Do Gun‐Purchase Waiting Periods Save Lives?

ABSTRACT

We study the impact of gun‐purchase waiting periods on suicide rates using county‐level mortality data from 1991–2019. We find that waiting periods are associated with a reduction in both firearm and overall suicide rates of approximately 5% and 2%, respectively. Novelly, we find that counties that are within 50 miles from a state without a waiting period experience no statistically significant reduction in suicides. Our findings reveal that the decrease in suicides under a waiting period is driven by counties that are more than 50 miles from a non‐restricted neighboring state.

1. Introduction

In 2021 suicides in the United States exceeded 48,000 and were the country's second‐leading cause of death among adults under the age of 45. Of suicides, more than 50% involve the use of a firearm (Garnett and Curtin 2023). Firearm‐related suicide attempts are the most lethal form of suicide attempts, with over 90% of firearm‐related suicide attempts being fatal (Elnour and Harrison 2008; Shenassa et al. 2003; Miller et al. 2013). Although gun‐control policies are a widely divisive topic in the US, a Gallup poll in 2022 found that more than 75% Americans would be in favor of a 30‐day mandatory purchase delay ¹ for all firearms. However, any gun‐control measure that potentially infringes on Second Amendment rights and impedes commerce requires a demonstration of efficacy to be justified and to withstand court challenges (Winkler 2011). Thus, a natural question for policymakers is whether waiting periods are effective in reducing suicide rates.

The impact of access to firearms on gun violence has been studied for decades, although identifying causal effects has proven difficult due to a variety of selection issues. Recent works by Vitt et al. (2018) and Balestra (2018) provide evidence that access to firearms is associated with higher suicide rates. Whether public policy can separate individuals who are at risk of attempting suicide from firearms has also been subject to debate. Work by Leung et al. (2019) finds that the more gun laws a state has, the lower the suicide rate, with the association mainly concentrated among young men. Andrés and Hempstead (2011) provide evidence that laws that restrict overall access to guns may be more effective in reducing suicides compared to laws that aim to limit access for “high‐risk” individuals.

Although previous evidence shows that gun control laws, in general, are associated with lower rates of gun violence, waiting periods may be particularly effective in reducing suicide rates for three reasons. First, interviews with survivors have shown that the decision to attempt suicide is often impulsive, with the time between the decision and attempt often being a matter of minutes or hours (Deisenhammer et al. 2009; Paashaus et al. 2021). Therefore a waiting period law may induce a “cooling‐off” period, resulting in individuals changing their minds (Simon et al. 2001). Second, firearms are the most lethal means of attempting suicide (Simon et al. 2022; Miller et al. 2012). For reference, drug overdose is fatal in less than 5% of attempted suicides; therefore, even if all suicide attempts with a firearm switched to other methods, overall suicides might still decrease. Finally, waiting periods may act as a general impediment to the purchase of firearms, resulting in a smaller general stock of guns, a condition shown to be associated with lower suicide rates (Miller et al. 2002; Edwards et al. 2018).

In this paper, we exploit the variation in waiting periods across US states and we find that gun purchase waiting periods are negatively associated with both overall suicide rates and firearm‐related suicide rates. In particular, we find that waiting periods reduce firearm‐related suicide by 5%. The magnitude of the effect of waiting periods is found to be more than 50% greater than when county‐level data are aggregated to the state level.

We find that a distance of 1–50 miles to a state without a waiting period has no statistically significant impact on firearm‐related suicides. Further, we find that at a distance greater than 50 miles from a no‐wait purchase opportunity, the reduction in firearm‐related suicide resulting from a waiting period increases to 5.2% and becomes statistically significant, suggesting that a distance of 50+ miles results in a “binding” waiting period and is the main driver behind any observed reduction in suicides in previous studies, as in ours. This finding implies that there are “non‐binding” situations, which depend on the distance to a neighboring state without a waiting period.

One concern is that implementing waiting period laws will result in individuals substituting to other means. We do not find evidence of a substitution effect, and, in several specifications, we find that waiting periods are associated with declines in non‐firearm suicides, although the magnitudes of the impact are smaller than the impact on firearm‐related suicides. This finding could suggest that suicides are socially “contagious.” There is evidence that suicides are socially contagious, using data from media reports (Stack 2005, 2009; Gould 2001; Romer et al. 2006; Gould et al. 2014) and peer groups (Baller and Richardson 2009; Abrutyn and Mueller 2014; Mueller and Abrutyn 2015; Fletcher 2017). That is, when the rate of suicide by the leading means falls, overall suicide becomes less prevalent, and rates of suicide by other means drop, too. This complementary effect is the opposite of what one might expect and implies that policies aimed at components of suicide prevention may have unexpected benefits.

Our results are robust to several alternative specifications. First, we limit our sample to a non‐staggered treatment by examining the passage of the Brady Act in 1994 and then again by exploiting the variation in the states that no longer had a waiting period when the Brady Act ended in 1999. In addition, we use the estimation strategy proposed in Chaisemartin and d’Haultfoeuille (2024) and employed in Myers (2024), which allows both for continuous treatment and for treatment to turn “on” and “off” as ours does when states enact and then repeal waiting periods. The results from this method reveal that being 50 or more miles from a non‐waiting‐period state leads to an approximate 5% reduction in firearm‐related suicides, consistent with our main findings.

