Manifesting Symbolic Representation through Collaborative Policymaking

Collaboration is a widely employed strategy for addressing complex policy problems that impact the interests of a diverse set of stakeholders. Whereas scholars have long recognized that including civil-society groups in collaborative policymaking processes leads to the generation of more impactful policy outputs, little known as to whether such inclusion impacts the belief systems of non-participating individuals who are represented by participating groups. In this article, we bridge and extend concepts from collaborative governance and representative bureaucracy literatures to argue that the representation of civil-society groups in collaborative policymaking-forums can have a transformative impact on the way non-participating community members perceive, and subsequently, interact with participating organizations. Empirically, we assess whether the inclusive representation of civil-society groups within a specific collaborative policing forum impacts citizens’ perceptions towards the main participating agency—the police department—finding that collaborative processes do have positive spillover effects. Our findings demonstrate that the importance of civil-society inclusion may be largely understated in the collaborative governance literature. Moreover, whereas the representative bureaucracy literature has solely considered how symbolic effects can arise from agency-wide and/or bureaucrat-specific demographics (e.g., race and gender composition), our findings demonstrate that alternative pathways exist.

The generation of collaborative forums-broadly defined as institutionalized decision-making venues where representatives from public, private, and/or non-profit stakeholder groups interact to resolve policy problems that impact their mutual interests-is a widely employed governance approach designed to tackle "wicked" policy problems (Emerson, Nabatchi, and Balogh 2012;Fischer and Leifeld 2015).
Through the facilitation of structured, repeated interactions, forums promote the capacity of participants to pool diverse resources, coordinate actions, and develop a shared understanding of policy problems, which may ultimately result in more legitimate, informed, and impactful policy outputs (Ansell and Gash 2008;Stern and Dietz 2008;Mewhirter and McLaughlin 2021).Recent research, however, has called to question the integrative character of forums-an attribute that serves as both the hallmark of collaboration, as well as the mechanism driving its ascribed benefits-demonstrating that less traditional, resource-poor organizations often lack the capacity to meaningfully participate in forums, resulting in lower levels of participation and (when they do participate) a lack of within-forum engagement (Angst et al. 2021;Scott and Thomas 2017;Wondolleck and Yaffee 2000).The exclusion and/or marginalization of such voices i) limits the diversity of financial, technical, and knowledge-based capital within forums, and ii) facilitates the generation of narrow policy solutions that disproportionately benefit select stakeholders, which, combined, iii) results in the development of less effective, less efficient, and less equitable solutions (Mewhirter, McLaughlin, and Fischer 2019;Wondolleck and Yaffee 2000).
Research demonstrates that while forums should strive to attain meaningful engagement from all relevant stakeholders, civil-society groups play a particularly impactful role in the development of successful forums, and thus, their participation should be prioritized (Ansell and Gash 2008;Dupuy and Defacqz 2021).In addition to providing valuable, non-redundant resources (e.g., distinct expertise) fundamental to development of impactful policy solutions, the meaningful inclusion of civil-society organizations (and at times, individual community members) in the decision-making process can minimize the "democratic deficit" associated policy-elite driven decision making by converting citizens (whose interests are either voiced directly, or through representative organizations) from passive consumers to co-producers of public goods and services (Doberstein 2016;Nabatchi 2010).That said, historically (and across substantive policy contexts), civil society groups have embodied the concept of "resource poor", often lacking the temporal, technical, and financial resources needed to both select into and meaningfully engage in forum activities (Ansell et al. 2020;Dobbin and Lubell 2019;Schilling 2009).Operating under the guise of inclusivity and integration while failing to attract and express the interests of civil-society groups can exacerbate, rather than mitigate, the existence of democratic deficits while simultaneously re-enforcing ex-ante power structures (Mewhirter, Lubell, and Berardo 2018;Scott and Thomas 2017;Swyngedouw 2006).
While the collaborative governance literature has long recognized the importance of integrating civil-society groups representative of the broader population into forum proceedings, to date, researchers have largely focused on how such factors influence the development of forum outputs (Nabatchi 2010;Torfing and Triantafillou 2011;Bidwell and Ryan 2006;Koontz and Moore Johnson 2004) and/or impact the way civil-society organizations perceive their role within a governance system (Dupuy and Defacqz 2021;Newman et al. 2004).While this research stream has had a profound impact on the literature's conceptualization of representation in collaborative settings (and the impacts thereof), little is known as to how and whether civil-society representation spills-over into the belief systems of everyday community members whose interests could be (but are often not) directly represented by relevant organizations.To partially address this gap, we bridge the collaborative governance and representative bureaucracy literatures, arguing that representation of civil-society groups in collaborative forums can have a symbolic and transformative effect on the way that non-participating community members perceive, and potentially, interact with all participating organizations.Specifically, we argue that the inclusion of civil-society groups in collaborative forums designed to formulate and codify policy