--------------------------------------------------------------------------------------------- log: D:\__PAPERS-CURRENT\prosecutors\regressions\oct22best.log log type: text opened on: 23 Oct 2008, 17:00:16 . * I started with the final.dta dataset. I put lots of variable transformations, dropping, > and renaming, into the data2.do file and apply it to final.dta to get final2.dta. Here, > we use final3, which is similarly created, but by Manu; . *------------------------------------------------------------------------; . * In constructing final2.dta, I already dropped the out-of-range winrates, but here's the > command anyway. ; . drop if winrate>100; (0 observations deleted) . drop if winrate<0; (0 observations deleted) . * We drop the top 5% of districts by population. This was not datamining-- they are just > too different for it to make sense to include them. If you do, everything is insignficant. > ; . generate inc = percapita*pop; (5 missing values generated) . _pctile pop, percentile(1(1)99); . scalar p95=r(r95); . scalar p5= r(r5); . scalar p10=r(r10); . scalar p98=r(r98); . drop if pop>p95; (94 observations deleted) . *Here we should compute summary statistics. ; . tabstat winrate felclosed budget cpsalary appointed inoffice term indexcrime m > etro pop pbush > blackper males1524 permale1524 peroccupied peremployed noschoolmale inc > localpay , stats(min p25 median mean p75 > max) f(%7.2f) columns(statistics) ; variable | min p25 p50 mean p75 max -------------+------------------------------------------------------------ winrate | 0.00 78.74 90.00 83.22 96.00 100.00 felclosed | 1.00 75.00 218.00 599.10 573.00 15639.00 budget | 0.01 0.13 0.30 0.78 0.75 15.53 cpsalary | 10.50 52.00 80.28 77.07 99.00 150.00 appointed | 0.00 0.00 0.00 0.01 0.00 1.00 inoffice | 0.00 4.00 7.00 9.14 14.00 40.00 term | 1.00 4.00 4.00 4.11 4.00 10.00 indexcrime | 0.00 0.03 0.13 0.46 0.46 8.86 metro | 0.00 0.00 0.00 0.20 0.00 1.00 pop | 0.00 0.01 0.03 0.07 0.08 0.54 pbush | 19.21 50.93 58.12 58.58 65.72 90.87 blackper | 0.00 0.00 0.00 0.01 0.01 0.08 males1524 | 24.00 839.00 2386.00 4969.50 5921.00 46233.00 permale1524 | 0.04 0.06 0.07 0.07 0.08 0.21 peroccupied | 0.23 0.83 0.89 0.86 0.93 0.98 peremployed | 0.24 0.42 0.46 0.45 0.49 0.64 noschoolmale | 0.00 24.00 82.00 235.46 260.00 7385.00 inc | 0.00 0.00 0.00 0.00 0.00 0.00 localpay | 5.81 1014.66 2574.78 5969.65 6797.96 73884.03 -------------------------------------------------------------------------- . *------------------------------------------------------------------------; . *This cosntructs logs for the instrumental variables.; . replace localpay=log(localpay); (1797 real changes made) . replace inc =log(inc); (1800 real changes made) . replace males1524 =log(males1524); males1524 was long now double (1805 real changes made) . replace permale1524 =log(permale1524); (1805 real changes made) . replace peroccupied =log(peroccupied); (1805 real changes made) . replace peremployed =log(peremployed); (1805 real changes made) . replace noschoolmale =log(noschoolmale); noschoolmale was long now double (1805 real changes made, 66 to missing) . *------------------------------------------------------------------------; . *Here is OLS for the prosecution regression, as a robustness check.; . reg felclosed budget pop indexcrime cpsalary appointed blackper pbush inoffice > term metro > blackper midwest northeast south , robust; Linear regression Number of obs = 1457 F( 13, 1443) = 37.48 Prob > F = 0.0000 R-squared = 0.5601 Root MSE = 735.3 ------------------------------------------------------------------------------ | Robust felclosed | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- budget | 