A back‐of‐the‐envelope calculation suggests that we would have expected to observe roughly 870 fewer firearm‐related suicides in 2019 had non‐waiting‐period states enacted waiting periods. Specifically, we estimate that there could have been up to 740 fewer suicides if non‐waiting period states adopted waiting periods and up to 130 fewer suicides if individuals within 50 miles of a non‐waiting‐period state were more than 50 miles away from a non‐waiting‐period state.

Seminal work by Ludwig and Cook (2000) leveraged the variation in state‐level gun laws caused by the passage of the 1994 Brady Act. They found that the adoption of stricter gun laws led to declines, though not statistically significant, in homicides and suicides. More recent research on waiting periods has broadly used evidence from changes in the state law of a single state and has found that these laws are associated with lower rates of gun‐related violence such as suicides (Oliphant 2022; Anestis et al. 2017) and homicides (Webster et al. 2014; Rudolph et al. 2015). Novelly, Edwards et al. (2018) created a nation‐wide panel of mandatory waiting period laws in the United States from 1990 to 2013 to exploit more state‐level variation and a longer time frame. Additionally, they layered federal and state laws, taking into account that states are affected by either or both. They find that purchase delays reduce suicide by approximately 2%. Unique to their identification strategy is the exploitation of within‐state variation in changes in purchase delays. Additionally, recent work by Donohue et al. (2023) has extended Edwards et al. (2018) to examine the impact of waiting periods on suicide rates by age. They find that the reduction in firearm‐related suicide is concentrated among young adults aged 21–34.

We are the first in the literature to exploit geographic variation in distance to a neighboring state without a waiting period to determine whether a statewide waiting period is “binding” or “non‐binding” on each county within the state. Understanding the impact of distance is useful because it allows us to identify whether a statewide regulation impacts all counties within that state or is, in practice, non‐binding on counties that are relatively close to states without a waiting period. We find that counties that lie within 50 miles of a no‐wait gun‐purchase opportunity have no statistically significant change in suicides. Therefore, we conclude that a state‐imposed waiting period is non‐binding for a county within 50 miles of a state without a waiting period. This information can help policymakers decide whether enacting a waiting period at the city, county, or state‐level would be expected to reduce suicides or if it would be more effective to devote resources to creating regional or national policies. Our findings suggest that coordinated, multi‐state regional laws imposing waiting periods are expected to see a greater reduction in suicide than would have been predicted by earlier estimates, as are national laws.

The remainder of the paper is organized as follows. In the next section, we discuss the legislative background on waiting periods and the legal data we use. Section 3 discusses the distance and mortality data and the data used for the control variables. Section 4 outlines the empirical strategy, and Section 5 presents empirical results. Section 6 concludes the paper.

2. Legislative Background and Legal Data

2.1. Background on the Brady Act and Waiting Period Data

After the failed assassination attempt against U.S. President Ronald Reagan in 1981 that left his press secretary, James Brady, permanently disabled, the federal legislature passed the Brady Act in 1993. The Brady Act imposed a five‐day waiting period for handgun purchases on all but six states that had qualified for exemptions. Before the implementation of the Brady Act, 23 states and the District had waiting periods already in place; ultimately 21 states were directly impacted by the Act. In 1998, the Brady‐imposed waiting period was eliminated with the passage of the NICS Improvement Amendments Act, which established the National Instant Criminal Background Check System (NICS), enabling instant background checks on firearm purchasers. Today, the NICS background check is conducted on all in‐store firearm purchases, including handguns, rifles, and shotguns, nationwide.

After the Brady Act's mandate ended in November 1998, 19 states repealed their waiting periods (with four states eliminating their waiting periods before November 1998). Over the next 2 decades, four more states repealed their waiting periods. By 2019, 17 states still had waiting periods of some kind.

We collect legal data for firearms at the state‐level from Siegel (2020), which provides detailed data on mandatory waiting periods and permit requirements for each state for the period 1991–2019. We cross reference these data with Edwards et al. (2018) and the firearm database from Cherney et al. (2022). When discrepancies in effective dates arise between the three sources, we identify the statute and use Westlaw to confirm or identify the correct effective date.

Following previous studies, we classify a state as having a waiting period if the state has a mandatory waiting period law or a permit requirement (Edwards et al. 2018). Although permits may not explicitly state a specific time frame to wait before one can obtain a firearm or handgun, in practice, permits result in some non‐zero wait time to comply with requirements. The legal data on states' mandatory waiting periods during the Brady Act are mainly from Edwards et al. (2018) for the period 1994–1998. We cross‐reference with Westlaw and perform exhaustive reconciliation to ensure the accuracy of our legal table.

In Table 1, we list each state, when it first enacted a waiting period law (with Pre‐1990 meaning sometime before 1990), and when it first repealed its mandatory waiting period law. Footnotes identify and explain any differences between our identification and Edwards et al. (2018).

TABLE 1.

Years mandatory waiting period laws were enacted and repealed.

