Preventive Medicine 79 (2015) 50–54
Contents lists available at ScienceDirect
Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed
Assessing views about gun violence reduction policy: A look at type of violence and expected effectiveness Susan B. Sorenson University of Pennsylvania, 3815 Walnut Street, Philadelphia, PA 19104, USA
a r t i c l e
i n f o
Available online 5 May 2015 Keywords: Cross-sectional surveys Experimental design Firearms Guns Policy Public opinion Public opinion poll Public policy Students Violence
a b s t r a c t Objective. Public opinion polling about gun policy is routinely conducted and often disregarded. The purpose of this research is to explore ways in which surveys can be made more useful to policy makers, researchers, and the general public. Methods. A stratified random sample of 1000 undergraduates at a private, urban university was recruited for an online survey about proposed gun policies. A total of 51.7% answered the questions analyzed herein. Including but going beyond typical assessments of agreement, the survey elicited respondent evaluations of the effectiveness of seven gun policies under two randomly assigned conditions: the type of gun violence (e.g., homicide, suicide, violent crime) and its magnitude. Participants were asked to estimate the effectiveness of each policy, including the possibility of making things worse. Results. Participants indicated strong support for all policies and expected each to be effective with one exception — a policy designed to increase the number of guns on the scene, that is, putting armed police in schools. Persons who did not support other policies, on average, did not expect them to make things worse. Telling participants about the scope of the violence did not but the type of gun violence did affect effectiveness ratings. Conclusions. Asking about expected effectiveness of (vs. general support for) a policy might identify some optimism: Even people who don't support a policy sometimes think it will be effective. Findings suggest that surveys about the effectiveness of gun violence policies likely assess views that exclude suicide, the most common form of gun-related mortality. © 2015 Elsevier Inc. All rights reserved.
Introduction Policymakers' disregard for the views of their constituents was perhaps on no better display than when, in 2013, roughly 90% of the American people favored universal background checks for the purchase of a gun (PollingReport.com, accessed 6 January 2015) and Congress failed to pass such legislation. In a representative government, the wishes of the people and the actions of their elected officials are expected to correspond relatively closely. However, that has not been the case when it comes to gun policy, perhaps the single most surveyed policy topic in recent years. Full consideration of the possible reasons for the discrepancy is beyond the scope of this article. Instead, we return to the surveys themselves and the importance of measuring views of the public, views that, in theory and expectation, can shape and inform policy. In addition, measuring views of the public is important in order to provide members of that public with a sense of how their individual opinions compare to those of their fellow citizens. Survey researchers typically present participants with a dichotomous choice between two abstractions. For example, the Pew Research Center
E-mail address: [email protected].
http://dx.doi.org/10.1016/j.ypmed.2015.04.025 0091-7435/© 2015 Elsevier Inc. All rights reserved.
(December, 2014) asked survey respondents whether they “… support gun rights or gun control”, Fox News (April, 2013) asked “Which is more important: protecting the constitutional right of citizens to own guns or protecting citizens from gun violence?”, and an ABC News/ Washington Post poll (April, 2013) asked which should be a higher priority: “…enacting new laws to try to reduce gun violence, or protecting the right to own guns?” Some polls and on-going surveys assess support for specific types of firearm-related policies that are either in place or under consideration (e.g., restricting ownership of high capacity clips/magazines) and find substantial public support for a wide range of policy options (Barry et al., 2013). Only a few take the next step and assess perceptions of the efficacy of such policies. And, when they do, respondents typically are asked whether they think the policy will be very effective, somewhat effective, not very effective, or not at all effective. The present investigation includes the possibility that respondents believe that certain policies could make things worse. Moreover, it is not clear what content surveys intend to assess when they ask respondents about “gun violence.” Surveys asking about policies related to “reducing automobile accidents” likely would be met with puzzlement; there are many types and sources of motor vehicle crashes (e.g., distracted driving, pedestrian-involved, speed-related) that require different interventions (e.g., passage and enforcement of
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
lower speed limits in school zones, ignition interlock [blow-and-go] devices for repeatedly drunk drivers, bike lanes). Similarly, gun violence takes multiple forms. Public health researchers note that there are more gun-related suicides than gun-related homicides each year in the U.S. (Centers for Disease Control and Prevention, 2014), that most people who attempt suicide do not go on to commit suicide (Bostwick and Pankratz, 2000), and that many unintentional gun-related deaths are eminently preventable (Vernick et al., 1999, 2003). Such optimism is not necessarily shared by the general public, however; for example, nearly three-quarters in the U.S. consider suicide to be largely not preventable (Miller et al., 2006). Presenting gun violence as a unitary and undefined concept might reflect – and perpetuate – the public's lack of knowledge and their unwillingness or inability to perceive distinctions among forms of gun violence. The present investigation, unlike most surveys: 1) asked respondents to estimate the effectiveness of the policies, 2) provided a frame of reference for ratings of effectiveness by specifying the outcome that might be affected by the policy, and 3) allowed respondents to indicate that they think the policy might make things worse, not just better. Four hypotheses were tested in a sample of college students: 1) Support for gun violence prevention policies will be high and policies will be perceived to be more effective than harmful; 2) Anticipated effectiveness of a policy will increase if information about the scope of the problem is provided; 3) Expected effectiveness of a policy will be similar for when the type of gun violence is not specified (as is the case with most opinion polls) and for when violent crime is specified; and 4) The anticipated efficacy of gun violence policies will differ by the type of gun violence, in particular, suicide will be viewed as different from other forms of gun violence. Methods Sample A stratified random sample of 1000 undergraduates (250 freshman, 250 sophomores, 250 juniors, and 250 seniors) at a northeastern university was recruited to participate in an online survey about gun policy in March 2013. Researchers were provided with only the e-mail addresses of the individuals sampled. To have reasonable confidence that the person who completed the questionnaire was the one invited to participate, at the conclusion of the survey, respondents were asked to provide the e-mail address at which the investigators contacted them. After this validity check was completed, e-mail addresses were removed from the data analysis file. A total of 517 respondents completed the survey questions analyzed herein and provided an e-mail address that was on the sample list. This yielded a participation rate that compares favorably to similar online surveys. The selfreported demographics (i.e., age, gender, ethnicity, international status, socioeconomic status, school year, and academic division) of the resulting sample were comparable to those of the undergraduate student body.