179.1518 70.17447 2.55 0.011 41.49689 316.8067 pop | 5588.569 1391.409 4.02 0.000 2859.168 8317.97 indexcrime | 69.47076 150.8975 0.46 0.645 -226.5311 365.4727 cpsalary | -.7862725 1.038924 -0.76 0.449 -2.824236 1.251691 appointed | -570.4642 456.7385 -1.25 0.212 -1466.407 325.4782 blackper | 18753.24 3464.766 5.41 0.000 11956.72 25549.76 pbush | 2.649936 2.241053 1.18 0.237 -1.746134 7.046006 inoffice | -4.286086 2.462682 -1.74 0.082 -9.116906 .5447347 term | 27.12519 32.80238 0.83 0.408 -37.22026 91.47063 metro | -68.38562 77.0421 -0.89 0.375 -219.5121 82.74088 blackper | (dropped) midwest | -92.30579 33.83815 -2.73 0.006 -158.683 -25.92857 northeast | -292.473 113.6594 -2.57 0.010 -515.4283 -69.51767 south | 43.68257 65.12092 0.67 0.502 -84.05925 171.4244 _cons | -192.919 183.5832 -1.05 0.294 -553.0376 167.1996 ------------------------------------------------------------------------------ . mfx, eyex nose; Elasticities after regress y = Fitted values (predict) = 575.29513 ------------------------------------------------------------------------------- variable | ey/ex X ---------------------------------+--------------------------------------------- budget | .2502102 .803479 pop | .656868 .067619 indexcrime | .0533704 .441966 cpsalary | -.1052401 77.0014 appointed | -.0095281 .009609 blackper | .1937945 .005945 pbush | .2697717 58.5668 inoffice | -.0674714 9.05628 term | .1943928 4.12286 metro | -.0243942 .205216 blackper | 0 .005945 midwest | -.060788 .378861 northeast | -.0352417 .069321 south | .019595 .258065 ------------------------------------------------------------------------------- . *------------------------------------------------------------------------; . *Here is IV instrumenting for both crime and the proseuction budget; . ivreg felclosed pop cpsalary appointed inoffice term metro blackper pbush mid > west northeast south (indexcrime budget = males1524 permale1524 peroccupied peremplo > yed noschoolmale inc > localpay ) , robust; Instrumental variables (2SLS) regression Number of obs = 1403 F( 13, 1389) = 37.70 Prob > F = 0.0000 R-squared = 0.5071 Root MSE = 787.4 ------------------------------------------------------------------------------ | Robust felclosed | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- indexcrime | 548.7475 716.0958 0.77 0.444 -855.9986 1953.494 budget | -82.36361 234.7954 -0.35 0.726 -542.9554 378.2282 pop | 4996.324 3749.715 1.33 0.183 -2359.393 12352.04 cpsalary | .5704517 2.132498 0.27 0.789 -3.612814 4.753717 appointed | 439.2121 1121.141 0.39 0.695 -1760.1 2638.525 inoffice | -4.577998 2.826157 -1.62 0.105 -10.122 .9659992 term | 3.675465 44.48482 0.08 0.934 -83.58922 90.94015 metro | -78.21893 85.30205 -0.92 0.359 -245.5537 89.11583 blackper | 14199.56 5875.651 2.42 0.016 2673.456 25725.67 pbush | .3413681 3.9006 0.09 0.930 -7.310336 7.993072 midwest | -97.58763 53.9934 -1.81 0.071 -203.505 8.329784 northeast | -363.7927 140.5676 -2.59 0.010 -639.5404 -88.04494 south | .0909869 75.38274 0.00 0.999 -147.7853 147.9673 _cons | 10.89953 318.858 0.03 0.973 -614.5957 636.3947 ------------------------------------------------------------------------------ Instrumented: indexcrime budget Instruments: pop cpsalary appointed inoffice term metro blackper pbush midwest northeast south males1524 permale1524 peroccupied peremployed noschoolmale inc localpay ------------------------------------------------------------------------------ . *------------------------------------------------------------------------; . *These two regressions check the R2 for the instruments. The R2 are high enough, even tho > ugh I am dubious of istrumenting for crime; . regress indexcrime males1524 permale1524 