Crime, Violence, and Political Participation

Abstract

The paper explores the relationship between crime victimization and political participation, making the argument that the observed (and oft-cited) relationship in surveys is the result of response bias. We show that people who are more extroverted, efficacious, and opinionated are more likely to respond positively to a question about victimization, and they are more likely to participate in politics. The implications of this finding are important and widespread—both for survey researchers interested in sensitive questions and for research on the relationship between crime and political engagement.

There is an emerging conventional wisdom in political science that experiences with violence and crime victimization may have a silver lining in terms of promoting political participation. In a much-cited paper in the American Political Science Review, Bateson (2012) makes the case people who have been victims of crime are more likely to participate in politics. She finds this relationship in surveys from regions around the world. Similarly, based on survey data, Blattman (2009) argues that experience with violence leads to higher levels of political engagement among ex-combatants. Both papers suggest that people who experience a violent event often react by becoming more politically active—attending local political meetings, protests, and political events. These studies are widely cited and contribute to a growing established belief that experiences with crime and violence have a positive impact on political participation. In this paper, we point to several empirical puzzles that raise doubts about the causal nature of this relationship and offer an alternative explanation for these findings. We worry that the relationship between victimization and political participation may be overstated. Bateson’s work presents a very large positive effect: In surveys from around the world, the estimated impact of crime victimization on participation “is generally equivalent to about 5-10 years of additional schooling” (2012, p. 575). At face value, this is difficult to accept. Although some people may respond to trauma by becoming involved in politics, it seems unlikely that almost everyone would.

We argue that the strong relationship between victimization and political participation may have another source: a particular form of response bias. If people who are more engaged in politics are more likely to respond to a question about crime victimization, we might see these results. We suspect that some people—particularly those who are more extroverted and efficacious—are both more likely to respond to survey questions about crime victimization and more likely to participate in politics. As a result, individual-level surveys show a strong relationship between crime victimization and political participation, but not because victimization is mobilizing people into politics. Rather, people who are more politically active—especially in nonvoting activities such as protest, contacting the government, and attending meetings—are more likely to report crime victimization in a face-to-face survey, compared with less politically active people, mainly because of personality differences. And, more troubling, there may be an unknown group of people who have been victimized who are unwilling to discuss it, and may be less participatory, making the true relationship very hard to uncover.

We show evidence that at least some of the bias in reporting crime victimization in face-to-face surveys is not the result of random underreporting. This is an important difference for several reasons. First, if positive responses to the crime victimization questions are biased in a way that is related to political participation, causal inference on the question of how crime victimization is related to political participation is very difficult using these survey tools. Second, although crime victimization is a particularly important example, it is likely that other sensitive questions face similar problems of bias related to political participation.

The paper proceeds as follows. First, we review recent scholarship on crime and political participation to demonstrate that the individual-level relationship between crime victimization and political participation found in surveys is at odds with much of the broader literature on the consequences of violent crime, which often finds crime suppresses political participation. We include a meta-analysis of the literature that shows a significant difference in the findings from studies that use survey data versus other types of data collection. We then make the case using descriptive statistics that bias in crime reporting on surveys may be partially to blame for these conflicting findings. The second part of the paper lays out our argument that political activity is directly associated with higher rates of willingness to answer questions about crime victimization. We detail several hypotheses consistent with our theory: First, we argue that people who are more efficacious, more opinionated, more extroverted, and more willing to answer sensitive survey questions are more likely to be politically active, and also more likely to say they have been a victim of a crime. Second, we argue that political activity is a strong predictor of those factors. Finally, we test these hypotheses using surveys from the AmericasBarometer.

Crime, Violence, and Participation: Debates and Conflicting Evidence

There are two very different stories about how experience with crime might affect political participation. For some, victimization is a source of activism. Out of anger or a sense of injustice, some people respond to crime with new energy for engaging with their political system to fight for change. Many activists explain their own path to political participation in these terms. For others, however, the terrifying experience of being a victim of a crime is a powerful deterrent to political participation as fear and caution win out. It is easy to imagine both these scenarios as reasonable responses to crime, and there most certainly are real-world examples of both. It is less clear which of these responses is more common, under what circumstances, and why.

