Do citizens prefer policies that the government has designed in collaboration with other stakeholders? Recent scholarship has found that collaborative governance may be more effective than government-only models of policymaking at addressing ‘wicked’ problems, or problems that are ‘dynamic and complex, with no clear definition and no obvious solution’ and that ‘involve multiple stakeholders in multiple organizations across multiple jurisdictions’ (Emerson and Nabatchi Reference Emerson and Nabatchi2015: 7). By working with, for example, local political leaders, civil society organizations and/or private-sector firms, national governments may harness additional resources, expertise and experience that can help forge more responsive policies (Ansell and Gash Reference Ansell and Gash2008; Emerson and Nabatchi Reference Emerson and Nabatchi2015; Scott Reference Scott2015; Stoker Reference Stoker2006).
Citizen support for collaboratively designed policy will be vital for sustaining this approach to governance. There are multiple reasons for this. For one, ‘wicked’ policy problems, such as immigration, climate change, illicit trafficking and global pandemics, are not going away. Additionally, governments qua policymakers face increasing pressures to innovate (Ansell and Torfing Reference Ansell and Torfing2021) and to ‘democratize’ policy governance (Heo Reference Heo2022).
Collaborative governance may be more effective than traditional governance models; in practice, however, collaborative governance is not easy. It requires extensive buy-in in terms of time, energy and resources. Governments may be too short-lived and short-sighted to earnestly engage in collaborative policymaking. Moreover, the benefits of collaboration may be too diffuse – the government may not reap the benefits of a policy that was designed in collaboration with other actors. Finally, power asymmetries among actors can inhibit effective collaboration (Mancilla García and Bodin Reference Mancilla García and Bodin2019). These barriers to collaborative governance are especially prominent in regions like Latin America (Cyr et al. Reference Cyr, Coda, Santamarina and Bianchiforthcoming).
Policymakers, therefore, face a conundrum. A collaborative governance model may design policies that are better poised to address the myriad complex problems that governments face today. But the costs of collaboration may be perceived as too high. If citizens prefer collaboratively designed policy, it may be possible to mitigate these costs and incentivize actors to take collaboration more seriously.
We know very little about how citizens perceive collaboratively designed policies (Cain et al. Reference Cain, Gerber and Hui2021: 901). This article seeks to fill this gap. It tests whether citizens prefer collaboratively designed policies compared to policies created by government alone. It also takes initial steps to theorize the variations in citizen support for collaborative policy design. In focusing on how citizens comparatively perceive collaboration, we answer Ellen Rogers and Edward Weber’s (Reference Rogers and Weber2010) call to think harder about the impact of collaborative governance. We go beyond the specific products of objective collaborative practices to measure whether citizens might actually prefer them – a specific outcome that is itself crucial for incentivizing further collaborative governance.
To measure perceptions of collaboratively designed policies, and specifically whether citizens prefer them to policies that have been created by government alone, we run a set of conjoint experiments with individuals in Argentina and Chile, asking them to rank and rate an education policy, a health policy and a policy to mitigate violence against women. Participants are exposed to different descriptions of those policies, which vary randomly in terms of whether and with whom the national government collaborates. Respondents must then rank and rate each policy pair. We therefore can measure whether and by how much individuals in each country will be more likely to prefer a policy designed in collaboration with non-national government stakeholders.
As far as we know, our article offers first-of-its-kind data on citizens’ relative perceptions of collaborative governance in Latin America – specifically, whether they prefer collaborative policy designs or government-only designs. We choose to focus on Latin American countries because in this region government institutions on their own tend to inspire low confidence (Ardanez et al. Reference Ardanaz, Otálvaro-Ramírez and Scartascini2023). Moreover, Latin America tends to be disproportionately vulnerable to the kinds of wicked problems that collaborative governance models are well poised to address (Blofield et al. Reference Blofield, Giambruno and Filgueira2020). The inability of Latin America governments to address complex policy challenges has led to at times extended social and political crises (Somma et al. Reference Somma, Bargsted, Disi Pavlic and Medel2021). Consequently, collaborative policy design and implementation can be particularly impactful in the region (Cyr et al. Reference Cyr, Coda, Santamarina and Bianchiforthcoming). Understanding how citizens perceive collaboratively designed policy is, therefore, especially important.
We argue that, in general, respondents in both countries will tend to prefer policies designed collaboratively. Citizens will prefer collaboratively designed policies because their trust in government alone is quite low. By contrast, non-national government political actors, including local government officials, civil society and the private sector, tend to elicit higher levels of trust. By involving these other, more trusted actors in policy design, the overall evaluation of those policies will improve.
