Abstract
Previous research on firearms has not adequately addressed a fundamental question about handgun ownership: Why do some people own handguns while most in the United States do not? We use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine adolescent and adult correlates of handgun ownership, including socialization, victimization and fear of crime, political ideology, and societal insecurities. We also investigate the differences between “typical” owners and “atypical” owners who own more handguns. We find that socialization, victimization, conservatism, and societal insecurity all independently increase the likelihood of handgun ownership, and atypical handgun owners are more likely to be conservative and to have experienced victimization than typical owners.
Introduction
America has a significantly higher rate of private gun ownership than other Western countries. In the United States, there are about 93 guns per 100 citizens compared with an average of about 25 guns per 100 citizens in other Western countries (Hepburn, Miller, Azrael, & Hemenway, 2007). The prevalence of handguns, which are justified mostly as weapons for personal protection, is even higher in the United States compared with peers than differences in overall gun ownership, which includes many long guns used for hunting and sport (Hemenway, 2004). For instance, Canada has a similar number of long guns (shotguns or rifles) but only had 1.2 million handguns in 1998 compared with an estimated 76 million handguns in the United States (Hemenway, 2004). Despite the large international disparity, concerns about comparatively high rates of violence in the United States, heated political debates about the “right to bear arms,” and disagreements about the relationship between guns and crime, there remains very little research, let alone consensus, on why people choose to own guns. Unfortunately, partisan political differences and a resultant lack of public funding and support for collecting and analyzing data on gun violence and gun ownership have left scholars struggling to thoroughly and satisfactorily understand the role of guns in America (Pew Research Center, 2012; Williams & Ferris, 2015).
In recent years, the increase in privately owned guns in the United States has largely been driven by handgun ownership (Azrael, Hepburn, Hemenway, & Miller, 2017). In 1994, 33% of the 192 million privately owned guns in the United States were handguns, but in 2015, 40% of the 265 million privately owned guns in the United States were handguns (Azrael et al., 2017). Who are these owners of handguns? From previous research, we know that White, middle-aged men are most likely to own handguns, and the most common reason given is for protection (Hepburn et al., 2007). Furthermore, despite high levels of gun availability, the rate of household firearm ownership appears to be slowly declining. In 2015, only 33% of households reported owning a firearm compared with 38% of households in 2004 and 42% of households in 1994 (Azrael et al., 2017; Cook & Ludwig, 1997; Hepburn et al., 2007). Who is this minority of Americans who owns guns and why do they choose to arm themselves? Existing research has not fully addressed these questions. A paucity of studies examine gun ownership at the individual level and (a) provide only cross-sectional demographic characteristics of gun owners (Azrael et al., 2017; Cook & Ludwig, 1997; Hepburn et al., 2007), (b) focus on a narrow analysis of one or two correlates of ownership (Cao, Cullen, & Link, 1997; Kleck, Kovandzic, Saber, & Hauser, 2011; Turner, Phillips, Tigri, Williams, & Hartman, 2016; Wallace, 2017), or (c) use nongeneralizable or qualitative methods (Carlson, 2015; Stroud, 2012).
In the present study, we use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine a variety of theoretical explanations for why young adults own handguns. Furthermore, we look at whether people who own many handguns differ from more typical handgun owners who own one or two and from people who do not own a handgun. The small literature examining gun owners provides several theoretical explanations for why individuals choose to own a handgun. The main theories are socialization (e.g., culture and parents; Cao et al., 1997), victimization and fear of crime (DeFronzo, 1979; Hill et al., 1985; Kleck et al., 2011; Turner et al., 2016; Wallace, 2017), political conservatism (Cook & Ludwig, 1997; Hepburn et al., 2007), and societal insecurities (Carlson, 2015). However, most prior studies are narrowly focused and unable to consider multiple explanatory factors simultaneously.
Our study has at least two advantages over prior research. First, Add Health data contain many rich measures, including contextual data that match census data to respondents, that enable a more rigorous examination of the many possible explanations for handgun ownership. Also, past studies were largely cross-sectional, making it more difficult to model temporal order with confidence. Second, the data contain a very precise measure of handgun ownership in which respondents are asked whether they personally own a handgun and how many. Past studies often used a more general and less clear measure of gun ownership, asking only if there was a (or any) gun in the household. We combine these two strengths to simultaneously test many of the theoretical explanations for handgun ownership and examine whether the theoretical explanations hold for individuals who own one or two handguns as well as those who own more.
