Firearm policy in the United States has long been a serious policy issue. Much of the previous research on crime and firearms focused on the effects of states’ passage of concealed handgun licensing (CHL) legislation. Today, given the proliferation of CHL legislation and growing strength of the “pro-gun” movement, the primary policy focus has changed. State legislators now face issues concerning whether and how to increase access to CHLs. Because of this transformation, this research moves away from the research tradition focused on the effect of a legislative change allowing CHLs. Instead, we consider two issues more policy relevant in the current era: What are the dynamics behind CHL licensing? Do increases in concealed handgun licensing affect crime rates? Using county-level data, we found that the density of gun dealers and other contextual variables, rather than changing crime rates, had a significant effect on increases of the rates at which CHLs were issued. We also found no significant effect of CHL increases on changes in crime rates. This research suggests that the rate at which CHLs are issued and crime rates are independent of one another—crime does not drive CHLs; CHLs do not drive crime.
America has the most heavily armed civilian population in the western world. Estimates from the most recent Small Arms Survey [
As interesting, inspiring, or frightening as these data may be, general data on firearm ownership have, for a single reason, been the focus of little criminological research in the United States. American criminals, in the vast majority of crimes involving firearms, use handguns. In 2011, for example, handguns were used in 72 percent of firearm homicides, while rifles and shotguns were used in slightly less than 8 percent of firearm homicides [
More specifically, for decades, researchers have investigated the impact of the passage of concealed handgun licensing (CHL) legislation in various American states. The debate concerning the effects of CHL legislation began in 1997 with the publication of John Lott’s work indicating that the passage of concealed handgun legislation significantly reduced crime [
However ambiguous the results of the research on this topic may be, the effect of the passage of CHL legislation is, in reality, no longer a pressing policy issue. Forty-six of the 50 states in the United States now issue licenses to citizens that allow them to carry concealed handguns [
Most recently, the state of Kansas passed legislation allowing any adult resident to carry a concealed handgun without acquiring a license or receiving any training on gun safety or relevant state law related to the use of deadly force [
Unfortunately, little research is currently available to ascertain just what the effects of such “liberalization” may be. At this point John Lott offers the currently available research related to changes in CHL permit density and crime. Lott reports the results of his preliminary analyses of state-level concealed handgun permit data and crime from 2007 and concludes that a one percentage point increase in the percent of the adults in a state holding CHL permits may generate roughly a 1.4 percent drop in the murder rate [
This research follows Lott’s lead in using CHL permit data as its main independent variable. However, this research differs from Lott’s research in that we used county-level CHL permit and serious crime data covering at least a decade after passage of CHL legislation in four states. In its report, the National Academy of Sciences indicated that new approaches and different data are needed to be used to develop a clearer picture of the relationship between concealed carry and crime [
One can argue that both crime and demographics may affect the demand for CHLs. The other side of the issue (analogous to supply) involves estimating the number of opportunities civilians have to acquire a CHL. Unfortunately, no previous research has investigated this issue. In this research, we use the number of federally licensed firearm dealers in a county to represent the supply of CHL providers. CHL applicants must have a handgun, so the supply of firearm dealers is an important factor in access to a handgun. The density of Federal Firearm Licensees (FFLs or licensed gun dealers) should be positively associated with increased opportunities to acquire a CHL.
In Texas, for example, CHL applicants must provide a photo and fingerprints for a background check, pass a written exam, and pass a handgun proficiency test, which involves the supervised and graded a marksmanship test on a firing range [
Hypothesized relationships in model estimating changes in county CHL rates.
In this research we take advantage of the longitudinal nature of the available data to allow each county to serve over time as its own “covariates” and estimate the model depicted in Figure
Hypothesized relationships in model estimating the relationship between CHLs and crime.
Our analyses used data on the number of CHLs issued from 1998 to 2010 in every county in Florida, Michigan, Pennsylvania, and Texas. After reviewing the publicly available data from each of the states with CHL legislation, only these four states publicly reported CHL and arrest data at the county level for at least a decade following the passage of the CHL legislation in the state. Table
Concealed handgun information for study states.
