Civil Society Organizations as Hindrances to Participation

Although we expect CSOs to positively affect the political participation of custodial citizens connected to them, reasons exist to be pessimistic about their influence. There has been a steady growth in the number of charitable nonprofit organizations and nonprofit advocacy organizations, which often do not mobilize clients and rely little on memberships to advance their interests. The growth of CSOs focused on charitable purposes and without mass memberships (e.g., American Society for the Prevention of Cruelty to Animals or the American Civil Liberties Union) shrinks stores of social capital for conversion to political capital.^{Footnote 56} Such organizations “focus on maximizing the number of people involved without developing their capacity for civic action.”^{Footnote 57} It likely undermines the political utility of CSOs for citizens,^{Footnote 58} be they custodial or noncustodial.

Also, custodial citizens, generally, are likely to reach out to and be contacted by social services CSOs that focus on meeting immediate needs (e.g., shelter, substance abuse recovery, employment).^{Footnote 59} Such organizations seek to support their own persistence first and foremost. Fearing that mobilization of clients could negatively affect their finances, and misunderstanding government regulations of political activities by nonprofits, social services CSOs tend to avoid politics, limiting their level of political engagement.^{Footnote 60}

Even when CSOs foster political engagement, many bias their mobilization. CSOs tend to target individuals who demonstrate greater activism^{Footnote 61} and who possess more positive social constructions and political capital.^{Footnote 62} Given the central tendencies of the racial and class demographics of custodial citizens, the political strategies of many CSOs disregard custodial citizens and their communities, preferring to mobilize the less marginalized.^{Footnote 63}

Civil Society Organizations as Pathways to Participation

Those realities notwithstanding, it is plausible that civil society positively affects the participation of custodial citizens. Political participation has associational anchors and “organizational roots” in civil society.^{Footnote 64} Routine functions and incentives of charities, associations, and other forms of CSOs strengthen the anchors and roots for political participation.^{Footnote 65} Moreover, many CSOs engage in a variety of activities consequential for the political behavior of clients, volunteers, members, and other stakeholders (figure 1).^{Footnote 66}

Figure 1 Civil society organizations pave ways for political participation

Many CSOs, inclusive of many faith-based/religious CSOs like black churches^{Footnote 67} and secular CSOs like labor unions, educate, train, and socialize individuals for political participation.^{Footnote 68} They teach people to develop opinions and perceive their interests, form group consciousness and identify shared grievances, speak their concerns, and amplify their voices.^{Footnote 69} They cultivate personal commitments to public issues and collective problem solving.^{Footnote 70} They help individuals develop political efficacy, education, and civic skills for participation. Political engagement by individuals connected to charitable CSOs can increase from their participation in the “ordinary and routine” practices and activities of the organizations—“activity that has nothing to do with politics or public issues, can develop organizational and communications skills that are relevant for politics and thus can facilitate political activity.”^{Footnote 71} For instance, the one-on-one conversations, opportunities for public speaking, and collective problem solving of sacred and secular CSOs influence civic development, even when they are indirectly political.^{Footnote 72} As a result, connections to CSOs by custodial citizens should reduce barriers to participation, including resource, efficacy, and recruitment barriers.^{Footnote 73}

Beyond dispositional and institutional reasons, people participate politically because others recruit and mobilize them,^{Footnote 74} and CSOs can channel people into opportunities for political participation. Personal connections to CSOs increase the likelihood that “political leaders” will attempt to mobilize them for political influence: “First, organizations mobilize their own members, often explicitly . . . . Second, organizations expose their members to mobilization by sympathetic politicians, activists, and other organizations.”^{Footnote 75} Thus, CSOs create bridges that connect the civic and the political spheres, which can positively affect resources, orientations, and recruitment for political participation through mobilization.^{Footnote 76} Hence, “people who belong to associations are more likely to be mobilized and more likely to participate than people who do not belong.”^{Footnote 77} As well, some CSOs are capable of activating and mobilizing clients, constituents, and members for political participation because of “reciprocal service provision” that produces patron/client relationships that benefit organizations, their clients, and even political elites.^{Footnote 78}