Literature Review
Who Owns Guns?
We know surprisingly little about why people own guns, but researchers have looked at the demographic characteristics of who owns guns in general and, to a lesser extent, handguns in particular. Men who are middle aged or older are more likely to own a handgun than either younger men or women (Hepburn et al., 2007). Also, political conservatives are more likely to own handguns (Cook & Ludwig, 1997; Hepburn et al., 2007). Individuals with higher educational attainment are less likely to own a handgun (Cook & Ludwig, 1997; Hepburn et al., 2007). Glaeser and Glendon (1998) looked at who owned guns and found that membership in progun social groups, rather than criminal behavior or victimization, was the strongest predictor of gun ownership. However, they found that the motivation for owning a handgun or pistol appeared to be different. Protection was the strongest predictor of owning only a handgun (Glaeser & Glendon, 1998). Therefore, handgun ownership appears to have different motivations than rifle and shotgun ownership.
Why Own Guns?
Only a few prior studies examine why people own guns. Past studies have examined what factors might motivate handgun ownership, but they have been limited by either the measure of handgun ownership or a lack of contextual variables. More specifically, most studies have been narrowly focused on a limited number of correlates, usually just one or two (Cao et al., 1997; DeFronzo, 1979; Kleck et al., 2011; Turner et al., 2016; Wallace, 2017). Another limitation of previous research is that no one has examined whether owning many handguns has different motivations than owning one or two handguns. In 2015, 8% of gun owners owned 39% of all guns, and half of all guns were owned by only 3% of the general population (Azrael et al., 2017).
Socialization
One proposed motivation for handgun ownership is that being socialized in a progun environment is positively related to the likelihood that an individual will own a handgun. Cao et al. (1997) studied Cincinnati residents to test whether being socialized in environments that normalized gun ownership was linked to the likelihood that an individual had a personal gun for protection. They found that growing up with a gun in the home was associated with owning a handgun for protection as an adult. Cao et al. (1997) and O’Connor and Lizotte (1978) looked at whether there is a Southern subculture of violence that manifests in an increased likelihood of gun ownership, but they did not find a statistically significant relationship. Cao et al. (1997) also did not find an association between region of residence and gun ownership, which suggests that it is likely the more intimate socialization effect of the home that influences gun ownership.
Other researchers however have found a connection between living in the South and handguns. Bankston et al. (1990) found that living in the South predicted carrying a handgun for protection. Although a fear of crime was not related to gun carrying for protection, being male and living in the South were significant correlates. Therefore, the role that parent socialization and regional effects of living in the South plays in handgun ownership needs to be further explored.
Victimization and fear of crime
One of the most studied motivations for handgun ownership is prior victimization and the fear of being a victim of crime. Several studies have found that a fear of crime is associated with handgun ownership (DeFronzo, 1979; Hill et al., 1985; Kleck et al., 2011). The relationship between fear of crime and handgun ownership is moderated by gender according to Hill et al. (1985) as men who had a fear of crime were more likely to own a handgun than women. Kleck et al. (2011) found that beyond the association between a fear of crime and handgun ownership, having a high-perceived risk of victimization was associated with plans to purchase a handgun in the near future. Arthur (1992) however found that a fear of crime was associated with handgun ownership in White individuals but that there was no such association for Black individuals.
A correlate of crime closely related to a fear of crime is having been victimized. Several studies have explored whether having been victimized is associated with an increased likelihood of owning a handgun (Arthur, 1992; Turner et al., 2016; Wallace, 2017). Findings from these studies have shown that victimization is positively associated with both handgun ownership as well as gun carrying.
Political conservatism
Hepburn et al. (2007) found a positive association between political conservatism and handgun ownership when they used a nationally representative survey to study private gun ownership in the United States. The study was cross-sectional, so they were unable to account for childhood factors that may have affected the decision to own a gun.
Societal insecurity
Carlson (2015) argues that the rise in “conceal and carry” gun permits in America is due to a feeling of economic and societal decline and a belief that individuals need to protect themselves and their communities. Carlson (2015) conducted in-depth interviews of 60 “conceal and carry” handgun owners in Michigan and found that most of the interviewees talked about how breakdowns in society and their bleak outlook of how much assistance the authorities and government would be able to provide led them to acquire “conceal and carry” gun permits. Men with “conceal and carry” permits self-identified as “protectors” (Carlson 2015; Stroud, 2012). Similarly, Jiobu and Curry (2001) tested the association between feelings of confidence in the federal government and gun ownership. They found support for their hypothesis that having societal insecurity or lack of faith in the federal government was associated with an increased likelihood of owning a gun.
