TABLE 6-6 Trend Model with Common Time Pattern, 2000 Data

 

Years

Controlsa

Violent Crime

Murder

Rape

0. Committee replication SE

1992b

Yes

–2.15

(0.39)**

–3.41

(0.62)**

–3.37

(0.48)**

1. Comm estimate w/ covariates SE

2000

Yes

–0.95

(0.18)**

–2.03

(0.26)**

–2.81

(0.20)**

2. Comm estimate w/o covariates SE

1992b

No

–1.41

(0.47)**

–1.52

(0.97)

–3.45

(0.82)**

3. Comm estimate w/o covariates SE

2000

No

–0.62

(0.17)**

0.12

(0.32)

–2.17

(0.30)**

aThe regressions use the covariates and specification from the original Lott and Mustard (1997) models that do not control for state poverty, unemployment, death penalty execution rates, or regional time trends. The controls include the arrest rate for the crime category in question (AOVIOICP), population density in the county, real per capita income variables (RPCPI RPCUI RPCIM RPCRPO), county population (POPC), and variables for the percentage of the population that is in each of many race x age x gender categories (e.g., PBM1019 is the percentage of the population that is black, male, and between ages 10 and 19). The “no controls” specification includes county fixed effects, year dummies, and th dummy for whether the state has a right-to-carry law.

earlier sample periods almost completely disappear with the extension of the sample to 2000. The committee views the failure of the original dummy variable model to generate robust predictions outside the original sample as important evidence of fragility of the model’s findings.12

These results are also substantially different from those found when using the expanded set of control variables first adopted by Lott (2000: Table 9.1). As described above, Ayres and Donohue (2003b) estimate a dummy variable model using the revised new data (see Table 6-3). As in Lott (2000, Table 9.1) and Plassmann and Whitley (2003), they modify the original specification to include additional covariates (i.e., state poverty, unemployment, and death penalty execution rates) and region-interacted time patterns, as opposed to a common time trend used in the original Lott models (Lott 2000:170). These seemingly minor adjustments cause sub-

12  

In light of the variability in the estimates, statistical tests might aid in determining whether particular specifications can be rejected by the data. It is not possible to test empirically whether a proposed set of explanatory variables is the correct one. It is possible to test for specification, given a set of controls (see Horowitz, Appendix D). None of the models examined by the committee passes a simple specification test (i.e., Ramsey’s 1969 RESET test).



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement