Abstract
Objectives. To examine the relation of “shall-issue” laws, in which permits must be issued if requisite criteria are met; “may-issue” laws, which give law enforcement officials wide discretion over whether to issue concealed firearm carry permits or not; and homicide rates.
Methods. We compared homicide rates in shall-issue and may-issue states and total, firearm, nonfirearm, handgun, and long-gun homicide rates in all 50 states during the 25-year period of 1991 to 2015. We included year and state fixed effects and numerous state-level factors in the analysis.
Results. Shall-issue laws were significantly associated with 6.5% higher total homicide rates, 8.6% higher firearm homicide rates, and 10.6% higher handgun homicide rates, but were not significantly associated with long-gun or nonfirearm homicide.
Conclusions. Shall-issue laws are associated with significantly higher rates of total, firearm-related, and handgun-related homicide.
Firearm violence is a major public health problem. In 2015, there were approximately 36 000 firearm-related deaths in the United States; 13 463 were homicides, 22 018 were suicides, and 489 were unintentional injuries.1 During the same year, 72.9% of homicides were firearm homicides1 and, of these, approximately 90% were committed with a handgun. A central question in the debate about public policies to reduce firearm violence is whether easier access to concealed handguns increases or decreases the rate of firearm-related homicides.2 Some have argued that the feared or actual presence of armed citizens may deter violent crime.3 Others have suggested that a higher prevalence of people carrying guns will increase the likelihood that an altercation results in a fatality.4 Thus, having a clear understanding of the impact of concealed-carry laws on firearm-related homicide would help guide policymakers who are aiming to reduce firearm violence.
As of the end of 2015, all states allowed certain persons to carry concealed handguns, but there were 3 major variations in permitting policy5 (Table 1). In 9 states, law enforcement officials had wide discretion over whether to issue concealed-carry permits; these are referred to as “may-issue” states. In 32 states, there was little or no discretion; these are referred to as “shall-issue” states because permits must be issued if requisite criteria are met. In an additional 9 states, there was no permit necessary to carry a concealed handgun; these are referred to as “permitless-carry” states. The wide variation in these policies between states and over time presents the opportunity to compare homicide rates between states with varying concealed-carry permitting policies to examine the impact of concealed-carry laws on homicide.
TABLE 1— Concealed-Carry Permitting Laws and Age-Adjusted Firearm Homicide Rates by US State, 2015, and Status of Laws During the Period of 1991 to 2015
| State | Age-Adjusted Firearm Homicide Rate,a 2015 (per 100 000) | Status of Concealed-Carry Permitting Law, 2015 | Effective Date of Current (as of 2015) Concealed-Carry Law |
| Hawaiib | 0.75 | May issue | Before 1991 |
| New Hampshire | 0.96 | Shall issue | Before 1991 |
| Rhode Island | 0.99 | May issue | Before 1991 |
| Maine | 1.14 | Shall issue | Before 1991 |
| Massachusetts | 1.26 | May issue | Before 1991 |
| Utah | 1.39 | Shall issue | 1995 |
| Idaho | 1.29 | Shall issue | Before 1991 |