peroccupied peremployed noschoolmale; Source | SS df MS Number of obs = 1739 -------------+------------------------------ F( 5, 1733) = 269.83 Model | 575.718543 5 115.143709 Prob > F = 0.0000 Residual | 739.506206 1733 .426720257 R-squared = 0.4377 -------------+------------------------------ Adj R-squared = 0.4361 Total | 1315.22475 1738 .756746116 Root MSE = .65324 ------------------------------------------------------------------------------ indexcrime | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- males1524 | .4137589 .0284485 14.54 0.000 .357962 .4695558 permale1524 | -.6280646 .0873808 -7.19 0.000 -.7994474 -.4566817 peroccupied | -.2982143 .1366841 -2.18 0.029 -.5662975 -.0301311 peremployed | .1403284 .1531262 0.92 0.360 -.1600032 .44066 noschoolmale | .0545734 .0219515 2.49 0.013 .0115192 .0976276 _cons | -4.59407 .3809961 -12.06 0.000 -5.34133 -3.846809 ------------------------------------------------------------------------------ . regress budget inc localpay ; Source | SS df MS Number of obs = 1778 -------------+------------------------------ F( 2, 1775) = 443.81 Model | 1243.47675 2 621.738373 Prob > F = 0.0000 Residual | 2486.59405 1775 1.40089806 R-squared = 0.3334 -------------+------------------------------ Adj R-squared = 0.3326 Total | 3730.0708 1777 2.09908317 Root MSE = 1.1836 ------------------------------------------------------------------------------ budget | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- inc | .3838782 .0383271 10.02 0.000 .3087072 .4590491 localpay | .229344 .0390116 5.88 0.000 .1528306 .3058574 _cons | 2.621691 .6462777 4.06 0.000 1.354146 3.889237 ------------------------------------------------------------------------------ . * We need to do a Hausman or Wald test for whether Crime is exogenous. ; . *------------------------------------------------------------------------; . *Here is IV instrumenting for just the proseuction budget. This is more like OLS and is > a good regression. ; . ivreg felclosed pop indexcrime cpsalary appointed inoffice term metro blackper > pbush midwest northeast south ( budget = inc localpay ) , robust; Instrumental variables (2SLS) regression Number of obs = 1448 F( 13, 1434) = 45.72 Prob > F = 0.0000 R-squared = 0.5359 Root MSE = 754.92 ------------------------------------------------------------------------------ | Robust felclosed | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- budget | 390.3319 229.2816 1.70 0.089 -59.43133 840.0951 pop | 3717.13 1719.714 2.16 0.031 343.7055 7090.555 indexcrime | -41.63276 180.3244 -0.23 0.817 -395.3606 312.0951 cpsalary | -1.077062 1.120092 -0.96 0.336 -3.274258 1.120133 appointed | -1263.962 941.9084 -1.34 0.180 -3111.628 583.704 inoffice | -4.883573 2.809926 -1.74 0.082 -10.39558 .628433 term | 39.61192 31.99957 1.24 0.216 -23.15907 102.3829 metro | -35.65155 73.31217 -0.49 0.627 -179.4621 108.1591 blackper | 20736.95 4224.665 4.91 0.000 12449.77 29024.14 pbush | 4.178463 3.272659 1.28 0.202 -2.241249 10.59818 midwest | -66.77325 46.70203 -1.43 0.153 -158.3849 24.83837 northeast | -225.5369 128.9566 -1.75 0.081 -478.5007 27.42695 south | 85.04183 89.92273 0.95 0.344 -91.35237 261.436 _cons | -336.7172 252.3399 -1.33 0.182 -831.712 158.2777 ------------------------------------------------------------------------------ Instrumented: budget Instruments: pop indexcrime cpsalary appointed inoffice term metro blackper pbush midwest northeast south inc localpay ------------------------------------------------------------------------------ . *------------------------------------------------------------------------; . * Here is a test of the last regression to see if the size of the district