The target of Kovandzic's July, 1988 comment is the 1997 paper by Kwon, Scott, Safranski and Bae. In this paper the authors attempt to "...empirically evaluate the effectiveness of gun control laws that have been adopted by states and municipalities" (p. 41). Kovandzic claims that the research by Kwon et al. is flawed in several ways and he concludes that "...no statistically significant and negative relationship between gun control laws and firearm related deaths can be found in this data set" (p. 362).
Upon review of the criticisms by Kovandzic, it is clear that several of his points are well justified. The trouble with the work of Kwon et al., however, does not end here. There are serious econometric questions that also arise concerning the behavior of the stochastic error term and the specification of the model. I begin by addressing some of the points raised by Kovandzic and the reply issued by Kwon et al.
Kovandzic's Critique and the Repy by Kwon et al.
The original article by Kwon et al. begins with a section entitled "Research Background" wherein they write, "In spite of charged emotional debates and passage of numerous laws and regulations, no empirical studies have been done to evaluate the effectiveness of gun control laws in this country" (p. 41). This statement is simply irresponsible. The authors give no response when Kovandzic provides evidence of numerous studies conducted on this topic. Indeed, a simple search on the familiar EconLit CD-ROM database using the key words "gun control laws" gave 15 hits of which two were clearly related research: Sommers (1980) and Magaddino and Medoff (1982).
Gun laws and measurement error.
Although Kwon et al. purport in their abstract that their research evaluates the effectiveness gun laws adopted by states and municipalities (p. 41), their analysis and data set has no information regarding municipalities. Kovandzic criticizes the authors for relying on a December 20, 1993 Time magazine article as a sole source of data on state gun control laws. In fact, in the Time article Handgun Control, Inc. (the Washington, D.C. based lobby group supporting gun control laws), was cited as the source of data. In any case, Kovandzic is correct in noting that Kwon et al. does not provide a clear explanation of the laws being tested. Given the focus on the Brady Bill in their paper, the authors perhaps should have consulted the (1994) article by DeFrances and Smith (both of the Bureau of Justice Statistics, U.S. Department of Justice). DeFrances and Smith provide a breakdown of those states in 1994 that already had laws complying with the 5-day waiting period set out in the Brady Law (1994, 74: Table 1).
Tests of statistical significance
Kovandzic and Kwon et al. clash over the issue of statistical significance, especially regarding the dichotomous variable labeled by Kwon et al. as "Law" which is intended to capture the effect of state gun control laws (as defined by the authors) on the number of firearm-related deaths. Kovandzic complains that, "They (Kwon et al.) found no statistically significant relationship between gun laws and firearm-related deaths, but they continually refer to the findings as if they did" (1998, 365). Kwon et al. point out in their reply that they did indeed report that the estimated coefficient was not statistically significant. They then proceed to discuss the merits of using the test of significance approach as a means of judging importance. On this issue both parties have legitimate points.
What appears to be at the root of this point of contention is a lack of careful writing and interpretation of statistical results. Kwon et al. seem to contradict themselves as they write in their abstract that "The study results indicate that gun control laws have a very mild effect on the number of gun related deaths while socioeconomic variables such as a state's poverty level, unemployment rate and alcohol consumption have significant impact on firearm related deaths" (1997, 41). …