State	Enacted	Repealed	State	Enacted	Repealed
Alabama	Pre‐1990	2000	Montana	1994	1999
Alaska	1994	1999	Nebraska	1991	—
Arizona a	1994	1994	Nevada	1994	1999
Arkansas a	1994	1997	New Hampshire	—	—
California	Pre‐1990	—	New Jersey	Pre‐1990	—
Colorado	—	—	New Mexico	1994	1999
Connecticut	Pre‐1990	—	New York	Pre‐1990	—
Delaware	—	—	North Carolina	Pre‐1990	—
Florida b	1991	—	North Dakota	1994	1999
Georgia a	1994	1996	Ohio	1994	1999
Hawaii	Pre‐1990	—	Oklahoma	1994	1999
Idaho	—	—	Oregon a	Pre‐1990	1996
Illinois	Pre‐1990	—	Pennsylvania	Pre‐1990	—
Indiana	Pre‐1990	1999	Rhode Island	Pre‐1990	—
Iowa	Pre‐1990	—	South Carolina	1994	1999
Kansas	1994	1999	South Dakota	Pre‐1990	2009
Kentucky	1994	1999	Tennessee	Pre‐1990	1999
Louisiana	1994	1999	Texas	1994	1999
Maine	1994	1999	Utah	—	—
Maryland	Pre‐1990	—	Vermont	1994	1999
Massachusetts	Pre‐1990	—	Virginia	—	—
Michigan	Pre‐1990	—	Washington	Pre‐1990	—
Minnesota	Pre‐1990	—	West Virginia	1994	1999
Mississippi	1994	1999	Wisconsin	Pre‐1990	2015
Missouri a	Pre‐1990	2007	Wyoming	1994	1999

Note: Legal data are from Edwards et al. (2018), Siegel (2020), and Cherney et al. (2022). When differences arise among the sources, we use Westlaw to identify the effective and repeal dates of the legal statute identified in Cherney et al. (2022).

^a Repeal dates differ from Edwards et al. (2018). (1) For Arizona, A.R.S. § 13‐3108.01 effectively repealed the mandatory waiting period and became effective October 1994. (2) Arkansas stopped implementing the Brady Act with the ruling Printz v United States, 521 U.S. 98, 117 S.Ct. 2365 (1997). (3) Georgia state law repealed the federal mandate with Ga. Code Ann. § 16‐11‐172 which became effective June 1996. (4) For Missouri, the permit law was repealed with Mo. Rev. Stat. § 571.090 effective August 28, 2007. (5) For Oregon, the waiting period law was repealed by O.R.S. 1995 c.729 § 13. effective July 1, 1996.

^b Effective and repeal dates differ from Edwards et al. (2018). Florida implemented a 3‐day waiting period with Fla. Stat. § 790.0655(1) (a), which became effective October 1, 1991, and was later amended in 2018. Since this law became effective at the end of the calendar year, we treat 1992 as the first treated year in our data.

3. Data

We examine whether the distance to a state without a waiting period impacts suicide rates. A priori, if a waiting period impacts suicide through a “cooling off” period, then we would expect counties in waiting‐period states that are relatively close to states without a waiting period to be less impacted than those that are farther away. For example, Washington state has a waiting period law, while Oregon repealed its law in 1996. Spokane County, WA, is less than 30 miles from Kootenai County, ID, so we expect Spokane County to be less impacted than King County, WA (Seattle), which is over 100 miles from a non‐waiting‐period state.

To calculate the distance to the nearest county without a waiting period, we use county‐to‐county distance for every US county from the US Census. Distance is measured as great circle distances and calculated using the Haversine formula based on internal points in the geographic area. ² For counties with waiting periods, we identify the closest county in a state without a waiting period law. We do not observe the actual location of firearms sales, so for counties that are in a state without a waiting period, we set the distance to zero. ³ Figure 1 shows the evolution of the distance to the nearest county without a waiting period in 5‐year increments.

FIGURE 1.

Average distance to a county without a waiting period over time. This figure shows the average distance to the nearest county without a waiting period over 5‐year increments. Data on distance between counties is from the NBER County Distance Database. Legal data are from Edwards et al. (2018), Siegel (2020), and Cherney et al. (2022). When differences arise among the sources, we use Westlaw to identify the effective and repeal dates of the legal statute cited in Cherney et al. (2022). See Table 1 for more detail on precise enactment and repeal dates.

We obtain county‐level mortality data from the Centers for Disease Control from 1991 to 2019. This database provides annual suicide mortality counts for firearm‐related and non‐firearm suicides. We identify suicides with cause codes U03, X60‐X84, and Y87.0, and firearm‐related suicides with cause codes X72‐X74. We estimate that we have firearm‐related suicide figures for 98.7% of the county‐years and for well over 99.9% of the U.S. population in each year.

Given that suicide rates and methods vary by age, sex, and race, we include county‐by‐year demographic controls for the gender, racial, and age composition compiled from the Surveillance of Epidemiology and End Results program (SEER, 2024). Specifically, for each county‐year we calculate the proportion of the county that is white, Black, male, and age 15 or older. We also include economic controls, the county‐year unemployment rate published by the Bureau of Labor Statistics Local Area Unemployment Statistics (LAUS), the median household income, and the proportion of the population that is below the federal poverty line published by the Census Bureau's Small Area Income and Poverty Estimates (SAIPE, 2024) program. ⁴

We also control for several state‐level firearm policies that may impact suicide rates using the database from Siegel (2020). The binary variable Background check takes the value of one if the state has either a standalone law or a background check is required to obtain a permit, and zero otherwise. The database separately identifies whether the background check requires a search of state mental health records, so we control for this separately, since Vars et al. (2022) found an impact on suicide rates. We also control for whether there is a minimum legal sale age for firearm purchases (Fridel et al. 2024).