Several key studies point to a positive relationship between victimization and participation. Bateson (2012), for example, argues that crime has a strong and positive effect on many kinds of political participation, including solving problems in the community, attending community meetings, protests, town meetings, and political meetings. She argues that crime leads to higher levels of political interest, willingness to talk about politics and try to persuade others, and willingness to act as a group leader. Bateson shows that there is a strong positive correlation between crime victimization and each of these types of participation, although voting is not included in the analysis. Crime victimization also translates into a greater likelihood of participating in cultural events (Reyes-Martínez et al., 2020). More extreme exposure to violence can also boost political activity. Blattman (2009) shows that after civil wars, people affected by violence often enter political life with new energy. In addition to greater involvement in community organizations, Blattman finds that male combatants who experienced the most violence in the Ugandan civil war were also the most likely to vote in national elections and serve as community leaders. Higher levels of altruism, community involvement, and political participation are also present in the villages that suffered the highest levels of violence during Nepal’s civil war (Gilligan et al., 2014). Similarly, Bellows and Miguel (2009) demonstrate that, in Sierra Leone, individuals who directly experienced intense war violence are more likely to attend community meetings, join local political and community groups, and vote than individuals in households that experienced less intense violence.

Crime and violence, however, are not always mobilizing forces. An abundance of evidence suggests the opposite. Victims of crime—and those living in high-crime areas—often feel insecure, alienated, and disconnected from their communities. Victims frequently fear the social stigma associated with victimization and are more likely to become withdrawn and unhappy. Victims of crime regularly suffer from paralyzing psychological issues, including fear, anxiety, and depression (e.g., Atkeson et al., 1982; Frieze et al., 1987; Norris & Kaniasty, 1992). Many are reluctant to speak out about their experiences due to fear of stigmatization or feelings of insecurity. This is especially true in places where violence may be linked to gang activity or organized crime. Victims not only fear judgment from neighbors that they brought the violence upon themselves, but also that speaking out will lead to reprisals from the perpetrator(s) (Villagrán, 2013). Unsurprisingly, victims of crime and those living in violent areas report lower levels of life satisfaction than peers in safer environments (Powdthavee, 2005) and are less likely to vote (Ley, 2018). This is particularly true for victims of violent crime. Victims of violent crime are less likely to vote, while victims of nonviolent crime are no less likely to vote than nonvictims (Berens & Dallendörfer, 2019). Frühling et al. (2003) suggest that victimization leads to feelings of alienation and helplessness, causing victims to disconnect from their communities.

Victims of crime and violence are also more likely to be mistrusting of political institutions, which may dampen some types of political participation. This is true for political institutions victims are more likely to interact with directly after crime, such as the police, attorney’s office, or prosecutor’s office, as well the national government and congress (Blanco, 2013; Blanco & Ruiz, 2013). Victims of crime also report lower levels of satisfaction with democratic institutions (Ceobanu et al., 2011; Fernandez & Kuenzi, 2010), and this relationship is exacerbated in places with police misconduct (Cruz, 2015). People who experience election irregularities, including violence or intimidation, are less likely to view elections as free or fair (Kerr, 2013). Feelings of insecurity erode interpersonal trust, an important predictor of community and political involvement (Putnam et al., 1993). More directly, most evidence that does not rely on individual-level measures of crime shows that voter turnout is depressed by crime. Trelles and Carreras (2012), for example, find that municipalities in Mexico with higher rates of crime consistently report lower voter turnout. Coleman (2002) reports that high crime rates are associated with lower turnout at the state and county levels in the United States. Cordova (2016) also shows that participation is lower in high-crime neighborhoods affected by gang activity in El Salvador. As this discussion makes clear, much of what we know about crime and participation does not fit well with the survey findings on the topic, which is a key motivation for this paper.