In what follows, we provide a brief overview of the literature from which our hypotheses are derived. We then explain our research design, before reporting and analysing the results of the conjoint experiments. We find that, on balance, collaboratively produced policies are preferred slightly over government-produced policies in Argentina and Chile. We also find that citizens prefer some combinations of collaboration over others. In particular, citizens seem to most consistently prefer policies that are the result of collaboration between national and local government. Support for collaboration with other stakeholders, including the private sector or civil society, is more unstable and driven by other factors, such as ideology. Despite these nuances, we provide initial evidence that citizens prefer collaborative governance models in Latin America – a fact that could help to incentivize greater investment into this type of policymaking.
Preferences for collaborative policy design
Extensive research has shown that citizen support for policy impacts the policy process. Citizen support can increase the likelihood that a policy will be implemented (Page and Shapiro Reference Page and Shapiro1983). A lack of support, by contrast, can lead to innovation and change (Ferretti et al. Reference Ferretti, Pluchinotta and Tsoukiàs2019). Politicians are especially responsive to the policy preferences of specific groups, including their supporters (Barberá et al. Reference Barberá, Casas, Nagler, Egan, Bonneau, Jost and Tucker2019) and higher-income constituents (Persson and Sundell Reference Persson and Sundell2024). Citizen support can also be quite complex and nuanced (Stoutenborough et al. Reference Stoutenborough, Bromley‐Trujillo and Vedlitz2014). Their views on different policy options can vary extensively (O’Connor et al. Reference O’Connor, Bord, Yarnal and Wiefek2002).
Just as citizen support can help determine policy content, it may also be crucial for shaping who is involved in policy design. Collaborative governance (CG) is ‘a collective decision-making process based on more or less institutionalized interactions between two or more actors that aims to establish common ground for joint problem solving and value creation’ (Douglas et al. Reference Douglas2020: 498).Footnote 1 Collaborative governance differs from polycentric governance. The latter recognizes that decision-making in governance is complex and involves multiple, (semi-)autonomous actors/organizations working in cooperation but also in competition (Carlisle and Gruby Reference Carlisle and Gruby2019: 928). Collaborative governance posits that deeper engagement among actors is required to address complex policy problems (Emerson Reference Emerson2018). Unlike network governance, which emphasizes the process by which different stakeholders address a policy problem (see e.g. Provan and Kenis Reference Provan and Kenis2008), CG is problem focused and oriented towards how collaboration yields policies that are different from those produced by any single actor. Collaborative governance can reasonably be juxtaposed with government-only designed policy when it comes to addressing a particular problem.
Indeed, public policy scholars have suggested that collaborative governance can be a more effective approach to creating policies that address ‘wicked’ problems (Emerson and Nabatchi Reference Emerson and Nabatchi2015), because no single public or private actor – including the national government itself (Abers and Keck Reference Abers and Keck2009) – can effectively address complex problems on their own (Ansell and Gash Reference Ansell and Gash2008; Emerson and Nabatchi Reference Emerson and Nabatchi2015; Huxham and Vangen Reference Huxham and Vangen2013). The COVID-19 pandemic, for example, was a concern for public health, education, the economy and several other policy areas. By pooling the knowledge, resources and experiences of actors from the public and private spheres, collaboratively designed policies can better address complex problems (Bryson et al. Reference Bryson, Crosby and Stone2006: 44). When carried out successfully, CG can lead to higher-quality, more transparent, and more legitimate policies (Ansell and Gash Reference Ansell and Gash2008; Johnston et al. Reference Johnston, Hicks, Nan and Auer2021). Given these outcomes, CG has been viewed as inclusive, deliberative and, therefore, democratic.Footnote 2
To be sure, the actual effectiveness of CG remains largely speculative. We lack the capacity to know if CG works, since the ‘empirical literature still struggles to produce robust generalizations and cumulative knowledge that link contextual, situational and institutional design factors to processes and outcomes’ (Douglas et al. Reference Douglas2020: 495). Still, governments and policymakers are increasingly turning to CG in response to growing frustration with top-down policy design processes. For example, the United Nations Development Programme (UNDP), together with the Andean Development Bank (CAF in Spanish), promotes collaborative governance as one way for national governments to achieve better development outcomes in Latin America and the Caribbean.Footnote 3 The heads of several multilateral development banks recently committed to promoting collaboration in their efforts to address multifaceted crises.Footnote 4 These trends suggest that collaborative governance is here to stay.
Do citizens prefer collaboratively designed policies over those crafted via more conventional (read: government-only) policymaking processes? A wealth of literature explains why citizens tend to support different types of (traditionally produced) policies. Policy content matters (Bechtel and Scheve Reference Bechtel and Scheve2013; Burgoon et al. Reference Burgoon, Nicoli and Vandenbroucke2022; Huber and Wicki Reference Huber and Wicki2021), as do a series of citizen-level traits, including ideological positioning and support for the government (e.g. Huber and Wicki Reference Huber and Wicki2021); proximity to the policy at hand (Bechtel et al. Reference Bechtel, Genovese and Scheve2019); knowledge about the policy (Rhodes et al. Reference Rhodes, Axsen and Jaccard2014); and level of altruism more generally (Bechtel et al. Reference Bechtel, Genovese and Scheve2019).
