Child Access Prevention Laws and Juvenile Firearm-Related Homicides

Abstract

Debate over safe-storage gun regulations has captured public attention in the aftermath of several high-profile shootings committed by minors. To date, the existing literature provides no evidence that these laws are effective at deterring gun crime, a conclusion that has prompted the National Rifle Association to assert that such regulations are “unnecessary” and “ineffective.” Using data from the FBI’s Supplementary Homicide Reports for the period 1985–2013, we find that child access prevention (CAP) laws are associated with a 17 percent reduction in firearm-related homicides committed by juveniles. The estimated effect is stronger among whites than nonwhites and is driven by states enforcing the strictest safe-storage standard. We find no evidence that CAP laws are associated with firearm-related homicides committed by adults or with non-firearm-related homicides committed by juveniles, suggesting that the observed relationship between CAP laws and juvenile firearm-related homicides is causal.

“Too often, minors have also used their families’ unsecured firearms to intentionally perpetrate violence against others.”

-Giffords Law Center to Prevent Gun Violence (2018a)

“This poorly thought out legislation is without any consideration for personal circumstances. It invades people’s homes and forces them to render their firearms useless in a self-defense situation by locking them up.”

-National Rifle Association-Institute for Legislative Action on Seattle’s recently passed safe-storage ordinance (NRA-ILA 2018).

1. Introduction

One of the deadliest school shootings in U.S. history recently intensified public discussion over the safe storage of firearms after it was learned that the guns were taken from the shooter’s home and belonged to his father (Coaston 2018; Mann 2018).⁴ This comes at a time of rising youth gun violence and increasing public support for gun restrictions.⁵ For instance, a 2017 U.S. survey found that approximately 60 percent of gun owners backed safe-storage requirements for guns in households with children (Barry et al. 2018).⁶ As states grapple with decisions on gun control, Americans prefer child access prevention (CAP) laws to more divisive policies such as bans on assault weapons and high-capacity magazines (Ingraham 2018).

CAP laws encourage the safe storage of firearms by imposing liability on adults who allow children unsupervised access to guns (Giffords Law Center to Prevent Gun Violence 2018a). Gun safety advocates support CAP laws as a way to limit firearm-related homicides, as well as a way to decrease unintentional shootings and suicides among minors (Jones 2017; Iannelli 2018). On the other hand, critics argue that safe-storage requirements impede a person’s ability to defend their home and family during a violent intrusion, and that these laws may actually increase incidences of murders, rapes, robberies, and other forms of violent crime (Kopel et al. 2000).

As public calls for safe storage grow louder, it is likely that an increasing number of state legislatures will come under pressure to pass CAP laws or toughen their existing CAP requirements. In fact, New York lawmakers passed a bill in March, 2019 to require “gun owners who live with someone under 16, or who know that someone under that age could access their gun, to lock up their firearm when it is not in use” (Wang 2019, para. 4). One of the few municipal-level ordinances requiring the safe storage of firearms passed in Seattle, Washington on July 9, 2018 (Norimine 2018).⁷ Recent estimates suggest that 7 percent of U.S. children (≈ 4.6 million) live in homes with an unlocked and loaded firearm (Azrael et al. 2018).

Uniformly, the existing literature finds no statistically significant evidence that CAP laws reduce violent crime (Cummings et al. 1997; Lott and Whitley 2001; Lott 2003; DeSimone et al. 2013; Anderson and Sabia 2018).⁸ Referencing Lott and Whitley (2001), the National Rifle Association has claimed that CAP laws are “unnecessary, ineffective, and endanger law-abiding gun owners” (Mortensen 2017, para. 2). We challenge this claim by providing evidence that CAP laws effectively deter gun crime among juveniles.

Using the FBI’s Supplementary Homicide Reports (SHR), this study is the first to explore the relationship between CAP laws and homicides committed by juveniles where a firearm was used in the commission of the crime. We focus on homicides, rather than other forms of violent crime, because information on the offender’s age is available and the laws are intended to affect households where one or more juveniles are present. Examining the period 1985–2013, a span when 26 states and the District of Columbia adopted CAP legislation, our estimates suggest that CAP laws are associated with a 17 percent reduction in the expected number of firearm-related homicides committed by juveniles, and this effect is driven by states enforcing a “negligent storage” standard, the strictest form of CAP legislation. Furthermore, we find that CAP laws are not associated with firearm-related homicides committed by adults nor are they associated with non-firearm-related homicides committed by juveniles, providing evidence that the relationship between CAP laws and juvenile firearm-related homicides is causal.

The remainder of the paper is organized as follows. In Section 2, we provide background on CAP laws and review the literature; in Section 3, we describe our data and methods; and in Section 4 we report our results and explore a potential mechanism through which CAP laws may affect juvenile firearm-related homicides, namely gun ownership. Section 5 concludes.

2. Background

The storage of firearms within the home was unregulated in the United States until 1981, when Missouri became the first state to pass a CAP law. Under the Missouri law, it is illegal to recklessly provide firearm access to a person under the age of 18 (Giffords Law Center to Prevent Gun Violence 2017). Table 1 lists the CAP laws passed during our sample period and Appendix Figure 1 illustrates their geographic evolution over time.⁹

Table 1.

Child Access Prevention Laws

	Effective Year	Type of CAP Law
California	1992	Negligent Storage
Colorado	2000	Reckless Endangerment
Connecticut	1990	Negligent Storage
Delaware	1994	Reckless Endangerment
D.C.	2009	Negligent Storage
Florida	1989	Negligent Storage
Georgia	1994	Reckless Endangerment
Hawaii	1992	Negligent Storage
Illinois	2000	Negligent Storage
Indiana	1994	Reckless Endangerment
Iowa	1990	Negligent Storage
Kentucky	1994	Reckless Endangerment
Maryland	1992	Negligent Storage
Massachusetts	1998	Negligent Storage
Minnesota	1993	Negligent Storage
Mississippi	1994	Reckless Endangerment
Missouri	1981	Reckless Endangerment
Nevada	1991	Reckless Endangerment
New Hampshire	2001	Negligent Storage
New Jersey	1992	Negligent Storage
North Carolina	1993	Negligent Storage
Oklahoma	1993	Reckless Endangerment
Rhode Island	1995	Negligent Storage
Tennessee	1994	Reckless Endangerment
Texas	1995	Negligent Storage
Utah	1993	Reckless Endangerment
Virginia	1992	Reckless Endangerment
Wisconsin	1992	Reckless Endangerment

Notes: Data on CAP laws were obtained from Lott and Whitley (2001), Webster et al. (2004), DeSimone et al. (2013), and Giffords Law Center to Prevent Gun Violence (2021a), and our own searches of legislative codes.