Meta-analysis

To explore these contradictions more systematically, we conducted a meta-analysis of the literature on crime, violence, and victimization. We split the studies included in the meta-analysis into two subgroups: (a) studies that use survey data and (b) studies that do not use survey data. To simplify comparison, each study included has been assembled into a standardized effect size index with a mean of zero and unit standard deviation. All control variables and weights employed by the studies included in this meta-analysis are taken at their stated values. Both random effect and fixed effect models are included in this meta-analysis. For our analysis, a random effect model is an appropriate model choice for comparison, as the underlying assumption of a fixed effect meta-analysis model is that there is a common effect generated by an identical underlying variable of interest for all studies included in the analysis. Random effect meta-analysis models relax this strict assumption of an identical underlying variable and allow the effect to vary across studies included in the sample (Borenstein et al., 2009). This is an important distinction, as variability in the nature and context of both crime and violence are likely to be heterogenous across studies.

The results, in both the fixed effect and random effect models, indicate that survey studies mostly exhibit positive effects of crime victimization and individuals’ probability of self-reporting pro-social behaviors. As Figure 1 shows, the studies that utilize survey data find much more positive and consistent findings than the studies that utilize nonsurvey data. In sum, studies that utilize individual-level survey responses as measures of both crime and political participation tend to find that crime victimization is associated with more participation while studies that use other data are more likely to find mixed results or that crime tends to depress participation. At the very least, there is good reason to believe that the effect of crime on participation would be conditional on other factors. In the next section, we offer a potential explanation for these contradictory findings about the effect of crime victimization on participation and the odd mismatch between the image of crime in Latin America described in surveys compared with what we know from other sources.

Personality, Political Engagement, and Willingness to Answer Sensitive Questions

There is good reason to be skeptical of self-reporting on an issue as fraught as crime victimization. We know, for example, that some groups are more likely to report crime than others and that often willingness to report is related to power and privilege. Latin Americans who live in poor neighborhoods are less likely to report instances of violence than their middle- and upper-class counterparts, even when levels of violence are comparable (Rubio, 1998). Although reporting to the police is different than talking about victimization to a survey researcher, it is possible that some of the same concerns or hesitations might apply. In fact, survey evidence paints quite a different picture from other types of data. If all we knew about crime in Latin America came from surveys, we would have an odd picture of what crime is like. Based on who answers, “yes,” to a question about recent crime victimization, we learn that wealthier, more educated, older men living in more peaceful countries are the people most often victimized in robberies, burglaries, assaults, fraud, blackmail, extortion, violent threats, or any other type of crime. This is puzzling considering most of what we know about crime in the region from other sources. Crime is almost always worse in poorer neighborhoods, and the bulk of crime is committed by and perpetrated against poor young people (United Nations Development Program, 2013). Poorer people are consistently shown to be at greater risk for victimization than their socioeconomically advantaged peers (Carrión et al., 2011).

Just looking at surveys, we would also learn that the most violent countries in Latin America are Peru, Argentina, Venezuela, and Uruguay, with over 20% of the population reporting victimization in the past 12 months. Except for Venezuela, this stands in stark contrast to the list of most dangerous countries based on country-level homicide rates. According to United Nations homicide rates, Honduras, El Salvador, Venezuela, and Guatemala, each with over 40 homicides per 100,000 people, are the most dangerous countries in Latin America. Conversely, Argentina and Uruguay have the lowest homicide rates, after Chile, and are well below the regional average of 23 homicides per 100,000. Although not perfect, homicide is an accepted proxy for violent crime and a widely used indicator of the level of security within a country (UNODC, 2013). As a robustness check, we also use measures of assault and theft from the UN Office on Drugs and Crime (https://dataunodc.un.org/), and the findings are very similar.

Women are also less likely to report victimization than men. An estimated 40% of women in Latin America become victims of gender motivated violence during their lifetime (Economic Commission for Latin America and the Caribbean, 2009), but women in the developing world often underreport crime victimization due to feelings of shame, perceived impunity, fear of retaliation, or because violence is simply viewed as normal (Palermo et al.,2013). Across all waves of the AmericasBarometer survey from 2006 to 2014, only 75 women reported victimization in the form of rape or sexual assault. Actual numbers are estimated to be far higher, as violence against women in the region is at an all-time high. It is likely that the surveys fail to capture the full extent of victimization, particularly for disadvantaged groups, which distorts the true relationship between victimization and political participation.