Evidence is strong and consistent that CSOs, even human services CSOs, can inform, activate, and mobilize marginalized people. CSOs help them overcome the greater resource, efficacy, and recruitment barriers to participation that they face relative to people who are not marginalized.^{Footnote 79} The provision and shaping of routine opportunities for engagement often allow CSOs to demonstrate that collective action produces symbolic and substantive rewards for marginalized people.^{Footnote 80} Accordingly, civic voluntarism among lower SES individuals is higher in communities with more and stronger CSOs than in neighborhoods with fewer and weaker ones.^{Footnote 81}

Furthermore, CSOs are central to social welfare delivery in communities with high rates of custodial citizens. Again, it is unusual for human services CSOs to explicitly emphasize political action and mobilization. Nevertheless, some CSOs deliberately enable custodial citizens to develop and practice political resistance.^{Footnote 82} Moreover, some CSOs exist to politically activate and mobilize custodial citizens and their communities.^{Footnote 83} They provide custodial citizens with a greater sense of civic and political worth for fostering “new citizenship” and political participation through associational life (see figure 2).^{Footnote 84}

Figure 2 Redeeming citizenship by/for people with criminal justice contact

Photo credit: Michael Leo Owens

In sum, CSOs can perform multiple roles that bear on political participation. From fostering group consciousness, solidarity, and social capital to spending resources to influencing elections, CSOs can develop the democratic capacities, sentiments, and activities of citizens, custodial or otherwise. Accordingly, there are more reasons than not to expect that personal connections between custodial citizens and CSOs mitigate—and possibly reverse—the negative effects of criminal justice contact on political participation by custodial citizens. Therefore, we predict that custodial citizens with personal connections to CSOs are more likely to participate by voting and nonvoting political activities than custodial citizens without personal connections to CSOs.

CAS Survey—Criminal Justice Contact and Personal Connections to Civil Society Organizations

Our key independent variables are criminal justice contact and personal connections to CSOs. Our measures of criminal justice contact come from a CAS question: We are interested in how much contact people have had with the police. In the past five years, have you . . . (please select all that apply)—been questioned by the police for any reason, been on probation or parole, served time in jail and/or prison. Eleven percent of respondents indicated that the police had questioned them and five percent reported being under correctional control in the past. We model self-reported levels of criminal justice contact in the past five years with dummy variables (Yes = 1, 0 = No) for detained (i.e., questioned by the police) and correctional control (i.e., prison, parole, or probation). The CAS data do not permit a refined assessment of the length of time either under correctional control or length of time since criminal justice contact.

The CAS queried respondents about their personal connections to CSOs: Some people participate in groups and organizations while others do not. Do you currently belong to, volunteer with, attend meetings of, or pay dues for any of the following types of groups? Types of groups included: (1) religious group; (2) neighborhood or community organization; (3) labor union; (4) a professional organization; (5) an ethnic/cultural organization; (6) political organization that focuses on a specific cause; (7) political party; (8) civic organization (e.g., Rotary Club); (9) other; and (10) none of these. Initially, we coded the types of CSOs that respondents identified having personal connections to into six dummy variables: religious, community, ethnic, labor, political, and other. We tested the independent effects of these dichotomous variables. Ultimately, we exclude them because the CAS provided insufficient data for rigorous tests and robust results. Consequently, our measure of personal CSO connections is a dichotomous variable, measuring personal or direct participation in any type of group (Yes = 1, No = 0). Forty-two percent of respondents reported connections to CSOs. Although we know that higher levels of personal involvement in organizations produce greater personal involvement “in governmental politics” and other arenas of political participation such as elections,^{Footnote 97} the CAS did not measure the degree of personal involvement in CSOs.