Table 2 summarizes the mortality and control variables used in the analysis that follows. Columns 1 and 2 present the mean and standard deviation for all counties over the timeframe. Columns 3 and 4 show the same, but for the counties more than 50 miles from a non‐waiting‐period state at the time of the observation. Columns 5 and 6 present the mean and standard deviation for the counties that, at one point in time, will be more than 50 miles away from a non‐waiting‐period state but are within 50 miles at the time of observation. The outcome variables presented are overall Suicide rate, Firearm‐related suicide rate, and Non‐firearm suicide rate. Column 1 shows that the overall suicide rate in the U.S. is 12.24 suicide deaths per year per 100,000 population and that the majority of suicides are firearm‐related. Columns 2 and 3 reveal that the overall and firearm‐related suicide rates in counties in states with active waiting periods are, on average, lower than in counties in states without waiting periods. Columns 5 and 6 show that both the overall suicide rate and firearm‐related suicide rate are higher in counties that will be “treated” by the waiting period, but do not have any active waiting period at the time of observation.

TABLE 2.

Summary statistics for outcome variables and for legal, economic, and demographic controls.

	All counties		Counties 50+ miles
	1991–2019		Waiting period		No waiting period
	Mean	SD	Mean	SD	Mean	SD
Outcomes (per 100k pop.)
Suicide rate	12.24	5.79	11.18	4.89	14.03	6.58
Firearm‐related suicide rate	6.50	4.64	5.46	3.93	8.41	5.03
Non‐firearm suicide rate	5.74	3.06	5.73	2.67	5.61	3.59
Other firearm laws
Background check	0.45	0.50	0.69	0.46	0.01	0.11
Mental health facility check	0.28	0.45	0.45	0.50	0.03	0.17
Minimum legal sale age	0.61	0.49	0.72	0.45	0.48	0.50
Economic controls
Unemployment rate	5.95	2.58	6.13	2.69	5.95	2.46
Proportion of HH $\le$ FPL	13.74	5.55	13.27	5.17	15.55	5.60
Median HH income (2019$)	62,606	16,250	64,350	15,569	56,205	12,813
Demographic controls
Proportion of the pop. that is:
Black	0.13	0.13	0.12	0.11	0.15	0.15
White	0.80	0.15	0.81	0.14	0.80	0.15
Hispanic ethnicity	0.14	0.16	0.16	0.16	0.14	0.18
Male	0.49	0.01	0.49	0.01	0.49	0.01
Under 19	0.27	0.03	0.27	0.03	0.28	0.03
Age 20–29	0.14	0.03	0.14	0.03	0.14	0.03
Age 30–39	0.14	0.02	0.15	0.02	0.14	0.02
Age 40–49	0.14	0.02	0.14	0.02	0.14	0.02
Age 50–64	0.17	0.03	0.17	0.03	0.17	0.03
Observations	90,755		37,917		40,151

Note: The data span the period 1991–2019. Legal data for waiting periods are from Edwards et al. (2018); Siegel (2020); Cherney et al. (2022) (see Table 1 for details). Mortality data are from the CDC. Suicide deaths are identified with cause codes U03, X60‐X84, and Y87.0. Other firearm laws are from Siegel (2020). The county‐year unemployment rate is from the Bureau of Labor Statistics Local Area Unemployment Statistics (LAUS). The median household income and the proportion of the population that is below the federal poverty line are from the Census Bureau's Small Area Income and Poverty Estimates (SAIPE, 2024) program. Demographic controls are from the Surveillance of Epidemiology and End Results program. Other firearm laws are binary variables used as controls. Economic controls are continuous variables. Demographic controls are binary variables.

Under Other Firearms Laws are listed the three legal controls that may impact suicide rates. The control for a universal background check is particularly important, as states implemented these policies during the Brady Act period. We have three economic controls, the unemployment rate, the proportion of households at or below the federal poverty line (Proportion of HH $\le$ FPL), and median household income (Median HH income (2019$)), since economic factors can influence suicide attempts (Elbogen et al. 2020) and rates (Huikari and Korhonen 2021; Marcotte et al. 2020). We see that the unemployment rate is fairly consistent across the different counties, though eventually‐treated counties that are not actively impacted by waiting periods have a slightly higher poverty rate and lower median household income.

Finally, Table 2 presents the demographic controls for race/ethnicity and for age, since older males disproportionately influence suicide rates (Garnett and Curtin 2023). We do not observe any major differences in demographics across the six columns, except for the variation in the proportion of the population that is Black.

4. Methodology

We exploit the variation in timing in the adoption and repeal of waiting periods. Given the presence of county‐year observations with zero suicides, we employ a fixed‐effects Poisson quasi‐maximum likelihood estimator. Specifically, we estimate:

$$E\left[Suicide{s}_{\mathit{cst}}|Wai{t}_{st},{\alpha }_c,{\gamma }_s,{\theta }_t,X_{\mathit{cst}}\right]=\exp \left({\beta }_{1}Wai{t}_{st}+{\alpha }_c+{\gamma }_s+{\theta }_t+X_{\mathit{cst}}\psi \right)$$ (1)

where $Suicides$ is the number of suicides (firearm‐related, non‐firearm, and overall, each in succession) in county $c$, of state $s$, at time $t$. Our primary variable of interest, $Wait$, is a binary variable equal to one if the state has adopted a waiting period and zero otherwise. We include county, state, and year fixed effects represented by ${\alpha }_c$, ${\gamma }_s{\theta }_t$, respectively. Vector $X$ includes our time‐varying state and county controls, such as state policies on background checks and county‐level demographic information, as described in Section 3.