“Atypical” Handgun Owners
Missing in this discussion of the correlates of handgun ownership is whether there are differences between owning one or two handguns, what we call a “typical” amount, and owning three or more handguns, which we refer to as “atypical.” According to Parker, Horowitz, Igelniek, Oliphnat, and Brwon (2017), 66% of gun owners own more than one gun with 29% owning more than five guns, but of those who only own 1 gun, 70% are handgun owners. We do not know how many handguns these multiple gun owners have; the population of “atypical” multiple handgun owners needs to be explored (Parker et al., 2017). To our knowledge, no research has explored what factors contribute to how many handguns a handgun owner owns. Our study moves the literature forward by distinguishing between typical and atypical handgun ownership in our exploration of how socialization, victimization and the violent crime rate, political ideology, and societal insecurity relate to gun ownership. This is especially important as gun “stockpilers” have become the focus of media accounts, and an improved understanding of atypical handgun owners will inform this public discourse (LaFrance, 2016).
Present Study
In the present study, we use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine a variety of theoretical explanations for why young adults own handguns. Furthermore, we look at whether people who own many handguns differ from more typical handgun owners who own one or two and from people who do not own a handgun.
Importantly, the richness of the Add Health data allows us to address many of the limitations of prior research. First, the data are nationally representative and longitudinal, enabling a more rigorous examination than was often possible in prior studies using more localized samples or cross-sectional designs. Second, the large sample and variety of high-quality measures enable an analysis in which we can examine many explanatory factors simultaneously to determine the strongest predictors. Furthermore, we are able to test quantitatively the concept of societal insecurity (Carlson, 2015) that heretofore only been examined using qualitative methods. Third, the measure of handgun ownership is very precise and asks specifically about personally owning a handgun. Prior studies often combined handguns with other kinds of guns which is problematic given what we know about the different purposes of ownership. Also, it was unclear who owned the gun as the question often referred only to the existence of a gun in the home. Fourth, we compare typical handgun owners who own one or two guns with owners who possess many handguns. No studies have examined the characteristics of people who feel the need to own more handguns than are required for personal protection. Although one possibility is that these “more extreme” owners are merely collectors and look similar to typical owners, it is also possible that they may be quite different demographically, socially, and contextually.
Hypotheses
Informed by prior empirical findings and the theoretical explanations derived from them, we examine the demographic correlates of handgun ownership in young adulthood as well as test the following hypotheses about why people own handguns. Testing these hypotheses also lends insight into why demographic differences in handgun ownership exist. Furthermore, we hypothesize about what differentiates people who own many more handguns than the average owner, a phenomenon that has not received much, if any, systematic attention:
Data
In the present study, we use data from Waves I and III of the National Longitudinal Study of Adolescent to Adult Health (Add Health). Wave I was conducted in 1994 to 1995, and Wave III was conducted in 2001 to 2002. The Wave III sample consists of 15,701 respondents. The sample design of Add Health is a stratified random sample of high schools in the United States. We also use contextual data that match census data to respondents from Wave III (Harris et al., 2009).
Dependent Variables
The first dependent variable, handgun ownership, is measured as a dichotomous variable. The respondent was asked if he or she owned a handgun and to not include one he or she owned for his or her employment.
Our second dependent variable is type of handgun owner which consists of a three category measure. This variable was created using a question that asked the respondent how many handguns you own that are not for work from zero to 10 or more. The first category is not owning a handgun, the second category is owning one to two handguns, and the third category is owning more than two handguns. We use one or two handguns versus three or more because Azrael et al. (2017) found that, on average, a gun owner owns five guns and most gun owners own both long guns and handguns. One to two handguns is the most likely average of handguns owned by a gun owner based on the overall number of guns. We also show supplemental findings using a count measure indicating the number of handguns. This variable was created using the response to the question of how many handguns does the respondent own. The response categories are 0 = no handgun, 1 = 1 handgun, 2 = 2 handguns, 3 = 3 handguns, 4 = 4 handguns, 5 = 5 handguns, 6 = 6 handguns, 7 = 7 handguns, 8 = 8 handguns, 9 = 9 handguns, 10 = 10 or more handguns.
