COMMITTEE’S ANALYSIS: ARE THE ESTIMATES ROBUST?

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 

This section presents the results of the committee’s own analysis of

Lott’s revised new data covering the period 1977-2000. The purpose of the

analysis is to clarify and illustrate some of the causes of the conflicting

results. The committee has not attempted to form our own estimates of the

effects of right-to-carry laws. Rather, our analysis is directed toward gaining

a better understanding of the fragility of the estimates. We begin by

illustrating the sensitivity of the findings to extending the sample period to

cover the years 1993-2000. We then demonstrate that the basic qualitative

results are sensitive to variations in the explanatory variables. In all cases,

we use the revised new data set. There is a consensus that these revised data,

covering the periods 1977-2000, are correct.

Horowitz discusses this problem in further detail and provides a statistical

explanation for the fragility in the estimates in Appendix D. This

appendix describes two fundamental sources of difficulty in causal inference

that are especially relevant to studies of right-to-carry laws. One is the

difficulty of choosing the right explanatory variables for a statistical model.

The second is the difficulty of estimating the relation among crime rates, a

large number of potential explanatory variables, and the adoption of rightto-

carry laws. Even if the correct explanatory variables were known, it

would be hard to specify a model correctly, especially in high dimensional

settings with many explanatory variables. The committee drew on some of

these ideas in our deliberations but did not adopt them in total as part of

our consensus report. This statistical argument is presented to stimulate

further discussion and dialogue on these issues.