changes can send a symbolic signal to community members that the participating organizations (often, public agencies, businesses, non-profits, etc.), will respect and amplify the viewpoints of the community, and are willing to consider community preferences when deciding how to devise policies that impact community interests.Representative processes can enhance public perception of the bureaucracy, impacting the way they perceive, and subsequently, interact with participating groups.This symbolic effect, however, may be contingent on community members believing that civil-society groups are meaningful, not-token, participants whose interests and concerns are both heard and ultimately expressed through policy decisions.We argue that the transformative effect of collaboration, of course, is contingent on citizens' knowledge of i) a forum's existence, ii) the composition of forum actors, and iii) forum processes and outcomes.Absent this knowledge, spillover effects (through the detailed mechanism) cannot occur.
We leverage 2017 survey data collected from Cincinnati residents to test our propositions.The survey asked residents several questions regarding their knowledge about a prominent policing forum, the Cincinnati Collaborative Agreement, as well as their beliefs regarding the legitimacy of and trust in the primary participating bureaucracy: the Cincinnati Police Department (CPD).Empirically, we assess how resident knowledge regarding the Collaborative Agreement-which features the meaningful participation of an array of civil-society organizations who have historically exerted considerable leverage over forum deliberations and outputs-shapes perceptions regarding CPD officers and the institution, itself.Results from a series of mixed ordered logistic regressions demonstrate that individuals who have a high degree of knowledge regarding forum representation, processes, and outcomes are more likely to believe that the CPD is legitimate (officers have legitimate authority, officers should be obeyed), and that CPD can be trusted (CPD can be trusted, officers are honest, comfort in dealing with officers).
This study holds important implications for collaborative governance, representative bureaucracy, and public policy scholars, as well policing practitioners.First, our findings demonstrate that the importance of meaningfully including civil-society groups into the collaborative making process is largely understated within the collaborative governance literature.While a rich literature extols how such inclusion leads to more effective, efficient, and importantly, equitable policy outputs, our findings demonstrate that this is only one of the (potentially) numerous benefits.
Second, we fill two important theoretical gaps in the representative bureaucracy literature.First, while research suggests that representation can lead to improved citizen perceptions towards and interactions with relevant bureaucracies, this research stream solely considers how the agency-wide and/or bureaucrat-specific demographics (e.g., race, gender, age, etc.) impacts citizen experiences with and perceptions towards bureaucrats and the institution, itself (see Headley et al. 2021).Our findings demonstrate that symbolic effects can manifest from alternate pathways: in this case, through collaborative partnerships with civil-society groups.As shown in our empirical analysis, these pathways are not mutually exclusive: symbolic effects can simultaneously manifest through both pathways, leading to compounding, positive effects.Second, while the translation from passive to active representation points to certain institutional and political preconditions that may bolster bureaucrats' tendencies or abilities to actively represent a population (Keiser et al. 2002;Meier 1993;McLaughlin et al. 2021), less is known about possible institutional factors that may cultivate or attenuate the path from passive to symbolic representation.Our findings align with recent symbolic representation work: passive representation, absent certain moderating institutional or organization variables (Headley et al. 2021), may not enough to produce symbolic representation.Here we show that a foundational tenet of the collaborative policy process, principled engagement (not tokenism), may be needed to facilitate translation from passive to symbolic representation.
Finally, and pragmatically, our work indicates that stakeholders may be faithfully represented without participating: an important finding given the fine line between the importance of diversity/inclusivity in collaboration and the reality of stagnation with too many interests (e.g., see Lubell, Mewhirter, and Berardo 2020).Such substantive effects are perhaps most important in the context of policing where potential outcomes of police-citizen interactions may have deleterious consequences.In fact, research shows that hostile interactions with police is a key determinant of poor (and at times, violent) outcomes (Worall et al. 2018).More so, demonstrating that the effects of collaborative processes seep into the broader community of stakeholders may stimulate greater reciprocity between police and community as coproducers of public safety.

Civil-Society Representation in Collaborative Settings
"Inclusivity, transparency, and diversity" serves as the mantra of collaboration and is often heralded as necessary condition for well-functioning forums (Nabatchi 2010;Torfing and Triantafillou 2011;Fung 2015;Koontz and Moore Johnson 2004;Bidwell and Ryan 2006).Whereas a well-established literature demonstrates how heightened representation fuels the development of more effective, efficient, and equitable outputs (Fischer and Leifeld 2015;Emerson, Nabatchi, and Balogh 2012), little is known about how and whether the way stakeholders are represented within policymaking proceedings impacts public perceptions regarding (and, potentially, interactions with) participating bureaucracies.In this section we augment and apply lessons from representative bureaucracy literatures to the collaborative governance literature to assess how such representation (or lack thereof) might spillover into stakeholder belief systems, manifesting in changes to perceptions regarding legitimacy and trust.

From Passive to Symbolic Representation within Collaborative Forums
Research within the representative bureaucracy literatures finds that the mere existence of passive representation-i.e., when the demographic characteristics of a relevant bureaucracy and/or individual bureaucrat mirror that of the community they serve-can have an empowering effect on citizens' behaviors and attitudes (Riccuci et al. 2015;Theobold and Haider-Markel 2008).Here, passive representation functions as a cue to citizens that the interests of specific (and often, marginalized) groups may be afforded equitable treatment (Theobold and Haider-Markel 2008;Headley et al. 2021).Passive representation can, through one's awareness of bureaucrat characteristics, lead to positive perceptions of the bureaucracy, and potentially, impact citizen-bureaucrat interactions: a phenomenon referred to as "symbolic representation" (Riccuci et al. 2015;Gade and Wilkins 2013).For example, Gade and Wilkins (2013) find that veterans perceive greater satisfaction with implementation of Veteran Affairs' programs when they believe their provider is also a veteran.Others find that both gender and race representation in police bureaucracies increases citizens' perceptions of job performance and legitimacy as well as reports of sexual assaults (Riccuci et al. 2014;Theobold and Haider-Markel 2008;McLaughlin et al. 2021;Meier and Nicholson-Crotty 2006).Symbolic representation manifests itself regardless of any actions taken by the bureaucracy or individual bureaucrats and thus intrinsically and symbolically enriches bureaucratic outcomes (Theobald and Haider-Markel 2008;Headley et al. 2021).
The collaborative governance literature also recognizes the importance of representation, largely focusing on how the inclusion of all stakeholder groups within policymaking forums facilitates the development of more effective, efficient, and importantly, equitable policy solutions (Ansell and Gash 2008;Dupuy and Defacqz 2021;Newman et al. 2004).Within forums, collaboration unfolds as stakeholders spend considerable time building relationships, working through collective differences, and engaging in good faith negotiations guided by fair and transparent processes (Fischer and Leifeld 2015;McLaughlin, Mewhirter and Lubell 2021;Mewhirter and Berardo 2019).Here, broad-based inclusion is "at the heart of the legitimation process" as stakeholders are involved during problem definition, solution design, and implementation stages (Ansell and Gash 2008 p. 556;Burger et al. 2001).Thoughtful and active inclusion of civil society stakeholders represents a more open democratic process where citizens have a real chance to have their viewpoints heard and actively shape decisions that address their needs (Dupuy and Defacqz 2021;Sorensen et al. 2020;Ansell and Gash 2008).
Merging these literatures, we argue that deliberately mobilizing often under-represented civilsociety organizations-for example, black advocacy groups, community-based groups, youthorganizations, homeowners associations, and education, religious and social service leaders-to participate in the development of public policy demonstrates a commitment by participating bureaucratic partners to inclusion and representation.These stakeholders, as consumers of public services in their own community, can provide local and contextual information and be direct partners during collaboration (Dupuy and Defacqz 2021;Simmons 2008).Here, the emphasis on inclusivity and existence of passive representation during collaborative policy processes signals to non-participating citizens that their interests are being faithfully represented and that stakeholders representing their viewpoints during collaboration are true coproducers of the output.As a result, we propose that beliefs about representation and inclusivity during the collaborative process may lead citizens to i) be trusting of the participating bureaucracy, and ii) perceive the participating bureaucracy as legitimate.

Why Passive Representation May Not be Enough: The Role of Meaningful Engagement
Abundant research has shown, however, that not all stakeholders have the resources needed to meaningfully engage in forum processes.Such issues are common to civil-society organizations who often lack technical, financial, and temporal based resources, relative to other organizations (e.g., public agencies, private entities, etc. : Arnstein 1969;Wondolleck and Yaffee 2000;Ansell and Gash 2008).
Absent institutions to counter these power and resource imbalances, ultimately, collaborative processes may be coopted by stronger actors which will continue to marginalize resource poor voices and generate inequitable decisions (Choi and Robertson 2014;Fung 2015;Scott and Thomas 2017).When this occurs, forums operate under the guise of inclusivity, using citizens as tokens to proffer the appearance of representation with no genuine intention of yielding any influence.In these cases, we argue that passive representation of civil society groups in collaboration may not be enough to warrant belief system changes in citizens (i.e., symbolic representation).
While collaboration may sometimes accentuate rather than mitigate power and resource imbalances between stakeholders, research points to certain features of the collaborative process that may counter this domineering.Here meaningful principled engagement, defined as ongoing efforts to i) develop shared understanding of the problem and others' interests, ii) build agreement on language, tasks, and rules used during collaboration, iii) engage in open, reasoned, and fair deliberation, and iv) make joint procedural and substantive decisions (Emerson, Nabatchi, and Balogh 2012) signals fairness, open and inclusive discourse, and representative decisions informed by all participants (Emerson, Nabatchi, and Balogh 2012;Ansell and Gash 2008;Leach 2006).As a result, we argue that if citizens perceive meaningful principled engagement occurred between all stakeholders and that civil society groups were integral in shaping policy decisions, then, citizens may i) be trusting of the participating bureaucracy, and ii) perceive the participating bureaucracy as legitimate.

