Chicago high school malmquist productivity index

8 S. Grosskopf, C. Moutray Economics of Education Review 20 2001 1–14 Table 1 Means for Chicago and Illinois high schools, 1989–1994 a 1989 1990 1991 1992 1993 1994 Chicago n = 60 n = 60 n = 60 n = 60 n = 61 n = 61 Illinois n = 677 n = 667 n = 664 n = 657 n = 648 n = 647 Percentage of white students Chicago 11.997 11.307 10.598 10.135 9.964 9.890 Illinois 84.036 83.691 83.001 83.042 82.560 82.162 Total school enrollment Chicago 1728.417 1647.250 1610.267 1624.167 1605.639 1584.919 Illinois 767.532 752.936 749.566 769.254 793.111 805.685 Percentage of LEP students Chicago 2.817 4.155 4.972 5.805 6.841 7.536 Illinois 0.694 0.933 1.085 1.160 1.324 1.430 Low-income students Chicago 36.295 37.650 43.415 57.302 56.792 69.282 Illinois 14.376 14.689 15.945 17.121 18.132 20.530 Mobility rate Chicago 27.872 28.352 27.895 27.215 28.439 24.221 Illinois 15.011 14.910 15.149 14.489 14.690 14.927 Attendance rate Chicago 80.483 79.982 79.707 79.879 77.656 77.656 Illinois 92.355 92.516 92.416 92.537 92.363 92.021 HS graduate rate Chicago 45.615 46.460 43.347 48.028 49.562 49.562 Illinois 85.301 84.600 85.284 85.226 85.606 80.666 Average ACT English score Chicago 15.050 b 15.363 15.063 14.898 14.618 14.618 Illinois 20.152 b 20.521 19.950 19.866 20.046 19.985 Average ACT math score Chicago 15.883 b 15.893 15.908 16.061 15.830 15.830 Illinois 19.318 b 19.818 19.814 20.005 20.185 20.075 Number of teachers c Chicago 110.443 107.642 107.492 106.802 105.962 94.780 Average teacher salary c Chicago 33044.754 34904.138 35018.49 37787.151 41887.728 41036.428 Number of admin. c Chicago 4.063 4.093 3.828 2.813 3.358 3.423 Average admin. salary c Chicago 43070.100 44128.171 45377.372 52057.411 53480.715 52834.667 a Sources: Compiled by the authors using data taken from the Illinois State Board of Education, Chicago Panel on Public School Policy and Finance, and ACT 1989. b Concordant value. c State averages not available. teacher and administrator data compiled from the Chicago Panel on Public School Policy and Finance. Table 1 illustrates the striking difference in character- istics between public high schools in Chicago and in the state as a whole. Chicago schools have relatively more nonwhite students, bigger enrollments, lower graduate rates, higher truancy, and lower scores. As shown in Appendix C, these differences are often statistically sig- nificant based on simple z-tests. 13 Turning to Table 2 which includes descriptive stat- istics of the variables included in our model, we see that the data are fairly stable over time. In addition, standard deviations are generally smaller than the means, suggest- ing some degree of homogeneity within the sample. 13 It should be noted at this point that since the American Collegiate Test ACT changed its format in 1989, the scores for 1989 are not automatically comparable with those for 1990– 1994. Thus, the earlier scores for each individual school have been adjusted using a concordance table distributed by ACT; the new scores approximate those consistent with the newer “enhanced” ACT test given in recent years. A copy of the con- cordance tables appears in Appendix B .