Data collection instrument The questionnaire asked about seven policies that were proposed by various sources (Vice President Biden's Commission on Gun Violence, the Johns Hopkins Gun Policy Research Center, and the National Rifle Association). See Table 1 for a list of the policies. Respondents were asked to respond yes or no to indicate whether they “do or do not” support each policy. The section on support was followed by an embedded experiment. Respondents were randomly assigned to one of five types of gun violence (homicide, suicide, rampage/mass shooting, violent crime, or accidental shooting). Respondents then were randomly assigned to the provision of information: half were presented with a statement of fact about the scope of the violence (see Table 2) and half were not. Fig. 1 shows the eleven conditions to which respondents were assigned — ten indicating a type of gun violence and one in which no form of gun violence was specified. The size of the groups ranged from 45 to 49. Respondents were asked to consider this information – the type and, if provided, the scope of the violence – when answering questions about effectiveness.
51
Table 1 Proposed policies examined. 1. Ban the sale of military-style, semiautomatic assault weapons that are capable of shooting more than 10 rounds of ammunition without reloading 2. Ban the sale of large-capacity ammunition clips or magazines that allow some guns to shoot more than 10 bullets before reloading 3. Require a background check system for all gun sales to make sure a purchaser is not legally prohibited from having a gun 4. Require states to report a person to the background-check system who is prohibited from buying a gun either because of involuntary commitment to a hospital for psychiatric treatment or because of being declared mentally incompetent by a court of law 5. Allow cities to sue licensed gun dealers when there is strong evidence that the gun dealer's careless sales practices allowed many criminals to obtain guns 6. Maximize enforcement efforts to prevent gun violence and prosecute gun crime 7. Put armed police officers in every school
The next section addressed expected effectiveness. Survey research that asks about anticipated effectiveness typically asks respondents to indicate whether they think a policy will be very effective, somewhat effective, not very effective, or not effective at all. This standard practice eliminates half of the potential distribution of responses; respondents do not have the option of indicating that the proposed policy might not be simply ineffective but might make things worse. Thus, after each policy, we provided a scale from − 100 (“a LOT worse”) to +100 (“a LOT better”) with “0” marked as “no effect” and asked respondents to slide a marker to the point on the scale that corresponded to their assessment of the influence the proposed policy would have on that type of gun violence. Respondents were asked to make such a rating for each policy and “All of these policies together.” The mean and median expected effectiveness ratings were compared and the findings were similar, suggesting that the ratings distributions were not badly skewed; means are reported herein. In order to compare findings to those obtained with current survey practice, we included one condition that did not specify the type of gun violence. In other words, standard practice does not provide a referent, thus, when answering a question, respondents could be thinking of homicide or violent crime or any one of several forms of gun violence.
Statistical analysis Simple frequencies were calculated for each question and by condition for expected efficacy. Descriptive results are displayed graphically. Multivariate regressions and their accompanying F-tests were conducted to examine the impact of type of violence on the expected effectiveness of each policy. And, finally, we calculated the difference between the expected efficacy ratings for suicide and the expected efficacy ratings for the other forms of gun violence in order to assess whether the anticipated effectiveness of the policies differed for suicide.
Table 2 Type of gun violence, with and without information about its magnitude, specified in questions about effectiveness. 1. How do you think each policy would affect suicide? 2. About half of the suicides in the U.S. are by a firearm. How do you think each policy would affect suicide? 3. How do you think each policy would affect homicide? 4. About 2/3s of the homicides in the U.S. are by a firearm. How do you think each policy would affect homicide? 5. How do you think each policy would affect violent crime? 6. About 23% of the violent crime in the U.S. involves a gun. How much do you think each policy would affect violent crime? 7. How do you think each policy would affect accidental shootings? 8. About 600 people are accidentally shot and killed each year in the U.S. How do you think each policy would affect accidental shootings? 9. How do you think each policy would affect rampage/mass shootings? 10. In 2012, 45 people were killed in rampage/mass shootings in the U.S. How do you think each policy would affect mass shootings? 11. How effective do you think the policy would be? (i.e., no type of gun violence was specified)
52
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
Participants were asked to indicate, on a scale of − 100 (a LOT worse) to +100 (a LOT better), how effective they expected each policy to be. As shown in Fig. 3 and as hypothesized, all but one of the individual policies were expected to be effective. For example, on a scale of −100 to +100 respondents thought that universal background checks merited an average of 54, which indicates that they think the background checks would reduce gun violence. The expected effectiveness of the policies ranged from 34 (cities can sue dealers) to 54 (universal background checks). Only one of the policies was expected to make the problem worse: putting armed police in schools. When support for the policy (yes–no) was taken into account, again except for putting armed police in schools, each policy was expected to improve the situation. When respondents supported a policy, the average effectiveness rating for each policy ranged between 36 (cities can sue dealers) to 54 (universal background checks). When respondents did not support a policy, the average effectiveness rating for each policy ranged from 2 (allow cities to sue certain gun dealers) to 39 (universal background checks). Even when respondents opposed the policies related to background checks, they thought they would be somewhat effective (24 for requiring states to report mental health records and 39 for universal background checks). Information about magnitude of the violence
Fig. 1. Question type randomly assigned to each participant, which was followed by ratings of expected effectiveness.