Knowledge of Forum Processes: A Critical Link
Mirroring the argument that passive representation may convert to symbolic representation only when one is aware of certain bureaucratic features (Headley et al. 2021), we contend that knowledge serves as the mechanism linking forum representation to increased perceptions about participating bureaucracies.Specifically, we contend that forums that include and maintain meaningful engagement from civil-society organizations can have a transformative (i.e., symbolic) effect on the way citizens perceive participating organizations.Such effects are only expected to hold if and when residents are aware of and understand the inner workings of the collaborative forum itself.While all citizens may (theoretically) benefit from policy outputs that arise from such collaboratives, only those who understand the origins of such policy changes should observe changes in beliefs through the ascribed mechanism.

Case & Data
We leverage data collected from a 2017 community survey of Cincinnati residents to examine how citizen knowledge regarding a high-profile policing forum-the Collaborative Agreement (CA)impacts citizens' beliefs towards the CPD: the primary public agency participating within the forum whose policies and practices are subject to forum debate.The CA, detailed below, maintains the collaborative features that we argue (above) are necessary to generate symbolic effects: specifically, the CA is i) predominately composed of civil-society organizations ii) who have exercised a large degree of control over forum processes.While the CA maintains a relatively high profile in the region-at least relative to other collaborative forums-citizens maintain a high degree of variability regarding their knowledge about it.We leverage this variability to test our theory, anticipating that citizen knowledge of the CA will be associated with more positive beliefs regarding the legitimacy of and trust in the CPD.
Below, we provide more details regarding the CA, as well as our data collection effort.

The Collaborative Agreement: Processes and Outputs
The CA-which emanated from a 2001 federal lawsuit by the Cincinnati Black United Front and the ACLU of Ohio against the CPD-forced CPD leadership to the collaborative table with a wide array of community groups (e.g., black advocacy groups, more-general community groups, business and education leaders, religious and social service leaders, youth organizations, and others) to regularly meet and discuss policing policies, community concerns, and find common ground for reform.Although the court mandate expired in 2007, the forum remains (voluntarily) in operation to this day.Since the agreement's inception, participating civil-society groups have exerted a tremendous amount of leverage over forum processes, ultimately resulting in a number of policy changes that coalesce with community interests (many of which restrict CPD autonomy and facilitate citizen oversight).Such changes include (but are not limited to): the establishment of the Citizen Complaint Authority (2002); the addition of disengagement ( 2003) and de-escalation language (2004) into use of force policy standards; the introduction of a Mental Health Response Team (2004); the development of a use of force tracking system ( 2006 I think our officers in Cincinnati are on the cutting edge, that we've adopted it… [The Collaborative Agreement] is a part of our culture.It's a part of our DNA, and that's reflected, I think, in the good relationships we have right now in the community between our police and our citizens (WCPO 2016).

Data Collection
The goal of our data collection effort is to identify and link Cincinnati-resident knowledge of the CA to their beliefs about the CPD.While ideally, we would leverage i) a probabilistic sampling method to obtain an unbiased sample, or alternatively, ii) a quota-based sampling procedure that assures representativeness on a subset of characteristics, the demographics of Cincinnati residents complicate such approaches.Here, survey research has long established that probabilistic-based (such as random digit dialing) and quota based (such as quota driven online panels) samples are unlikely to generate sufficient responses from non-white, urban-poor populations (often referred to as "hard to reach populations": see Faugier and Sargeant 1997).While such issues can be mitigated (through adjustments and weighting) when sampling from a large population, they are exacerbated when sampling from smaller, constrained geographic areas (such as a city: Boas, Christenson, and Click 2020).Given that 42.3% of the ~308,000 city residents are Black and that 34.8% of Black residents live under the poverty line (Census 2019), such approaches are unlikely to yield sufficient responses from Cincinnati's Black, urban-poor population: a population who i) is most likely to encounter CPD officers on a regular basis, and ii) whose relationship with the CPD provided the catalyst for the CA's establishment.
Consistent with well-established survey-practices involving hard to reach populations, we instead employ a venue-based sampling approach (Farzana et al. 2001), coupled with chain-referral method (or "snowballing") procedure (Shaghaghi, Bhopal and Sheikh 2011) to attain a more representative sample.
We facilitated survey taking opportunities at nine free-to-attend, large-scale, community events using laptops and iPads.These events were civic and/or entertainment in nature, did not have a political theme, and took place within the city's urban core.Data collection occurred during June 2 and September 6, 2017, via a weblink to the Qualtrics-embedded survey.Respondents could participate if they lived in the Cincinnati area (i.e., Hamilton County, Ohio) and were at least 18 years of age.After completing the survey, respondents were asked to provide contact information for those in their social network.Emails were sent to those respondents soliciting their participation.The survey unfolded by asking respondents a variety of questions regarding their knowledge of the CA, perceptions of the CPD legitimacy, trust in the CPD, as well as a battery of traditional demographic and social variables.
Note that while such a strategy allows us to obtain a "less-biased" sample (relative to the alternatives), it is not without its limitations.Specifically, such a procedure may result in an under sampling of more affluent, white populations that live further away from the urban core.Summary statistics for all variables are provided in the Table A1 in the Appendix.