is correlated ii > th the errors. It is not, by construction, since POP was a RHS variable. ; . predict v1, resid; (348 missing values generated) . regress v1 pop; Source | SS df MS Number of obs = 1457 -------------+------------------------------ F( 1, 1455) = 0.00 Model | 34.4015343 1 34.4015343 Prob > F = 0.9938 Residual | 818231506 1455 562358.424 R-squared = 0.0000 -------------+------------------------------ Adj R-squared = -0.0007 Total | 818231541 1456 561972.212 Root MSE = 749.91 ------------------------------------------------------------------------------ v1 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- pop | 1.692752 216.4268 0.01 0.994 -422.8491 426.2346 _cons | -.379975 24.49775 -0.02 0.988 -48.43467 47.67472 ------------------------------------------------------------------------------ . *------------------------------------------------------------------------; . *This is a tobit for the winrate regression, as a robusntess check. It comes out with budg > et insignfiicant. ; . tobit winrate indexcrime budget felclosed cpsalary appointed inoffice term met > ro > blackper midwest northeast south, ul(100) ; Tobit regression Number of obs = 1561 LR chi2(12) = 88.96 Prob > chi2 = 0.0000 Log likelihood = -6238.4021 Pseudo R2 = 0.0071 ------------------------------------------------------------------------------ winrate | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- indexcrime | .1331057 1.023293 0.13 0.897 -1.87408 2.140292 budget | .1253574 .6048518 0.21 0.836 -1.061057 1.311772 felclosed | -.0027224 .0006881 -3.96 0.000 -.0040722 -.0013727 cpsalary | -.0564526 .023207 -2.43 0.015 -.1019731 -.0109321 appointed | -4.562003 6.410653 -0.71 0.477 -17.13648 8.012472 inoffice | .1488772 .0759892 1.96 0.050 -.0001754 .2979298 term | 1.055121 .6617764 1.59 0.111 -.2429512 2.353193 metro | -1.26499 1.452172 -0.87 0.384 -4.113421 1.58344 blackper | -74.88437 56.43543 -1.33 0.185 -185.5823 35.81353 midwest | 1.721679 1.356208 1.27 0.204 -.9385186 4.381876 northeast | .0926906 2.43226 0.04 0.970 -4.678179 4.863561 south | .3356204 1.530712 0.22 0.826 -2.666866 3.338106 _cons | 85.36668 3.308042 25.81 0.000 78.87796 91.85539 -------------+---------------------------------------------------------------- /sigma | 20.40859 .3985044 19.62692 21.19025 ------------------------------------------------------------------------------ Obs. summary: 0 left-censored observations 1367 uncensored observations 194 right-censored observations at winrate>=100 . *------------------------------------------------------------------------; . * This is the winrate regression with both crime and budget instrumented. The Wald test ca > nnot reject exogeneity, thoguh.; . ivtobit winrate felclosed cpsalary appointed inoffice term metro > blackper midwest northeast south (indexcrime budget = males1524 permale1524 pe > roccupied peremployed noschoolmale inc > localpay ), ul(100) twostep; Two-step tobit with endogenous regressors Number of obs = 1497 Wald chi2(12) = 66.47 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ winrate | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- indexcrime | 7.393918 7.284232 1.02 0.310 -6.882914 21.67075 budget | -7.071619 6.008579 -1.18 0.239 -18.84822 4.704979 felclosed | -.0007111 .0015712 -0.45 0.651 -.0037907 .0023684 cpsalary | .0132132 .0442312 0.30 0.765 -.0734783 .0999046 appointed | 21.39403 23.82832 0.90 0.369 -25.30861 68.09667 inoffice | .181491 .0869223 2.09 0.037 .0111264 .3518556 term | .7949113 .7281685 1.09 0.275 -.6322727 2.222095 metro | -1.456109 1.665572 -0.87 0.382 -4.720571 1.808353 blackper | -211.3983 122.3795 -1.73 0.084 -451.2577 28.46111 midwest | 1.463497 1.434251 