Contrary to what was hypothesized, informing respondents about the scope of a particular type of gun violence did not affect – that is, did not increase or decrease – their perceptions of the efficacy of the policy. This finding (not tabled) was consistent in regressions taking into account the type of gun violence specified, whether information was provided, and the interaction of the type of gun violence by whether information about the magnitude of the problem was provided. The discussion section further addresses this finding. Type of violence
Results Support and expected effectiveness As shown in Fig. 2, with one exception (armed guards in schools) support was high for each of the proposed policies. Six of the seven policies garnered the support of more than 80% of the participants (range: 84.1%–97.5%).
As hypothesized, gun suicide was perceived as being different from the other forms of intentional and unintentional gun violence. Table 3 reports differences in the mean expected efficacy ratings between suicide and each other form of violence. The differences were generally consistent; banning assault weapons, banning large capacity clips, allowing cities to sue gun dealers, and maximizing enforcement and prosecution were seen as more effective in reducing homicide, violent crime, unintentional gun deaths, and mass shootings than in reducing
ban assault weapons
ban large capacity clips
universal background checks
states report mh records
cities can sue dealers
enforce & prosecute
armed officers in schools 0
10
20
30
40
50
60
70
80
Fig. 2. Support for seven proposed gun violence reduction policies, 517 undergraduates, March 2013, %.
90
100
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
53
Ban assault weapons Ban large capacity clips Universal background checks States report mh records Cities can sue dealers Enforce and prosecute Armed officers in schools -20
-10
0
10
20
supported
30
40
50
60
did not support
Fig. 3. Expected effectiveness of seven proposed gun violence reduction policies, 517 undergraduates, March 2013, a LOT worse (−100) to a LOT better (+100).
suicide. For example, respondents expected banning assault weapons to be more effective in reducing homicide than in reducing suicide (mean rating difference = 23). The expected effectiveness of requiring states to report their mental health records and placing armed officers in schools largely did not differ for suicide and the other types of gun violence. Moreover, findings suggested that when other types of gun violence (e.g., homicide, violent crime) are specified and when the type of gun violence is not specified (as is the case in most current survey research), respondents appear to exclude suicide from consideration of the expected effectiveness of a policy. Subsequent regressions indicated that, regardless of whether information about the scope of the violence was provided, participants' expectations of each policy's effectiveness were substantially lower for suicide than for other forms of gun violence. Discussion Current opinion polls about gun-related policies typically ask respondents whether they agree or disagree (or support or don't support) a given policy. They ignore context, that is, they rarely specify exactly what type of violence a policy is designed to address nor do they ask respondents about their expectations of the effectiveness of a policy. In this paper we looked at both of those factors and this is what we found: Consistent with surveys of the general public (e.g., Barry et al., 2013), the vast majority of respondents supported a wide range of gun policies. The exception was the proposal to put armed police officers in schools, the sole policy that would increase the number of guns on the scene. The
lack of support for having armed officers in schools might be related to the fact that the sample is comprised of students; although there clearly is overlap between enforcement and protection, students might perceive armed officers in educational settings as less protective than supervisory and patrolling. Again with the exception of putting armed officers in schools, policies were expected to be more helpful than harmful when it comes to reducing gun violence. A finding that might not be a surprise is that policies that are supported are more likely to be deemed effective than are policies that are not supported. It is worth noting, however, that when people support a policy they don't necessarily expect high levels of effectiveness. This is in marked contrast to how gun-related policy is framed as of late: proposed policies to reduce gun violence typically are held to a high and somewhat simplistic bar — does it or will it work. Participants' expectations were relatively modest and perhaps more realistic; they do not expect a single policy to eliminate or substantially reduce gun violence. Findings indicate that the association between support and expected effectiveness isn't necessarily as might be anticipated: a policy can be opposed and still be deemed effective or at least not to make things worse. Universal background checks are one such example; those who did not support the policy nonetheless rated it as a 39 (−100 = a LOT worse; 0 = no effect; 100 = a LOT better). And, conversely, people can think a policy could reduce gun violence and still be against it. They likely give priority to countervailing factors (e.g., concerns about government “over reach”) rather than to concerns about gun violence. Thus, assessments of support for a policy should not be equated with assessments of its expected effectiveness.
Table 3 Expected effectiveness of seven proposed gun violence policies, mean ratings for five forms of gun violence compared to the mean rating for suicide.
Ban assault weapons Ban large capacity clips Universal background checks States report mh records Cities can sue dealers Enforce & prosecute Armed officers in schools
n
Homicide
Violent crime
488 467 501 504 480 482 428
23⁎⁎⁎ 20⁎⁎⁎ 16⁎⁎⁎
25⁎⁎⁎ 26⁎⁎⁎ 18⁎⁎⁎
21⁎⁎⁎ 20⁎⁎⁎ 12⁎
33⁎⁎⁎ 30⁎⁎⁎ 13⁎
32⁎⁎⁎ 28⁎⁎⁎ 19⁎⁎⁎
–6 21⁎⁎⁎ 25⁎⁎⁎ 15⁎
–1 13⁎⁎ 22⁎⁎⁎ 3
−9 18⁎⁎⁎ 16⁎⁎⁎
−3 14⁎⁎ 24⁎⁎⁎ 24⁎⁎⁎
2 18⁎⁎ 20⁎⁎⁎ 1
Note. Expected effectiveness ranged from −100 (a LOT worse) to +100 (a LOT better). ⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.