CAP laws take a variety of forms. Fourteen states and the District of Columbia impose criminal liability on individuals who negligently store firearms. Generally, these laws apply whenever a person “knows or reasonably should know” that a minor “is likely to gain access” to a firearm (Giffords Law Center to Prevent Gun Violence 2021a, para. 9). In these states, if a minor gains access to a firearm that was not properly stored, the gun owner faces potential fines, imprisonment, or some combination of both. For instance, violation of Minnesota’s negligent storage CAP law is punishable by up to a $3,000 fine and one year in jail (Peters 2013). The remaining states listed in Table 1 levy a weaker standard for criminal liability and impose penalties only in the event of reckless, knowing or intentional conduct by the adult. These laws exclude negligence, which is the lowest threshold for liability (Giffords Law Center to Prevent Gun Violence 2021a).¹⁰ In some cases, CAP laws have been used to punish dealers and manufacturers who failed to include the appropriate safety devices with the sale of their firearms (Shaffer 2000).

CAP laws vary along other margins as well. For example, the broadest form of negligent storage laws applies regardless of whether a minor actually gains possession of a firearm. In these states, criminal liability is imposed simply when a child “may” or “is likely to” gain access. In other states, criminal liability applies only if a child is found using or carrying a firearm. Common exceptions to these laws include when access is gained via illegal means (e.g., breaking and entering), the firearm is used in self-defense, or the firearm is used to aid law enforcement (Giffords Law Center to Prevent Gun Violence 2021a). Additionally, the definition of a “minor” varies from state to state (Giffords Law Center to Prevent Gun Violence 2018a).¹¹ For evidence on individuals being charged with unsafe gun storage in CAP law states, see Borden (1995), James (1996), “Parents Charged” (2009, 2017), Associated Press (2010), Young (2012), Harmacinski (2013), Ly (2013), Amaral (2014), Lopez and Goff (2014), Angst (2016), Bell (2016), Cutts and Majchrowicz (2016), Spies (2016), Boren (2017), “Father Charged” (2017), Stevens (2017), City News Service (2019), Rabin (2019), Press Staff (2020), Raddatz (2020), and Aliyu (2021). A recent review of cases in which children under the age of 12 either shot and killed themselves or were shot and killed by another child found that approximately half of the deaths resulted in a criminal charge. If the parent involved was a felon, the case almost always resulted in a criminal charge (Penzenstadler et al. 2017).

Consistent with the hypothesis that CAP laws raise the cost of unsafely storing firearms in households with minors, there is evidence that these laws reduce youth firearm access. For instance, Lott and Whitley (2001) and Prickett et al. (2014) found that CAP laws were associated with higher rates of safe-gun storage. Yet, because both of these studies are based on cross-sectional variation in CAP laws, they should be interpreted as descriptive.¹² More recent research, however, exploits within-state policy variation to causally identify CAP law effects. Using data from the Youth Risk Behavior Surveys for the period 1993–2013, Anderson and Sabia (2018) explored the relationship between CAP laws and gun carrying among high school students under the age of 18. Their results suggest that CAP laws were associated with an almost 20 percent decrease in the rate of past-month gun carrying, and these effects were driven by states that enforce a negligent storage standard. To our knowledge, these findings represent the best “first-stage” evidence that CAP laws reduce gun carrying among youths.

Despite the results from the studies listed above, there has been little discussion on the particular salience of CAP laws. As a simple test of salience, we collected Google trends data to explore whether searches on “gun safe” spiked in the Seattle-Tacoma metropolitan area after Seattle passed their recent safe-storage bill (Figure 1). On July 10, 2018, the day immediately after the bill passed, the volume of searches on “gun safe” hit its peak during the period June 15, 2018 through August 15, 2018. There was no spike at the same time in Portland, a reasonable counterfactual city (Appendix Figure 2). At a minimum, these results suggest that the population covered by Seattle’s law is aware of the policy change and may be taking steps to comply.¹³

Figure 1. Google Searches for “Gun Safe” in Seattle-Tacoma, Washington Pre- and Post-Seattle CAP Law.

Notes: On July 9, 2018, Seattle, Washington passed a bill mandating the safe storage of firearms. The bill created civil infractions for failing to safely store a gun when the owner should reasonably know that the gun could be accessed by a minor (Groover 2018). Google search data are from the Seattle-Tacoma metropolitan area. Values on the vertical axis represent Google search interest relative to the highest point during the period June 15, 2018 to August 15, 2018. A value equal to 100 indicates peak search volume, whereas a value of 50 indicates a day where the term was half as popular. A value of 0 means there were not enough data for this day.

Only a handful of previous studies have explored the relationship between CAP laws and some form of violent crime. Using data from the Compressed Mortality Files of the National Center for Health Statistics for the period 1979–1994, Cummings et al. (1997) found no evidence that CAP laws deter gun-related homicides among victims under the age of 15. Using data from the FBI’s Uniform Crime Reports (UCR) for the periods 1979–1996 and 1977–1998, Lott and Whitley (2001) and Lott (2003), respectively, found that CAP laws were associated with increases in homicides, rapes, robberies, and burglaries. However, Pepper (2005) showed that Lott’s results are sensitive to model specification and that some of the reported estimates are not replicable.¹⁴ Using hospital discharge data from 11 states for the period 1988–2003, DeSimone et al. (2013) found that CAP laws were associated with a 5 percent reduction in non-self-inflicted gun injuries, which included injuries from assaults. When injuries from assaults were considered as a separate outcome, the estimated CAP law coefficient was not consistently statistically significant across model specifications. Finally, Anderson and Sabia (2018, p. 517) assembled the first comprehensive data set of school-associated shooting deaths in the United States and concluded that CAP laws “do not have an observable impact” on juvenile firearm-related homicides that are committed on school property.¹⁵ In summary, no study of which we are aware has found that CAP laws generate a statistically significant reduction in gun crime, a conclusion that may undermine the efficacy of these laws.¹⁶

Our research extends the literature in at least four important ways. This study is the first to estimate the effects of CAP laws on firearm-related homicides committed by juveniles, a contribution made possible because the SHR data include information on the age of the offender and whether a firearm was used in the commission of the crime. Information on the age of the offender was unavailable in the data used by Cummings et al. (1997), Lott and Whitely (2001), and Lott (2003), preventing these authors from estimating the juvenile-gun-crime effects of a policy that targets households with minors. By using data on homicides committed by offenders of all ages, one could easily fail to detect an effect that is concentrated among minors. Indeed, none of these studies found that CAP laws were associated with fewer homicides.¹⁷ Consequently, our study provides the strongest evidence to date that CAP laws improve public safety by reducing violent crime.