We use two measures of voting from the CAS—whether a respondent was registered to vote and whether they reported voting in the 2012 presidential election. Among registered voters in the CAS, 89% reported voting in 2012.^{Footnote 98} We measure nonvoting political participation by self-reports of seven political activities, inclusive of a few electoral activities besides voting, respondents performed within 12 months prior to the survey: (1) signed a petition; (2) shared political information via social media; (3) attended a protest; (4) wrote a letter to an elected official; (5) donated to a political cause; (6) volunteered for a political campaign; or (7) issued a political opinion publicly in the form of an op-ed or calling into a radio show. We scaled the items, creating a conventional nonvoting political participation index. It ranges from zero to seven, with a mean of 1.53 activities.^{Footnote 99} The variety of activities in our index raises the possibility that one activity (e.g., protesting) drives observed relationships between criminal justice contact and nonvoting political participation. To evaluate the appropriateness of the use of the index we model the independent effects of each type of criminal justice contact and CSO connections on each item in the index. Figure 3 displays the marginal effects.

Figure 3 Marginal effects of criminal justice contact on participation

Notes: The figure reflects the marginal effect of correctional control, being detained by the police, and CSO connections on each item in the nonvoting participation index. Coefficients reflect fully specified models, located in tables A6 and A7 of the online appendix.

There is no evidence that correctional control negatively correlates with any item in the nonvoting political participation index. Instead, we observe that criminal justice contact positively correlates with contacting a government official, volunteering, and donating money to a cause. It is otherwise unrelated to nonvoting political participation. Figure 3 also does not reveal that any one activity or type of activity underlies a positive association between police detentions for questioning and nonvoting political participation. Finally, figure 3 illustrates that CSO connections statistically increase participation in all activities but one—sharing opinions on TV, radio, or in a newspaper. Given the general absence of a pattern of criminal justice contact affecting nonvoting political participation, we retain the seven-item scale as our measure of political participation beyond voting and registering to vote.

We control for race, gender, age, education, income, political interest, efficacy, party identification, marriage, and unemployment. The CAS stratified its sample by race. It did not stratify by other characteristics co-occurring with criminal justice contact (e.g., income and education). There may be an underrepresentation of respondents with criminal justice contact because of under-sampling some subpopulations or respondents concealing criminal justice contact. To mitigate potential data biases arising from sampling error, we apply weights to the CAS data where possible, using U.S. Census Bureau estimates of the demographics of the Chicago metropolitan area.^{Footnote 100}

Aggregate Data—Criminal Justice Contact and Density of Civil Society Organizations

We rely on a unique set of public records of criminal convictions to measure criminal justice contact at the community level. Our data from the Chicago Justice Project (CJP), originally collected by the Office of the Chief Judge of the Circuit Court of Cook County (CCCC), includes records of criminal convictions and sentences by the Criminal Division of the CCCC between 2005 and 2009.^{Footnote 101} The records are cases the State’s Attorney brought against 173,204 individuals it charged with felonies. For our measure and analysis, we retained felony conviction records for residents of the city of Chicago, relying on physical home address the court provided for each defendant. We successfully geocoded 90% of the 42,200 cases in the dataset (N=37,980). We corrected for an extremely right-skewed raw rate of felony convictions by logging conviction rates (log(Convictions)).^{Footnote 102}

We also measure density of CSOs at the community level, which somewhat parallels or proxies for personal connections to CSOs. Extant research on civil society, especially in Chicago, tells us important things about the measure, beyond the fact that it is an indicator of the organizational foundation and milieu of communities.^{Footnote 103} First, CSO densities are relatively stable by decade. Second, CSO density is positively correlated with events for civic engagement and action, especially “charity events, community festivals, public meetings, recreational activities, and workshops,” and political protests and rallies.^{Footnote 104} Third, CSO density is a strong predictor of the propensity of “collective action” in or by neighborhoods in Chicago.^{Footnote 105}