We argue that survey responses to questions about something as personal and troubling as crime victimization might not be reliable. We also argue that problem is not just one of random unwillingness to answer questions about being victimized but that willingness to respond to such questions might be strongly correlated with political engagement. People who feel more efficacious and empowered are often more participatory (Finkel, 1985; Verba et al., 1996; Wolak, 2018). We suspect they might also be more willing to discuss a potentially shame-inducing situation related to victimization. People who are less involved in their communities or in political life may also be less willing to respond to questions about crime—or more likely to underreport their own experiences of crime. Survey questions about crime victimization can be sensitive for all respondents, not just disadvantaged groups. Respondents frequently underreport experiences or behaviors that may be seen as socially undesirable (Tourangeau & Yan, 2007), and crime victimization is an experience that tends to be underreported in face-to-face surveys (for a review see Krumpal, 2011).

What if victims of crime who are politically active were much more likely to report their victimization when asked on a survey? The mismatch between official crime statistics and the crime victimization responses suggests that there may be some problems with the survey responses. Instead of random underreporting, we argue that the likelihood of reporting crime victimization in face-to-face surveys is associated with political activity and engagement through the mechanism of personality, giving false positive correlations that may not reflect causal relationships or true patterns. In any population, there is a true number of people who have experienced crime firsthand. Given that being victimized can be traumatic and embarrassing, some people might prefer to keep their experiences to themselves. Others share freely and answer with little hesitation. If underreporting is essentially random, we have a measurement problem that might affect the coefficient estimates in an analysis, but not a problem that seriously affects inference. We argue that people who are more confident, have higher feelings of efficacy, and are more likely to have extroverted personalities are more likely to be politically active and are also more likely to feel confident disclosing their own personal history with crime.

To test our argument, we consider the determinants of answering “yes” on a direct question about crime victimization to see whether political activity makes reporting more likely. These tests should not be read as a traditional regression analysis—to be clear, we are not arguing that these models serve as evidence of a causal relationship between political activity and actual crime victimization. Rather, we are using these models as a diagnostic, looking for evidence that political activity influences the willingness to positively report crime victimization. Necessarily, these models include a lot of error—it is unlikely that people who have never been victims would give a false positive, so our ability to use this measure as a true estimate of crime is limited. With that in mind, we hypothesize that if people who are more politically active are more likely to report crime, we should observe the following relationships:

First, we expect that more politically active people respond positively to crime victimization questions on surveys at higher rates than their less politically active peers. Second, we expect that people who engage in more intensive, higher effort political activities are the most likely to report crime victimization. Voting in most countries in Latin America is compulsory, and rates of voter turnout are very high. Since voting is an occasional, relatively low-cost activity that the majority of people engage in, we expect a stronger relationship between other, more costly, types of participation and willingness to respond positively to a crime victimization question. In sum, we hypothesize that more politically active people are more likely to respond positively to survey questions about being a victim of a crime.

Our argument is based on the idea that the same characteristics that are more common among people who participate in politics are also more common among those willing to talk openly about difficulty issues to a survey interviewer. However, these characteristics should not predict true experiences of victimization. We focus on four factors: (a) efficacy, (b) extroversion, (c) opinionation, and (d) willingness to answer sensitive questions unrelated to crime. People who feel more efficacious are more confident that their opinion matters. People who are generally extroverted may be less likely to withdraw in response to victimization and thus more likely to openly share. We expect that people with these characteristics are more likely to report being a victim of a crime—again, with the understanding that underreporting may be common and over-reporting very rare, meaning that even if we observe strong relationships, we are probably not estimating accurate models of crime victimization since the core argument here is that the measure of crime victimization is unreliable. Our second hypothesis is that people who are more efficacious, more extroverted, more opinionated, and more willing to answer sensitive questions are also more likely to respond positively to survey questions about being a victim of a crime.