5. Chicago high school malmquist productivity index

In this section we explicitly account for changes in the performance of Chicago high schools over time by estimating productivity change using the Malmquist pro- ductivity index described earlier. In calculating this index for Chicago high schools, only the sixty observations that appear in all six fiscal years are included. Although we estimate indexes for each high school for each pair of years, we summarize the results in Table 3. Included are the geometric means of the indirect Malmquist productivity index and all of its components, the efficiency change and the technology change indices for each pair of fiscal years. The last col- umn, which is outlined, shows the overall changes between fiscal year 1989 and 1994, averaged over the 60 schools. Recall that values equal to one represent no change for the pairs of years; improvement is signified by values greater than one. In glancing over the numbers in the last column, since the technical innovation change index for all Chicago public high schools is 0.977, the pro- 9 S. Grosskopf, C. Moutray Economics of Education Review 20 2001 1–14 Table 2 Means for Chicago high school inputs and outputs, 1989–1994 Standard deviations in parentheses a 1989 1990 1991 1992 1993 1994 Inputs: Number of teachers 110.443 107.642 107.492 106.802 105.962 94.780 38.212 35.839 36.892 35.838 37.203 31.463 Avg. teacher salary 33044.754 34904.138 35018.149 37787.151 41887.728 41036.428 899.388 933.360 1019.953 1240.670 1195.129 1124.551 Number of administrators 4.063 4.093 3.828 2.813 3.358 3.423 1.517 0.948 1.048 0.869 1.309 1.336 Avg. admin. salary 43070.100 44128.171 45377.372 52057.411 53480.715 52834.667 2526.711 3445.006 3368.020 4326.771 4115.021 4330.407 XTEST English ACT 26439.294 25837.269 24845.766 24681.069 24450.248 23651.386 12318.535 12349.963 12366.053 12366.053 12670.123 12359.530 XTEST Math ACT 27788.364 26621.969 26087.068 26518.869 26170.135 25591.733 12015.385 12045.309 12292.520 12394.690 12820.066 12607.778 Outputs: ATT 1405.999 1335.103 1302.979 1315.845 1283.241 1251.470 578.836 565.805 579.766 579.182 593.876 583.523 HSG 823.128 791.992 740.131 818.316 805.533 819.209 559.918 483.784 532.303 562.745 517.026 547.751 YTEST English ACT 2242.069 2266.313 2202.348 2166.948 2208.623 2130.208 2046.988 1799.375 1775.004 1749.573 1737.715 1848.956 YTEST Math ACT 2352.207 2336.398 2327.326 2334.563 2371.899 2304.535 2064.253 1819.594 1887.275 1866.835 1864.369 1955.770 a Sources: Compiled by the authors using data taken from the Illinois State Board of Education, Chicago Panel on Public School Policy and Finance, and ACT 1989. Table 3 Geometric means of Chicago high school Malmquist productivity index results a Fiscal years Fiscal years Fiscal years Fiscal years Fiscal Years Fiscal years 1989 and 1990 1990 and 1991 1991 and 1992 1992 and 1993 1993 and 1994 1989 and 1994 Malmquist productivity 0.973 1.005 1.006 0.987 1.013 0.982 index n = 60 n = 60 n = 60 n = 60 n = 60 n = 60 Malmquist ,1 0.941 0.971 0.946 0.985 0.952 0.931 n = 37 n = 37 n = 26 n = 39 n = 23 n = 35 Malmquist .1 1.027 1.053 1.054 1.042 1.053 1.058 n = 23 n = 26 n = 34 n = 21 n = 37 n = 25 Efficiency change index 0.995 1.011 1.006 0.994 0.999 1.006 n = 60 n = 60 n = 60 n = 60 n = 60 n = 60 Efficiency change ,1 0.953 0.970 0.956 0.960 0.955 0.941 n = 26 n = 21 n = 24 n = 29 n = 27 n = 24 Efficiency change .1 1.028 1.034 1.040 1.028 1.037 1.051 n = 34 n = 39 n = 36 n = 31 n = 33 n = 36 Technical change index 0.978 0.995 1.000 0.992 1.013 0.977 n = 60 n = 60 n = 60 n = 60 n = 60 n = 60 Technical change ,1 0.967 0.980 0.971 0.983 0.991 0.964 n = 46 n = 38 n = 23 n = 44 n = 26 n = 48 Technical change .1 1.015 1.020 1.018 1.018 1.031 1.029 n = 14 n = 22 n = 37 n = 16 n = 34 n = 12 a Sources: Compiled by the authors using data taken from the Illinois State Board of Education, Chicago Panel on Public School Policy and Finance, and ACT 1989. 10 S. Grosskopf, C. Moutray Economics of Education Review 20 2001 1–14 Table 4 OLS results of time trend equations for the Malmquist productivity index, 1989–1994 Standard errors in parenthesis a Dependent variable Intercept Time change in change in change in Adjusted R 2 F-value Durbin- adm.teacher personnel the of Watson ratio expenditures white per student Students Malmquist 1.005 b 0.002 0.00002 20.001 b 0.00002 0.0721 6.812 1.933 productivity index 0.004 0.003 0.0001 0.0003 0.00007 Efficiency change 1.005 b 20.002 0.00003 20.0005 d 0.00001 index 0.003 0.002 0.00001 0.0003 0.00006 0.0001 1.008 2.012 Technical change 1.002 c 0.004 c 20.00001 20.001 c 0.000006 0.1553 14.745 1.921 index 0.002 0.002 0.00006 0.0002 0.00004 a Sources: Compiled by the authors using data taken from the Illinois State Board of Education, Chicago Panel on Public School Policy and Finance, and ACT 1989. b Significant at the 1 percent level of significance. c Significant at the 5 percent level of significance. d Significant at the 10 percent level of significance. duction possibilities frontier for these schools has shifted inward slightly on average between 1989 and 1994. There is, however, a slight increase in efficiency on aver- age at the high schools. The overall average change in productivity, however, is negative. Looking at the more disaggregated results, we see that there is a nearly even split between schools that improved in terms of productivity and those that declined. In addition, the average productivity vacillates around one on average over this time period. The same pattern occurs for the two components of productivity change. Note, however, that over half of the schools con- sistently achieve improvements in the efficiency change; i.e., they are catching up to the frontier over time. To determine if there are any significant changes, we include a regression model in which our productivity change, technical change, and efficiency change meas- ures are dependent variables. Independent variables include a time trend, the change in the ratio of adminis- trators to teachers, the change in per pupil expenditures, and the change in the percent of students who are white. The results are summarized in Table 4. Although we find no significant time trend in the pro- ductivity index or the efficiency change index, there is a small but significant improvement in the technical change index. The only other significant variable is the per pupil expenditure variable, which is consistently small but negative in all three regressions, i.e., schools with greater per pupil expenditure are associated with lower productivity. Finally, if we interpret the intercept term as the average value of our measures after con- trolling for our independent variables, we find that we can reject the hypothesis that the efficiency change inter- cept is equal to one at the ten-percent level of signifi- cance, based on a simple t-test. 14 This suggests that there has been a measured improvement in the efficiency of Chicago high schools on average over this time period. On the other hand, we cannot reject this hypothesis for the other index measures.

6. Concluding remarks