Study state | CHL laws [ |
Estimated number of CHL holders [ |
CHLs per population of 100,00 |
Percent of population with CHL |
State laws allowing open carry of handguns |
---|---|---|---|---|---|
Florida | Shall Issue 1987 | 1,278,246 | 6,779 | 7% | No |
Michigan | Shall Issue 2000 | 430,095 | 4,325 | 4% | Yes |
Pennsylvania | Shall Issue 1989 | 872,227 | 6,876 | 7% | Yes |
Texas | Shall Issue 1995 | 798,048 | 2,816 | 3% | No |
United States | — | 11,113,013 | 3,599 | 4% |
As Table
Two of the four states also allowed open carry of handguns. However, both states place restrictions on such activities. In Michigan, a citizen is required to obtain a permit to purchase a handgun for open carry and that purchase is registered with state officials [
To investigate what factors affect the rates at which CHLs are issued, the research team was able to use data from three of our four states (Florida, Pennsylvania, and Texas). The analytic database included data taken from five county-level datasets: (1) concealed handgun licenses (CHL), (2) arrest data from the Federal Bureau of Investigation (UCR), (3) Federal Firearms Licensees (FFLs), (4) US census data, and (5) the Area Resource Files (ARF). The CHL and UCR data sources are discussed in Section
The Bureau of Alcohol, Tobacco, Firearms and Explosive (ATF) posted information on all Federal Firearms Licensees (federally licensed gun dealers) in our study counties. We downloaded the publicly available 2010 and 2011 data for three of the states from the ATF website [
The research team obtained data for 388 counties in our three-state database. However, we excluded three counties in Texas; one county had a CHL rate that was an extraordinary outlier; two counties had missing data. The final database contained 385 counties. Texas counties constituted 65.2 percent of the counties, while Pennsylvania and Florida each accounted for roughly half of the remaining counties. County-level covariates included in the analysis were county demographic characteristics gathered from the U.S. Census Bureau [
This analysis only used data from two sources. CHL data were obtained from publicly available state data, and crime (arrest) data from the publicly available Uniform Crime Reports were used. The CHL data were obtained from the Florida Department of Agriculture and Consumer Services [
The dependent variable for this analysis was the annual rate at which concealed handgun licenses were issued in a county, and the major independent variable was the change in the arrest rate for each type of index crime: seven individual crime arrests, violent crime arrests, property crime arrests, and total UCR arrests in the previous year.
To calculate the concealed handgun license (CHL) rate and the Federal Firearm Licensee (FFL) rate, we divided the total number of CHLs issued in 2011 in each county and the total licensed federal firearm dealers in the county by the total number of people aged 20 and over in the county times 10,000 people. We used the population aged 20 and over in our analysis because the available census data did not provide total population aged 21 and over. In most states, a person must be at least 21 years of age to apply for concealed handgun license. In Texas, however, a person at least 18 years of age can apply for concealed handgun license if he/she is currently serving in or honorably discharged from the military [
We calculated the change in arrest rates per 100,000 from 2009 to 2010 and used these figures in the construction of our lagged arrest/crime variable. We separately analyzed data for each of the seven crimes, for violent crimes, for property crimes, and for total crimes. Violent crimes include murder, rape, robbery, and aggravated assault, while property crimes include burglary, larceny, and motor vehicle theft. We calculated the density of FFLs per 10,000 persons in the county.
We performed natural log transformations on the CHL rate and FFLs rate because these variables exhibited positive skews. Using the data on general firearm ownership, we included a number of county-level covariates (e.g., age structure, average income) in our models.
Our analyses of the effect of CHL rates on crime included annual county arrest rates for murder, rape, robbery, aggravated assault, burglary, larceny, and motor vehicle theft. We further categorized these crimes into two broader categories: violent crime (murder, rape, robbery, and aggravated assault) and property crime (burglary, larceny, and motor vehicle theft).
The research team used U.S. Census Bureau county-level population data to develop county-level crime rates. For 2000 and 2010, we used the actual census estimates [
We converted the number of CHLs issued in a year into rates per 10,000 county residents and changes in crime from one year to another into rates per 100,000 county residents. We used total county level population estimates, rather than population over 21 years of age, to calculate the rates of CHL issuance in these analyses, in contrast to our other analyses. We did so because intercensal population estimates were not available for the above-21 population.
We used bivariate analyses to examine the relationships among log CHL rate, log FFL rate, and arrest rates for different crimes. Unadjusted and adjusted regression models were estimated. Since one cannot expect an instantaneous change in CHL rates due to changes in crime rates, we lagged the change in crime rate, our main independent variable, by one year. County covariates were included in the models. Covariates that had no significant effect were pruned from the final model.
