Prof. Robert Ehrlich’s review of the first edition of my book, More Guns, Less Crime, is well-written and it is interesting to know that he owns a gun despite his concerns about research on the benefits of doing so. Unfortunately, however, his discussion is incomplete and simply inaccurate. Below are responses to some of the more important claims he makes.
"Lott neglects to tell the reader that all his plots are not the actual FBI data (downloadable from their Web site), but merely his fits to the data."
There are several places in my book that discuss how the diagrams show how crime rates change before and after right-to-carry laws are adopted once other factors have been taken into account. It is important to distinguish not just whether there was a decline in crime rates, but whether there was a decline relative to other states that did not adopt the right-to-carry laws. The second edition of More Guns, Less Crime, which was published in 2000, was also clear on this point and the graphs showed the changes in crime relative to other states that did not change their laws and were in the same region of the country.
"Lott has used the data from 10 states in his book."
I used data from the entire United States. The first edition used state-level data from all the states and the District of Columbia, as well as county-level data for the entire country from 1977 through 1992 (and, in some estimates, up to 1994). The second edition of the book not only updated the county and state data through 1996, but also used city-level data for the largest 2,000 cities.
Possibly what Prof. Ehrlich means here is that only 10 states (with a total of 718 counties) adopted right-to-carry laws during the 1977-1992 period. The point of examining all counties in all the states was to make a year-by-year comparison of how the crime rates had changed in the counties with the right-to-carry laws relative to the counties in states without the laws. In the second edition of my book, a total of 20 states, representing 1,432 counties, adopted right-to-carry laws between 1977 and 1996.
"The actual data are much more irregular with lots of ups and downs, and they show nothing special happening at time t=0."
My book reports the year-to-year changes in crime rates (see pages 136-7), and these results are consistent with the before-and-after trends. One of the benefits of examining the change in trends is that there are straightforward statistical tests to see if the change is statistically significant.
"Overall, averaging the 10 states, there is a small but not statistically significant increase in the robbery rate at t=0, certainly not the dramatic decrease Lott’s fits show."
Prof. Ehrlich has examined state-level robbery rates for the 10 states that had adopted right-to-carry laws between 1977 and 1992, using data extended up until 1995 for the four years on either side of adoption. He finds that there is no statistically significant change in before-and-after trends. He claims to use data up until 1997, but that is not possible since he limited the sample to only four years after adoption, and the first full year these states had the law in effect was 1992. I have tried to replicate his results, but have been unable to do so: Robbery rates are declining after adoption relative to how they were changing prior to adoption.
Yet, even if his data analysis had been correct, his approach has a lot of problems. The main difficulty is that there is no comparison of what is going on in the states that do not adopt right-to-carry laws. When such a comparison is made, the drop in crime is about twice as large in right-to-carry state and twice as statistically significant. Accounting for other factors (e.g., the arrest rate for robbery) also increases the statistical significance of the drop. Many aspects of what he did are unclear, such as whether he weighted each state equally or weighted them by population (as is normally done). But neither approach altered the final result.
"What [Lott] does is to fit a smooth curve (actually a parabola) to the data earlier than t=0, and a separate curve to the data later than t=0."
This is only one of several different approaches reported in my book. The first edition also presented actual data on the number of permits issued per county over time for several states where the data were available. The second edition further examined whether differences in right-to-carry laws can affect the number of people who get permits (e.g., the permitting fees, the length of the training requirement, and how many years the law has been in effect), and whether this in turn can explain the changes in crime rates.
"Given a completely random set of data, Lott’s fitting procedure is virtually guaranteed to yield either a drop or a rise near time t=0."
This is not literally true. Besides a flat line, other possibilities very obviously include the crime rate first rising and then falling after adoption--or falling and then rising. The question is also not whether there is a change in trends, but also whether those changes are statistically significant.