Unintentional
6
Mass shooting
No type specified
54
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
Empirically informing respondents is rare in surveys about gun violence so we experimentally tested whether providing information about the scope of gun violence would increase perceptions of policy efficacy. It did not. Political science researchers have documented that providing information affects respondents' support for a hypothetical policy. It is possible that information would affect agreement with (but not expected effectiveness of) a gun-related policy, a possibility that was not examined in this study. Perhaps information other than or in addition to the magnitude of the problem would affect assessments of expected effectiveness. For example, respondents might respond differently if told “Nearly 80 of 100 suicide attempts with a gun are fatal but only about 3 of 100 with pills are fatal. About half of the suicides in the U.S. are by a firearm.” rather than just “About half of the suicides in the U.S. are by a firearm.” In our experiment, ratings of expected effectiveness were very similar for gun homicide, violent crime, unintentional gun death, and mass shootings as well as when the type of gun violence was unspecified. We expected “violent crime” and no specification to receive similar ratings, which did occur, however, there was little difference in the ratings for all five of the categories. Whether current survey practice of not specifying a type of gun violence when asking about support for a policy (vs. expected effectiveness of a policy, as we asked) is affected by the designation of a type of violence has not, to our knowledge, been examined. By contrast, suicide is viewed differently. A range of policies believed to be potentially effective in reducing a diverse set of types of gun violence are not expected to be as effective for suicide. Thus, it appears that respondents thought of suicide as separate from and, at least with the seven policies examined, less preventable. Most similar for all types of gun violence is the policy related to mental health, that is, requiring states to report to the background check system the names of those involuntarily hospitalized for psychiatric treatment or declared mentally incompetent by a court. Thus, keeping guns from those with severe mental health problems are expected to have a similar influence on suicide, homicide, violent crime, unintentional gun deaths, mass shootings, and when a type of gun violence is not specified.
Conclusions Aside from the idea of putting armed police in schools, support for policies designed to reduce gun violence was high among the sampled college students. The expected effectiveness of a policy was higher among respondents who supported a policy but sometimes even those who did not support a policy thought it could be effective. Telling respondents about the magnitude of the violence did not influence their ratings of expected effectiveness; perhaps more or different information than simply the scope of the problem would influence their assessments. One could reasonably expect that the perceived efficacy of a firearmrelated policy would differ by the type of firearm violence but our findings suggest that people might actually think of a unitary “gun violence problem” when anticipating the effectiveness of a policy. At least that appears to be the case for the seven potential policies and five of the six types of violence examined herein. The sole exception was suicide. The policies generally were perceived as least efficacious in reducing suicide. This suggests two avenues for the future: Suicide, the most common form of mortality by guns in the U.S., might be brought into conversations and thinking about gun violence more generally, and policies that might reduce suicide by gun merit additional consideration. Notably, one policy – universal background checks – was expected to reduce all of the forms of violence. Funding None. Human participant protection The research was reviewed and approved by the institutional review board of the University of Pennsylvania. Conflict of interest statement The author has no conflicts of interest.
References Strengths and limitations In terms of strengths, the present investigation used an experimental design to assess cause and effect in assessments of expected effectiveness of a policy. In addition, the survey, unlike most opinion surveys about gun policies: 1) asked respondents to estimate the effectiveness of the policies, 2) allowed respondents to indicate that they think the policy might make things worse, not just better or to have no impact, and 3) provided a frame of reference for ratings of effectiveness by specifying the outcome (i.e., the type of gun violence) that might be affected by the policy. The response rate was relatively high for an online survey. The sample is both a strength and a limitation. Relatively little information is available about college student attitudes despite the recent expansion of gun rights activities to college campuses (e.g., open carry on campus, separate residence halls for gun owners). Moreover, current college students are the first generation to have grown up with school shootings as a semi-regular part of their educational careers; thus, their perspectives are of particular interest. The stratified random sample helped us to effectively depict the perspectives of the students. However, given the narrow age range and the single setting, whether findings hold for other age cohorts and persons in other locales merits investigation.
ABC News/Washington Post Poll, April 11–14 2013. . Available from:, http://www. pollingreport.com/guns.htm (Accessed 6 January 2015). Barry, C.L., Mcginty, E.E., Vernick, J.S., Webster, D.W., 2013. After Newtown — public opinion on gun policy and mental illness. N. Engl. J. Med. 368 (12), 1077–1081. http://dx.doi.org/10.1056/NEJMp1300512. Bostwick, J.M., Pankratz, V.S., 2000. Affective disorders and suicide risk: a reexamination. Am. J. Psychiatry 157 (12), 1925–1932. http://dx.doi.org/10.1176/appi.ajp.157.12. 1925. Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, 2014. Web-based Injury Statistics Query and Reporting System (WISQARS). [online]. Available from:, www.cdc.gov/ncipc/wisqars (Accessed 3 December 2014). FOX News Poll, April 20–22 2013. . Available from:, http://www.pollingreport.com/guns. htm (Accessed 6 January 2015). Miller, M., Azrael, D., Hemenway, D., 2006. Belief in the inevitability of suicide: results from a national survey. Suicide Life Threat. Behav. 36 (1), 1–11. http://dx.doi.org/ 10.1521/suli.2006.36.1.1. Pew Research Center, 2014. Growing Public Support for Gun Rights. December. Available from:, http://www.people-press.org/files/2014/12/12-10-14-Gunsrelease.pdf (Accessed 6 January 2015). Pollingreport.com, d. Guns. [Online] Available from:, http://www.pollingreport.com/guns. htm [Accessed 6 January 2015]. Vernick, J.S., Meisel, Z.F., Teret, S.P., Milne, J.S., Hargarten, S.W., 1999. “I didn't know the gun was loaded”: an examination of two safety devices that can reduce the risk of unintentional firearm injuries. J. Public Health Policy 20 (4), 427–440. http://dx.doi. org/10.2307/3343129 http://www.jstor.org/stable/3343129. Vernick, J.S., O'Brien, M., Hepburn, L.M., Johnson, S.B., Webster, D.W., Hargarten, S.W., 2003. Unintentional and undetermined firearm related deaths: a preventable death analysis for three safety devices. Inj. Prev. 9 (4), 307–311. http://dx.doi.org/10. 1136/ip.9.4.307.