Second, given the sample time frame under study, we exploit a considerable amount of CAP law variation relative to previous research. For instance, Cummings et al. (1997), Lott and Whitley (2001), Lott (2003), and DeSimone et al. (2013) observed pre- and post-treatment data for 12, 15, 16, and 8 states, respectively. We observe pre- and post-treatment data for 26 states and the District of Columbia. Third, given prevailing racial disparities in gun violence, it is important to consider heterogeneous effects by race (Bindu et al. 2018). To our knowledge, no previous studies on CAP laws have estimated effects separately for whites versus nonwhites. Finally, because these studies predate the recent uptick in youth gun violence, a fresh investigation is needed.

3. Data and Empirical Framework

State-level homicide data come from the FBI’s Supplementary Homicide Reports (SHR) for the period 1985–2013 (Puzzanchera et al. 2020).¹⁸ All 50 states and the District of Columbia contribute observations to the analysis. The SHR data are part of the Uniform Crime Reporting (UCR) program and are based on information from law enforcement agencies that is compiled by state authorities and forwarded to the FBI. Unlike the standard data made available by the UCR, the SHR data provide details on each incident, such as offender demographics and whether a firearm was used in the commission of the crime (U.S. Department of Justice 2014).¹⁹

To explore the relationship between CAP laws and juvenile firearm-related homicides, we estimate a Poisson regression that takes the following form:

$$ln\mathit{\text{Juvenile Firearm Homicide}}{s}_{st}={\beta }_0+{\beta }_{1}CAPLa{w}_{st}+X_{st}{\beta }_2+v_s+w_t+r_s•w_t,$$ (1)

where Juvenile Firearm Homicides_st represents the expected number of firearm-related homicides committed by under-18-year-olds in state s during year t.²⁰ The natural logarithm of the state population of under-18-year-olds is used as an offset variable.²¹ One advantage of the Poisson is that it accommodates values equal to 0.²²

The independent variable of interest, CAP Law_st, is equal to 1 if state s was enforcing a CAP law during year t, and equal to 0 otherwise.²³ The vector X_st includes state-level controls for demographics (% Nonwhite, % Under 18, % Male), economic conditions (Unemployment Rate, Per Capita Income), policing resources (Police Expenditures), political preferences (Democrat), mental health coverage (Mental Health Parity Law), and other gun laws (Shall Issue Law, Stand Your Ground Law, Minimum Possession Age, Background Check Law, Trigger Lock Law).²⁴ Table 2 provides weighted means and definitions for the variables in X_st.²⁵ The vectors v_s and w_t represent state and year fixed effects, respectively. Importantly, the year fixed effects control for the strong secular trends in homicide that occurred during the sample period (Cooper and Smith 2011). Following Cheng and Hoekstra (2013), we also include region-by-year fixed effects, denoted by r_s • w_t.²⁶ These allow us to control for differential shocks by region over time. Lastly, in most specifications, we include state-specific linear time trends to control for state-level unobservables that evolve smoothly over time, such as attitudes towards gun control. Standard errors are corrected for clustering at the state level (Bertrand et al. 2004).²⁷

Table 2.

Descriptive Statistics for Juvenile Firearm-Homicides and CAP Law Analysis, 1985–2013

	CAP Law = 1a	CAP Law = 0	Full Sample	Description
Juvenile Firearm Homicides	57.4 (77.1)	49.8 (70.1)	53.4 (73.6)	Number of firearm-related homicides committed by under-18-year-olds
Independent variables
% Nonwhite	0.198 (0.089)	0.169 (0.084)	0.183 (0.088)	Percent of the state population that is nonwhite
% Under 18	0.256 (0.019)	0.254 (0.020)	0.255 (0.020)	Percent of the state population that is under 18 years of age
% Male	0.492 (0.006)	0.488 (0.007)	0.490 (0.007)	Percent of the state population that is male
Unemployment Rate	0.130 (0.036)	0.129 (0.031)	0.130 (0.034)	State youth unemployment rate
Per Capita Income	39,213 (6,145)	34,561 (5,914)	36,772 (6,455)	State real income per capita (2010 dollars)
Police Expenditures	270 (73.9)	230 (80.0)	240 (80.0)	State police expenditures per capita (2010 dollars)
Democrat	0.417 (0.490)	0.498 (0.497)	0.460 (0.495)	= 1 if state has a Democratic governor, = 0 otherwise
Mental Health Parity Law	0.547 (0.494)	0.183 (0.385)	0.356 (0.476)	= 1 if state has a mental health parity law, = 0 otherwise
Shall Issue Law	0.507 (0.500)	0.375 (0.484)	0.438 (0.496)	= 1 if state has a shall issue gun law, = 0 otherwise
Stand Your Ground Law	0.174 (0.373)	0.090 (0.280)	0.130 (0.330)	= 1 if state has a Stand Your Ground gun law, = 0 otherwise
Minimum Possession Age	0.958 (0.191)	0.594 (0.485)	0.767 (0.417)	State minimum age requirement to possess a handgun
Background Check Law	0.535 (0.499)	0.355 (0.479)	0.441 (0.497)	= 1 if state requires background checks for private sales on firearms, = 0 otherwise
Trigger Lock Law	0.229 (0.420)	0.000 (0.000)	0.109 (0.311)	= 1 if state requires trigger locks to accompany dealer and private firearm sales, = 0 otherwise
N	529	853	1,382

^aIf a CAP law is in effect for any portion of the year, the observation is included in this column.

Notes: Means are weighted and standard deviations are in parentheses.

Beyond controlling for the right-hand side variables listed above, we take four approaches to address the potential endogeneity of CAP laws. First, we estimate event-study regressions to investigate whether juvenile firearm-related homicides were trending systematically prior to CAP law adoption. Second, we conduct falsification-type tests where we consider firearm-related homicides committed by 18+ year-olds and juvenile non-firearm-related homicides as outcomes. While spillovers across age groups and substitution towards non-firearm weapons may exist, we would be worried that large estimated effects for these outcomes simply reflect unobserved and confounding state-level efforts to curb general gun violence and youth crime. Third, we explore the sensitivity of β₁ to a range of alternative specifications, including controlling for a state’s underlying general crime trend. Finally, we assess how much of our identifying variation comes from using always-treated or early-adopting CAP law states as counterfactuals (Goodman-Bacon 2021).

4. Results

The baseline results of our analysis are presented in Table 3. The estimate of β₁ reported in column (1) comes from a model that does not control for any of the state-level covariates listed in Table 2. It suggests that CAP laws are associated with a 24 (e^−0.272 – 1 = −0.238) percent reduction in firearm-related homicides committed by juveniles. Including state-level demographics, economic conditions, political and mental health controls, and other gun policies on the right-hand side reduces the size of the estimated coefficient on CAP Law by 8.2 log points.²⁸ Controlling for state-specific linear time trends reduces the estimate of β₁ further, but by only 0.6 log points.²⁹ Specifically, the estimate reported in column (4) suggests that CAP laws lead to a 17 percent reduction in juvenile firearm homicides. While we do not observe every policy or state-level characteristic that may be simultaneously correlated with our outcome of interest and CAP laws, the stability of the estimates in columns (1) through (4) is reassuring.³⁰

Table 3.