Measurement: Dependent Variables
Our theory contends that community member knowledge regarding representation during collaborative processes can have a transformative effect on their beliefs regarding participating bureaucracies.Concurrent with past symbolic representation research, we anticipate that this could impact residents' perceptions regarding the legitimacy of and trust in participating bureaucracies (see Headley et al. 2021).While several government agencies participated (to some extent) in forum meetings, the CPD serves as the primary bureaucracy whose policies and practices are subject to forum debates.Given the CPD's unique role within the forum, we focus our attention on how citizens view CPD bureaucrats (i.e., officers), as well as the institution, itself.
We leverage three variables to capture beliefs about legitimacy.The first, Legitimacy, attempts to broadly capture the concept in theory, whereas the latter two-Obey and Ignore-capture the concept in practice.To create these variables, each respondent was asked the extent to which they agree with the following statements: i) "Overall, the CPD officers in my community are legitimate authorities and people should obey the decisions they make"; ii) "You should do what the CPD officers in your community tell you to do, even when you disagree with their decisions"; iii) "There are times when it is okay to ignore what CPD officers in your community tell you to do".Responses range from 1-5, where 1= "strongly disagree", 2="disagree".3="neither agree nor disagree", 4="agree", and 5="strongly agree".
The variable capturing obedience is reverse coded, so that higher values correspond to greater legitimacy.
Similarly, we include three variables to operationalize the concept of trust in CPD officers: the first, Trust captures the concept more broadly, whereas the latter two-Honesty and Comfort-capture its application.The variables Honesty and Comfort use the same response scale detailed above.To create these two measures, respondents were asked to respond to the following statements: "The CPD officers in my community are often dishonest" and "If I get stopped by the CPD, I feel comfortable talking with the CPD".Responses to the first statement were reverse coded, so that higher values correspond with greater trust.To create Trust, respondents were asked the extent to which the Cincinnati Police Department can be trusted.Responses range from 1-4, where 1="not at all", 2= "a little", 3="some" and 4="a lot".
The percentage of respondents that selected each value of the dependent variables is described in

Measurement: Independent Variables
We hypothesize that citizens' perceptions of CPD are partially a function of their knowledge regarding the CA.Given that civil-society groups are both well represented within the forum and exert considerable influence over forum deliberations, we anticipate that resident knowledge regarding the CA will be positively associated with the perceptions of trust and legitimacy.To capture this, respondents were asked the extent to which they agree with the following statement: "I am familiar with the Collaborative Agreement."Responses range from 1-5, where 1= "strongly disagree", 2="disagree".
3="neither agree nor disagree", 4="agree", and 5="strongly agree." Given that we argue that legitimacy and trust are enhanced only when citizens are aware of both forum membership and true principled engagement during collaborative processes-a relatively high level of knowledge about the CA-we anticipate that the effect of knowledge on outcomes may only be present when individuals are in fact aware of the forum (=4 or =5 on our scale), and potentially, when they maintain a high degree of knowledge (=5 only).We restructure our variable to better capture the knowledge construct.Our main independent variable, CA Knowledge, takes the value of 0, if they "strongly disagree", "disagree", or "neither agree nor disagree" with the prompt (i.e., low knowledge), the value of 1 if they "agree" with the prompt (i.e., medium knowledge), and 2 if they "strongly agree" (i.e., "high knowledge").Given this operationalization, we treat CA Knowledge as a nominal variable, assigning values of 1 (medium knowledge) as the reference category.
The distributions of initial responses and final measure are presented in Figures 1 and 2, below.
Note that while we feel that this operationalization best reflects our construct of interest, it is possible that the measure is better capturing "level of knowledge", with higher value indicating greater knowledge (as opposed to the not knowledgeable (=0), knowledgeable (=1), highly knowledgeable (=2) scale we rely on).In the Appendix, we re-estimate our models (described below) using this alternate operationalization.
As shown in Table A2, findings are consistent across five of the six regressions.We include the variable, Passive Officer, which captures the extent to which an individual believes they are being "passively represented" by CPD officers.Specifically, our question asks individuals the extent to which they agree that CPD officers "look like me": responses range from 1-5, where 1= "strongly disagree", 2="disagree", 3="neither agree nor disagree", 4="agree", and 5="strongly agree."As shown in the Table A1 in the Appendix, respondent, on average, do not tend to exhibit strong beliefs regarding the extent to which they share common features with CPD officers.
We include this variable for two reasons.First, linking such perceptions to beliefs about legitimacy and trust allows us to observe whether passive representation indeed activates symbolic representation as indicated in the literatures.Second, and (for the purposes of this paper) more importantly, the inclusion of this variable allows us to simultaneously test for the presence of both forms of symbolic representation-that emanating from i) sharing features with officers (as normally discussed in the representative bureaucracy literatures), and ii) representation in collaborative forums-in a single regression.In doing so, we can assess whether symbolic effects emanating from alternative pathways can simultaneously manifest within individuals, resulting in compounding, positive effects.