We prefer the Poisson fixed effects model to other nonlinear models, such as the negative binomial model, as it is known to be robust to most assumption failures. Additionally, it allows for any type of variance‐mean relationship (including both over‐ and underdispersion), and it can accommodate arbitrary serial correlation when standard errors are appropriately clustered. In contrast, the fixed‐effects negative binomial model imposes a very specific form of overdispersion, requires conditional independence (ruling out serial correlation), and is not known to be robust to violations of its assumptions, among other shortcomings. Thus, we proceed with the Poisson Fixed Effect model with clustered standard errors, to have the most flexible framework for our analysis (Cameron and Trivedi 2013; Wooldridge 1999).

In this first specification, though our primary policy variable (waiting period) varies at the state level; using county‐level observations still provides a few advantages for our analysis. Suicide rates vary substantially within states, and county‐level demographic and socioeconomic factors are important predictors of suicide risk that cannot be adequately captured by state‐level aggregates. By using county‐level data, we can control for specific local characteristics known to correlate with suicide rates, such as the age distribution, race, gender, and economic conditions, which improves the precision of our estimates. Additionally, county‐level analysis better reflects the local conditions that influence suicide risk, allowing for more accurate measurement of the impact of waiting period laws based on the unique demographic and socioeconomic profile of each county. We found that the use of county‐level data resulted in a statistically significant greater magnitude in the estimate of the effect compared to using state‐level controls.

In our primary analysis, we exploit the unique feature that the adoption (and repeal) of waiting periods affects some counties within a state more than others. To illustrate this variation, we map the average travel distance (in 25‐mile bins) to a neighboring state without a waiting period every five years (1993–2018). Here, we can visually see that in 1998 (the year before the Brady Act was repealed), most of the US population lived in counties that were far from a state without a waiting period. By 2003, many states had repealed their waiting period laws, so there were more counties in states that had waiting periods that were within 50 miles of a non‐waiting period state.

To exploit this within‐county‐level variation in distance, we estimate the following Poisson model:

$$E\left[Suicide{s}_{\mathit{cst}}|Dis{t}_{\mathit{cst}},{\alpha }_c,{\gamma }_s,{\theta }_t,X_{\mathit{cst}}\right]=\exp \left(Dis{t}_{\mathit{cst}},{\alpha }_c,{\gamma }_s,{\theta }_t,X_{\mathit{cst}}\right]$$ (2)

Where $Dist$ measures the number of miles to the nearest neighboring state without a waiting period. We include a quadratic term for distance, as the effects likely are nonlinear. All Poisson specifications include the county population as the exposure variable. Similar to Equation (1) above, all specifications include time‐varying socioeconomic demographic controls, county, state, and year‐fixed effects. The county, state, and time‐fixed effects control for all time‐invariant county and state‐level factors and overall time effects that might influence suicides. Here, the underlying assumption is that suicides in counties would have evolved similarly absent changes to waiting period status.

Although the underlying assumption is inherently untestable, we examine the trends in suicide rates by affected counties in Figure 2. Panel A presents the trends in suicide rates over time as states enact waiting period laws. We center our event time around zero and show trends from 3 years before to 4 years after a state enacts a waiting period law. Since the first round of repeals occurred in 1999, we restrict our post‐period to four years after the waiting period. The dashed navy line shows the suicide rates for counties that were not impacted by the waiting period law. In contrast, the solid navy line shows the trends for counties that will be impacted by the MWP, which we define as an increase in distance of 50 miles or more. The trends follow each other fairly closely, with the highly impacted counties appearing to be on a smooth downward trend post‐MWP enactment.

FIGURE 2.

Average suicide rates over relative time. This figure plots the average suicide rates over relative time. Panel A shows the average suicide rates around the time of enacting a waiting period. The navy solid line shows average suicide rate for counties that will enact a waiting period, the dashed line shows the rate for counties that never enact a waiting period. For never‐treated counties, year 1994 is used to center around zero. Panel B shows the average suicide rates around the repeal of a waiting period law. The navy solid line shows the trend for counties in a state that will eventually repeal a waiting period. The dashed line shows the trend for counties that will never repeal (“always treated”) in our sample period. We use the year 2000 for our “always treated” group to center around zero. Data on suicides and population are from the CDC.

Figure 2 Panel B presents the trends in suicide rates around the time of repealing a waiting period. The dashed navy line shows the suicide rates for counties that are always 50 miles or farther from a neighboring state without a waiting period. The solid blue line shows the trends in suicide rates for counties that are 50 miles or more away from a neighboring state with no waiting period, but the state repeals the waiting period. Similar to Panel A, the trends follow each other fairly closely, and then the suicide rate increases around year two following the repeal.

5. Results

We begin with Table 3, which presents the results from Equation (1), first without any additional state or county‐level controls. Here, our primary variable of interest is Waiting period, a binary variable equal to one if a state has a waiting period, either through a mandatory purchase delay or through the presence of permit requirements. The results in Column 1 imply that a state that adopts a waiting period reduces its overall suicides by approximately 5.3%. ⁵ Column 2 shows that after a state enacts a waiting period, firearm‐related suicides fall by approximately 8.8%. Interestingly, we also see a reduction in non‐firearm‐related suicides. To contextualize the findings, we present the Mean of the dependent variable for the state‐level analysis, which is the average number of suicide deaths, firearm‐ and non‐firearm‐related, in a county when a waiting period is not present. For example, the 8.8% decrease in firearm‐related suicides would translate to roughly $0.088\times 5.43$, or approximately 0.47 fewer firearm‐related suicides per 100,000 population.