| Iowa | 1.62 | Shall issue | Before 1991 |
| North Dakota | 1.69 | Shall issue | Before 1991 |
| Vermont | 1.76 | Permitless carry | Before 1991 |
| Minnesota | 1.77 | Shall issue | 2003 |
| South Dakota | 1.97 | Shall issue | Before 1991 |
| New York | 2.07 | May issue | Before 1991 |
| Wyoming | 2.16 | Permitless carry | 2011c |
| Montana | 2.17 | Shall issue | Before 1991 |
| Washington | 2.32 | Shall issue | Before 1991 |
| Oregon | 2.35 | Shall issue | Before 1991 |
| Connecticut | 2.43 | May issue | Before 1991 |
| Colorado | 2.46 | Shall issue | 2003 |
| Nebraska | 2.67 | Shall issue | 2007 |
| West Virginia | 2.89 | Shall issue | Before 1991 |
| Wisconsin | 3.18 | Shall issue | 2011 |
| New Jersey | 3.22 | May issue | Before 1991 |
| Virginia | 3.29 | Shall issue | 1995 |
| Kansas | 3.35 | Shall issue | 2007 |
| California | 3.52 | May issue | Before 1991 |
| Arizona | 3.56 | Permitless carry | 2010c |
| Kentucky | 3.96 | Shall issue | 1996 |
| Texas | 4.04 | Shall issue | 1995 |
| Pennsylvania | 4.34 | Shall issue | Before 1991 |
| Ohio | 4.38 | Shall issue | 2004 |
| Nevada | 4.49 | Shall issue | 1995 |
| North Carolina | 4.54 | Shall issue | 1995 |
| Indiana | 4.61 | Shall issue | Before 1991 |
| Florida | 4.66 | Shall issue | Before 1991 |
| Michigan | 4.74 | Shall issue | 2001 |
| New Mexico | 4.79 | Shall issue | 2001 |
| Alaska | 5.22 | Permitless carry | 2003c |
| Arkansas | 5.34 | Shall issue | 1995 |
| Illinois | 5.45 | Shall issue | 2013 |
| Tennessee | 5.51 | Shall issue | 1994 |
| Georgia | 5.73 | Shall issue | Before 1991 |
| Oklahoma | 5.87 | Shall issue | 1995 |
| Delaware | 6.12 | May issue | Before 1991 |
| South Carolina | 7.55 | Shall issue | 1996 |
| Maryland | 7.69 | May issue | Before 1991 |
| Missouri | 7.92 | Shall issue | 2003 |
| Alabama | 8.43 | Shall issue | 2013 |
| Mississippi | 9.11 | Shall issue | 1991 |
| Louisiana | 9.96 | Shall issue | 1996 |
Note. “May-issue” states are those in which law enforcement officials had wide discretion over whether to issue concealed-carry permits. “Shall-issue” states are those in which there was little or no discretion; permits must be issued if requisite criteria are met. “Permitless-carry” states are those in which there was no permit necessary to carry a concealed handgun.
a From Centers for Disease Control and Prevention (CDC).1
b Data for Hawaii are unavailable for the years 2010 to 2015 because the CDC’s Web-Based Injury Statistics Query and Reporting Systems does not report homicide counts fewer than 10. The data here are from 2009.
c Changed from “may issue” to “shall issue” in 1994.
The critical difference between may-issue and shall-issue laws is that in may-issue states, law enforcement officials may use their judgment in making decisions about whether to approve or deny a permit application, whereas in shall-issue states, no judgment is involved—the application must be approved unless the applicant is categorically prohibited from concealed handgun possession. In may-issue states, the element of discretion allotted to law enforcement is typically a judgment regarding the “suitability” or “need” of a person to carry a concealed weapon (Table 2). Law enforcement officials have a wide degree of latitude in making these judgments. In shall-issue states, the categorical prohibitions consist of a list of specific criminal convictions.