Contents lists available at ScienceDirect
Preventive Medicine journal homepage: www.elsevier.com/locate/ypmed
Assessing views about gun violence reduction policy: A look at type of violence and expected effectiveness Susan B. Sorenson University of Pennsylvania, 3815 Walnut Street, Philadelphia, PA 19104, USA
a r t i c l e
i n f o
Available online 5 May 2015 Keywords: Cross-sectional surveys Experimental design Firearms Guns Policy Public opinion Public opinion poll Public policy Students Violence
a b s t r a c t Objective. Public opinion polling about gun policy is routinely conducted and often disregarded. The purpose of this research is to explore ways in which surveys can be made more useful to policy makers, researchers, and the general public. Methods. A stratified random sample of 1000 undergraduates at a private, urban university was recruited for an online survey about proposed gun policies. A total of 51.7% answered the questions analyzed herein. Including but going beyond typical assessments of agreement, the survey elicited respondent evaluations of the effectiveness of seven gun policies under two randomly assigned conditions: the type of gun violence (e.g., homicide, suicide, violent crime) and its magnitude. Participants were asked to estimate the effectiveness of each policy, including the possibility of making things worse. Results. Participants indicated strong support for all policies and expected each to be effective with one exception — a policy designed to increase the number of guns on the scene, that is, putting armed police in schools. Persons who did not support other policies, on average, did not expect them to make things worse. Telling participants about the scope of the violence did not but the type of gun violence did affect effectiveness ratings. Conclusions. Asking about expected effectiveness of (vs. general support for) a policy might identify some optimism: Even people who don't support a policy sometimes think it will be effective. Findings suggest that surveys about the effectiveness of gun violence policies likely assess views that exclude suicide, the most common form of gun-related mortality. © 2015 Elsevier Inc. All rights reserved.
Introduction Policymakers' disregard for the views of their constituents was perhaps on no better display than when, in 2013, roughly 90% of the American people favored universal background checks for the purchase of a gun (PollingReport.com, accessed 6 January 2015) and Congress failed to pass such legislation. In a representative government, the wishes of the people and the actions of their elected officials are expected to correspond relatively closely. However, that has not been the case when it comes to gun policy, perhaps the single most surveyed policy topic in recent years. Full consideration of the possible reasons for the discrepancy is beyond the scope of this article. Instead, we return to the surveys themselves and the importance of measuring views of the public, views that, in theory and expectation, can shape and inform policy. In addition, measuring views of the public is important in order to provide members of that public with a sense of how their individual opinions compare to those of their fellow citizens. Survey researchers typically present participants with a dichotomous choice between two abstractions. For example, the Pew Research Center
E-mail address: [email protected].
http://dx.doi.org/10.1016/j.ypmed.2015.04.025 0091-7435/© 2015 Elsevier Inc. All rights reserved.
(December, 2014) asked survey respondents whether they “… support gun rights or gun control”, Fox News (April, 2013) asked “Which is more important: protecting the constitutional right of citizens to own guns or protecting citizens from gun violence?”, and an ABC News/ Washington Post poll (April, 2013) asked which should be a higher priority: “…enacting new laws to try to reduce gun violence, or protecting the right to own guns?” Some polls and on-going surveys assess support for specific types of firearm-related policies that are either in place or under consideration (e.g., restricting ownership of high capacity clips/magazines) and find substantial public support for a wide range of policy options (Barry et al., 2013). Only a few take the next step and assess perceptions of the efficacy of such policies. And, when they do, respondents typically are asked whether they think the policy will be very effective, somewhat effective, not very effective, or not at all effective. The present investigation includes the possibility that respondents believe that certain policies could make things worse. Moreover, it is not clear what content surveys intend to assess when they ask respondents about “gun violence.” Surveys asking about policies related to “reducing automobile accidents” likely would be met with puzzlement; there are many types and sources of motor vehicle crashes (e.g., distracted driving, pedestrian-involved, speed-related) that require different interventions (e.g., passage and enforcement of
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
lower speed limits in school zones, ignition interlock [blow-and-go] devices for repeatedly drunk drivers, bike lanes). Similarly, gun violence takes multiple forms. Public health researchers note that there are more gun-related suicides than gun-related homicides each year in the U.S. (Centers for Disease Control and Prevention, 2014), that most people who attempt suicide do not go on to commit suicide (Bostwick and Pankratz, 2000), and that many unintentional gun-related deaths are eminently preventable (Vernick et al., 1999, 2003). Such optimism is not necessarily shared by the general public, however; for example, nearly three-quarters in the U.S. consider suicide to be largely not preventable (Miller et al., 2006). Presenting gun violence as a unitary and undefined concept might reflect – and perpetuate – the public's lack of knowledge and their unwillingness or inability to perceive distinctions among forms of gun violence. The present investigation, unlike most surveys: 1) asked respondents to estimate the effectiveness of the policies, 2) provided a frame of reference for ratings of effectiveness by specifying the outcome that might be affected by the policy, and 3) allowed respondents to indicate that they think the policy might make things worse, not just better. Four hypotheses were tested in a sample of college students: 1) Support for gun violence prevention policies will be high and policies will be perceived to be more effective than harmful; 2) Anticipated effectiveness of a policy will increase if information about the scope of the problem is provided; 3) Expected effectiveness of a policy will be similar for when the type of gun violence is not specified (as is the case with most opinion polls) and for when violent crime is specified; and 4) The anticipated efficacy of gun violence policies will differ by the type of gun violence, in particular, suicide will be viewed as different from other forms of gun violence. Methods Sample A stratified random sample of 1000 undergraduates (250 freshman, 250 sophomores, 250 juniors, and 250 seniors) at a northeastern university was recruited to participate in an online survey about gun policy in March 2013. Researchers were provided with only the e-mail addresses of the individuals sampled. To have reasonable confidence that the person who completed the questionnaire was the one invited to participate, at the conclusion of the survey, respondents were asked to provide the e-mail address at which the investigators contacted them. After this validity check was completed, e-mail addresses were removed from the data analysis file. A total of 517 respondents completed the survey questions analyzed herein and provided an e-mail address that was on the sample list. This yielded a participation rate that compares favorably to similar online surveys. The selfreported demographics (i.e., age, gender, ethnicity, international status, socioeconomic status, school year, and academic division) of the resulting sample were comparable to those of the undergraduate student body.