Finally, if political participation is directly related to bias in crime reporting on surveys because more politically active people are generally more confident, we expect political activity to be a good predictor of the characteristics that make crime reporting more likely: efficacy, extroversion, opinionation, and willingness to answer sensitive questions. It is important to note that these hypotheses remove crime victimization from the equation entirely, and instead focus on whether the qualities that predict crime victimization are predicted by political activity. More politically active people are more efficacious, more extroverted, more opinionated, and more willing to answer other sensitive questions.

Research Strategy and Findings

To test our hypotheses, we primarily use a dataset composed of five waves of the AmericasBarometer survey from the Latin American Public Opinion Project (LAPOP), and additional country-level indicators. We also check our key findings using data from the Afrobarometer and the Asian Barometer surveys. The AmericasBarometer dataset covers 18 Latin American countries every other year from 2006 to 2014. There are between 1,000 and 3,000 respondents per country-year. Consistency in question wording across waves of the AmericasBarometer allows us to be confident that our findings are robust across time and countries.

The dependent variable, crime victimization, is measured using the survey question, “Have you been a victim of any type of crime in the past 12 months? That is, have you been a victim of robbery, burglary, assault, fraud, blackmail, extortion, violent threats, or any other type of crime in the past 12 months?” Of the those surveyed across Latin America between 2006 and 2014, roughly 19% reported they experienced crime in the last year. Crime victimization is measured dichotomously—one if the individual experienced crime and zero otherwise.

To capture a broad range of participatory actions, we estimate models using individual political acts, as well as an additive index of political actions. Our additive index is useful for two reasons. First, we can examine whether there is a difference between the number of acts victims engage in verses nonvictims. Second, because the results are the same whether we run individual models or use the additive index (see Online Appendix Table 1), the additive model is less cumbersome for the reader to understand and interpret. In our index, we examine the same acts of political engagement as Bateson (2012), but also expand our analysis to include voting, which other studies suggest is likely to be depressed by crime (Coleman, 2002; Trelles & Carreras, 2012), and participation that occurred prior to victimization.

To measure political participation, we use questions from the AmericasBarometer that document the respondents’ self-reported participation in a variety of activities. Vote is a dichotomous measure of whether the respondent voted in the last presidential election. Since presidential elections do not occur every year, models using vote are constrained to a period when the respondent could have voted and been a victim of crime in the past 12 months. Our political participation index includes the seven political activities used in Bateson (2012). The index is a count of the number of political activities the respondent engaged in during the same 12-month period when they could have experienced victimization and whether they are interested in politics. Activities include protesting, attending a town meeting, attending a community meeting, attending the meeting of a political party, helping to solve a community project, interest in politics, and trying to persuade someone to support a certain candidate or party. A zero indicates no participation, and seven indicates they participated in every activity.

We also examine the relationship between working on a campaign during the last presidential election and crime victimization. The model is restricted to cases where the individual could not have worked on a presidential campaign and been a victim of crime in the same 12-month period. This allows us to test whether prior political participation predicts reporting of victimization. Similarly, we test whether contacting a local official, a federal office, or a congressional representative impacts victimization. These survey questions ask respondents if they have ever engaged in these acts and are coded dichotomously. There is a possibility that respondents engaged in these acts after victimization, but we believe it is more likely that respondents are reflecting on participation that occurred in recent years, rather than months.

To test our second hypothesis, we construct measures of efficacy, extroversion, opinionation, and sensitive questions using LAPOP questions. Efficacy is the extent to which the respondents feel they “understand the most important political issues of this country,” and it is measured on a 7-point scale. In our sample, efficacy is normally distributed with the largest percentage of respondents reporting that they feel they have a mid-level understanding of political issues. Extroversion is the average of the degree to which the respondent rated themselves as “quiet and shy” and “social and active.” Most respondents in the sample are moderately to very extroverted.