Juvenile Firearm-Related Homicides and CAP Laws

	(1)	(2)	(3)	(4)
	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides
CAP Law	−0.272** (0.131)	−0.195* (0.116)	−0.190* (0.112)	−0.184** (0.081)
Mean	53.4	53.4	53.4	53.4
N	1,382	1,382	1,382	1,382
Demographic, economic, political, and mental health controls	No	Yes	Yes	Yes
Other gun laws	No	No	Yes	Yes
State-specific linear trends	No	No	No	Yes

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate Poisson regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Weighted means for the dependent variable are reported. Demographic, economic, political, and mental health controls: % Nonwhite, % Under 18, % Male, Unemployment Rate, Per Capita Income, Police Expenditures, Democrat, and Mental Health Parity Law. Other gun laws: Shall Issue Law, Stand Your Ground Law, Minimum Possession Age, Background Check Law, and Trigger Lock Law. All models control for state fixed effects, year fixed effects, and region-by-year fixed effects. Standard errors, corrected for clustering at the state level, are in parentheses.

In column (1) of Table 4, we investigate the parallel trends assumption by adding a lead on CAP Law to the model, equal to 1 if a CAP law was passed in year t + 1, and equal to 0 otherwise. The estimated coefficient on the lead is small, positive, and nowhere near statistically significant. In columns (2) and (3) of Table 4, we add a series of leads to the model. They are, without exception, statistically indistinguishable from zero. Importantly, we observe no clear systematic trend in juvenile firearm-related homicides leading up to the passage of CAP laws, providing further evidence that the parallel trends assumption is satisfied.

Table 4.

Leads and Lags of CAP Law

	(1)	(2)	(3)	(4)	(5)
	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides
5 Years Prior to CAP Law	…	…	…	…	0.064 (0.057)
4 Years Prior to CAP Law	…	…	…	…	0.113 (0.077)
3 Years Prior to CAP Law	…	…	0.066 (0.057)	0.042 (0.067)	0.119 (0.090)
2 Years Prior to CAP Law	…	0.001 (0.068)	0.025 (0.077)	0.0003 (0.094)	0.081 (0.121)
Year Prior to CAP Law	0.015 (0.058)	0.015 (0.071)	0.040 (0.084)	0.006 (0.098)	0.094 (0.120)
CAP Law	−0.178** (0.087)	−0.177* (0.092)	−0.148 (0.097)	…	…
Year of Law Change	…	…	…	−0.071 (0.111)	0.018 (0.136)
1 Year After CAP Law	…	…	…	−0.063 (0.110)	0.031 (0.130)
2 Years After CAP Law	…	…	…	−0.301** (0.133)	−0.198 (0.145)
3 Years After CAP Law	…	…	…	…	−0.194 (0.145)
3+ Years After CAP Law	…	…	…	−0.310** (0.134)	…
4 Years After CAP Law	…	…	…	…	−0.223 (0.162)
5+ Years After CAP Law	…	…	…	…	−0.158 (0.187)
p-value (joint significance of lags)	…	…	…	0.004	0.002
Mean	53.4	53.4	53.4	53.4	53.4
N	1,382	1,382	1,382	1,382	1,382

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate Poisson regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Weighted means for the dependent variable are reported. All models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors, corrected for clustering at the state level, are in parentheses.

Next, in column (4) of Table 4, we replace CAP Law with an indicator that is equal to 1 the year in which a CAP law went into effect, 3 leads of this indicator, and 3 lags. In column (5), we consider 5 leads and 5 lags.³¹ Again, there is no evidence that juvenile firearm-related homicides began trending prior to the adoption of CAP laws. In addition, we observe that the estimated effect of CAP laws appears 2 years after the law change and is relatively constant thereafter. The lagged effect is consistent with the notion that it may take some time for the laws to become salient and for gun owners to adjust their behavior (Gift 2019). Figures 2a and 2b plot the estimates shown in columns (4) and (5), respectively.

Figure 2a. Pre- and Post-CAP Law Trends in Juvenile Firearm-Related Homicides.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) are reported, where the omitted category is four or more years before treatment. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

Figure 2b. Pre- and Post-CAP Law Trends in Juvenile Firearm-Related Homicides.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) are reported, where the omitted category is six or more years before treatment. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

4.1. Adult Firearm-Related Homicides, Juvenile Non-Firearm-Related Homicides, and Firearm-Related Homicides Committed by an Unknown Offender

In the first two columns of Table 5, we replace juvenile firearm-related homicides with firearm-related homicides committed by adults. Specifically, in column (1), we consider the number of firearm-related homicides committed by 18+ year-olds. In the second column, we restrict this age range and consider the number of firearm-related homicides committed by 18- to 24-year-olds. Because these laws may have spillover effects across individuals within households (e.g., siblings), we are hesitant to refer to these as true falsification tests. However, we do expect CAP laws to bind less for these older age groups. The estimated coefficients indeed suggest this is the case, as both are small in magnitude and statistically insignificant.³² Moreover, an event-study analysis of adult firearm-related homicides (Appendix Figure 4) shows little evidence of pre-CAP law differences between treatment and control states, suggesting that CAP laws were not simply passed in the midst of downward trends in firearm-related homicide rates or as a reactionary response to increasing gun violence. Relatedly, Anderson and Sabia (2018) found no evidence to suggest that CAP laws were passed in response to general trends in violent crime or school shooting events where a death was involved.³³

Table 5.

Adult Firearm-Related Homicides, Juvenile Non-Firearm-Related Homicides, and Firearm-Related Homicides where Offender is Unknown

	(1)	(2)	(3)	(4)
	Adult Firearm Homicides (18+ year-olds)	Adult Firearm Homicides (18- to 24-year-olds)	Juvenile Non-Firearm Homicides	Firearm Homicides, Offender Unknown
CAP Law	0.011 (0.042)	−0.046 (0.054)	−0.001 (0.068)	0.030 (0.091)
Mean	352.0	170.9	18.5	222.2
N	1,382	1,382	1,382	1,382

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate Poisson regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. The dependent variable is equal to the number of specified homicides in state s during year t. Weighted means for the dependent variable are reported. All models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors, corrected for clustering at the state level, are in parentheses.