We constructed our CSO density measure from the Exempt Organizations Business Master File Extracts^{Footnote 106} of the Internal Revenue Service (IRS). The IRS identifies all active, tax-exempt, CSOs in the United States with annual incomes greater than $25,000 that are registered with the federal government and are required to file annual financial disclosure reports. Also, the IRS dataset classifies all nonprofit organizations by missions, describes the types of services they provide, and identifies their physical addresses. We chose all CSOs operating in Chicago that the IRS categorized as 501(c)(3), 501(c)(4), 501(c)(5), 501(c)(7), and 501(c)(9) nonprofit organizations (N=13,932).^{Footnote 107} We aggregated CSOs to police beats, standardizing counts by rates of CSOs per 1,000 people. Like the convictions data, densities of CSOs skewed rightward,^{Footnote 108} requiring logging CSO densities (log(CSO)).

We use two measures of nonvoting political participation that involve contacting public officials, either of a community or particularistic bent.^{Footnote 109} The first measure is a multiyear variable of public attendance at police beat meetings, using proprietary CAPS data on the number of civilian attendees per beat per month from 2013 through 2015.^{Footnote 110} We regard attendance at police beat meetings as an aggregate equivalent of survey items measuring attendance at community or political meetings. We calculated both the mean average attendance across all years by police beat and rates of meeting attendance per 1,000 people. Attendance rates ranged from .55 to 36.2, with a mean of 3.6.

Influenced by a resurgence in the study of nonvoting political participation, particularly citizen-initiated contact with government,^{Footnote 111} our second measure of contact operationalizes nonvoting political participation by instances of requests for nonemergency services from the city of Chicago via its 24-hour municipal 311 call system.^{Footnote 112} We treat a 311 call as an analog to the conventional survey measure of contacting a public official. Furthermore, 311 callers demonstrate “custodianship” for the commons, contribute to the maintenance of public goods, and reveal a “civic disposition—that is manifest in a broader pattern of political participation, including behaviors like voting and volunteering.”^{Footnote 113} Theoretically, variation in criminal justice contact and CSO densities should influence 311 calls. If higher levels of criminal justice contact, for instance, degrade trust in government, whereby less engagement with political life results, eroded trust should diminish the propensity of communities to contact governments to address neighborhood issues.^{Footnote 114} We geocoded all 311 calls in Chicago in 2014 (N = 584,644) and calculated calls per 1,000 people at the level of police beats. Rates ranged from 1.9 calls to 221.1 calls, averaging 87.5 calls.^{Footnote 115}

Turning to voting participation in the municipality of Chicago, our two measures are turnout by registered voters in the 2014 Illinois general election (2014 voter turnout) and voter turnout in the 2015 mayoral election (2015 voter turnout), derived from electoral precinct-level data from the municipal Board of Election Commissioners. Estimated mean voter turnout for the 2014 election is 47% and 38% for 2015. Descriptive statistics and correlation matrices for all variables in our individual-level and community-level analyses are available from tables A1–A4 of the online appendix.^{Footnote 116}

Individual-Level Analysis

To get a baseline estimate of the relationship between criminal justice contact and political participation, we assess the relationships of criminal justice contact and measures of voting participation and nonvoting participation at the individual level. We would expect, based on extant studies, that voting participation by custodial citizens would be lower than voting participation by noncustodial citizens, with more intense criminal justice contact being associated with greater reductions in voting. The results in table 1 partially support the expectations.

Table 1 Effects of criminal justice contact on political participation

Notes: ^aWe model registered to vote and voted in 2012 with logistic regression.

b We model nonvoting political participation with Poisson regression. A dispersion test, using the R package “AER,” yielded an estimate of .662, suggesting the data are not burdened by overdispersion.

* p<.05; **p<.01; ***p<.001.

More intense criminal justice contact, measured by correctional control via imprisonment or jailing or community supervision through parole or probation, is negatively associated with individuals being registered to vote and having voted. However, the relationship is not statistically significant. Conversely, and curiously, being detained by the police for questioning is associated with an increased likelihood of voter registration. Yet, like having been under correctional control, we failed to observe any association between police detentions for questioning and voting in 2012.