Independent Variables
The first focal independent variable is socialization. As we cannot directly measure all aspects of socialization, we use two dimensions of socialization that represent how socialization may shape handgun ownership. The first dimension of socialization is living in the South. We use Wave III contextual data to create a dummy variable based on whether the respondent lived in the South. The second dimension of socialization is whether the respondent had access to a firearm in adolescence. The respondents were asked in Wave I whether they had access to a handgun, whether they had access to a shotgun, and whether they had access to a rifle. We created four mutually exclusive categories of adolescent gun exposure. The first category is whether the respondent had access to a handgun only. The second category is whether the respondent had access to a shotgun or rifle only. The third measure is having access to both a handgun and a long gun. The reference category is not having access to any guns in adolescence.
Victimization and the county-level violent crime rate are also focal independent variables. The violent crime rate is measured using the Uniform Crime Report of violent crimes per 100,000 people in the respondent’s county of residence at Wave III. We consider the local violent crime rate to represent a potential form of indirect victimization. Victimization is a sum scale created using the Wave III responses about victimization in the past 12 months for the following circumstances: Have you ever seen someone shoot or stab another person, Have you ever seen someone pulled a gun on you, Have you ever seen someone pulled a knife on you, Have you ever seen someone shot you, Have you ever seen someone stabbed you, You were beaten up but nothing was stolen, and You were beaten up and something was stolen from you? The response categories were 0 = no and 1 = yes for each question.
The next focal variable is political ideology, which is measured at Wave III using the variable that asks “in terms of politics, do you consider yourself conservative, liberal, or middle-of-the-road.” Response categories are 1 = very conservative, 2 = conservative, 3 = middle-of-the-road, 4 = liberal, and 5 = very liberal. We reverse-coded responses such that higher values indicate more political conservatism.
The last focal independent variable is societal insecurity. This is measured at Wave III as a sum scale of responses to three questions. The three questions were “how much do you agree or disagree with the following statements: I trust the federal government, the state government, and the local government.” The response categories are 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, and 5 = strongly disagree.
Control Variables
The first set of control variables are for demographic and adolescent characteristics. The first is gender. This is coded as 1 = female and 0 = male. We control for race/ethnicity using dummy variables for those who indicated they were non-Hispanic White, African American, Asian American, American Indian, Other race, or Hispanic of any race. Each race/ethnicity category is mutually exclusive. Another control is family socioeconomic status which is measured in adolescence and is based on the Bearman and Moody (2004) scale. Another control variable is age at Wave III.
We also control for delinquency in adolescence. Delinquency is a mean scale that consists of 11 measures. The measures indicate how many times in the last 12 months the respondent engaged in a variety of behaviors: paint graffiti; damage property; take something from store without paying for it; get into serious physical fight; hurt someone badly enough to need bandages or care from a doctor; run away from home; drive a car without its owner’s permission; steal something worth more than US$50; go into a house or building to steal something; use or threaten to use a weapon; sell Marijuana or other drugs; steal something worth less than US$50; take part in a group fight; and act loud, rowdy, or unruly in a public place. The response categories for each item were 0 = never, 1 = 1 or 2 times, 2 = 3 or 4 times, and 3 = 5 or more times.
The second set of control variables are the adult controls. The first adult control variable is the education of the respondent by Wave III. The categories are no high school degree, high school degree, associate’s degree, and bachelor’s degree. Yearly income of the respondent’s household was measured at Wave III, which was measured by using the actual yearly income of the respondent or the best guess of the respondent with what their yearly income is.
We also control for adult offending. Adult offending is a sum scale comprised of 11 measures of different offenses. Respondents are asked how many times in the last 12 months they did the following: steal something worth more than US$50; go into a house or building to steal something; use or threaten to use a weapon; sell marijuana or other drugs; steal something worth less than US$50; take part in a group fight; buy, sell, or hold stolen property; use someone else’s credit card and so on; deliberately write a bad check; pulled a gun or knife; and shot or stabbed someone. The response categories for each item were 0 = never, 1 = 1 or 2 times, 2 = 3 or 4 times, and 3 = 5 or more times.
Another adult control variable is family structure. There are six mutually exclusive dummy categories of family structure at Wave III: Married and with at least one child (the reference category), married and with no children, cohabiting and with at least one child, cohabiting and with no children, single and with at least one child, and single and with no children. Finally, we also include a measure of urbanicity at Wave III. This is a dichotomous variable, that is, 1 = lives in an urban area and 0 = does not live in an urban area.