1.02 0.308 -1.347582 4.274577 northeast | .2256489 2.598587 0.09 0.931 -4.867489 5.318787 south | -2.222178 2.474124 -0.90 0.369 -7.071372 2.627016 _cons | 82.9467 3.486001 23.79 0.000 76.11427 89.77914 ------------------------------------------------------------------------------ Instrumented: indexcrime budget Instruments: felclosed cpsalary appointed inoffice term metro blackper midwest northeast south males1524 permale1524 peroccupied peremployed noschoolmale inc localpay ------------------------------------------------------------------------------ Wald test of exogeneity: chi2(2) = 0.49 Prob > chi2 = 0.7816 Obs. summary: 1335 uncensored observations 162 right-censored observations at winrate>=100 . *------------------------------------------------------------------------; . *This is the winrate regression with just budget endogenous. ; . ivtobit winrate indexcrime felclosed cpsalary appointed inoffice term > > metro blackper midwest northeast south (budget = inc > localpay ), ul(100) robust ; Fitting exogenous tobit model Fitting full model Iteration 0: log pseudolikelihood = -7899.3712 (not concave) Iteration 1: log pseudolikelihood = -7868.8014 (not concave) Iteration 2: log pseudolikelihood = -7863.6382 (not concave) Iteration 3: log pseudolikelihood = -7859.5217 (not concave) Iteration 4: log pseudolikelihood = -7856.6793 (not concave) Iteration 5: log pseudolikelihood = -7854.1242 (not concave) Iteration 6: log pseudolikelihood = -7849.1082 (not concave) Iteration 7: log pseudolikelihood = -7848.6073 (not concave) Iteration 8: log pseudolikelihood = -7848.3173 (not concave) Iteration 9: log pseudolikelihood = -7847.9493 (not concave) Iteration 10: log pseudolikelihood = -7847.8307 Iteration 11: log pseudolikelihood = -7842.6265 (backed up) Iteration 12: log pseudolikelihood = -7839.9534 Iteration 13: log pseudolikelihood = -7839.7112 Iteration 14: log pseudolikelihood = -7839.7031 Iteration 15: log pseudolikelihood = -7839.7031 Tobit model with endogenous regressors Number of obs = 1551 Wald chi2(12) = 11.94 Log pseudolikelihood = -7839.7031 Prob > chi2 = 0.4504 ------------------------------------------------------------------------------ | Robust | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- winrate | budget | 27.3293 11.82627 2.31 0.021 4.150239 50.50836 indexcrime | -23.54985 11.13703 -2.11 0.034 -45.37803 -1.721677 felclosed | -.0105557 .0049169 -2.15 0.032 -.0201926 -.0009188 cpsalary | -.1033284 .0816696 -1.27 0.206 -.2633979 .0567412 appointed | -114.533 49.95868 -2.29 0.022 -212.4502 -16.61579 inoffice | -.1103782 .1205994 -0.92 0.360 -.3467487 .1259922 term | 1.702548 1.030251 1.65 0.098 -.3167056 3.721802 metro | -2.759829 2.412 -1.14 0.253 -7.487261 1.967603 blackper | 349.5396 230.5106 1.52 0.129 -102.2529 801.3321 midwest | 1.717489 1.854315 0.93 0.354 -1.916902 5.35188 northeast | -.3038923 3.853855 -0.08 0.937 -7.85731 7.249525 south | 8.101869 4.197668 1.93 0.054 -.1254102 16.32915 _cons | 75.59456 4.424667 17.08 0.000 66.92237 84.26675 -------------+---------------------------------------------------------------- /alpha | -27.25121 11.77424 -2.31 0.021 -50.3283 -4.174124 /lns | 2.920836 .0345573 84.52 0.000 2.853105 2.988567 /lnv | -.1612927 .0637887 -2.53 0.011 -.2863163 -.0362691 -------------+---------------------------------------------------------------- s | 18.55679 .6412719 17.34154 19.8572 v | .8510429 .0542869 .751025 .9643807 ------------------------------------------------------------------------------ Instrumented: budget Instruments: indexcrime felclosed cpsalary appointed inoffice term metro blackper midwest northeast south inc localpay ------------------------------------------------------------------------------ Wald test of exogeneity (/alpha = 0): chi2(1) = 5.36 Prob > chi2 = 0.0206 Obs. summary: 1357 uncensored observations 194 right-censored observations at winrate>=100 . mfx, eyex nose at(median); Elasticities after ivtobit y = Fitted values (predict) = 76.381495 ------------------------------------------------------------------------------- variable | ey/ex X ---------------------------------+--------------------------------------------- budget | .10734 .3 indexcrime | -.0422397 .137 felclosed | -.0312325 226 cpsalary | -.1086672 80.328 appointed | 0 0 inoffice | -.0101156 7 term | .0891603 4 metro | 0 0 blackper | .0060574 .001324 midwest | 0 0 northeast | 0 0 south | 0 0 inc | 0 -9.39752 localpay | 0 7.86788 ------------------------------------------------------------------------------- . *------------------------------------------------------------------------; . *Here is a test for whether the residuals are correlated with population They are not.; . predict v2, ystar(0,100); (244 missing values generated) . gen v3 = v2 - winrate; (244 missing values generated) . regress v3 pop; Source | SS df MS Number of obs = 1561 -------------+------------------------------ F( 1, 1559) = 2.07 Model | 994.049391 1 994.049391 Prob > F = 0.1507 Residual | 749596.04 1559 480.818499 R-squared = 0.0013 -------------+------------------------------ Adj R-squared = 0.0007 Total | 750590.09 1560 481.147493 Root MSE = 21.928 ------------------------------------------------------------------------------ v3 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- pop | 8.771184 6.100211 1.44 0.151 -3.194298 20.73667 _cons | -9.500423 .6976848 -13.62 0.000 -10.86892 -8.131924 ------------------------------------------------------------------------------ . *------------------------------------------------------------------------; . * This is the winrate regression with just crime instrumented. The Wald test cannot rej > ect exogeneity ; . ivtobit winrate budget felclosed cpsalary appointed inoffice term metro > blackper midwest northeast south (indexcrime = males1524 permale1524 peroccup > ied peremployed noschoolmale ), ul(100) twostep; Two-step tobit with endogenous regressors Number of obs = 1507 Wald chi2(12) = 77.01 Prob > chi2 = 0.0000 ------------------------------------------------------------------------------ winrate | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- indexcrime | .9939758 3.254423 0.31 0.760 -5.384576 7.372528 budget | -.1660165 1.113682 -0.15 0.881 -2.348793 2.01676 felclosed | -.0029311 .0009633 -3.04 0.002 -.0048192 -.0010431 cpsalary | -.0332942 .0235661 -1.41 0.158 -.0794829 .0128945 appointed | -4.343843 7.171025 -0.61 0.545 -18.39879 9.711108 inoffice | .1354903 .0774288 1.75 0.080 -.0162674 .2872481 term | 1.060645 .6592226 1.61 0.108 -.2314073 2.352698 metro | -1.564989 1.601023 -0.98 0.328 -4.702937 1.572959 blackper | -84.52367 61.32395 -1.38 0.168 -204.7164 35.66905 midwest | 1.780046 1.365637 1.30 0.192 -.8965533 4.456646 northeast | -.0990681 2.466866 -0.04 0.968 -4.934037 4.7359 south | .2694343 1.545699 0.17 0.862 -2.76008 3.298948 _cons | 83.3849 3.328238 25.05 0.000 76.86168 89.90813 ------------------------------------------------------------------------------ Instrumented: indexcrime Instruments: budget felclosed cpsalary appointed inoffice term metro blackper midwest northeast south males1524 permale1524 peroccupied peremployed noschoolmale ------------------------------------------------------------------------------ Wald test of exogeneity: chi2(1) = 0.07 Prob > chi2 = 0.7872 Obs. summary: 1345 uncensored observations 162 right-censored observations