TABLE 3.

The effect of waiting periods on suicides.

	Model 1: No controls			Model 2: Model 1 + controls
	All suicides	Firearm	Non‐firearm	All suicides	Firearm	Non‐firearm
Outcome variable	(1)	(2)	(3)	(4)	(5)	(6)
Panel A: County‐level data (our full 29‐year period)
Waiting period	−0.054**	−0.092**	−0.078***	−0.023*	−0.051***	−0.039**
	(0.026)	(0.036)	(0.026)	(0.014)	(0.015)	(0.019)
N	90,697	90,639	89,772	88,182	88,003	87,377
Panel B: State‐level data (our full 29‐year period)
Waiting period	−0.057**	−0.098***	−0.079***	−0.010	−0.022*	−0.032*
	(0.026)	(0.034)	(0.027)	(0.011)	(0.011)	(0.017)
Mean (count)	565	334	230	565	334	230
N	1470	1470	1470	1441	1441	1441
Panel C: State‐level, 1991–2013 (same period as Edwards et al. (2018))
Waiting period	−0.042*	−0.071**	−0.079***	−0.016	−0.022**	−0.042***
	(0.022)	(0.028)	(0.024)	(0.011)	(0.011)	(0.015)
Mean (count)	565	334	230	565	334	230
N	1164	1164	1164	1141	1141	1141

Note: Estimates are based on a Poisson model evaluating the expected suicide count using annual county‐level data for the period 1991–2019. Data on suicides are from the CDC. All regressions cluster standard errors at the state level. Model 1 includes county (Panel A), state (Panels B and C), and year fixed effects. Model 2 includes legal controls from Siegel (2020), economic controls from the BLS and SAIPE, and demographic controls from SEER. Firearm, economic, and demographic controls are those listed in Table 2. Panel B aggregates the data to the state level. Panel C aggregates the data to the state level and restricts the years to 2013 and earlier to facilitate comparison with Edwards et al. (2018). To aid interpretation of the coefficients, we present the mean of the corresponding county (Panel A) or state (Panels B and C) suicide rate for those counties (states) that do not have a waiting period.

*p < 0.10.

**p < 0.05.

***p < 0.01.

Columns 4 through 6 add the county‐specific controls that may correlate with suicides, as discussed in Section 3. Here, we see that waiting periods lead to an approximate 2.3% reduction in overall suicides. We continue to see significant effects on both firearm and non‐firearm suicide rates. The findings imply that the presence of a waiting period reduces firearm and non‐firearm suicides by approximately 5% and 3.8%, respectively. The findings from Column 5 translate to roughly 740 fewer firearm‐related suicides annually if every county in a non‐waiting period state instead had a waiting period. ⁶

Panel B of Table 3 presents the findings from Panel A but aggregates the data to the state‐level to more closely follow Edwards et al. (2018). Columns 2 and 5 suggest a significant and negative reduction in firearm‐related suicides, consistent with Edwards et al. (2018), who find that waiting periods lead to a decrease in firearm‐related suicides of 2.2%–9.3%, depending on the specification. Columns 1 and 4 both find a negative impact on overall suicide, although the effect is not statistically significant once we add county‐level controls. We again see a negative impact of waiting periods on non‐firearm‐related suicides as well.

Panel C of Table 3 presents the results when we aggregate to the state‐level and restrict to the same years of Edwards et al. (2018). Columns 1–3 of Panel C are most similar to Column 1 in Table 4 of Edwards et al. (2018), which also does not include any controls, though our primary estimation strategy uses a Poisson model while Edwards et al. (2018) uses the natural log of the different suicide rates. ⁷ They find a significant reduction in overall suicides of 4.9%, 9.1% for firearm‐related suicides, and 5.9% for non‐firearm suicides; all estimates are well within the 95% confidence interval of our findings.

TABLE 4.

The effect of distance to a neighboring state with no waiting period on suicides.

	Linear distance			Quadratic distance
	All suicides	Firearm	Non‐firearm	All suicides	Firearm	Non‐firearm
Outcome	(1)	(2)	(3)	(4)	(5)	(6)
Distance (100s miles)	−0.002	−0.006**	−0.002	−0.008**	−0.014***	−0.006
	(0.002)	(0.003)	(0.003)	(0.004)	(0.005)	(0.006)
Distance squared (100s miles)				0.001**	0.001**	0.001
				(0.000)	(0.000)	(0.001)
N	88,182	88,003	87,377	88,182	88,003	87,377

Note: Estimates are based on a Poisson model evaluating the expected suicide count using annual county‐level mortality data for the period 1991–2019. $Distance$ is a continuous variable equal to the distance (in hundreds of miles) from a county in a state with a waiting period to the nearest county in a state without a waiting period. For example, a county in a state with a waiting period that is 150 miles from a county without a waiting period will have a distance equal to 1.5. $Distance$ is set to zero for every county in a state with no waiting period. Data on suicides are from the CDC. See Table 2 for a list of the control variables. Standard errors are clustered at the county level.