Analytic Strategy and Results
First, we examine descriptive statistics to compare the social, cultural, and demographic characteristics of adults who own and do not own handguns. Second, we use logistic regression to simultaneously test how socialization, victimization and the violent crime rate, political ideology, and societal insecurities are related to the likelihood of gun ownership in young adulthood. Third, we use multinomial logistic regression to determine whether any characteristics differentiate those who own one or two typical handguns versus a larger, more atypical number. The multinomial analysis is limited to men because there are few women who own a large number of handguns. We also provide supplemental findings showing that we considered other ways of modeling gun ownership (e.g., as a count measure) and whether the correlates of handgun ownership differed by gender. We used multiple imputation on all of the independent variables, which allowed us to account for missing data. Also, we used sample weights in the multivariate analysis to make the sample nationally representative and to correct for the complex sampling design.
Descriptive Statistics
In Table 1, we show descriptive statistics for the full sample partitioned by handgun ownership and note statistically significant differences at thep < .05 level. Consistent with past research, handgun owners are disproportionately males; 75% of handgun owners are men, whereas nonowners are mostly (54%) women. The heavily male nature of handgun ownership colors the overall portrait of owner characteristics. For example, handgun owners have higher levels of juvenile delinquency and adult offending than nonowners, but this might be because men are overrepresented among gun owners. Similarly, handgun owners are slightly older than nonowners, and this might explain why owners are more likely to be married or cohabiting and with kids and to have higher incomes despite having less education.
Means and Proportions of Independent Variables by Handgun Ownership.
Note. SES = socioeconomic status.
p < .05.
Turning to the primary characteristics of interest, consistent with past research on gun culture and socialization, handgun owners more often live in rural areas and in the South. Similarly, there appears to be an intergenerational relationship surrounding ownership; while 47% of handgun owners grew up with a gun in the home, only 22% of nonowners grew up with guns. Handgun owners also have higher levels of victimization than nonowners. Again, it remains unclear whether this is an artifact of gender composition. Handgun owners live in less violent counties than nonowners (which could be explained by higher levels of rural living among gun owners). Handgun owners are more socially conservative than those who do not own handguns. Also, handgun owners feel more socially insecure than nonowners.
Multivariate Results
Table 2 shows the unstandardized coefficients for a series of nested logistic regression models examining the adolescent and adult correlates of handgun ownership. Model 1 includes demographic and adolescent control variables. Model 2 adds the adult control variables. In Models 3 through 6, we add each of the four hypothesized focal variables separately to examine the effects of socialization; victimization; and the county-level violent crime rate, conservatism, and societal insecurity on handgun ownership, respectively. This also enables us to explore whether any of these hypothesized correlates mediate, or “explain,” the relationship between control variables and the likelihood of owning a handgun. Model 7 is a full model that includes all of the variables.
Logistic Regression of Handgun Ownership on Adolescent and Adult Predictors (Unstandardized Coefficients;n = 14,322).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Several findings stand out in Table 2, Model 1. Women are considerably less likely to own a handgun than men. There is no Black–White difference in handgun ownership, but other racial/ethnic groups are less likely to own handguns than White and Black adults. Asians, in particular, are the least likely to own a handgun. Age is positively associated with handgun ownership, and family socioeconomic status (SES) in adolescence is negatively associated with ownership. Adolescents who engaged in more delinquency were more likely to own a handgun as adults than less delinquent youth.
In Model 2, we add adult covariates to the demographic and adolescent controls presented in Model 1. Examining two indicators of socioeconomic status in adulthood, higher levels of education are associated with lower levels of handgun ownership, but income is positively related to ownership. Similar to delinquency, adult offending is positively related to handgun ownership. Adults living in single-person households are less likely to own a handgun than adults living in married households. This might be because there are fewer opportunities for ownership with fewer potential buyers in the household. Similarly, age is positively related to gun ownership. Growing older likely provides more opportunities for the accumulation of weapons. Consistent with prior research, adults living in urban areas are less likely to own a handgun than adults living in rural areas.
In Models 3 to 6, we show the effects of the focal variables on handgun ownership. In Model 3, we examine socialization. The odds of owning a handgun are 71% higher for adults living in the South than for adults living in other parts of the country (odds = exp[0.534]). Similarly, people who grew up with guns in the home are more likely to own handguns as adults. People who grew up with only a handgun or only a long gun (i.e., a rifle or shotgun) were similarly more likely to own a handgun in adulthood than people who grew up with no guns in the household (odds = 1.67 and 1.56, respectively). People with the highest likelihood of handgun ownership in adulthood are those who grew up with both handguns and long guns in the home; the odds of owning a handgun in adulthood are 3.4 times greater for those who grew up with both handguns and long guns compared with adults who did not have any guns in the home as adolescents.