In the third column of Table 5, we consider the relationship between juvenile non-firearm-related homicides and CAP laws. If CAP laws are associated with large reductions in juvenile non-firearm-related homicides, we would be worried that the estimates in Table 3 simply reflect an unobserved and confounding factor. This turns out to not be the case, as the estimated coefficient on CAP Law is small in magnitude and statistically indistinguishable from zero. This result also suggests that juveniles do not turn to weapons other than firearms to commit homicides after a CAP law goes into effect.

Approximately 30 percent of SHR cases have an unknown offender (U.S. Department of Justice 2014). To the extent that state-level rates of missing information on the offender are correlated with CAP laws, our results could be biased. To address this issue, we regress the number of firearm-related homicides where information on the offender is unknown on CAP Law and the full set of controls. The estimated coefficient in the final column of Table 5 suggests this type of measurement error is not systematic to CAP laws, as it is small in magnitude and statistically insignificant. In general, the Table 5 results support the notion that the observed relationship between CAP laws and juvenile firearm-related homicides is causal.

4.2. Heterogeneous Effects

We explore heterogeneous effects in Table 6. In columns (1) and (2), we consider firearm-related homicides committed by white and nonwhite juveniles, respectively.³⁴ The estimated effect for whites indicates that CAP laws are associated with a 23 percent decrease in firearm-related homicides, while the estimate for nonwhites indicates an 11 percent decrease. The latter estimate, however, is not statistically significant at conventional levels (p-value = 0.160). Given that white Americans owned guns at significantly higher rates than nonwhites throughout the sample period, this pattern of results is perhaps not surprising (Cook and Ludwig 1997; Smith and Son 2015; Parker et al. 2017).³⁵ It is also possible that disparities in institutional trust (which may affect legal compliance) or information availability are responsible for the observed differences by race.³⁶ These explanations are, however, admittedly speculative. When we restrict our attention to male juvenile offenders (column (3)), the estimated coefficient suggests that CAP laws lead to a 17 percent decrease in firearm-related homicides.³⁷

Table 6.

Heterogeneity

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
	White Juvenile Firearm Homicides	Nonwhite Juvenile Firearm Homicides	Male Juvenile Firearm Homicides	Juvenile Firearm Homicides, Excluding Long Guns	Juvenile Firearm Homicides, Excluding Multiple-Victim Events	Juvenile Firearm Homicides, Excluding School Shootings	Juvenile Firearm Homicides, Victim a Stranger	Juvenile Firearm Homicides, Known Victim	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides
CAP Law	−0.267** (0.110)	−0.121 (0.086)	−0.181** (0.077)	−0.168** (0.083)	−0.182** (0.080)	−0.179** (0.081)	−0.369*** (0.124)	0.008 (0.091)	…	−0.193** (0.082)	…
CAP Law × pre-1995	…	…	…	…	…	…	…	…	−0.110 (0.081)	…	…
CAP Law × post-1995	…	…	…	…	…	…	…	…	−0.269** (0.105)	…	…
CAP Law × Democrat	…	…	…	…	…	…	…	…	…	0.020 (0.059)	…
Negligent Storage	…	…	…	…	…	…	…	…	…	…	−0.298*** (0.091)
Reckless Endangerment	…	…	…	…	…	…	…	…	…	…	−0.011 (0.110)
Mean	21.8	31.0	51.0	49.7	51.8	53.3	18.8	26.9	53.4	53.4	53.4
N	1,382	1,382	1,382	1,145	1,382	1,382	1,382	1,382	1,382	1,382	1,382

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate Poisson regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. The dependent variable is equal to the number of specified homicides in state s during year t. Weighted means for the dependent variable are reported. All models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors, corrected for clustering at the state level, are in parentheses.

In column (4) of Table 6, we exclude homicides that were committed with a long gun (e.g., a long rifle or a shot gun). While homicides committed with a long gun represent less than 15 percent of all juvenile firearm homicides in the data, they are often the weapon of choice in high-profile events, such as school shootings.³⁸ The estimated coefficient in column (4) indicates that long gun-related homicides are not driving the observed relationship between CAP laws and total juvenile firearm-related homicides. In columns (5) and (6), we exclude multiple-victim events and school-associated shooting deaths, respectively, from the juvenile firearm homicide count.³⁹ The estimated coefficient on CAP Law changes little when considering either of these alternative definitions of the dependent variable, indicating that neither mass nor school shootings are particularly important for the observed reduction in juvenile firearm-related homicides.⁴⁰ In sum, the results in columns (4) through (6) indicate that CAP laws are successful at deterring the most common type of juvenile firearm-related homicides (i.e., single-victim events where a firearm other than a long gun was used to commit the crime). We also see that the observed effect is concentrated among homicides where the offender and victim were categorized as strangers, rather than acquaintances or family members (columns (7) and (8)).⁴¹

In column (9) of Table 6, we interact CAP Law with pre- and post-1995 indicators.⁴² Cummings et al. (1997) focused on the period 1979 through 1994 and found that CAP laws were associated with a (statistically insignificant) decrease in gun-related homicides among victims under the age of 15. Our results are similar to those in Cummings et al. (1997) despite our focus on the age of the offender, rather than the age of the victim.⁴³ Specifically, we find that CAP laws are associated with a (statistically insignificant) 10 percent decrease in juvenile firearm homicides for the pre-1995 period. For the post-1995 period, the estimated effect is larger in magnitude and statistically significant at the 5 percent level.⁴⁴

Next, we test for heterogeneous effects by populations that are perhaps less likely to be aware of the existing gun legislation in their state. Specifically, we interact CAP Law with Democrat, which is equal 1 if state s had a Democratic governor during year t. According to national polls conducted by the Pew Research Center, Republicans are “much more focused on gun rights” than their Democratic counterparts, and this gap has grown over time (Bacon, Jr. 2019). The estimated coefficient on this interaction term is small, positive, and statistically indistinguishable from zero.⁴⁵

In the final column of Table 6, we replace CAP Law with two mutually exclusive indicators, Negligent Storage and Reckless Endangerment. As mentioned above, negligent storage laws are the stricter form of CAP legislation and impose criminal liability on individuals who allow a minor access to a firearm that was not properly stored. On the other hand, reckless endangerment laws only impose criminal liability when an individual “intentionally, knowingly, and/or recklessly” provides a firearm to a minor (Giffords Law Center to Prevent Gun Violence 2018a). The results in column (11) suggest that the observed CAP law effects are driven by the stricter negligent storage laws. These results are consistent with those in Anderson and Sabia (2018), who found that negligent storage laws were more effective than reckless endangerment laws at reducing gun carrying among minors.⁴⁶ Figure 3 shows no evidence that juvenile firearm-related homicides began trending prior to the adoption of negligent storage laws.

