September 28, 2001
This README file accompanies the STATA file "jcr01.dta," which is meant to help replicate the results in Michael Horowitz and Dan Reiter, "When Does Aerial Bombing Work? Quantitative Empirical Tests, 1917-1999," Journal of Conflict Resolution 45 (April 2001): 147-173.
The variables and methods are described in detail in the article itself. The variable labels used in the data set should be self-explanatory. The two probit runs below replicate Models 1 and 3 in Table 1 on page 159 of the article. Note that for the analysis which produced the results for Model 3, different variables are used for the outcome and denial variables, reflecting the suggested corrections of Barry Watt. Note also that the "warid" variable refers to the war in which the coercion campaign is attempted, and we cluster on "warid" in order to control for possible within-war autocorrelation. Finally, the data set includes a few variables not used in the analysis which describe the cases, "date," "attacker," and "defender."
. clear
. use "E:\Program Files\Macromedia\Dreamweaver 3\webpage\jcr01.dta", clear
. *This is a replication of column 1, Table 1.
. *This is executed in STATA 6.0.
. probit outcome punish denial demand democ industry othrfrce sanction nucle
> ar, robust cluster(warid)
Iteration 0: log likelihood = -34.584873
Iteration 1: log likelihood = -15.105229
Iteration 2: log likelihood = -11.309473
Iteration 3: log likelihood = -9.9511302
Iteration 4: log likelihood = -9.5105944
Iteration 5: log likelihood = -9.4276605
Iteration 6: log likelihood = -9.4230465
Iteration 7: log likelihood = -9.4230272
Probit estimates Number of obs = 53
Wald chi2(8) = 39.52
Prob > chi2 = 0.0000
Log likelihood = -9.4230272 Pseudo R2 = 0.7275
(standard errors adjusted for clustering on warid)
------------------------------------------------------------------------------
| Robust
outcome | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------+--------------------------------------------------------------------
punish | .1415185 .3321195 0.426 0.670 -.5094237 .7924607
denial | 2.686 .4904869 5.476 0.000 1.724663 3.647337
demand | -2.440988 .8808502 -2.771 0.006 -4.167423 -.7145534
democ | -.3416109 1.339955 -0.255 0.799 -2.967875 2.284653
industry | -.0616905 .1184552 -0.521 0.603 -.2938586 .1704775
othrfrce | 1.51443 .9876601 1.533 0.125 -.4213486 3.450208
sanction | -3.304687 .8386596 -3.940 0.000 -4.94843 -1.660945
nuclear | 3.379216 1.147654 2.944 0.003 1.129855 5.628577
_cons | -9.667408 2.12726 -4.545 0.000 -13.83676 -5.498055
------------------------------------------------------------------------------
Note: 4 failures and 0 successes completely determined.
. *This is a replication of Column 3, Table 1.
. probit wattout punish wattdenv demand democ industry othrfrce sanction nu
> clear, robust cluster(warid)
Iteration 0: log likelihood = -33.254163
Iteration 1: log likelihood = -21.15556
Iteration 2: log likelihood = -19.564205
Iteration 3: log likelihood = -19.310605
Iteration 4: log likelihood = -19.296913
Iteration 5: log likelihood = -19.296856
Iteration 6: log likelihood = -19.296856
Probit estimates Number of obs = 53
Wald chi2(8) = 40.76
Prob > chi2 = 0.0000
Log likelihood = -19.296856 Pseudo R2 = 0.4197
(standard errors adjusted for clustering on warid)
------------------------------------------------------------------------------
| Robust
wattout | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------+--------------------------------------------------------------------
punish | .0886617 .2462197 0.360 0.719 -.3939201 .5712435
wattdenv | 1.197892 .2138805 5.601 0.000 .7786936 1.61709
demand | -.0462876 .5921437 -0.078 0.938 -1.206868 1.114293
democ | .6556095 .6879186 0.953 0.341 -.6926862 2.003905
industry | .0457161 .0784257 0.583 0.560 -.1079954 .1994276
othrfrce | .5744768 .6075075 0.946 0.344 -.616216 1.765169
sanction | -1.267941 .5812003 -2.182 0.029 -2.407073 -.1288094
nuclear | 1.911232 .7098693 2.692 0.007 .519914 3.30255
_cons | -5.837682 1.172169 -4.980 0.000 -8.135091 -3.540272
------------------------------------------------------------------------------
. clear
.