TABLE 2— Elements of Discretion in Law Enforcement Decisions to Approve or Deny Concealed Handgun Carry Permits: “May-Issue” US States, 2015
| State | Elements of Discretion | Citation |
| California | Applicant must be of “good moral character” and must have “good cause” for issuance of the license. | California Penal Code § 26150, § 26155 |
| Connecticut | Applicant must intend only to make “legal use” of the handgun and must be a “suitable person to receive such permit.” | Connecticut General Statutes § 29-28 |
| Delaware | Applicant must be “of good moral character,” must desire the handgun for “personal protection” or “protection of the person's property,” and must submit signed, written statements of 5 “respectable citizens” of the county who testify that the applicant is a person “of sobriety and good moral character” and “bears a good reputation for peace and good order in the community” and that a handgun is “necessary for the protection of the applicant or the applicant's property.” The Superior Court has discretion to approve or deny the application. | Delaware Code § 1441 |
| Hawaii | Must be “an exceptional case,” the applicant must show “reason to fear injury to the applicant's person or property,” the applicant must be “a suitable person” to be licensed, and the chief of police must determine that the person “is qualified to use the firearm in a safe manner.” | Hawaii Revised Statutes § 134-9 |
| Maryland | Applicant must have a “good and substantial reason to wear, carry, or transport a handgun, such as a finding that the permit is necessary as a reasonable precaution against apprehended danger,” and the applicant must not have “exhibited a propensity for violence or instability that may reasonably render the person's possession of a handgun a danger to the person or to another.” | Maryland Public Safety Code § 5-306 |
| Massachusetts | Applicant must be a “suitable” person and must not be judged to potentially create a risk to public safety. | Massachusetts General Laws 140 § 131 |
| New Jersey | Applicant must demonstrate a “justifiable need to carry a handgun” and must submit endorsements by 3 individuals who have known the applicant for at least 3 years that the applicant is “a person of good moral character and behavior.” | New Jersey Statutes § 2C:58–4 |
| New York | Applicant must be “of good moral character,” must be “of good character, competency, and integrity,” and there must be no “good cause” for denial of the license. | New York Penal Law § 400.00 |
| Rhode Island | Applicant must have “good reason to fear an injury to his or her person or property” or have “any other proper reason” for carrying a handgun and must be a “suitable person to be so licensed.” | General Laws of Rhode Island § 11-47-11 |
Note. “May-issue” states are those in which law enforcement officials had wide discretion over whether to issue concealed-carry permits.
Unfortunately, the existing literature on the impact of concealed carry laws is inconsistent. At least 10 national studies have examined the relationship between shall-issue concealed-carry laws and firearm-related or total homicide rates at the state level (Table A, available as a supplement to the online version of this article at http://www.ajph.org).3,6–14 In 2 studies, shall-issue laws were found to decrease homicide rates.3,6 In 2 studies, these laws were found to increase homicide rates.7,8 Six studies reported no clear impact of shall-issue laws on homicide rates.9–14 The inconsistency of these results has understandably created some confusion about what approach is most effective to address the firearm violence problem.
Most of the published literature on this topic includes data that are more than a decade old: the most recent year of data analyzed was 2010, and only 3 of the 10 studies examined data past the year 1998 (Table A, available as a supplement to the online version of this article at http://www.ajph.org). Since 1998, 11 additional states have enacted shall-issue laws.5 This provides more variation over time and a longer follow-up period to examine this research question. Moreover, Ayres and Donohue15 and Hepburn et al.11 have suggested that the relationship between concealed-carry laws and homicide rates may have been different during the period before and after the early 1990s. In addition, studies that included homicide rates from before 1994 were examining a trend that was increasing, whereas studies examining homicide rates after 1994 were capturing declining trends. For these reasons, a reexamination of this research question with more recent data is needed.
One limitation of the existing literature is that no previously published research has examined the specific impact of concealed-carry laws on handgun versus long-gun homicide rates. This is important because if such laws increase homicide by making it easier for people at high risk for violence to carry handguns, this effect should only be observed in relation to handgun-related homicides, not homicides committed with long guns. On the other hand, if permissive concealed-carry laws deter crime by generating fear among potential perpetrators of encountering an armed individual, then all crime including handgun, long-gun, and nonfirearm homicide should decrease.
Another limitation of previous studies is that nearly all of them used linear models. However, homicide rates represent count data, and the distribution of homicide rates across states is highly skewed16 (Figure A, available as a supplement to the online version of this article at http://www.ajph.org). Plassmann and Tideman argued that a count model (such as a Poisson or negative binomial model) is the most reliable for analyzing crimes, such as homicides, with low occurrence rates.16 Beyond the Plassmann and Tideman study, only 1 other study11 used a count model.