Figure 3. Pre- and Post-Negligent Storage Law Trends in Juvenile Firearm-Related Homicides.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) are reported, where the omitted category is six or more years before treatment. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, Reckless Endangerment, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

4.3. Robustness Checks

We report the results of various robustness checks in Table 7. In the first column, we list our preferred estimate from column (4) in Table 3 for comparison. In the second and third columns, we drop states with 10 or more and 5 or more missing years of data, respectively.⁴⁷ Some states did not report any data to the SHR program in some years, while other state-year cells are so severely underreported that they have been made unavailable by the Office of Juvenile Justice and Delinquency Prevention.⁴⁸ When dropping states with missing years of data the estimated coefficient on CAP Law changes little in magnitude and remains statistically significant at the 5 percent level.

Table 7.

Robustness of Relationship between Juvenile Firearm-Related Homicides and CAP Laws

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Estimate from column (4), Table 3 for comparison	Drop states with 10+ years of missing data	Drop states with 5+ years of missing data	Drop region-by-year fixed effects	Control for adult property crime rate	Unweighted	OLS (levels)	OLS (log)
CAP Law	−0.184** (0.081)	−0.173** (0.083)	−0.181** (0.083)	−0.167* (0.087)	−0.176** (0.081)	−0.184** (0.081)	−0.558* (0.326)	−0.089 (0.074)
Mean	53.4	53.8	54.7	53.4	53.4	20.9	1.52	2.52
N	1,382	1,328	1,287	1,382	1,382	1,382	1,382	1,382

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. In columns (1)–(6), the dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. In column (7), the dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds per 100,000 population of this age group in state s during year t. In column (8), the dependent variable is equal to the natural log of the number of firearm-related homicides committed by under-18-year-olds per 100,000 population of this age group in state s during year t plus 1. Unless stated otherwise, weighted means for the dependent variable are reported. Unless stated otherwise, all models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors, corrected for clustering at the state level, are in parentheses.

Next, we drop the region-by-year fixed effects from the right-hand-side of the estimating equation. The estimated coefficient from this exercise suggests that CAP laws are associated with a 15 percent decrease in juvenile firearm-related homicides. In column (5), we include the adult property crime rate on the right-hand side of the estimating equation. The adult property crime rate is unlikely to be affected by CAP laws, but should capture a state’s general crime trend.⁴⁹ In column (6), we consider unweighted estimates by not controlling for the population exposure variable. The estimated coefficient on CAP Law changes little under these alternative specifications.

Finally, we estimate equation (1) with OLS rather than modeling homicides as a count process. In column (7), we consider a level-level specification, while results from a log-level specification are shown in column (8). The estimated coefficient on CAP Law is negative, large in magnitude, and statistically significant at the 10 percent level under the level-level specification, but loses precision when the dependent variable is defined as the natural log of the juvenile firearm homicide rate.⁵⁰

In Table 8, we repeat the robustness checks listed above to examine the sensitivity of the Negligent Storage estimated reported in the final column of Table 6. In general, the estimated coefficient on Negligent Storage is quite robust across the alternative specifications under consideration. Again, there is little evidence to suggest that the weaker reckless endangerment laws are effective at reducing gun violence among juveniles.

Table 8.

Robustness of Relationship between Juvenile Firearm-Related Homicides and Negligent Storage Laws

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Estimates from column (8), Table 6 for comparison	Drop states with 10+ years of missing data	Drop states with 5+ years of missing data	Drop region-by-year fixed effects	Control for adult property crime rate	Unweighted	OLS (levels)	OLS (log)
Negligent Storage	−0.298*** (0.091)	−0.282*** (0.092)	−0.278*** (0.090)	−0.246*** (0.092)	−0.284*** (0.086)	−0.302*** (0.092)	−0.829** (0.388)	−0.151* (0.086)
Reckless Endangerment	−0.011 (0.110)	−0.011 (0.113)	−0.034 (0.113)	−0.009 (0.113)	−0.015 (0.110)	−0.005 (0.109)	−0.045 (0.305)	0.030 (0.090)
Mean	53.4	53.8	54.7	53.4	53.4	20.9	1.52	2.52
N	1,382	1,328	1,287	1,382	1,382	1,382	1,382	1,382

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. In columns (1)–(6), the dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. In column (7), the dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds per 100,000 population of this age group in state s during year t. In column (8), the dependent variable is equal to the natural log of the number of firearm-related homicides committed by under-18-year-olds per 100,000 population of this age group in state s during year t plus 1. Unless stated otherwise, weighted means for the dependent variable are reported. Unless stated otherwise, all models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors, corrected for clustering at the state level, are in parentheses.

In Figure 4, we assess the robustness of the estimated coefficient on CAP Law to dropping one CAP law state at a time. The effect sizes range from −6.3 log points when we drop California to −22.6 log points when we drop Indiana. As indicated above, populous states enforcing a negligent storage standard (e.g., California, Illinois, and Texas) contribute important weight to the estimated effect of CAP laws.⁵¹ We repeat this exercise in Figure 5 to examine the robustness of the estimated coefficient on Negligent Storage. Here, the estimated effects range from −18.0 log points when we drop California to −35.4 log points when we drop New Jersey. In all cases, the estimated coefficient on Negligent Storage is statistically significant at the 5 percent level.⁵²

Figure 4. Robustness of Estimated Coefficient on CAP Law to Dropping One CAP Law State at a Time.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) come from separate regressions where one CAP law state is dropped at a time. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

Figure 5. Robustness of Estimated Coefficient on Negligent Storage to Dropping One Negligent Storage Law State at a Time.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) come from separate regressions where one negligent storage state is dropped at a time. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, Reckless Endangerment, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

To conclude this section, we consider a diagnostic test recently developed by Goodman-Bacon (2021), who showed that the standard two-way fixed effects estimator can be decomposed into a weighted average of all potential 2×2 difference-in-differences (DD) estimates where the weights are based on treatment group size and variance in treatment (Cunningham 2021). If treatment effects are heterogeneous over time, then comparing early-adopting CAP law states to late-adopting CAP law states can lead to biased estimates. Encouragingly, most of our identifying variation comes from comparisons between treated states and states that never pass a CAP law (Appendix Table 5).⁵³

4.4. Background Check Analysis

While CAP laws may affect juvenile firearm-related homicides by encouraging safe storage, they also increase the costs of firearm ownership. Disentangling these mechanisms is important from a policy perspective because most Americans support safe storage, yet gun lobby groups and their supporters generally oppose laws that make owning a firearm more difficult (BBC 2020, Dunn 2020). Consequently, one can imagine stronger bipartisan support for a policy that encourages gun safety but does not decrease gun ownership. Alternatively, if CAP laws reduce firearm ownership, the evidence presented above would lend support to the general “less guns-less crime” hypothesis.

