1.153E7 Equal variances not .085 3.762E7 2.140E7 -5397637.784 Equal variances assumed .119 2.107E7 1.338E7 -5548588.428 Equal variances not .075 2.107E7 1.158E7 -2215804.472 Equal variances assumed .167 2598919.757 1865408.434 -1113361.334 Equal variances not .115 2598919.757 1619960.060 -656006.187 Equal variances assumed .131 2.812E7 1.841E7 -8516315.469 Equal variances not .084 2.812E7 1.595E7 -3942946.300 assumed TCURSVC assumed TCUROTH assumed TSAL assumed Expenditures and Achievement Independent Samples Test t-test for Equality of Means 95% Confidence Interval of the Difference Upper TCURINST Equal variances assumed 8.677E7 Equal variances not 8.063E7 assumed TCURSVC Equal variances assumed 4.769E7 Equal variances not 4.435E7 assumed TCUROTH Equal variances assumed 6311200.848 Equal variances not 5853845.701 assumed TSAL Equal variances assumed 6.475E7 Equal variances not 6.018E7 assumed 54