After adding socialization measures in Model 3, there are two notable changes in coefficients from Model 2. First, the Asian–White gap in handgun ownership has been almost halved when the socialization measures are included. None of the other focal variables modify the race–handgun ownership relationship. Supplemental analysis (not shown) shows that the gun in the home measures are mediating the Asian–White gap. That is, Asians are less likely to grow up with guns in the home than are Whites. Second, the effect of living in an urban area is reduced by approximately 40% suggesting that part of the “urban effect” is compositional. That is, people living in urban areas are less likely to own handguns because those same people are also less likely to live in the South and less likely to have grown up with guns in the home.
The results in Model 4 indicate that people living in counties with more violent crime are less likely to own a handgun. This does not support the idea that gun ownership results from an objective threat of crime and might be an artifact of unmeasured differences in urbanization or population density. However, personally experiencing victimization is positively related to the likelihood of owning a handgun. Interestingly, after adding county-level violence and victimization to the model, the effect of adult offending on handgun ownership is reduced by approximately 35%. Supplemental analysis (not shown) shows that adolescent delinquency, adult offending, and adult victimization are strongly related factors correlated with an increased likelihood of gun ownership and suggest a relationship with stable, long-term involvement in risky behavior.
Model 5 shows that political ideology is related to handgun ownership. People who hold more politically conservative beliefs are more likely to own a handgun. Results in Model 6 indicate that feelings of societal insecurity also are associated with an increased likelihood of owning a handgun. Neither conservatism nor societal insecurity appears to mediate any of the effects of the control variables on handgun ownership.
Model 7 shows the effects of all the controls and all of the focal independent variables. In this full model, all the focal variables remain unchanged except that the county-level violent crime rate is no longer a significant predictor of owning a handgun. Living in the South, growing up with guns in the home, experiencing victimization, being politically conservative, and feeling societally insecure are all independently associated with an increased likelihood of owning a handgun. To assess the relative strength of these associations, I examine the standardized coefficients of the full model.
Tables 3 and 4 show the same models as Table 2 split be gender. A Chow (1960) test using an ordinary least squares (OLS) regression model showed that the models splitting men and women are not significantly different from each other, so we are justified in presenting the combined model in Table 2.
Logistic Regression of Handgun Ownership on Adolescent and Adult Predictors Among Men (Unstandardized Coefficients; n = 6,680).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Logistic Regression of Handgun Ownership on Adolescent and Adult Predictors Among Women (Unstandardized Coefficients; n = 7,503).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Table 5 shows the standardized coefficients for the full model. To ease comparisons across variables, we combined groups of categorical variables like those used for race/ethnicity, family structure, and growing up with a gun in the home into composite variables (Bollen & Bauldry, 2011; Heise, 1972). Composite variables are constructed by summing the component variables (indicators) weighted by their unstandardized coefficients from the full model. This enables us to compare the relative strength of the blocks of categorical variables against the other individual focal variables in the model.
Standardized Coefficients for Predictors of Handgun Ownership (n = 14,322).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Gender is the strongest correlate of owning a handgun even after accounting for a number of other demographic, social, and contextual factors. Among control variables, adolescent delinquency is the second strongest correlate. Looking at the focal variables, growing up with a gun in the home is the strongest correlate of owning a handgun in young adulthood (B = .200), followed by living in the South (B = .150) and experiencing victimization (B = .134). Conservatism (B = .061) and societal insecurity (B = .069) are relatively weaker correlates of owning a handgun. Therefore, although victimization, conservatism, and societal insecurity are associated with increased likelihoods of handgun ownership, the results show that socialization clearly has the strongest effect on the likelihood of owning a handgun.
In Table 6, we show the results of a multinomial logistic regression examining possible differences among handgun owners to determine whether any characteristics differentiate those who own one or two typical handguns versus a larger, more atypical number. The analysis is limited to men because there are few women who own a large number of handguns. Out of 7,505 women in the sample, only 20 owned three or more handguns. Also, due to small cell sizes with some family structure groups, we now operationalize family structure as single, cohabiting, and married instead of separating out whether they have children. Married is the new reference group for the family structure variables.