Measurement: Control Variables
We include a number of control variables that could bias our estimates.We employ a modelbased inference approach (Eggers et al. 2020), whereby variable inclusion is justified only if there is a plausible pathway by which a given factor can simultaneously impact our independent and dependent variables.
We include a wide array of demographic variables meant to capture the socio-economic status of a respondent.Here, we recognize that social and economic factors impact i) the likelihood and manner in which an individual interacts with officers (and in turn, their views of CPD), as well as ii) the likelihood that one has the time, energy, and capacity to seek information regarding how and to what extent their interests are being represented in collaborative decisions.Specifically, we adjust for whether an individual has a college degree (=1) or not (=0), their employment status (0= not employed; 1=employed), their sex (0=male; 1=female)1 , age2 , and income3 .
We also include variables Stopped, Called, and Victim, which capture whether an individual has been stopped by CPD (=1), called CPD (=1), or was a victim of a crime (=1) or not (=0), in the last three years.Here, we expect that individuals who have these experiences may be more likely to look into CPD policies and practices (and thus, have a higher likelihood of knowing about the CA), and that such experiences may also impact their view about the CPD and CPD officers.As shown in Table A1 in the Appendix, the individuals in our sample have, on average, interacted with CPD at relatively high rates over the last three years: 38.96% of respondents had been stopped by CPD, 42.21% had called CPD, and 45.27% had been a victim of a crime.
The variable, Resident Years, captures how long a respondent has been a resident of the area.We expect that those who have lived and experienced policing in the area prior to the CA establishment (an era defined by racial tension between CPD and residents), may have different views regarding CPD, relative to those who did not.Simultaneously, we anticipate residents who lived in the area when the CA was initially established should be more knowledgeable about its existence.Variable values range from 1-8, where: 1=1 year or less; 2=2-5 years; 3=6-10 years; 4=11-15 years; 5=16-20 years; 6=21-30 years; 7=31-40 years; 8= 41 years or more.
While a wide range of civil-society groups operate within the CA, several groups (e.g., the Black United Front) have played a particularly powerful role in shaping forum deliberations and outputs.While many civil-society groups attempt to represent the interests of the community, by and large, community members may better identify with a particular (or subset of) organization(s).Individuals who are better represented by relevant groups may, because of the stakes involved, be better informed about the CA.
Moreover, the group in which an individual perceives themselves likely have different histories with the CPD, and thus, may have varying perceptions toward it and its officers.To capture this, respondents were asked which of the following stakeholder groups they primarily identify with: the Black community; the City of Cincinnati (employee or other); the police; white residents; other minority communities (including LGBTQ); youths; businesses; religious groups.Each stakeholder group is represented by a dummy variable: in our analysis, the "Black community" is used as our reference category.
Note that our inclusion of stakeholder groups precludes us from including respondent race given the presence of multicollinearity.For instance, in our data, roughly 79% of Black respondents identified "Black community" as their primary stakeholder group; around 66% of white respondents selected "white resident" as their primary group.While we prefer the inclusion of the stakeholder terms-given that this allows respondents to self-identify the group with whom they primarily identify-we recognize that race may play a powerful role in shaping citizen perceptions of police.In the Appendix, we re-estimate our models (described in the following section) excluding the stakeholder term and including a dummy variable indicating whether a respondent is Black (=1) or not (=0).Note that we include as the vast majority of our sample identifies as either Black (43.6%) or white (52.13%), which is largely consistent with US Census population estimates (41.3% and 50.3%, respectively).As shown in Table A3, the findings are consistent across variable operationalizations.