As the effect of receiving a CHL license on crime is expected to materialize over time, we conducted longitudinal analyses. We estimated the effects of changes in CHL density on crimes using three sets of time lag models: A one-year lagged CHL rate model. A two-year lagged CHL rate model. A third model with both one- and two-year time lags for CHLs.
To account for the highly skewed nature of CHL rates, we used the natural log of CHL rates in our analysis. Analyses were completed using STATA statistical software (version 12.1) [
Another advantage of the time lag model is that it allows for easy identification of any significant latent association between measures over time. We accounted for the within-county correlation for crime rates and CHL license rates over time by computing cluster-robust standard errors at the county level [
In each of these models we included indicator variables for each state. As Table
In supplementary analysis, data from each state were analyzed separately. This was done to assure that pooling the data from the four states did not obscure differences among the states. In other analyses, annual county crime data points indicating a zero rate of a crime were deleted from analyses. This strategy served dual purposes. It removed smaller counties with potentially questionable UCR data from the analysis. It also protected the model estimates from any bias toward a positive relationship between CHLs and crime due to the fact that any change in the crime rate in a year following a year with no crime could only mean an increase in crime.
The results of these supplementary data analyses did not differ significantly from the results of the more general analyses. In few instances in these analyses, a positive relationship was observed between CHL licensing and crime. These results did not involve the violent crimes that one expects CHL to affect, and the results were not consistent.
For example, in these more finely grained analyses, the only significant negative result came in Florida. For larceny, the one-year lag for CHLs had a significant negative coefficient (
In addition, one might be concerned about the effects of spatial correlation on the results. The effect of such correlation is to increase the likelihood of Type 1 error. The results presented below seem to be in no danger of such a bias.
Table
Variables used in analyses of rates at which CHLs are issued in counties
Variables | Average ( |
SD |
---|---|---|
CHLs issued in 2011 | ||
Log concealed handgun license (CHL) rate (per 10,000, aged 20+) in 2011 | 4.72 | 0.65 |
|
||
Change in arrest rate (per 100,000) from 2009 to 2010 | ||
Murder | −0.09 | 0.88 |
Rape | −0.04 | 1.85 |
Robbery | −0.22 | 1.73 |
Assault | 0.93 | 27.45 |
Burglary | −1.45 | 6.29 |
Larceny | −2.10 | 10.00 |
Motor vehicle theft | 1.96 | 41.94 |
Violent crimes | 0.59 | 27.66 |
Property crimes | −1.58 | 44.71 |
Total crimes | −0.99 | 54.18 |
Log federal firearms license (FFL) rate (per 10,000) in 2011 | 1.75 | 0.62 |
Percent female divorced | 10.83 | 2.75 |
Percent urban population | 49.17 | 31.82 |
Percent unemployed rate, aged 16+ | 8.13 | 2.25 |
Percent person over 25 with less than HS education | 21.23 | 8.36 |
|
||
Counties | ||
State |
|
% |
|
||
Texas | 251 | 65.2 |
Pennsylvania | 67 | 17.4 |
Florida | 67 | 17.4 |
Table
Parameters for lagged county crime rates and logged federal firearms license rate (FFL) in models estimating county CHL rate.
Modeling CHL rates | Crime coefficient (SE) | Log FFL coefficient (SE) |
---|---|---|
Murder | 0.034 (0.026) | 0.196 |
Rape | −0.005 (0.013) | 0.194 |
Robbery | 0.006 (0.014) | 0.196 |
Assault | −0.002 (0.0008) | 0.192 |
Burglary | 0.002 (0.004) | 0.190 |
Larceny | −0.003 (0.002) | 0.194 |
Motor vehicle theft | −0.001 |
0.194 |
Total crimes | −0.001 |
0.194 |
Table
Average annual change in crime rates and CHL rates per county population of 100,000 (
Average annual change per population of 100,000 |
Years | |||
---|---|---|---|---|
1998–2010 | ||||
FL | MI | PA | TX | |
Concealed handgun licenses | 598.4 | 604.8 | 1,469.1 | 386.1 |
Personal crime rate | −9.15 | 0.14 | 1.36 | 1.36 |
Property crime rate | −6.12 | −1.69 | −0.79 | −8.65 |
Murder rate | 0.09 | −0.43 | 0.03 | −0.11 |
Rape rate | −1.34 | −0.42 | −0.22 | −0.35 |
Robbery rate | 0.03 | 0.08 | 0.54 | −0.264 |
Assault rate | −7.75 | 0.37 | 1.05 | 1.62 |
Burglary rate | −1.08 | −0.90 | −1.49 | −3.12 |
Larceny rate | −1.13 | 0.34 | 2.39 | −3.04 |
Motor theft rate | −3.73 | −1.38 | −2.08 | 0.26 |
The changes in crime rates exhibit a number of patterns. Rape and burglary showed decreases across all four of our study states. Robbery, aggravated assault, and larceny increased in Michigan and Pennsylvania, with Texas and Florida showing mixed results for the rates of changes in these crimes. Murder decreased in Michigan and Texas, but it increased in Florida and Pennsylvania.