Multinomial Logistic Regression of Handgun Ownership on Adolescent and Adult (Unstandardized Coefficients) Predictors for Men (1-2 Handguns is Reference; n = 6,738).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Table 6 shows the results of the full model with the “typical” gun owner as the comparison group. Notably, of the focal independent variables, only conservative ideology and experiencing victimization were predictive of owning three or more handguns as compared with owning one or two handguns. Some other interesting findings do emerge comparing typical and atypical handgun owners. Hispanics are less likely to be “atypical” handgun owners than Whites. Also, adult offending is positively associated with atypical handgun ownership. This suggests that owning several handguns could be seen as necessary to either protect oneself while offending or using a handgun in the offending behavior. Thus, “atypical” handgun owners may not be the “good guys” with guns that Kleck and Gertz (1995) are describing, but rather may be engaged in riskier activities.
We also examined whether a count, ordinal, or multinomial measure better examined the relationship between the amount of handguns owned and the correlates of handgun ownership. Table 7 shows the results of an ordered logistic regression using the same three handgun groups as Table 6. We find that the ordered logistic regression results are very similar to those using the binary measure of handgun ownership. We believe this is because the association between the correlates and whether an individual owns no handguns compared with owning one to two handguns is a stronger relationship that between owning one to two handguns and owning three or more. Therefore, we use the multinomial approach because it fits better with other theoretical conceptualization emphasizing different types of gun owners. Similarly, Table 8 shows the results for a count measure of handgun ownership using negative binomial regression. The findings shown in Table 8 are consistent with the results in Table 6, but the relationship between groups is clearer in Table 6. We believe multinomial logistic regression to be the most appropriate analysis for explaining differences between those who own a few handguns and those who own more.
Ordered Logistic Regression of Handgun Ownership on Adolescent and Adult Predictors (Unstandardized Coefficients; n = 6,758).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Negative Binomial Regression of Handgun Ownership on Adolescent and Adult Predictors (Unstandardized Coefficients; n = 6,758).
Note. SES = socioeconomic status.
p < .05. **p < .01. ***p < .001.
Discussion
Previous research on why people own handguns is surprisingly limited. Few prior studies examine multiple predictors of handgun ownership at the same time, and there is little to no research on why people own large, atypical numbers of weapons. In the present study, we used a nationally representative sample and a longitudinal design to examine simultaneously several of the predictors for handgun ownership outlined in the prior literature: socialization (Cao et al., 1997), victimization and fear of crime (DeFronzo, 1979; Hill et al., 1985; Kleck et al., 2011; Turner et al., 2016; Wallace, 2017), political ideology (Cook & Ludwig, 1997; Hepburn et al., 2007), and societal insecurities (Carlson, 2015; Jiobu & Curry, 2001).
We found support for most of our hypotheses. For Hypothesis 1, we found that both socialization measures, growing up with a gun in the house and living in the South, are related to increased likelihoods of being a handgun owner in young adulthood. In fact, we found that of all the focal independent variables, the socialization measures had the strongest effect on the likelihood of owning a handgun. Hypothesis 2 was partially supported in that victimization was positively related to the likelihood of owning a handgun, but the rate of county-level violent crime had no effect in the full models. We also found support for Hypothesis 3. Political conservatism is positively associated with increased handgun ownership. Hypothesis 4 was also supported as societal insecurity that was positively related to owning a handgun. We established only partial support for Hypothesis 5, which predicted that the correlates would differentiate owning an atypical amount of handguns from owning a typical amount of handguns. Experiencing victimization and being more conservative were the only focal independent variables that distinguished being an atypical handgun owner versus being a typical handgun owner. This suggests that although the strongest predictors of handgun ownership in general are the socialization measures, the difference between owning a typical amount and owning an atypical amount is driven by conservative ideology, victimization, and offending.
Previous research has shown that most handgun owners claim they own their handguns for protection (Cook & Ludwig, 1997; Hemenway, 2011; Hepburn et al., 2007). Our finding that victimization is positively associated with the likelihood of owning a handgun provides support for this idea. However, offending behavior in adulthood is also positively related to handgun ownership, suggesting that there is more going on here. If victimization is the reason that people become handgun owners and we know that those who offend are more likely to be victims, then we might expect that offending would not predict handgun ownership when victimization was added to the model (Lauritsen, Sampson, & Laub, 1991). Victimization did reduce the effect of offending but did not explain all of the relationship between offending and handgun ownership. The socialization measures also support the idea of an intergenerational transmission of the need for a handgun for protection. Living in the South is associated with an increased likelihood of handgun ownership, signaling this protection may be seen as necessary or an accepted way to avoid victimization.