Opinionation is another measure that can be used to determine if there is bias in the reporting of victimization. Some survey respondents are more likely than others to answer questions that require them to take a stance or state an opinion. We know that propensity to answer opinion-based survey questions about politics may be due to factors positively correlated with political knowledge and participation (Delli Carpini & Keeter 1996; Krosnick & Milburn 1990), and posit that opinionation should similarly predict propensity to truthfully answer sensitive survey questions. However, opinionation should not be correlated with experiences of victimization if the measure is unbiased. We adopt Carreras’s (2014) measure of opinionation that uses 10 random opinion questions from the 2008 AmericasBarometer survey to build an index capturing how likely individuals are to take a stance (i.e., did not respond “don’t know”) to opinion questions. Each question is coded one if they responded and zero if not. These questions are then combined into an opinion index ranging from 0 (answered none of the questions) to 10 (answered all). Roughly 70% of respondents answered all the randomly chosen opinion questions.

To capture propensity to answer sensitive questions, we use responses to questions about abortion and clientelism. Abortion is controversial and also illegal in most Latin American countries. In 2014, respondents were asked whether abortion should be permitted when a pregnancy poses a risk to the mother’s health. Fifty-six percent of respondents agreed that abortion should be permitted in such cases, while 43% felt that maternal health did not justify an abortion (Jelen & Bradley 2014). Similarly, we tested whether victims were more likely than nonvictims to report being offered a good or service in exchange for their vote for a political candidate or party (i.e., clientelism). Questions about clientelism or vote-buying may be similarly controversial, and there is some evidence from list experiments that people underreport vote-buying experiences in direct questions (Gonzalez-Ocantos et al., 2012). Only 11% of all respondents report being offered something in exchange for their political support.

Estimation Strategy

Due to the nested structure of the data, we use a multilevel estimation strategy (Snijders & Bosker, 2011). In each model, we include controls for common individual-level attributes known to impact political participation. These include gender, age, years of education, whether the respondent lives in a rural area, and a measure of asset-based wealth quintile. We also control for homicide rates as a proxy for the level of crime in each country using data from the United Nations Office on Drugs and Crime.

Findings

We find that political acts and engagement are among the strongest predictors of answering “yes” to the crime victimization question. The first model in Table 1 shows that for every additional act an individual participates in, they are more likely to report crime victimization. In Model 3, we find that individuals who volunteered for a campaign during the last presidential race are also more likely to report victimization. Because campaigning occurred prior to victimization, it most clearly suggests that participation in politics may make individuals more likely to report victimization, rather than victimization as a driver of participation. In Models 4–6, we find that respondents who have contacted a local official, federal office, or a congress member at any point in their lives are also more likely to report victimization. As expected, we find a null relationship between voting and reporting victimization. This supports our hypothesis that higher effort activities, such as working for a campaign and participating in multiple political activities, will be related to reporting victimization.

These results call into question the causal relationship between crime victimization and political participation. It is particularly telling that individuals who worked on a political campaign outside of the 12-month period when they could have been victimized are more likely to report victimization. Although individuals involved in politics are sometimes the targets of violence, there is little reason to believe that the targeting of politically active individuals can explain the strong relationship between crime victimization and political participation across individuals and contexts. It is more likely that individuals who are politically engaged are willing to discuss their experiences of victimization with a stranger. It also makes sense that voting does not have the same effect. In Latin America, where voting is often compulsory, it is a low-effort activity that most citizens engage in because they are legally obligated to do so. It is individuals who participate in higher-effort activities that overcome the fear of reporting their victimization experience on a face-to-face survey. We provide more evidence for this claim in the next section by demonstrating the strong relationship between reporting crime victimization and people who have attributes consistent with high levels of political participation.

In Table 2, we find that efficacy, extroversion, opinionation, and answering sensitive questions are all predictive of an individual’s propensity to report crime. Individuals who are more efficacious and extroverted are more likely to report on a face-to-face survey that they were victims of crime. Similarly, the more opinion questions a respondent is willing to answer instead of stating that they “do not know,” the more likely they are to report an experience with crime. Finally, respondents who are willing to state that they are pro-abortion, or that they have been offered a good or service in exchange for their vote (i.e., clientelism) are also more likely to report crime.