Turning to the effect of contact on nonvoting participation, we would expect a positive relationship, as most previous studies conclude. As table 1 shows, the results from Chicago partially support the expectation.^{Footnote 117} Respondents who report that the police had detained them for questioning are likely to have greater levels of nonvoting political participation. The finding confirms results from other studies that show less intense criminal justice contact may increase nonvoting participation.^{Footnote 118} Yet there is no evidence from Chicago that incarceration and community supervision increase or decrease nonvoting participation. This, too, is relatively consistent with prior research.^{Footnote 119}

So far, we have evidence that criminal justice contact may have varied effects on voting and nonvoting political participation. The positive associations we observe for civic voluntarism through personal connections to CSOs—controlling for other factors, including correctional control—are, however, consistent for voting and nonvoting participation (table 1). The relationship achieves statistical significance across all three models. Personal connections to CSOs are associated with increased participation, as we theorized. This strengthens our expectation that personal connections to CSOs are positively associated with voting and nonvoting political participation by custodial citizens.

Turning our attention to examining the moderating effect of personal connections to CSOs on the political participation of individuals with criminal justice contact, we explore the relationship between criminal justice experiences with personal connections to CSOs. The results in table 2 suggest that personal connections to CSOs correlate with a greater propensity of nonvoting participation by custodial citizens, relative to those with criminal justice contact who lack CSO connections. The interaction of correctional control and personal connections to CSOs produced results that accord with our hypothesis about political participation beyond registering to vote and voting. Personal connections to CSOs are positively related to greater nonvoting political participation by people who have been under correctional control, after controlling for factors that conventionally correlate with civic voluntarism.

Table 2 The interactive effects of contact and CSO connections on forms of participation

Notes: ^a We model registered to vote and voted in 2012 with logistic regression.

^b We model nonvoting political participation with Poisson regression. A dispersion test, using the R package “AER,” yielded an estimate of .660, suggesting an absence of overdispersion.

* p<.05; **p<.01; ***p<.001. Reported coefficients are from fully specified models, located in table A8 of the appendix.

To better interpret the relationships among criminal justice contact, connections to CSOs, and nonvoting political participation, we calculated the expected score on the nonvoting political participation index by degree of criminal justice contact. We did this for custodial citizens with and without connections to CSOs. Regardless of the degree of contact, personal connections to CSOs improve nonvoting participation by custodial citizens relative to those lacking connections to CSOs (figure 4). Absent CSO connections, criminal justice contact diminishes the expected level of nonvoting political participation. Those without criminal justice contact and without CSO connections have an expected score of 1.25 activities on the index. The expected score on the index shrinks to one activity for those who have been under correctional control and lack CSO connections. Among similarly situated custodial citizens with CSO connections who have been under correctional control, however, the expected score on the index increases to 2.25 activities.^{Footnote 120}

Figure 4 The impact of criminal justice contact and CSO connections on participation

Notes: The figure reflects the interactive effect of criminal justice contact and CSO connections on participation among CAS respondents in the Chicago Metropolitan Area. Expected values derived from the nonvoting political participation model (refer to table 2).

Remarkably, the size of the positive association of personal connections to CSOs on nonvoting political participation is larger for those with criminal justice contact than for those without it. For noncustodial citizens, personal connections to CSOs improve nonvoting political participation by about .3 political activities. Among Chicagoans the police have detained, CSO connections improve the expected score on the index by about .6 political activities. Personal connections to CSOs also increase the expected score by 1.25 political activities among respondents who have experienced some form of correctional control. Moreover, the expected value of nonvoting participation for custodial citizens with personal ties to CSOs exceeds that of their counterparts without criminal justice contact.