Our research on who owns handguns, and attempts to better understand why, may help inform public debates and policy about guns. Our findings show that both adolescent and adult socialization processes are linked to handgun ownership. The fact that growing up with a gun in the home predicts adult handgun ownership suggests an intergenerational transmission of handgun ownership. Also, living in the South as an adult was linked to an increased likelihood of owning a handgun even after taking into account growing up with a gun in the home. Therefore, the normalizing of handgun ownership occurs outside of the home as well. In this way, gun ownership is seen as common, shedding light on why gun owners are so resistant to policies that limit handgun ownership. When you combine this normalization of handgun ownership with the effect of societal insecurity, which shows a lack of faith in the government, and the effect of conservatism, which has an engrained resistance to regulation coupled with a literal understanding of the Constitution, it becomes more clear why handgun ownership is such a tightly held right by handgun owners. Policies that threaten their status of “handgun owner” could be construed as attacks on what they believe to be normal, by a government they do not have faith in, and ultimately seen as a violation of their rights. Any legislation to restrict handgun ownership would have to address these concerns.
Limitations
This study does have limitations. We did not have a subjective measure of fear of crime or victimization. We used a measure of county-level violent crime to attempt to examine the effect of exposure to violence in context, an indirect form of victimization, on handgun ownership. Even this “objective” measure would have been stronger if it measured violent crime levels at the neighborhood level. An individual may not be aware of the actual violent crime rate or may have unique perceptions of crime risk that we are unable to tap in the present study.
Our measure of societal insecurity is also limited because it is only a measure of trust in the local, state, and federal governments and does not involve insecurity in ways beyond trust in government. A more nuanced measure would include how the respondent feels about police or the state of their city or neighborhood.
Also, our study relies on a school-based sample. It may be missing respondents who were unable to regularly attend school due to homelessness or other factors such as truancy or antisocial behavior. As such, our sample might be “lower risk” and our findings may not be generalizable to the entire young adult population. Also, our sample is young for measuring handgun ownership. Cook and Ludwig (1997) claimed that the most common handgun owners were over 40 years of age. Since respondents in Wave III of Add Health are relatively young, only 18 to 24 years of age, the predictors of ownership may differ in young adulthood compared with older adulthood when behaviors and attitudes might be different and possibly more stable. However, examining handgun ownership in young adulthood may be advantageous in that there has been less time to accumulate guns as gifts or inheritance, and it more accurately reflects the active choice to own guns.
Despite these limitations, we believe that our study provides unique breadth by examining many of the correlates of handgun ownership found in previous studies as well as the precision of a more nuanced measure of handgun ownership. As such, we were able to examine standardized coefficients to determine which focal variables had the strongest effects on the likelihood of handgun ownership. We were also able to explore differences between typical owners and more extreme owners, a distinction that has not been examined before but holds considerable public interest.
Conclusion
The rise in the U.S. gun stock since 1994 has largely been driven by handguns (Azrael et al., 2017). This is despite a decrease in the percentage of U.S. households that own a firearm (Azrael et al., 2017). It is important to understand who these handgun owners are and why they own a handgun. Previous research suggested that socialization, fear of crime and victimization, political ideology, and societal insecurity can predict who owns a handgun. Besides gender, the strongest correlate overall, we found that socialization is the strongest correlate of owning a handgun which signals an intergenerational transmission of handgun ownership. However, our results showed that socialization does not predict owning three or more handguns compared with owning one to two handguns.
Future research should further explore the differences between typical and atypical handgun owners. Although we found only a few differences between the two groups, specifically conservatism and societal insecurity, disentangling how socialization, indirect and direct victimization, political ideology, and societal insecurity hang together could help to explain differences between typical and atypical handgun owners. Another way to explore typical and atypical handgun owners would be to examine rural and urban handgun owners separately. We controlled for urbanicity in this study, but examining the two groups separately may reveal distinct cultures of handgun ownership.
Footnotes
Acknowledgements
The authors thank Raymond Swisher, Jorge Chavez, and Susan Brown for their helpful feedback of an earlier draft of this work. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by the Center for Family and Demographic Research, Bowling Green State University,which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD050959). Also this research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the Uni- versity of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Information on how to obtain the Add Health data files is available on the Add Health website (
). No direct support was received from grant P01-HD31921 for this analysis.