We examined the relationship between shall-issue concealed-carry laws and total, firearm-related, and non–firearm-related homicide rates, as well as handgun versus long-gun homicide rates across all 50 states during the 25-year time period of 1991 to 2015 with both count and linear regression models. We examined the specificity of the relationship between concealed-carry laws and homicide rates by separately modeling firearm versus nonfirearm homicide rates and then within firearm-related homicides by modeling handgun versus long-gun homicide rates. We analyzed the relationship between shall-issue concealed-carry laws and homicide rates by using both a count and a linear regression model, thus examining the robustness of results to the type of model used.
We used a quasi-experimental panel design, taking advantage of changes in state concealed-carry permitting laws over time, to explore the relationship between these laws and total, firearm-related, and non–firearm-related homicide rates in the 50 states over a 25-year period, 1991 to 2015. We modeled homicide rates in 2 ways: (1) using a negative binomial regression with homicide rates as the outcome variable and (2) using linear regression with log-transformed homicide rates as the outcome variable. In both cases, we included year and state fixed effects and controlled for a range of time-varying, state-level factors.
The main outcome variable was the age-adjusted firearm homicide rate in each year analyzed. For example, Missouri’s shall-issue law went into effect in 2003; thus, we analyzed homicide rates associated with Missouri’s shall-issue law for the years 2004 to 2015. We obtained homicide rates from the Centers for Disease Control and Prevention’s (CDC’s) Web-Based Injury Statistics Query and Reporting Systems (WISQARS) database.1 This is the ideal source for homicide data because there is complete annual reporting from all 50 states and because the data are extracted from the Vital Statistics death registry maintained by the National Center for Health Statistics, which is based on standardized death certificates. The completeness of reporting is approximately 99%.17 The CDC age-adjusted the rates to the 2000 standard population.
The second outcome variable was the handgun or long-gun homicide rate, obtained from the Federal Bureau of Investigation’s Uniform Crime Reports, Supplemental Homicide Reports (SHR).18 Although WISQARS does provide mortality data from International Classification of Diseases, Ninth Revision and Tenth Revision, codes that can list handgun and long gun as the cause of death, unfortunately, most death certificates involving a firearm homicide do not specify the type of weapon used. Therefore, most firearm homicide deaths in WISQARS are classified as “other and unspecified” firearm, and it is not possible to use these data to disaggregate handgun and long-gun homicides.19 By contrast, the SHR is missing data on the type of weapon used in firearm homicides in just 13.4% of cases. Thus, the SHR is the best, if not only, source for state-specific, firearm type–specific homicide data.
The SHR disaggregates firearm homicides into handgun, rifle, shotgun, and other (and unknown). We used the handgun deaths to generate handgun homicide rates and the sum of rifle, shotgun, and other gun deaths to generate long-gun homicide rates for each state and year. Although SHR data may include listing of multiple weapons in an incident, only 1 weapon may be associated with a homicide death.20 Because of missing data on weapon type, we excluded 13.4% of firearm homicide cases in estimating handgun homicide rates. Nevertheless, there was little discrepancy between the firearm homicide totals from WISQARS and the SHR, which were correlated at r = 0.98.
Because not all local law enforcement agencies complete the supplemental reports, the SHR data set excludes approximately 10% of all homicides.21 This problem was addressed by applying weights that adjusted each state- and year-specific estimate up to the overall number of homicides reported in the Uniform Crime Report for that state and year. Fox kindly provided us with updated SHR files that added previously missing data for Florida and included data through 2015.21
Using Thomson Reuters Westlaw to access historical state statutes and session laws, we developed a database indicating the presence or absence of 100 provisions of firearm laws in each state over the 25-year period.5 We coded laws by the year they went into effect, regardless of the month of the effective date. However, in the analytic models, we lagged the state laws by 1 year, which ensured that all laws were in effect during the year in which their impact was being assessed. Following Lott and Mustard,22 we assessed the impact of laws starting in the first full year they were in effect.
We examined the potential impact of shall-issue laws, comparing them to may-issue laws. In other words, using the may-issue states as the reference group, we estimated the impact of shall-issue laws on homicide rates. Because only 4 states had permitless-carry laws in place during the study period, there were not enough observations to allow any meaningful analyses of these laws. Therefore, we deleted state–year observations in which a permitless-carry law was in effect.