Nevertheless, personal connections to CSOs appear to not moderate the depressive effects of criminal justice contact on voting. Our finding, which is important, may reflect that the material and attitudinal barriers to voting custodial citizens face remain high despite any capacity of CSOs to pave the way for voting. Even in places like metropolitan Chicago, where all custodial citizens are eligible to vote, and even as their connections to CSOs may ease participation in other forms of civic engagement, barriers to voting may remain. Lastly, the results reveal that personal connections to CSOs by custodial citizens the police have detained for questioning may not be statistically significant for voting or nonvoting political participation.^{Footnote 121}

One implication of our findings is that the empirical claims of the literature about the negative democratic effect of the carceral state are too general to adequately describe civic voluntarism by custodial citizens. Certainly, custodial citizens may withdraw from or neglect voting. Nonetheless, many continue to participate outside the ballot box, especially when they are connected to CSOs. Moreover, some custodial citizens will begin participating beyond voting because they have personal connections to CSOs. Such connections may make their participation more likely and frequent because the CSOs create opportunities for education, activation, and mobilization, as they do for noncustodial citizens.

Aggregate-Level Data

Based on our results from the individual-level analysis, combined with the broader empirical findings in the literature on the political behavior and attitudes of custodial citizens, we explore a simple set of predictions about criminal justice contact at the community-level: greater criminal justice contact is associated with decreases in electoral participation, measured by voting and greater criminal justice contact is associated with increases in nonvoting political participation. Recall that our aggregate criminal justice contact measure is felony conviction rate. It is, as table 3 displays, statistically associated with lower turnout in the 2015 mayoral election. We logged the rate of felony convictions such that the coefficient estimate is interpreted as the absolute change in 2015 turnout given a percent change in felony convictions. Thus, a 5% increase in the felony conviction rate decreases expected voter turnout at the community-level in 2015 by approximately four percentage points. Also, while the relationship is not statistically significant, the felony conviction rate is negatively associated with the 2014 general election turnout.

Table 3 Effect of conviction rates and CSO densities on voting and nonvoting participation

Note: All dependent variables are continuous, and are modeled using ordinary least squares regression.

* p<.05; **p<.01; ***p<.001.

Independent of CSO density, the rate of felony convictions is associated with increased nonvoting political participation, measured by 311 calls. Moving from the minimum value of felony conviction rate (.018) to its mean value (7.5) increases the rate of nonemergency calls in Chicago from 24 calls per 1,000 residents in police beats to just over 100 calls. Similarly, the felony conviction rate positively relates to meeting attendance, with the relationship approaching statistical significance. The positive effect that may exist between the felony conviction rate on both types of nonvoting political participation perhaps indicates neighborhood disorder and maybe lower (or higher) levels of community efficacy.^{Footnote 122} That would mean felony conviction rates co-occur with needs for nonemergency assistance/antagonistic policing practices in need of redress.^{Footnote 123}

Furthermore, one would expect that the scale of CSO presence in communities would particularly influence political participation by custodial citizens, whereby CSOs mitigate the demobilizing effects of criminal justice contact. We test this expectation at the community-level by interacting felony conviction rates and CSO density. Parallel to the results from our individual level analysis, political participation should be greater in communities where conviction rates are higher and the presence of CSOs are denser, compared to communities with lower CSO densities. Comparatively, lesser CSO densities and lower conviction rates should matter less for political participation outcomes at the community-level. While CSO density and conviction rates independently increase requests for nonemergency assistance, we do not observe in table 4 a statistically significant interactive association with nonvoting political participation or with voting.

Table 4 Interactive effect of conviction rates and CSO densities on voting and nonvoting political participation

Notes: All dependent variables are continuous, and are modeled using ordinary least squares regression.

* p<.05; **p<.01; ***p<.001. Coefficients reflect fully specified models, located in table A12 of the online appendix.