Analysis and Results
Our empirical analysis seeks to examine the impact of an individual's knowledge about CA with perceptions regarding the legitimacy of and trust in the CPD and its officers.While the CPD maintains authority over all citizens in our survey, the frequency, manner and context in which individuals encounter police (an in turn, beliefs regarding police) may largely be impacted by the areas in which they primarily reside.Given the ordered nature of our dependent variable, coupled with the expectation that resident beliefs about police may be geographically clustered, we employ a mixed ordered logistic regression, including random intercepts that account for community specific features (Rabe-Hesketh and Skrondal 2008).Specifically, respondents were asked which of the 52 neighborhoods (or other area) they spend "most of their time."Each neighborhood is assigned a random intercept: individuals who spend most of their time outside of Cincinnati are assigned the same intercept.
Results are presented in Table 2, below.As shown, moving from medium-levels of knowledge about the Collaborative Agreement (=1) to low levels of knowledge (=0) does not have a significant impact on any of the dependent variables.However, moving from medium to high levels of knowledge, has a positive, and statistically significant effect across all models.Consistent with our theory, such findings demonstrate symbolic representation manifests from collaborative forums only when there is a strong understanding of the forum, itself.
The reported effects yield substantively important changes in individual perceptions.In Figure 2, we compare the predicted probabilities that a respondent reports the highest value of each dependent

Robustness Checks & Limitations
While we attempt to capture the causal impact of CA Knowledge on our outcomes of interest by adjusting for all relevant confounders, a number of limitations remain.First, it is possible that our dependent variable is not solely capturing the concept of "beliefs about CPD", but rather, is capturing sentiment towards political institutions more generally.If so, then our analyses would simply be revealing that there is a serial connection between one's i) understanding of governing institutions and ii) their trust in institutions.To test for this, we conduct two placebo-outcome tests, using beliefs about two, unrelated institutions as alternate outcome variables: Cincinnati schools and Cincinnati local government.The variable is captured in an identical manner as Trust, with values ranging from 1-4.If CA Knowledge yields a significant effect, then this critique is valid; if not, we can be confident that the effect of such knowledge is solely linked to beliefs about the CPD (Eggers et al. 2020).The results, presented in Table A4 in the Appendix, demonstrate that CA Knowledge does not impact trust in these institutions.
An alternate concern is that our sample-which includes individuals who feel that their interests are best represented by both i) civil-society groups that have been historically excluded from police policymaking as well as ii) traditionally enfranchised groups-may not be appropriate given that our theory concerns those represented by the former.While, for the purposes of comparison, we prefer the full sample, we accept that this could be impacting our estimates.To test for this, we re-estimate our regressions, excluding those who identify their stakeholder group as the i) police or ii) city employees (~10.46% of the sample).The results, presented in Table A5 in the Appendix, are entirely consistent with our primary model.
Similarly, it is possible that the observed effect for CA Knowledge is heterogeneous across stakeholder groups.Specifically, the effect of CA Knowledge may only hold for groups who already have, on average, more amicable relations with the CPD: if true, the null (or even negative) effect observed for some groups may be washed out by the positive effect observed among others.Given that the tumultuous relationship between CPD and the Black community is what gave rise to the CA's establishment, we anticipate that effects may vary between this group and others.To test for this, we again re-estimate our regressions, with a number of modifications: we eliminate the stakeholder group dummy variables; we include a new dummy variable indicating whether an individual primarily identifies with the Black community stakeholder group (=1) or not (=0); we include an interaction term between the CA Knowledge and the newly created dummy.The augmented model allows for us to examine whether i) the effect of CA Knowledge holds given the new specification, and, importantly ii) whether this effect varies between Black stakeholder groups, as compared to other groups.The results, presented in Table A6 in the Appendix, demonstrate that high levels of CA Knowledge (=2) is associated with heightened perceptions across all models, and ii) this effect does not vary between groups.
While our robustness checks provide greater confidence in our findings, our reliance on crosssectional, observational data precludes us from making causal claims.First, while we attempted to model all potentially problematic confounders, it is possible that the presence of omitted variable bias is affecting our estimates.Moreover, the potential for simultaneity bias remains.Here, individuals who have more favorable views of the police may be more likely to investigate policing policies (due to their satisfaction) and practice, and because of this, be more knowledgeable about the CA.While we would actually argue that, intuitively, the opposite is likely true (dissatisfaction should drive knowledge seeking behavior), we cannot observe, and hence, rule out this possibility.