We controlled for 12 state-level factors that (1) were found in the previous literature3,6–14 to be significantly related to homicide rates and (2) were significantly related to the presence of shall-issue laws in our data set (i.e., the regression coefficient for the variable was significant at a level of P = .05 in a logistic regression with shall-issue law as the dependent variable): household firearm ownership (using the standard proxy, which is the percentage of all suicides committed with a firearm), proportion of Blacks, proportion of young adults (aged 18 to 29 years), proportion of men among young adults, proportion of the population living in urban areas, total population, population density, per capita alcohol consumption, the nonhomicide violent crime rate (aggravated assault, robbery, and forcible rape), the poverty rate, unemployment rate, median household income, per capita disposable income, incarceration rate, and per capita number of law enforcement officers. Variable definitions and data sources are provided in Table B, available as a supplement to the online version of this article at http://www.ajph.org. We also controlled for the following state firearm laws that could serve as alternative explanations for changes in homicide during the study period: (1) universal background checks required for all handgun purchases, (2) waiting periods required for all handgun purchases, and (3) permits required to purchase or possess firearms.
Because homicide rates are not normally distributed but skewed and overdispersed, we modeled this outcome by using a negative binomial distribution. To control for clustering in our data by year (25 levels) and by state (50 levels), we entered year and state as fixed effects in the regression models. We used robust standard errors that account for the clustering of observations, serial autocorrelation, and heteroskedasticity.23
Our final model was as follows:
where 
is the probability that state s in year t has a homicide rate equal to 
, 
, and 
.
The mean homicide rate was then modeled as follows:
where CCst is a dummy variable for the presence of a shall-issue law, C is a vector of control variables, S represents state fixed effects, and T represents year fixed effects.
The negative binomial regression coefficients are reported as incidence rate ratios (IRRs). The IRR indicates the percentage difference in homicide rate for states with a shall-issue concealed-carry law compared with states with a may-issue law.
To check the robustness of our findings, we repeated the analyses with a linear regression model, with the log-transformed homicide rate as the outcome variable, again by using robust standard errors.23 As with the negative binomial models, we included year and state fixed effects, and we included the same state-level control variables.
We conducted analyses with Stata version 14.1 (StataCorp LP, College Station, TX). We evaluated the significance of regression coefficients by using a Wald test at α = 0.05.
We checked the robustness of our results by conducting several sensitivity analyses, including Restricting the analysis to the 23 states in which shall-issue laws were adopted during the study period, Using raw count data instead of homicide rates, Restricting the analysis to states with population greater than 1 000 000, Restricting the analysis to the period 1991 to 2002, Restricting the analysis to the period 2003 to 2015, and Using SHR instead of WISQARS homicide data (thus avoiding the problem of missing data for some smaller states after 1998).
During the study period, 23 states adopted shall-issue laws (Table 1). By 2015, 37 states had such laws. In the same year, the average firearm homicide rate in the states with shall-issue laws was 4.11 per 100 000, compared with 3.41 per 100 000 in the may-issue states. The number of states that had permitless-carry laws in effect at all during the study period was small (n = 4), as was the number of observations (n = 46), limiting our ability to analyze the impact of these laws. Because CDC does not report homicide counts of fewer than 10 in years after 1998, we were missing outcome data for several years for 6 states (Hawaii, New Hampshire, North Dakota, South Dakota, Vermont, and Wyoming); a sensitivity analysis with SHR data revealed that these omissions do not affect our findings.
In negative binomial regression models, shall-issue concealed-carry permitting laws were significantly associated with 6.5% higher total homicide rates compared with may-issue states (IRR = 1.065; 95% confidence interval [CI] = 1.032, 1.099; Table 3). The association was specific to firearm homicide rates, which were 8.6% higher in shall-issue states (IRR = 1.086; 95% CI = 1.047, 1.126). There was no significant association between shall-issue laws and nonfirearm homicide rates (IRR = 1.014; 95% CI = 0.963, 1.068). Further disaggregation within firearm homicides showed that the association between shall-issue laws and firearm homicide rates was specific to handgun homicide. Shall-issue states had handgun homicide rates that were 10.6% higher (IRR = 1.106; 95% CI = 1.039, 1.177), but there was no significant association with long-gun homicide rates (IRR = 0.999; 95% CI = 0.915, 1.090).