Figure 5 displays the marginal effects of CSO density, conviction rate, and the interaction between the two on each outcome of interest. These visualizations highlight that, independent of the felony conviction rate, the density of CSOs are positively related to nonvoting political participation. That is true, too, for the felony conviction rate in relation to 311 calls, regardless of CSO density. We interpret those results to mean that CSO density positively increases requests for nonemergency assistance in communities with either low or high felony conviction rates. Our interpretation is plausible, given that “organizational resources predict collective efficacy and [CSOs] produce externalities that foster collective action” in Chicago.^{Footnote 124}

Figure 5 Conviction rate, CSO density, and participation among Chicago police beats

Notes: The marginal effect of conviction rate, CSOs per 1,000 in the population and their interaction on voting, attendance at police beat meetings, and requests for nonemergency assistance. Coefficient estimates reflect models presented in table 4.

However, we must reiterate two points. First, higher rates of 311 calls may result from greater needs for nonemergency assistance in high criminal justice contact (i.e., high felony conviction rate) communities. Second, higher attendance at police beat meetings may stem from greater degrees of police-community antagonisms, if policing is more concentrated, “hot spot” oriented, and aggressive. Put another way, higher rates of citizen-initiated contact with government in one form or another in communities with more custodial citizens could result from a need in such communities for greater attention from public officials, a need that may be absent in communities with fewer custodial citizens.

To address those potentialities, we employed a matching causal inference strategy that allowed us to compare the impact of CSO density among similarly situated communities that differ primarily by felony conviction rates. The strategy allowed us to compare low conviction rate communities to otherwise similarly situated high conviction rate communities. We describe the matching strategy and report results in full in the online appendix. Results from the matched analysis generally corroborate conclusions we drew from our analysis of the full sample before matching communities by rates of felony convictions. Specifically, greater CSO density is associated with greater requests for nonemergency assistance and attendance at police beat meetings among communities with either low or high felony conviction rates (refer to table 5 and figure 6).

Table 5 Matched analysis: Interactive effect of conviction rates and CSO densities on voting and nonvoting political participation

Notes: All dependent variables are continuous, and modeled using ordinary least squares regression.

*p<.05; **p<.01; ***p<.001.

Figure 6 Conviction rate, CSO density, and participation among matched police beats

Notes: The marginal effects of conviction rate, CSOs per 1,000 in the population, and their interaction on voting, attendance at police beat meetings, and requests for nonemergency assistance, among matched police beats. Coefficient estimates reflect models presented in table 5.

To be clear, greater CSO density is associated with greater attendance at police beat meetings and requests for nonemergency assistance among low-conviction communities that are comparable in other ways to their high-conviction counterparts. Neither do conviction rates by themselves appear to impact nonvoting political behavior, nor does CSO density operate differently for low- or high-conviction communities. The overriding factor influencing nonvoting political participation appears to be CSO density. Results from the matching strategy, then, confirm that our broader findings about CSO density and nonvoting political participation are not due to need alone in those communities.

Voting, however, is different. While increasing CSO density is associated with increasing voter turnout in high and low criminal justice contact communities in Chicago in 2014, the slope of increase is greater for low conviction communities than for high ones. This illuminates how we might interpret the effects of CSOs on nonvoting political participation in communities. Although the interaction term for convictions rates and CSO density does not achieve statistical significance for either meeting attendance or 311 calls per capita, as is the case with the results of the unmatched analysis presented earlier in table 4 and figure 5, the nature of the relationships are similar: CSO density increases nonvoting political behavior in high-conviction and low-conviction police beats, even if the size of the effect is larger in low-conviction communities.

Overall, CSOs are associated with more nonvoting political participation in high conviction communities, which parallels what we observed in the individual-level analysis, whereby personal connections to CSOs were associated with increased nonvoting political participation by custodial citizens. Also, contacting government, either through attendance at police beat meetings or making requests for nonemergency assistance, is about more than need. CSO density matters as there is an association between CSO density and increased nonvoting political participation, regardless of community-level conviction rates. However, as we generally observed in the individual-level analysis, where personal connections to CSOs were not associated with changes in voting, barriers may remain that hinder higher densities of CSOs from yielding greater degrees of voting by communities with high rates of criminal justice contact.^{Footnote 125}