Data collection instrument The questionnaire asked about seven policies that were proposed by various sources (Vice President Biden's Commission on Gun Violence, the Johns Hopkins Gun Policy Research Center, and the National Rifle Association). See Table 1 for a list of the policies. Respondents were asked to respond yes or no to indicate whether they “do or do not” support each policy. The section on support was followed by an embedded experiment. Respondents were randomly assigned to one of five types of gun violence (homicide, suicide, rampage/mass shooting, violent crime, or accidental shooting). Respondents then were randomly assigned to the provision of information: half were presented with a statement of fact about the scope of the violence (see Table 2) and half were not. Fig. 1 shows the eleven conditions to which respondents were assigned — ten indicating a type of gun violence and one in which no form of gun violence was specified. The size of the groups ranged from 45 to 49. Respondents were asked to consider this information – the type and, if provided, the scope of the violence – when answering questions about effectiveness.
51
Table 1 Proposed policies examined. 1. Ban the sale of military-style, semiautomatic assault weapons that are capable of shooting more than 10 rounds of ammunition without reloading 2. Ban the sale of large-capacity ammunition clips or magazines that allow some guns to shoot more than 10 bullets before reloading 3. Require a background check system for all gun sales to make sure a purchaser is not legally prohibited from having a gun 4. Require states to report a person to the background-check system who is prohibited from buying a gun either because of involuntary commitment to a hospital for psychiatric treatment or because of being declared mentally incompetent by a court of law 5. Allow cities to sue licensed gun dealers when there is strong evidence that the gun dealer's careless sales practices allowed many criminals to obtain guns 6. Maximize enforcement efforts to prevent gun violence and prosecute gun crime 7. Put armed police officers in every school
The next section addressed expected effectiveness. Survey research that asks about anticipated effectiveness typically asks respondents to indicate whether they think a policy will be very effective, somewhat effective, not very effective, or not effective at all. This standard practice eliminates half of the potential distribution of responses; respondents do not have the option of indicating that the proposed policy might not be simply ineffective but might make things worse. Thus, after each policy, we provided a scale from − 100 (“a LOT worse”) to +100 (“a LOT better”) with “0” marked as “no effect” and asked respondents to slide a marker to the point on the scale that corresponded to their assessment of the influence the proposed policy would have on that type of gun violence. Respondents were asked to make such a rating for each policy and “All of these policies together.” The mean and median expected effectiveness ratings were compared and the findings were similar, suggesting that the ratings distributions were not badly skewed; means are reported herein. In order to compare findings to those obtained with current survey practice, we included one condition that did not specify the type of gun violence. In other words, standard practice does not provide a referent, thus, when answering a question, respondents could be thinking of homicide or violent crime or any one of several forms of gun violence.
Statistical analysis Simple frequencies were calculated for each question and by condition for expected efficacy. Descriptive results are displayed graphically. Multivariate regressions and their accompanying F-tests were conducted to examine the impact of type of violence on the expected effectiveness of each policy. And, finally, we calculated the difference between the expected efficacy ratings for suicide and the expected efficacy ratings for the other forms of gun violence in order to assess whether the anticipated effectiveness of the policies differed for suicide.
Table 2 Type of gun violence, with and without information about its magnitude, specified in questions about effectiveness. 1. How do you think each policy would affect suicide? 2. About half of the suicides in the U.S. are by a firearm. How do you think each policy would affect suicide? 3. How do you think each policy would affect homicide? 4. About 2/3s of the homicides in the U.S. are by a firearm. How do you think each policy would affect homicide? 5. How do you think each policy would affect violent crime? 6. About 23% of the violent crime in the U.S. involves a gun. How much do you think each policy would affect violent crime? 7. How do you think each policy would affect accidental shootings? 8. About 600 people are accidentally shot and killed each year in the U.S. How do you think each policy would affect accidental shootings? 9. How do you think each policy would affect rampage/mass shootings? 10. In 2012, 45 people were killed in rampage/mass shootings in the U.S. How do you think each policy would affect mass shootings? 11. How effective do you think the policy would be? (i.e., no type of gun violence was specified)
52
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
Participants were asked to indicate, on a scale of − 100 (a LOT worse) to +100 (a LOT better), how effective they expected each policy to be. As shown in Fig. 3 and as hypothesized, all but one of the individual policies were expected to be effective. For example, on a scale of −100 to +100 respondents thought that universal background checks merited an average of 54, which indicates that they think the background checks would reduce gun violence. The expected effectiveness of the policies ranged from 34 (cities can sue dealers) to 54 (universal background checks). Only one of the policies was expected to make the problem worse: putting armed police in schools. When support for the policy (yes–no) was taken into account, again except for putting armed police in schools, each policy was expected to improve the situation. When respondents supported a policy, the average effectiveness rating for each policy ranged between 36 (cities can sue dealers) to 54 (universal background checks). When respondents did not support a policy, the average effectiveness rating for each policy ranged from 2 (allow cities to sue certain gun dealers) to 39 (universal background checks). Even when respondents opposed the policies related to background checks, they thought they would be somewhat effective (24 for requiring states to report mental health records and 39 for universal background checks). Information about magnitude of the violence
Fig. 1. Question type randomly assigned to each participant, which was followed by ratings of expected effectiveness.