*p < 0.10.

**p < 0.05.

***p < 0.01.

We now move to Table 4, which reports the findings of the specification that exploits the variation in distance to a neighboring state without a mandatory waiting period. Columns 1 through 3 present the results on the effect of distance, represented as a linear function in miles to the nearest no‐waiting‐period county, on suicides. We see in Column 2 that increasing the distance to the closest county without a waiting period by 100 miles leads to a 0.6% reduction in firearm‐related suicides.

Columns 4 through 6 present the results on the impact of distance, represented as a quadratic function in miles to the nearest no‐waiting‐period county, on suicides. The results in Column 5 imply that a 100‐mile increase leads to a 1.4% reduction in firearm‐related suicides. Our preferred specification is the quadratic specification, which allows the marginal impact of distance to diminish as distance increases. The results in Column 4 imply that increasing distance by 100 miles leads to a 0.8% reduction in total suicides.

To further explore the impact of distance on suicides, we use the estimation from Columns 5 and 6 and present the predicted firearm and non‐firearm suicides as the distance to the nearest county without a waiting period increases in 25‐mile increments. The findings from this exercise are presented in Figure 3. The filled, navy circles and the left‐hand axis present the predicted firearm‐related suicides, while the orange, hollow circles and the right axis correspond to the non‐firearm suicides. The corresponding 95% confidence interval is plotted along with each prediction. We see a decline in the predicted firearm and non‐firearm suicides, though the decline for firearm‐related suicides is much steeper. For firearm‐related suicides, we see that once a county is more than 50 miles from a state without a waiting period, the predicted firearm‐related suicide rate and corresponding confidence intervals fall outside the confidence interval prediction when there is no law.

FIGURE 3.

Impact of distance on suicide rates, by type of suicide. This figure shows the number of predicted suicides for counties at various distances after estimating Equation (1) for firearm‐related suicides, in navy with solid circles and the left axis, and non‐firearm suicides, in orange with hollow circles and the right axis. Data on suicides and population are from the CDC. All regressions include county and year fixed effects. See Table 3 for a description of the control variables.

Next, we present our main findings in Table 5. Similar to Table 3, our primary variable of interest is Waiting period, a binary variable equal to one if a state has a waiting period, either through a mandatory purchase delay or through the presence of permit requirements. However, now motivated by the findings in Table 4, we restrict our sample of counties to those we anticipate may be relatively “less” (or “more”) impacted by a waiting period. Given that the psychology and sociology literature have found that the decision to attempt suicide is often impulsive, with the time between the decision and attempt often being a matter of minutes or hours (Deisenhammer et al. 2009; Paashaus et al. 2021), we define counties as “less” impacted by waiting period laws if they are within 50 miles of a state without a waiting period law. Panel A of Table 5 presents the results from this analysis. We then separately examine counties that may be “more” impacted by the waiting period, defined as the distance to the nearest county without a waiting period being either 50 miles or 150 miles or more away, which are Panels B and C in Table 5, respectively.

TABLE 5.

The effect of waiting periods on suicides by distance.

Outcome	All suicides	Firearm	Non‐firearm
	(1)	(2)	(3)
Panel A: Nearest county without a waiting period is $\le$ 50 miles
Waiting period	−0.014	−0.025	−0.031
	(0.016)	(0.018)	(0.026)
N	57,665	57,488	57,006
Panel B: Nearest county without a waiting period is $>$ 50 miles
Waiting period	−0.022**	−0.052***	−0.031**
	(0.009)	(0.011)	(0.012)
N	79,581	79,425	78,773
Panel C: Nearest county without a waiting period is $>$ 150 miles
Waiting period	−0.033***	−0.057***	−0.038**
	(0.011)	(0.013)	(0.016)
N	64,963	64,814	64,243

Note: Estimates are based on a Poisson model evaluating the expected suicide count using annual county‐level mortality data spanning 1991–2019. Panel A restricts observations to those counties that are 50 miles or closer to a state without a waiting period. Panel B restricts the observations to those counties that are either never treated or 50 miles or farther to a state without a waiting period. Panel C restricts to the counties that are either never treated or 150 miles or farther to a state without a waiting period. Data on suicides are from the CDC. See Table 2 for a list of the control variables. All regressions include county and year fixed effects. Standard errors are clustered at the county level.

*p < 0.10.

**p < 0.05.

***p < 0.01.

Panel A of Table 5 presents the results where we restrict our sample to counties that are 50 miles (or less) from a county in a state without a waiting period. The results show that counties that are within 50 miles of a county that is located in a state without a waiting period show no statistically significant impact on any suicide outcome.

Next, we restrict our sample to counties that are either not treated or where the nearest county in a state without a waiting period is 50+ miles away, as presented in Panel B of Table 3. We find that when a purchase delay is implemented, a county that is 50+ miles from a non‐waiting‐period state experiences statistically significant negative effects on firearm‐related suicide of over 5%, and on overall suicide of about 2%. When we restrict to counties that are 150+ miles away, we see a larger decline in all types of suicides, though the coefficients are within the 95% confidence interval suggested in Panel B. Overall, we find that when a county is closer to a state without a waiting period, the effect of implementing a waiting period is of lesser magnitude than when a county is farther away.