As mentioned, little is known about the relationship between CAP laws and the safe storage of firearms due to the lack of state panel data on household gun storage behavior. However, it is possible to indirectly estimate the relationship between CAP laws and rates of gun ownership by using data on firearm background checks. Following Lang (2013a, 2016) and using data from the FBI’s National Instant Check System for the period 1999–2013, we estimate an OLS regression where the dependent variable is defined as the natural log of the firearm background check rate per 100,000 population in state s during year t.⁵⁴ The right-hand side of the estimating equation is identical to that of equation (1). The results in Table 9 show that CAP laws are associated with a 9.3 percent decrease in the total background check rate. While this estimate is statistically insignificant at conventional levels (p-value = 0.149), the sign of the relationship is consistent with the argument that CAP laws increase the costs of owning a firearm. The size of the effect is larger in absolute magnitude if the analysis is restricted to background checks for handguns only (column (2)) or long guns only (column (3)).⁵⁵ However, these estimates are also statistically insignificant.

Table 9.

Background Checks and CAP Laws

	(1)	(2)	(3)
	Total Background Checks	Background Checks for Handguns	Background Checks for Long Guns
CAP Law	−0.098 (0.067)	−0.355 (0.243)	−0.186 (0.149)
Mean	4,612	1,115	2,119
N	765	765	765

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate OLS regression based on data from the FBI’s National Instant Check System for the period 1999–2013. The dependent variable is equal to the natural log of the number of specified background checks per 100,000 population in state s during year t. Weighted means for the dependent variable are reported. All models control for the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Regressions are weighted by state populations. Standard errors, corrected for clustering at the state level, are in parentheses.

5. Conclusion

While the majority of gun owners in the United States do not safely store all of their firearms (Crifasi et al. 2018), we know very little about the causal effects of gun-storage laws on gun violence. This policy question has taken on increased salience in the wake of several high-profile school shootings carried out by minors who obtained their guns from home (or the home of a relative). To better understand how safe-storage laws affect gun crime, the current study exploits state-level variation in safe-storage requirements. Specifically, using data from the FBI’s Supplementary Homicide Reports for the period 1985–2013, we examine the relationship between child access prevention laws and firearm-related homicides committed by juveniles.

Our results suggest that CAP laws lead to a 17 percent reduction in the expected number of firearm-related homicides committed by juveniles (i.e., under-18-year-olds), an effect size that is similar to those reported in previous studies on unintentional shooting deaths (Cummings et al. 1997), self-inflicted gun injuries (DeSimone et al. 2013), and youth suicides (Gius 2015).⁵⁶ The estimated effect is stronger for whites, as opposed to nonwhites, and is driven by states enforcing a negligent storage standard, the strictest form of CAP legislation. Negligent storage laws impose criminal liability on individuals who allow a minor access to a firearm that was not properly stored.

We also find that CAP laws are not associated with firearm-related homicides committed by adults or with non-firearm-related homicides committed by juveniles, providing evidence that the observed relationship between CAP laws and juvenile firearm-related homicides is not simply being driven by confounding trends in gun crime or juvenile violence.⁵⁷

Although the welfare gains attributable to CAP laws are difficult to gauge, their benefits to society can be calculated by combining the estimates reported in Table 3 with cost-of-crime figures published by McCollister et al. (2010). For states with CAP laws, our results suggest that the annual value of homicides avoided ranges between $41 million and $113 million (in 2021 dollars).⁵⁸

From a policy perspective, understanding the effects of CAP laws is vital as youth gun violence rises alongside public support for gun control (Parsons et al. 2018; Clement 2018). The juvenile homicide arrest rate has been increasing since 2012, while support for gun restrictions among Americans recently reached its highest point in the last 25 years (Clement 2018; Office of Juvenile Justice and Delinquency Prevention 2018). We view the results above as the most credible estimates of the relationship between CAP laws and youth gun violence, and the strongest evidence to date that CAP laws reduce juvenile firearm-related homicides. Because the majority of gun owners support safe-storage requirements for guns in households with children (Barry et al. 2018), our findings are particularly relevant for policymakers who are looking to find a middle ground on gun control.

Appendix

Appendix Figure 1.

Child Access Prevention Laws Over Time

Appendix Figure 2. Google Searches for “Gun Safe” in Portland, Oregon Pre- and Post-Seattle CAP Law.

Notes: On July 9, 2018, Seattle, Washington passed a bill mandating the safe storage of firearms. The bill created civil infractions for failing to safely store a gun when the owner should reasonably know that the gun could be accessed by a minor (Groover 2018). Google search data are from Portland, Oregon. Values on the vertical axis represent Google search interest relative to the highest point during the period June 15, 2018 to August 15, 2018. A value equal to 100 indicates peak search volume, whereas a value of 50 indicates a day where the term was half as popular. A value of 0 means there were not enough data for this day.

Appendix Figure 3. Distribution of Juvenile Firearm-Related Homicides.

Notes: Juvenile firearm-related homicide counts are measured at the state-year level. Weighted by state populations of under-18-year-olds.

Appendix Figure 4. Pre- and Post-CAP Law Trends in Adult Firearm-Related Homicides.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) are reported, where the omitted category is six or more years before treatment. The dependent variable is equal to the number of firearm-related homicides committed by 18+ year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

Appendix Figure 5. Robustness of Estimated Coefficient on CAP Lam to Dropping One Year at a Time.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) come from separate regressions where one year is dropped at a time. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

Appendix Figure 6. Robustness of Estimated Coefficient on Negligent Storage to Dropping One Year at a Time.

Notes: Poisson coefficient estimates (and their 90% confidence intervals) come from separate regressions where one year is dropped at a time. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Controls include the covariates listed in Table 2, state fixed effects, year fixed effects, region-by-year fixed effects, and state-specific linear time trends. Standard errors are corrected for clustering at the state level.

Appendix Table 1.

Data Sources for State-Level Covariates

	Data Source
% Nonwhite	National Cancer Institute’s SEER population data
% Under 18	National Cancer Institute’s SEER population data
% Male	National Cancer Institute’s SEER population data
Unemployment Rate	Bureau of Labor Statistics
Per Capita Income	Bureau of Economic Analysis
Police Expenditures	Bureau of Justice Statistics
Democrat	Authors’ own internet searches
Mental Health Parity Law	Lang (2013b)a and ParityTrack.org
Shall Issue Law	Grossman and Lee (2008), Donohue and Ayers (2009), Aneja et al. (2012), Hinkston (2012), United States Government Accountability Office (2012), Arnold (2015), and USA Carry (2015)
Stand Your Ground Law	McClellan and Tekin (2017)
Minimum Possession Age	Marvell (2001) and Gius (2015)
Background Check Law	Vernick and Hepburn (2003), Webster et al. (2014), and Giffords Law Center to Prevent Gun Violence (2018b)
Trigger Lock Law	Giffords Law Center to Prevent Gun Violence (2018c) and authors’ own searches of state legislative codes

^aUpdates to Lang (2013b) were provided via personal correspondence with the author.