Contrary to what was hypothesized, informing respondents about the scope of a particular type of gun violence did not affect – that is, did not increase or decrease – their perceptions of the efficacy of the policy. This finding (not tabled) was consistent in regressions taking into account the type of gun violence specified, whether information was provided, and the interaction of the type of gun violence by whether information about the magnitude of the problem was provided. The discussion section further addresses this finding. Type of violence
Results Support and expected effectiveness As shown in Fig. 2, with one exception (armed guards in schools) support was high for each of the proposed policies. Six of the seven policies garnered the support of more than 80% of the participants (range: 84.1%–97.5%).
As hypothesized, gun suicide was perceived as being different from the other forms of intentional and unintentional gun violence. Table 3 reports differences in the mean expected efficacy ratings between suicide and each other form of violence. The differences were generally consistent; banning assault weapons, banning large capacity clips, allowing cities to sue gun dealers, and maximizing enforcement and prosecution were seen as more effective in reducing homicide, violent crime, unintentional gun deaths, and mass shootings than in reducing
ban assault weapons
ban large capacity clips
universal background checks
states report mh records
cities can sue dealers
enforce & prosecute
armed officers in schools 0
10
20
30
40
50
60
70
80
Fig. 2. Support for seven proposed gun violence reduction policies, 517 undergraduates, March 2013, %.
90
100
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
53
Ban assault weapons Ban large capacity clips Universal background checks States report mh records Cities can sue dealers Enforce and prosecute Armed officers in schools -20
-10
0
10
20
supported
30
40
50
60
did not support
Fig. 3. Expected effectiveness of seven proposed gun violence reduction policies, 517 undergraduates, March 2013, a LOT worse (−100) to a LOT better (+100).
suicide. For example, respondents expected banning assault weapons to be more effective in reducing homicide than in reducing suicide (mean rating difference = 23). The expected effectiveness of requiring states to report their mental health records and placing armed officers in schools largely did not differ for suicide and the other types of gun violence. Moreover, findings suggested that when other types of gun violence (e.g., homicide, violent crime) are specified and when the type of gun violence is not specified (as is the case in most current survey research), respondents appear to exclude suicide from consideration of the expected effectiveness of a policy. Subsequent regressions indicated that, regardless of whether information about the scope of the violence was provided, participants' expectations of each policy's effectiveness were substantially lower for suicide than for other forms of gun violence. Discussion Current opinion polls about gun-related policies typically ask respondents whether they agree or disagree (or support or don't support) a given policy. They ignore context, that is, they rarely specify exactly what type of violence a policy is designed to address nor do they ask respondents about their expectations of the effectiveness of a policy. In this paper we looked at both of those factors and this is what we found: Consistent with surveys of the general public (e.g., Barry et al., 2013), the vast majority of respondents supported a wide range of gun policies. The exception was the proposal to put armed police officers in schools, the sole policy that would increase the number of guns on the scene. The
lack of support for having armed officers in schools might be related to the fact that the sample is comprised of students; although there clearly is overlap between enforcement and protection, students might perceive armed officers in educational settings as less protective than supervisory and patrolling. Again with the exception of putting armed officers in schools, policies were expected to be more helpful than harmful when it comes to reducing gun violence. A finding that might not be a surprise is that policies that are supported are more likely to be deemed effective than are policies that are not supported. It is worth noting, however, that when people support a policy they don't necessarily expect high levels of effectiveness. This is in marked contrast to how gun-related policy is framed as of late: proposed policies to reduce gun violence typically are held to a high and somewhat simplistic bar — does it or will it work. Participants' expectations were relatively modest and perhaps more realistic; they do not expect a single policy to eliminate or substantially reduce gun violence. Findings indicate that the association between support and expected effectiveness isn't necessarily as might be anticipated: a policy can be opposed and still be deemed effective or at least not to make things worse. Universal background checks are one such example; those who did not support the policy nonetheless rated it as a 39 (−100 = a LOT worse; 0 = no effect; 100 = a LOT better). And, conversely, people can think a policy could reduce gun violence and still be against it. They likely give priority to countervailing factors (e.g., concerns about government “over reach”) rather than to concerns about gun violence. Thus, assessments of support for a policy should not be equated with assessments of its expected effectiveness.
Table 3 Expected effectiveness of seven proposed gun violence policies, mean ratings for five forms of gun violence compared to the mean rating for suicide.
Ban assault weapons Ban large capacity clips Universal background checks States report mh records Cities can sue dealers Enforce & prosecute Armed officers in schools
n
Homicide
Violent crime
488 467 501 504 480 482 428
23⁎⁎⁎ 20⁎⁎⁎ 16⁎⁎⁎
25⁎⁎⁎ 26⁎⁎⁎ 18⁎⁎⁎
21⁎⁎⁎ 20⁎⁎⁎ 12⁎
33⁎⁎⁎ 30⁎⁎⁎ 13⁎
32⁎⁎⁎ 28⁎⁎⁎ 19⁎⁎⁎
–6 21⁎⁎⁎ 25⁎⁎⁎ 15⁎
–1 13⁎⁎ 22⁎⁎⁎ 3
−9 18⁎⁎⁎ 16⁎⁎⁎
−3 14⁎⁎ 24⁎⁎⁎ 24⁎⁎⁎
2 18⁎⁎ 20⁎⁎⁎ 1
Note. Expected effectiveness ranged from −100 (a LOT worse) to +100 (a LOT better). ⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.