at winrate>=100 . *------------------------------------------------------------------------; . *These correlations look for whether win rate actually is negatively correlated with budget > , as we point out as a theoretical possibility. It's a very small effect.; . correlate winrate budpop budcrime; (obs=1614) | winrate budpop budcrime -------------+--------------------------- winrate | 1.0000 budpop | -0.0285 1.0000 budcrime | 0.0142 0.8681 1.0000 . *------------------------------------------------------------------------; . *Here's another look at whether winrate and budget are inversely correlated when we don't c > ondition on prosecutions. Thsi omits felclosed, and so looks for whether one woudl see a > negative relation between winrate and budget; . ivtobit winrate indexcrime cpsalary appointed inoffice term > metro blackper midwest northeast south (budget = inc > localpay ), ul(100) robust ; Fitting exogenous tobit model Fitting full model Iteration 0: log pseudolikelihood = -7961.0904 (not concave) Iteration 1: log pseudolikelihood = -7931.46 (not concave) Iteration 2: log pseudolikelihood = -7924.7836 (not concave) Iteration 3: log pseudolikelihood = -7921.6253 (not concave) Iteration 4: log pseudolikelihood = -7915.857 (not concave) Iteration 5: log pseudolikelihood = -7913.6428 Iteration 6: log pseudolikelihood = -7907.8173 Iteration 7: log pseudolikelihood = -7903.2654 Iteration 8: log pseudolikelihood = -7902.9863 Iteration 9: log pseudolikelihood = -7902.9785 Iteration 10: log pseudolikelihood = -7902.9785 Tobit model with endogenous regressors Number of obs = 1551 Wald chi2(11) = 15.92 Log pseudolikelihood = -7902.9785 Prob > chi2 = 0.1442 ------------------------------------------------------------------------------ | Robust | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- winrate | budget | 19.47326 8.363513 2.33 0.020 3.081078 35.86545 indexcrime | -22.89105 9.978023 -2.29 0.022 -42.44762 -3.334485 cpsalary | -.0751083 .0689879 -1.09 0.276 -.210322 .0601054 appointed | -88.328 37.97569 -2.33 0.020 -162.759 -13.89701 inoffice | -.0269568 .0984014 -0.27 0.784 -.21982 .1659065 term | 1.363846 .8493496 1.61 0.108 -.3008486 3.028541 metro | -3.486866 2.187932 -1.59 0.111 -7.775134 .8014025 blackper | 98.57314 129.9579 0.76 0.448 -156.1397 353.286 midwest | 2.125826 1.634283 1.30 0.193 -1.07731 5.328961 northeast | 1.954608 3.490913 0.56 0.576 -4.887456 8.796672 south | 5.644639 3.21605 1.76 0.079 -.6587026 11.94798 _cons | 75.39424 4.01642 18.77 0.000 67.5222 83.26628 -------------+---------------------------------------------------------------- /alpha | -19.93306 8.344149 -2.39 0.017 -36.28729 -3.578826 /lns | 2.926072 .0342423 85.45 0.000 2.858958 2.993186 /lnv | -.1250786 .0633075 -1.98 0.048 -.249159 -.0009981 -------------+---------------------------------------------------------------- s | 18.65421 .6387637 17.44334 19.94913 v | .8824276 .0558643 .779456 .9990024 ------------------------------------------------------------------------------ Instrumented: budget Instruments: indexcrime cpsalary appointed inoffice term metro blackper midwest northeast south inc localpay ------------------------------------------------------------------------------ Wald test of exogeneity (/alpha = 0): chi2(1) = 5.71 Prob > chi2 = 0.0169 Obs. summary: 1357 uncensored observations 194 right-censored observations at winrate>=100 . *------------------------------------------------------------------------; . log close; log: D:\__PAPERS-CURRENT\prosecutors\regressions\oct22best.log log type: text closed on: 23 Oct 2008, 17:00:34 ---------------------------------------------------------------------------------------------