TABLE 3— Sensitivity Analyses of Relationship Between “Shall-Issue” Concealed-Carry Permitting Laws and Homicide Rates: United States, 1991–2015
| Homicide Rate, IRR (95% CI) | |||
| Type of Analysis | Total | Firearm | Handgun |
| Main analysis | 1.065 (1.032, 1.099) | 1.086 (1.047, 1.126) | 1.106 (1.039, 1.177) |
| Analysis restricted to states that adopted shall-issue concealed-carry laws during study period | 1.063 (1.028, 1.099) | 1.068 (1.030, 1.108) | 1.074 (1.002, 1.150) |
| Analysis using raw count of homicides with population as the exposure variable | 1.051 (1.020, 1.083) | 1.079 (1.039, 1.120) | 1.139 (1.067, 1.217) |
| Analysis restricted to states with population > 1 million | 1.055 (1.023, 1.087) | 1.067 (1.030, 1.105) | 1.095 (1.029, 1.166) |
| Analysis restricted to years before 2003 (1991–2002) | 1.058 (1.014, 1.104) | 1.067 (1.019, 1.116) | 1.107 (1.037, 1.180) |
| Analysis restricted to years after 2002 (2003–2015) | 1.064 (1.009, 1.122) | 1.100 (1.028, 1.176) | 1.274 (1.092, 1.488) |
| Analysis using Supplemental Homicide Report data instead of Vital Statistics data | 1.044 (1.006, 1.083) | 1.094 (1.047, 1.143) | 1.106 (1.039, 1.177) |
Note. “Shall-issue” states are those in which there was little or no discretion; permits must be issued if requisite criteria are met. CI = confidence interval; IRR = incidence rate ratio. All models include year and state fixed effects and control for the following time-varying, state-level factors: household gun-ownership levels, proportion of young men, proportion of young adults, proportion of Blacks, proportion living in an urban area, total population, population density, median household income, poverty rate, unemployment rate, per capita disposable income, per capita alcohol consumption, violent crime rate, incarceration rate, per capita law enforcement officers, universal background check laws for all handguns, waiting periods for all handguns, and permits required for all firearms.
The results of the linear regression analyses were similar. Here, shall-issue laws were significantly associated with 6.6% higher total homicide rates compared with may-issue states (95% CI = 3.0%, 10.4%; data not shown). The association was specific to firearm homicide rates, which were 11.7% higher in “shall issue” states (95% CI = 6.4%, 17.2%); there was no significant association between these laws and nonfirearm homicide rates. Further disaggregation within firearm homicides showed that the association between shall-issue laws and firearm homicide rates was specific to handgun homicide. Shall-issue states had handgun homicide rates that were 19.8% higher (95% CI = 10.3%, 30.1%), but rates of long-gun homicide were not significantly different in states with shall-issue compared with may-issue laws.
The significant association between shall-issue laws and higher total, firearm, and handgun-related homicide rates remained when we restricted the analysis to the 23 states in which these laws were adopted during the study period (Table 3). This pattern of results was robust to a series of additional sensitivity checks, including using raw count data, restricting the analysis to states with a population of more than 1 000 000, restricting the analysis to the period 1991 to 2002, restricting the analysis to the period 2003 to 2015, and using SHR instead of WISQARS homicide data.
To the best of our knowledge, this is the first study to examine the relationship between concealed-carry permitting laws and handgun-specific homicide rates. We found that, when we used both count and linear models and after we controlled for a range of time-varying state factors and for unobserved time-invariant state factors by using a fixed-effects model, shall-issue concealed-carry permitting laws were significantly associated with 6.5% higher total homicide rates, 8.6% higher firearm-related homicide rates, and 10.6% higher handgun-specific homicide rates compared with may-issue states.