Unintentional
6
Mass shooting
No type specified
54
S.B. Sorenson / Preventive Medicine 79 (2015) 50–54
Empirically informing respondents is rare in surveys about gun violence so we experimentally tested whether providing information about the scope of gun violence would increase perceptions of policy efficacy. It did not. Political science researchers have documented that providing information affects respondents' support for a hypothetical policy. It is possible that information would affect agreement with (but not expected effectiveness of) a gun-related policy, a possibility that was not examined in this study. Perhaps information other than or in addition to the magnitude of the problem would affect assessments of expected effectiveness. For example, respondents might respond differently if told “Nearly 80 of 100 suicide attempts with a gun are fatal but only about 3 of 100 with pills are fatal. About half of the suicides in the U.S. are by a firearm.” rather than just “About half of the suicides in the U.S. are by a firearm.” In our experiment, ratings of expected effectiveness were very similar for gun homicide, violent crime, unintentional gun death, and mass shootings as well as when the type of gun violence was unspecified. We expected “violent crime” and no specification to receive similar ratings, which did occur, however, there was little difference in the ratings for all five of the categories. Whether current survey practice of not specifying a type of gun violence when asking about support for a policy (vs. expected effectiveness of a policy, as we asked) is affected by the designation of a type of violence has not, to our knowledge, been examined. By contrast, suicide is viewed differently. A range of policies believed to be potentially effective in reducing a diverse set of types of gun violence are not expected to be as effective for suicide. Thus, it appears that respondents thought of suicide as separate from and, at least with the seven policies examined, less preventable. Most similar for all types of gun violence is the policy related to mental health, that is, requiring states to report to the background check system the names of those involuntarily hospitalized for psychiatric treatment or declared mentally incompetent by a court. Thus, keeping guns from those with severe mental health problems are expected to have a similar influence on suicide, homicide, violent crime, unintentional gun deaths, mass shootings, and when a type of gun violence is not specified.
Conclusions Aside from the idea of putting armed police in schools, support for policies designed to reduce gun violence was high among the sampled college students. The expected effectiveness of a policy was higher among respondents who supported a policy but sometimes even those who did not support a policy thought it could be effective. Telling respondents about the magnitude of the violence did not influence their ratings of expected effectiveness; perhaps more or different information than simply the scope of the problem would influence their assessments. One could reasonably expect that the perceived efficacy of a firearmrelated policy would differ by the type of firearm violence but our findings suggest that people might actually think of a unitary “gun violence problem” when anticipating the effectiveness of a policy. At least that appears to be the case for the seven potential policies and five of the six types of violence examined herein. The sole exception was suicide. The policies generally were perceived as least efficacious in reducing suicide. This suggests two avenues for the future: Suicide, the most common form of mortality by guns in the U.S., might be brought into conversations and thinking about gun violence more generally, and policies that might reduce suicide by gun merit additional consideration. Notably, one policy – universal background checks – was expected to reduce all of the forms of violence. Funding None. Human participant protection The research was reviewed and approved by the institutional review board of the University of Pennsylvania. Conflict of interest statement The author has no conflicts of interest.
References Strengths and limitations In terms of strengths, the present investigation used an experimental design to assess cause and effect in assessments of expected effectiveness of a policy. In addition, the survey, unlike most opinion surveys about gun policies: 1) asked respondents to estimate the effectiveness of the policies, 2) allowed respondents to indicate that they think the policy might make things worse, not just better or to have no impact, and 3) provided a frame of reference for ratings of effectiveness by specifying the outcome (i.e., the type of gun violence) that might be affected by the policy. The response rate was relatively high for an online survey. The sample is both a strength and a limitation. Relatively little information is available about college student attitudes despite the recent expansion of gun rights activities to college campuses (e.g., open carry on campus, separate residence halls for gun owners). Moreover, current college students are the first generation to have grown up with school shootings as a semi-regular part of their educational careers; thus, their perspectives are of particular interest. The stratified random sample helped us to effectively depict the perspectives of the students. However, given the narrow age range and the single setting, whether findings hold for other age cohorts and persons in other locales merits investigation.
ABC News/Washington Post Poll, April 11–14 2013. . Available from:, http://www. pollingreport.com/guns.htm (Accessed 6 January 2015). Barry, C.L., Mcginty, E.E., Vernick, J.S., Webster, D.W., 2013. After Newtown — public opinion on gun policy and mental illness. N. Engl. J. Med. 368 (12), 1077–1081. http://dx.doi.org/10.1056/NEJMp1300512. Bostwick, J.M., Pankratz, V.S., 2000. Affective disorders and suicide risk: a reexamination. Am. J. Psychiatry 157 (12), 1925–1932. http://dx.doi.org/10.1176/appi.ajp.157.12. 1925. Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, 2014. Web-based Injury Statistics Query and Reporting System (WISQARS). [online]. Available from:, www.cdc.gov/ncipc/wisqars (Accessed 3 December 2014). FOX News Poll, April 20–22 2013. . Available from:, http://www.pollingreport.com/guns. htm (Accessed 6 January 2015). Miller, M., Azrael, D., Hemenway, D., 2006. Belief in the inevitability of suicide: results from a national survey. Suicide Life Threat. Behav. 36 (1), 1–11. http://dx.doi.org/ 10.1521/suli.2006.36.1.1. Pew Research Center, 2014. Growing Public Support for Gun Rights. December. Available from:, http://www.people-press.org/files/2014/12/12-10-14-Gunsrelease.pdf (Accessed 6 January 2015). Pollingreport.com, d. Guns. [Online] Available from:, http://www.pollingreport.com/guns. htm [Accessed 6 January 2015]. Vernick, J.S., Meisel, Z.F., Teret, S.P., Milne, J.S., Hargarten, S.W., 1999. “I didn't know the gun was loaded”: an examination of two safety devices that can reduce the risk of unintentional firearm injuries. J. Public Health Policy 20 (4), 427–440. http://dx.doi. org/10.2307/3343129 http://www.jstor.org/stable/3343129. Vernick, J.S., O'Brien, M., Hepburn, L.M., Johnson, S.B., Webster, D.W., Hargarten, S.W., 2003. Unintentional and undetermined firearm related deaths: a preventable death analysis for three safety devices. Inj. Prev. 9 (4), 307–311. http://dx.doi.org/10. 1136/ip.9.4.307.