Taken together, these results suggest that the robust relationship between crime victimization and political participation is likely due, in part, to response bias. These political experiences and attitudes are theoretically unrelated to victimization but are likely closely tied to political participation. Our final analysis turns to the relationship between more politically active people and the qualities of being more efficacious, more extroverted, more opinionated, and more willing to answer other sensitive questions.

Figure 2 plots coefficients that show that political participation is, in fact, correlated with efficacy, extroversion, opinionation, and propensity to answer sensitive survey questions. Individuals who participate in high-cost activities, such as protesting and contacting representatives, and the lower effort act of voting are all more likely to be efficacious and extroverted. They are also more likely to offer an opinion on a variety of issues when asked, and to answer in the affirmative to sensitive questions. Taken together, these findings suggest that the strong relationship between crime victimization and political participation found in survey research may not be evidence that crime victimization increases political engagement. Instead, more extroverted and efficacious people may be more likely to report victimization and more likely to participation in politics.

Finally, we estimate two sets of models: (a) the relationship between crime victimization and political participation and (b) the same models with the inclusion of our measures of efficacy and extroversion. Figure 3 shows the marginal effects for both models for comparison. Clearly, the inclusion of the personality measures reduces the size of the effect of crime victimization on political participation adding further evidence to our claim that personality may be, at least in part, driving both willingness to talk about victimization and political participation.

Conclusions

We argue that the strong relationship between crime victimization and political participation at the individual level may be spurious due to bias in survey responses. Individuals who are more politically engaged are more likely to report victimization in a face-to-face survey than their less politically active peers. The strong correlation between crime victimization and political participation found in survey-based research is, in part, because response bias is directly related to political activity. Response bias is difficult to prove. However, a variety of evidence supports our claim that political engaged individuals are more likely to report their experiences of victimization than their less participatory peers. First, we provided descriptive evidence that the demographics of victims in the survey did not match the demographics of those most likely to experience crime in Latin America. Poor, uneducated, young men are the most likely to experience violence in the region, but the AmericasBarometer sample of victims is largely educated moderately wealthy men. Similarly, the most violent countries according to rates of victimization in the survey include Peru, Argentina, and Uruguay. These are very different from what are conventionally thought of as the most dangerous countries in Latin America based on homicide rates (i.e., Honduras, El Salvador, Venezuela, and Guatemala).

Second, we demonstrated that victimization is strongly predicted by acts of political participation that occurred prior to the period of victimization. If the experience of victimization were the impetus for engagement, this relationship would not exist. Similarly, reporting of victimization is strongly correlated with a respondent’s propensity to answer opinion questions and questions about sensitive subjects on the survey. Taken together, these results suggest that survey questions measuring crime victimization may in part capture propensity to report victimization rather than victimization itself. To talk about a victimization experience with a stranger requires risking stigmatization and discomfort. Politically active people, because they are more empowered, are more likely to be open about their experiences of victimization than their less engaged peers. Our findings suggest a need for caution in our conclusions and further research on the individual-level effects of crime and victimization on participation. We believe it is likely that experiences with crime and violence lead to political engagement in some cases, but it is difficult to understand why when the sample is affected by response bias.

In many ways, this is a frustrating paper to write. As researchers, we are always walking a fine line between trying to answer important questions and not overstepping the limits of what is feasible in terms of data and measurement. Measurement of sensitive and important variables is often where these two goals clash, and very often we rightly err on the side of making the best of limited data. In our minds, this only becomes seriously problematic when we have evidence that the problems of poor data are leading to false conclusions. Unfortunately, that appears to be the case in the observed relationship between crime victimization and political participation using survey data. More broadly, the question remains as to the true relationship between victimization and political participation, something that should be high on the agenda of things to figure out.

Supplementary Data

Supplementary Data are available at IJPOR online.

Conflicts of interest: None declared.

References

Carew Boulding is an associate professor in the Political Science Department at the University of Colorado Boulder.

Shawnna Mullenax received her PhD from Political Science Department at the University of Colorado in 2018 and is a survey researcher based in Washington, DC.

Kathryn Schauer is a graduate student at the Political Science Department in University of Colorado Boulder and working towards her PhD in political science.