A major reason for inconsistent results in the existing literature on the effects of concealed-carry laws may be that the relationship between concealed-carry laws and homicide rates was different during the period before and after the early 1990s.11,15 It is possible that despite the enactment of early shall-issue laws in the 1970s and 1980s, the demand for handgun permits in those states was modest. There has been a striking increase in the demand for pistols, especially those designed for concealed carry, during the past decade.24 Recently, Steidley found that the adoption of shall-issue laws during the period 1999 to 2013 was associated with a persistent, long-term increase in handgun sales in all 7 states studied.25 Our analysis provides further support for the hypothesis that the relationship between shall-issue laws and higher homicide rates increased over time, as the regression coefficients for these laws was higher for the second half of the study period (2003–2015) compared with the first half (1991–2002).
Our finding that the association between shall-issue laws and homicide rates is specific to handgun homicides adds plausibility to the observed relationship. If the relationship between shall-issue laws and homicide rates were spurious, one might expect to see the relationship hold for long-gun as well as handgun homicide rates. Moreover, this finding is inconsistent with the hypothesis that permissive concealed-carry laws deter crime by increasing the presence of armed individuals. Were that the case, one would expect to see lower handgun, nonhandgun, and nonfirearm homicide rates in shall-issue compared with may-issue states. The lack of an association between shall-issue laws and long-gun homicide rates is also inconsistent with the hypothesis that the presence of more concealed weapons escalates the level of violence in encounters that may involve a long gun.
This study has several novel strengths, including the use of both count and linear models, the use of recent data (through 2015), and the disaggregation of homicide rates. Nevertheless, caution should be exercised in assessing causality from an ecological study such as this one. In particular, these results should be interpreted with caution because of the possibility that they reflect a reverse association. That is, it is possible that the adoption of shall-issue concealed carry laws is associated with higher baseline homicide rates so that we are picking up not a causal effect of these laws on homicide but a systematic difference in baseline homicide rates between states that do or do not have these laws. However, our findings hold even when the analysis is restricted to states that started with may-issue laws at the beginning of the study period and adopted shall-issue laws during the study period.
An additional limitation of this study is that we could not consider the enforcement of concealed-carry laws.26 Enforcement of these laws may vary not only among states, but also among counties in the same state.11 In addition, we did not have information on the number of concealed-carry permits issued in each state or the number of homicides committed by concealed-carry permittees.
It is also important to note that we examined only fatal firearm injuries. Further research should investigate potential effects of concealed-carry laws on nonfatal firearm injuries.
Finally, we were unable to analyze the impact of permitless-carry laws because of the small number of observations. Only 4 states had permitless-carry laws in place during the study period. However, in the past 2 years, an additional 5 states have enacted such laws. Elucidating the impact of permitless-carry laws will require follow-up for the 9 states that now have such laws in effect.
Despite these limitations, this study suggests that there is a robust association between shall-issue laws and higher rates of firearm homicides. The trend toward increasingly permissive concealed-carry laws is inconsistent with public opinion, which tends to oppose the carrying of guns in public.27 Our findings suggest that these laws may also be inconsistent with the promotion of public safety.
See also Donohue, p.
ACKNOWLEDGMENTS
Support for this research was provided by the Robert Wood Johnson Foundation Evidence for Action Program (grant 73337).
We gratefully acknowledge the assistance of James Alan Fox, PhD, the Lipman Family Professor of Criminology, Law, and Public Policy at the School of Criminology and Criminal Justice at Northeastern University, who kindly provided the Multiply-Imputed Supplemental Homicide Reports File, 1976–2015, including the data sets and a codebook.
Note. The views expressed here do not necessarily reflect those of the Robert Wood Johnson Foundation.
HUMAN PARTICIPANT PROTECTION
This study made use of secondary data only and did not require institutional review board approval.