Discussion & Conclusion
In this paper, we examine whether civil society representation in a formal collaborative policymaking forum spills-over into the belief systems of everyday community members.By bridging the collaborative governance and representative bureaucracy literatures, we argue that representation of civilsociety groups in collaborative forums can have a transformative impact on the way non-participating community members perceive, and potentially, interact with CPD officers.Specifically, we argue that deliberately mobilizing under-represented civil service stakeholders may signal to non-participating citizens that their interests are being faithfully represented and that stakeholders representing their viewpoints during collaboration are true coproducers of the output.We find that the combination of passive representation and true principled engagement in collaborative forums does have a symbolic and transformative effect on the perceptions of non-participating stakeholders regarding CPD.
These findings hold important implications for both collaborative governance and representative bureaucracy literatures broadly as well as policing scholarship specifically.First, we show that the true hallmark attributes of collaboration (e.g., inclusivity, diversity, representation) can, through principled engagement that actually provides all stakeholders opportunities to be heard and shape decisions, offer symbolic and representative qualities outside of the stakeholders participating in the collaborative effort.
More so, deliberately mobilizing stakeholders who lack the resources and capacity to compete against powerful stakeholders can provide far reaching outcomes that align well with the democratic goals of collaboration.While some research still shows that collaborative forums may function as venues for powerful actors to achieve their goals (Angst et al. 2021;Choi and Robertson 2014;Fung 2015), our findings reinforce the importance of well-agreed upon standards of successful collaboration: inclusivity, representation, and principled engagement (Ansell and Gash 2008;Emerson, Nabatchi, and Balogh 2012).
Absent these core principles, passive representation may serve no other purposes but tokenism (Headley et al. 2021).Key leaders would be best served to avoid utilizing citizen engagement efforts under the guise of inclusivity without offering citizens real opportunities to influence policy process and outputs.
Our findings also build on recent scholarship examining dynamics that may transform passive to symbolic representation.Like scholarship that points to certain institutional preconditions necessary for the transformation of passive to active representation (e.g., discretion, salient policy benefits, critical mass, etc.: Keiser et al. 2002;Meier and Bohte 2001;Meier 1993)  While our results provide evidence regarding the symbolic effects of collaborative processes and outcomes, limitations remain.First, our data precludes us from disentangling whether citizens may perceive increasing levels of trust and legitimacy in CPD due to i) success of the actual collaborative process, ii) the representation of important civil society stakeholders, and/or iii) both of these dynamics operating simultaneously.In other words, we provide evidence that passive representation during a collaborative process was successful in increasing regular citizens' perceptions of police; however, the actual causal mechanism remains somewhat unclear.
); the initiation of the Cincinnati Initiative to Reduce Violence (2007); restrictions on the use of beanbag rounds (2010) and batons (2013); the introduction of body worn cameras (2016); and numerous updates to body worn camera technology (2019, 2020: City of Cincinnati 2020).The transformative effect that the CA had on CPD policing practices and CPD-community relations is perhaps best summarized by former NAACP President (and a regular participant in CA proceedings), Christopher Smitherman:

Figure
Figure 1: Distribution of Raw Responses

Finally, our results
are perhaps most important in the context of policing where the potential outcomes of police-citizen interactions may have deleterious consequences.Providing evidence that passive representation in collaborative processes may seep into belief systems of the broader community of stakeholders may stimulate greater reciprocity between police and community as coproducers of public safety.By enhancing perceptions of CPD trust and legitimacy in the broader community, citizens may be more compliant with police, fostering a model akin to that of community policing (Theobold and Haider-Markel 2008; Tyler 2001).

Table 1 ,
below.As shown, very few respondents reported the lowest possible values, and the majority of respondents reported favorable views.

Table 1 : Distribution of the Dependent Variables Response Legitimacy Trust
variable, when CA Knowledge equals low (=0) and high (=2)4.As shown, moving from low to high values of CA Knowledge increases the probability of selecting the highest level of Legitimacy,Obey,   Ignore, Trust, Honest, and Comfort by 56.59, 45.58, 17.69, 52.92, 66.80, and 73.04 percent, respectively.Importantly, the variable Passive Officer attains statistical significance across 5 of 6 regressions, providing further evidence that passive representation can, in this context, transform into symbolic representation.The finding that Passive Officer and CA Knowledge are simultaneously significant demonstrates that two forms of symbolic representation emanating at different stages of the policy process can concurrently manifest within individuals.A number of alternate covariates also attain statistical significance.For instance, being stopped by police, and being a victim of a crime tend to result in more negative perceptions, whereas age and income result in more positive perceptions.The positive, statistically significant coefficients for the stakeholder variables demonstrate that individuals who primarily identify as Black community stakeholders have consistently lower evaluations of the CPD relative to other stakeholder groups.

05 Figure 2: Predicted Probability that Response Equals Highest Value
, our work draws directly on core theoretical principles of collaborative governance to identify principled engagement as a necessary moderating variable to transform passive to symbolic representation.Future work should continue to dissect the types of institutional, political, or social variables necessary to cultivate this important transition.Additionally, while the representative bureaucracy literatures have solely considered how agency-wide and bureaucrat-specific demographics can manifest into symbolic effects, we show that this is just one (of potentially many) pathways.Notably, our findings show that these pathways can complement one another, resulting in compounding, positive effects.

Table A2 : Results with Alternate IV Operationalization
While we cannot disentangle these mechanisms, it is clear from prior research that if CPD bureaucrats did not provide meaningful opportunities for civil service groups to be heard during collaboration, cooption of the forum may likely have resulted in negative ramifications.This provides further evidence of the importance of the marriage of representation and principled engagement for symbolic representation.Ultimately, future experimental work may be needed to interrogate this mechanism in a more causal and nuanced manner.Exploring bias in police shooting decisions with real shoot/don't shoot cases.Crime & Delinquency 64 (9): 1171-1192.