Appendix Table 2.

Unweighted Means

	CAP Law = 1a	CAP Law = 0	Full Sample	Description
Juvenile Firearm Homicides	23.0 (40.9)	19.6 (38.6)	20.9 (39.5)	Number of firearm-related homicides committed by under-18-year-olds
Independent variables
% Nonwhite	0.202 (0.150)	0.148 (0.124)	0.169 (0.137)	Percent of the state population that is nonwhite
% Under 18	0.250 (0.022)	0.257 (0.025)	0.254 (0.024)	Percent of the state population that is under 18 years of age
% Male	0.491 (0.007)	0.491 (0.010)	0.491 (0.009)	Percent of the state population that is male
Unemployment Rate	0.122 (0.037)	0.125 (0.036)	0.124 (0.037)	State youth unemployment rate
Per Capita Income	39,086 (7,224)	33,405 (5,928)	35,580 (7,019)	State real income per capita (2010 dollars)
Police Expenditures	256 (87.3)	215 (91.9)	231 (92.2)	State police expenditures per capita (2010 dollars)
Democrat	0.468 (0.496)	0.519 (0.496)	0.499 (0.497)	= 1 if state has a Democratic governor, = 0 otherwise
Mental Health Parity Law	0.489 (0.498)	0.241 (0.426)	0.336 (0.470)	= 1 if state has a mental health parity law, = 0 otherwise
Shall Issue Law	0.548 (0.498)	0.498 (0.500)	0.517 (0.500)	= 1 if state has a shall issue gun law, = 0 otherwise
Stand Your Ground Law	0.167 (0.367)	0.117 (0.315)	0.136 (0.336)	= 1 if state has a Stand Your Ground gun law, = 0 otherwise
Minimum Possession Age	0.934 (0.235)	0.601 (0.484)	0.729 (0.438)	State minimum age requirement to possess a handgun
Background Check Law	0.457 (0.499)	0.184 (0.388)	0.289 (0.453)	= 1 if state requires background checks for private sales on firearms, = 0 otherwise
Trigger Lock Law	0.117 (0.322)	0.000 (0.000)	0.045 (0.207)	= 1 if state requires trigger locks to accompany dealer and private firearm sales, = 0 otherwise
N	529	853	1,382

^aIf a CAP law is in effect for any portion of the year, the observation is included in this column.

Notes: Means are unweighted and standard deviations are in parentheses.

Appendix Table 3.

Do Other Gun Control Policies Predict CAP Laws?

	(1)	(2)	(3)
	CAP Law	CAP Law	CAP Law
Shall Issue Law	−0.067 (0.115)	0.014 (0.096)	0.045 (0.063)
Stand Your Ground Law	0.062 (0.076)	0.134* (0.079)	0.199*** (0.060)
Minimum Possession Age	−0.094 (0.159)	−0.066 (0.140)	0.043 (0.111)
Background Check Law	0.148 (0.097)	0.079 (0.087)	0.119 (0.086)
Trigger Lock Law	0.324** (0.155)	0.115 (0.164)	−0.045 (0.162)
Mean	0.466	0.466	0.466
N	1,382	1,382	1,382
Demographic, economic, political, and mental health controls	No	Yes	Yes
State-specific linear trends	No	No	Yes

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate OLS regression based on data for the period 1985–2013. The dependent variable is equal to is equal to 1 if state s was enforcing a CAP law during year t, and equal to 0 otherwise. Weighted means for the dependent variable are reported. Demographic, economic, political, and mental health controls: % Nonwhite, % Under 18, % Male, Unemployment Rate, Per Capita Income, Police Expenditures, Democrat, and Mental Health Parity Law. All models control for state fixed effects, year fixed effects, and region-by-year fixed effects. Regressions are weighted by state populations. Standard errors, corrected for clustering at the state level, are in parentheses.

Appendix Table 4.

Juvenile Firearm-Related Homicides and Other Gun Control Policies

	(1)	(2)	(3)
	Juvenile Firearm Homicides	Juvenile Firearm Homicides	Juvenile Firearm Homicides
CAP Law	−0.242* (0.135)	−0.190* (0.112)	−0.184** (0.081)
Shall Issue Law	0.137 (0.114)	0.080 (0.101)	0.125** (0.061)
Stand Your Ground Law	−0.054 (0.136)	−0.036 (0.113)	0.153 (0.119)
Minimum Possession Age	−0.082 (0.149)	−0.074 (0.122)	−0.095 (0.137)
Background Check Law	0.055 (0.124)	0.116 (0.106)	0.006 (0.078)
Trigger Lock Law	−0.147 (0.118)	−0.003 (0.130)	−0.118 (0.143)
Mean	53.4	53.4	53.4
N	1,382	1,382	1,382
Demographic, economic, political, and mental health controls	No	Yes	Yes
State-specific linear trends	No	No	Yes

^*Statistically significant at 10% level;

^**at 5% level;

^***at 1% level.

Notes: Each column represents results from a separate Poisson regression based on data from the FBI’s Supplementary Homicide Reports for the period 1985–2013. The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds in state s during year t. Weighted means for the dependent variable are reported. Demographic, economic, political, and mental health controls: % Nonwhite, % Under 18, % Male, Unemployment Rate, Per Capita Income, Police Expenditures, Democrat, and Mental Health Parity Law. All models control for state fixed effects, year fixed effects, and region-by-year fixed effects. Standard errors, corrected for clustering at the state level, are in parentheses.

Appendix Table 5.

Bacon Decomposition

	Weight	Average DD estimate
Panel A. Juvenile Firearm-Related Homicides and CAP Laws
Earlier-treated states vs. Not-yet-treated states as controls	0.064	−1.112
Later-treated states vs. Earlier-treated states as controls	0.130	−0.740
Treated states vs. Never-treated states as controls	0.770	−0.265
Treated states vs. Always-treated states as controls	0.036	−0.446
Panel B. Juvenile Firearm-Related Homicides and Negligent Storage Laws
Earlier-treated states vs. Not-yet-treated states as controls	0.048	−1.858
Later-treated states vs. Earlier-treated states as controls	0.092	−1.221
Treated states vs. Never-treated states as controls	0.860	−0.552
Treated states vs. Always-treated states as controls	0.000	…

Notes: Each panel represents results from a Bacon Decomposition (Goodman-Bacon 2021), produced using the STATA command ddtiming (Goldring 2019). The dependent variable is equal to the number of firearm-related homicides committed by under-18-year-olds per 100,000 population of this age group in state s during year t. Both models control for state and year fixed effects. In panel B, states that passed a reckless endangerment law are excluded from the sample.