In Table
Multivariate analysis of changes in county crime rates per population of 100,000, using both one-year lagged and two-year lagged CHL (
Crimes | Change in CHL rate |
Change in CHL rate |
Model |
|
---|---|---|---|---|
|
|
|||
Personal | −4.35 (3.76) | 3.32 (3.79) | 2.9 | 0.000 |
Property | −4.52 (7.55) | 6.82 (7.89) | 4.5 | 0.000 |
Murder | 0.02 (0.50) | −0.03 (0.51) | 1.5 | 0.094 |
Rape | −0.27 (0.82) | 0.62 (0.92) | 4.0 | 0.000 |
Robbery | 0.85 (0.96) | 0.34 (1.00) | 3.2 | 0.000 |
Assault | −4.74 (3.62) | 2.01 (3.54) | 3.3 | 0.000 |
Burglary | −4.22 (4.71) | 5.22 (4.70) | 1.9 | 0.024 |
Larceny | 4.15 (5.69) | −4.10 (6.17) | 5.4 | 0.000 |
Auto theft | −0.68 (1.35) | −0.35 (1.45) | 3.6 | 0.000 |
Though the focus here was on CHL licensing, one might have expected that the state indicators for those states with the greatest density of concealed handguns, as Lott’s preliminary analysis at the state-level seems to suggest, would display significant negative relationships with changes in the crime rate. Our results provided no evidence to support this conclusion [
The basic question underlying the hypotheses investigated in this research is simple—Is CHL licensing related in any way to crime rates? The results of this research indicate that no such relationships exist. For our study states, during the time period covered by our data, changes in crime rates did not affect subsequent CHL licensing rates. In addition, CHL licensing rates did not have a significant, negative or positive, effect on subsequent crime rates.
These results have some implications for the current policy debates concerning concealed handguns. The logic of relaxing requirements for concealed carry for the purposes of public safety implies that such legislation should reduce crime rates. However, our results indicate that more concealed handgun licensees in our four states had no significant negative effect on crime rates. This lack of a significant negative relationship is especially noteworthy for violent crimes such as murder, rape, robbery, and aggravated assault.
Intertwined with much of the discussion of easing restrictions on the accessibility of CHLs is the idea that citizens seek access to concealed weapons as a result of the risk of victimization. Our analyses of real changes in individuals’ risk of being victimized did not appear to be a driving factor affecting increases in concealed handgun licensing.
Instead, our results indicated that contextual factors drive the acquisition of concealed carry permits. The age distribution of the county population, the degree of urbanization, the level of educational attainment, and the specific state of residence had significant effects on CHL acquisition. In addition, more people acquired concealed carry permits in counties where more businesses or individuals sold firearms. The number of CHLs issued was driven more by the number of individuals or businesses offering handguns for sale (the supply of handguns) than by changes in the real threat of victimization as measured by county crime rates.
This study has two limitations. First, analysis focused on the change in the rate of CHLs, not rate of gun ownership per se, or rate of unlawful concealed carry. Second, we analyzed data from only four states within a limited time span. With these limitations in mind, the research team believes that these results do provide information that may be useful when considering the easing of restrictions on CHL access. Our results imply that such changes, to the degree that they increase legal concealed carry rates, will not have an effect on crime rates. The results also suggest that increases in carry rates resulting from easing access to concealed carry licenses will be driven more by the supply of firearms dealers in an area than by changes in crime rates. As interesting as these results may be, further research using the types of data utilized in this research will determine the robustness of our results.
The authors declare that there is no conflict of interests regarding the publication of this paper.