Figure 4 shows the counties that are within 50 miles of a non‐waiting‐period state (in orange), the counties that are more than 50 miles from a non‐waiting‐period state (in blue), and the counties with no waiting period (in white) in 1998 and then in 2018. We can see that as states repeal waiting periods, not only are there more individuals in a state without a waiting period, but the proportion of the population that is (1) in a state with a waiting period but (2) is in a county that is 50‐miles or closer to a neighboring state without a waiting period has increased. In 2019, there were 2682 deaths from firearm‐related suicides in counties that were in waiting‐period states but within 50 miles of a non‐waiting‐period state. Using estimates from Table 5 Panel B Column 2, A quick, back‐of‐the‐envelope calculation suggests that had these individuals been slightly farther away from a non‐waiting‐period state (over 50 miles), we would expect to have seen up to 130 fewer firearm‐related suicides. ⁸

FIGURE 4.

Counties under and over 50 miles to the nearest non‐waiting‐period state. This figure shows counties that are over 50 miles from a state without a waiting period law (in blue) and 50 miles or less from a state without a waiting period law (in orange). Data on the distance between counties is from the NBER County Distance Database. See Table 1 for more detail on the legal data.

One common concern about implementing waiting period laws, or other laws that may discourage one particular type of suicide, is that individuals will substitute to other means. Interestingly, throughout Column 3 of Table 5, we find that waiting periods are associated with declines in non‐firearm suicides, although the magnitudes are smaller than the impact on firearm‐related suicides. This finding could be suggestive evidence that suicides are socially “contagious.” If suicides in a county are contagious, and a county has significant reductions in firearm‐related suicides, it may also bring down non‐firearm suicides. There is evidence that suicides are contagious, using evidence from media reports (Stack 2005, 2009; Gould 2001; Romer et al. 2006; Gould et al. 2014) and peer groups (Baller and Richardson 2009; Abrutyn and Mueller 2014; Mueller and Abrutyn 2015; Fletcher 2017).

A growing body of methodological literature identifies potential sources of confounders when using two‐way fixed effects (TWFE) when treatment is staggered and the impacts are heterogeneous, and proposes alternative estimators (Goodman‐Bacon 2021; Callaway and Sant’Anna 2021; Borusyak et al. 2024; Chaisemartin and d’Haultfoeuille 2023; Roth et al. 2023; Chaisemartin and d’Haultfoeuille 2024). The known biases are particularly exaggerated when considering a static two‐way fixed effects difference‐in‐difference estimator.

We conduct several analyses to address this problem in the Supporting Information S1. In one specification, we restrict the analysis to exploit a one‐time change that was caused by the implementation of the 1994 Brady Act; thus, treatment is not staggered. Finally, we use the estimation strategy proposed by Chaisemartin and d’Haultfoeuille (2024) and employed in Myers (2024). This approach differs from standard two‐way fixed effects (TWFE) difference‐in‐difference (DID) models by avoiding the potential bias that occurs when treatment effects are heterogeneous across time or groups and when treatment timing varies. Unlike traditional TWFE‐DID, which can produce estimates that are weighted averages of potentially negative and positive treatment effects, this estimator only uses “clean” comparisons between newly treated units and not‐yet‐treated units. This approach is particularly useful for our research design, as it allows for treatment to turn “on” and “off” multiple times– which occurs in our context when states enact and repeal waiting periods. While this approach limits us to using linear models rather than our preferred Poisson specification, the results from our OLS regressions (presented in Tables A1‐A3 in the Supporting Information S1) confirm that our main findings are robust to this alternative functional form. Using this method, we estimate that being 50 or more miles from a non‐waiting‐period state leads to an approximate 5% reduction in firearm‐related suicides, consistent with our main results from the Poisson specification. Figure A1 in the Supporting Information S1 presents these dynamic estimates, which also provide support for the parallel trends assumption through insignificant placebo effects in pre‐treatment periods.

6. Conclusion

Our study finds a larger effect of waiting periods than previously identified, as we obtain county‐level suicide data for 1991 through 2019. Further, we are able to isolate counties that are plausibly most affected by waiting periods, those counties that are relatively far (50+ miles) from a non‐waiting‐period state. We find that enacting waiting periods has a significant, negative effect, 5%. Counties close to a non‐waiting‐period state (within 50 miles) are unaffected by their own state's waiting‐period laws, as reflected in suicide rates, with no statistically significant change in response to such laws.

A back‐of‐the‐envelope calculation suggests that we would have expected to see roughly 870 fewer firearm‐related suicides in 2019 had non‐waiting‐period states enacted waiting periods. This 870 estimate is from approximately 740 fewer suicides from non‐waiting‐period states adopting waiting periods and 130 fewer suicides from individuals within 50 miles of a non‐waiting‐period state now being more than 50 miles away. Overall, the results are potentially informative to policymakers, as the findings suggest that national policies or regional policies, where adjacent states enact policies together, thereby minimizing the number of counties close to non‐waiting‐period states, would have a larger effect on suicide rates.

Author Contributions

Grace E. Arnold: data curation (equal), formal analysis (equal), methodology (equal), project administration (equal), supervision, validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Mitchell B. Priestley: conceptualization, data curation (equal), formal analysis (equal), methodology (equal), project administration (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available from the CDC's National Center for Health Statistics (NCHS). Restrictions apply to the availability of these data, which were used under license for this study. Data are available from https://www.cdc.gov/nchs/data/nvss/nchs‐research‐review‐application.pdf with the permission of the CDC's National Center for Health Statistics (NCHS).