Ian Ayres & John Donohue

6/7/03

In a letter to the editor in today's Columbus Dispatch, John Donohue claims my research is "fatally flawed" and that:

Trying to revive his discredited thesis, Lott now cites the work of
others rather than his own research. But neither of the studies he
refers to, one of which was written by his own co-author, David
Mustard, actually supports Lott's view. Mustard's work follows the
same flawed methodology that Lott employs, and the author of the
other paper (as Lott well knows) has withdrawn the paper after
concluding that the data on which it (and Lott's work) was based is
unreliable.

Donohue is responding to a piece that I had in the Dispatch in April where I wrote:

One particular fear that some police have is that right-to-carry laws
would actually make their jobs more dangerous by making it more
likely that they would be shot. Yet, research has shown that the laws
make police safer. Professor David Mustard at the University of
Georgia found that right-to-carry laws reduced the rate that officers
were killed by about 2 percent per year for each additional year
that the laws were in effect.

Other research, by David Olson at Loyola University and Michael
Maltz at the University of Illinois, found that when law-abiding
citizens carried concealed handguns, criminals were much less
likely to carry guns. In fact, they found gun murders fell by 20
percent. Fewer criminals carrying guns makes the jobs of police
less dangerous. By contrast, while law-abiding permit holders
have come to the aid of police, they have never killed a police officer.

There are problems with Donohue's claims here. 1) The Olson and Maltz paper was never "withdrawn." It was published in the October 2001 Journal of Law and Economics. 2) While Michael Maltz has indeed written a piece critically discussing the measurement error in county level data, the lead author on this paper, David Olson, has never written anything after his paper was published critical of his paper and Donohue does not cite anything to the contrary. 3) John Whitley and I have a piece that has just appeared in the June 2003 issue of the Journal of Quantitative Criminology where we point out that all data has measurement error, that the error in the county level crime data is not particularly large, and, most importantly, there is not systematic error.

Donohue has not examined any of David Mustard's data studying the impact of right-to-carry laws on police fatalities. Therefore it is difficult to understand how Donohue can claim that Mustard's methodology was "flawed". Donohue's apprehensions about there being a temporary increase in crime before there is a decline are straightforward empirical questions and can only be examined with data. As an aside, David Mustard and I have indeed co-authored one paper in the past, but the passage is written ambiguously enough to suggest that David may be my co-author on the paper in question ("which was written by his own co-author, David Mustard"), which would be an incorrect inferrence.

Donohue's piece concludes by claiming:

Legislators may feel a modest increase in the number of dead Ohioans in
exchange for the ability to carry hidden handguns is an acceptable trade-off.
But don't believe anyone who says concealed carry laws will reduce crime.
There is no credible support for that view.

Of course, despite Donohue's attempt to imply that there is only my work as well as the work cited by Mustard and Olson and Maltz, there are many other papers that find that concealed handgun laws reduce violent crime. Five papers in addition to the ones by Mustard as well as Olson and Maltz that were published in the Journal of Law and Economics during October 2001 also find significant benefits from concealed handgun laws. There are also papers by Plassmann and Whitley in the same issue of the Stanford Law Review as Ayres and Donohue's piece (an earlier version of the Plassmann and Whitley paper with me as a coauthor is available here and see also the data and corrected results are available at www.johnlott.org ) and work by people such as Bartley and Cohen as well as papers that academics such as Steve Bronars and William Landes have written with me. Unlike all of these papers the work by Ayres and Donohue was not refereed.

The Appendix in Plassmann and Whitley's paper (p. 1366, as well as a similar breakdown in appendix 1 of my book) shows that even the harshest critics find at least as much support that the right-to-carry laws lower violent crime as they have no impact and essentially no evidence that they increase violent crime.

Possibly a more important point is that Plassmann and Whitley show that the vast majority of Ayres and Donohue's own results show that right-to-carry laws reduce violent crime rates and produce overall benefits to society. Even their so-called hybrid results on a state-by-state basis are reversed when a more general year-by-year analysis of the law is broken down on a state-by-state basis. The Ayres and Donohue results using the state-by-state breakdown is driven by two factors: 1) that the hybrid approach of fitting a straight line with an intercept shift to nonlinear data overestimates the crime rates in the early years and 2) Ayres and Donohue then only focusing on the first five years after the law.

Finally, even Ayres and Donohue?s county data results using their hybrid specification do not show a statistically significant increase in crime. Their mistake is a simple one since they are only looking at the coefficient on the intercept shift by itself when the crime rate during first year that the law is in effect is measured by both the intercept shift as well as the trend variable. When you do that none of the net effects of the law is statistically significantly different from zero during the Law?s first year is zero. Just to see the point estimates during the first year of the law, Ayres and Donohue?s own results from their Table 10:

	Murder	Rape	Robbery	Aggravated Assault
Intercept Shift	6.9%	5.9%	5.9%	3.6%
Trend Effect	-3.5%	-3.4%	-3.4%	-4.1%
Net Effect During the First Year Using the Hybrid Approach	3.4%	2.5%	2.5%	-0.5%
Net Effect During the Second Year Using the Hybrid Approach	-0.1%	-0.9%	-0.9%	-4.6%

Of course, by the second year of the law all the point estimates are negative and they become more negative with each successive year. Again, the positive (if statistically insignificant) point estimates during the first year of the law are an artifact of trying to fit an intercept shift with a straight line trend to nonlinear data (see Plassmann and Whitley?s discussion on page 1328).