Understanding why citizens might prefer collaboratively designed policy involves considering an additional factor: governments created the policy in conjunction with other actors. Might the other actors’ participation impact citizen support? The answer is not given. Citizens may reject policies designed in collaboration with actors outside of national government. There are clear concerns about transparency and accountability. Collaborative governance includes actors who are not publicly vetted in an election (see e.g. Norris Reference Norris2011). Concerns that certain groups disproportionately influence policy design can compel citizens to view collaboration with suspicion and/or mistrust (Nabatchi Reference Nabatchi2010). Citizens in the United States, for example, viewed collaboration with organized groups negatively (Cain et al. Reference Cain, Gerber and Hui2021).
Despite these concerns about the non-democratic nature of collaborative policy design, it is nevertheless the case that contextual features are likely to shape how citizens perceive collaboratively designed policy. In Bolivia, for example, where social movements have used the politics of the street to pressure the government (Valdivia Rivera Reference Valdivia Rivera2021), citizens may be more supportive of policies designed in collaboration with social movements. Despite the importance of context, the great majority of work on CG has occurred in the Global North, especially the United States (see Koontz and Thomas Reference Koontz and Thomas2006 for an overview). We know much less about how collaborative governance operates in regions like Latin America, including whether citizens might prefer collaboratively designed policy.Footnote 5
We suspect that, despite identified concerns about transparency, accountability and the undue influence of unelected actors in policymaking, citizens in countries like Argentina and Chile will nonetheless prefer policies designed in collaboration over those designed by government alone. One key driver of this preference is the fact that national political institutions on their own tend to inspire very low levels of confidence.
Take, for example, the presidency, Congress and political parties – three institutions that participate in national policymaking (Figure 1). In 2020, only 32% of citizens in Latin America expressed a lot (14.1%) or some (17%) trust in the presidency; 65.5%, by contrast, had little (25.7%) or no (39.9%) trust. Regarding Congress, 19.7% of citizens in the region had a lot (5.0%) or some (14.7%) trust, while 76.1% had little (34.1%) or no (42.0%) trust. Finally, 13.5% had a lot (2.9%) or some (10.3%) trust in political parties, and 84% had little (28.9%) or no (55.1%) trust (Latinobarómetro Reference Latinobarómetro2020). Large majorities of citizens have little to no trust in several national policymaking institutions.
Figure 1. Levels of Citizen Trust in Latin America (2020)
Non-state and local actors, by contrast, are more trustworthy, and far fewer citizens express little or no trust in these groups. Using the same 2020 Latinobarómetro (Reference Latinobarómetro2020) survey, we find that 38.2% of Latin American citizens have a lot (11.0%) or some (27.2%) trust in NGOs, while 49.3% had little (30.2%) to no trust (19.1%) in these groups. There was limited data on citizen trust in the private sector, and so we use trust in banks as a proxy. Here, 41.9% had a lot (12.6%) or some (29.3%) trust in banks, while 54.6% had little (33.5%) to no trust (21.1%). While no question on trust in local government was included in the 2020 Latinobarómetro (Reference Latinobarómetro2020) survey, a similar question asked in the same Latin American countries in the 2021 Latin American Public Opinion Project (LAPOP Reference LAPOP2021) survey reveals that, on a scale of 1 to 7, where 1 denotes no trust and 7 denotes a lot of trust, 43.3% of citizens responded with 1, 2 or 3, while 42.9% responded with 5, 6 or 7, suggesting that nearly as many citizens have some or a lot of trust in local government as does not.
Overall, then, Latin American national government institutions elicit low levels of trust. At the same time, they have historically been perceived as highly corrupt (Canache and Allison Reference Canache and Allison2005). Non-national-government actors, by contrast, elicit higher levels of trust. If low levels of trust can negatively impact support for government actions and policies (Agley Reference Agley2020; Alessandro et al. Reference Alessandro, Lagomarsino, Scartascini, Streb and Torrealday2021), it follows that, in countries where large majorities of the population do not trust national government, collaboration may actually be less risky when it comes to governance accountability. Deeply distrusted policymakers have little to lose when it comes to collaborating with others – especially when these ‘others’ elicit higher levels of trust. This leads to our first hypothesis:
Hypothesis 1: People will prefer collaboratively designed policies over those produced by government alone.
Non-state and local actors have influence across a wide variety of policy domains in Latin America, including in education, health and policies against gender-based violence (Birn et al. Reference Birn, Nervi and Siqueira2016; Chambers-Ju Reference Chambers-Ju2021; Fahlberg et al. Reference Fahlberg, Velasquez, Wise and Simon2023). A focus on these three policy sectors is theoretically and empirically appropriate. For one, collaborative governance models have been used to address education (Eldridge et al. Reference Eldridge, Larry, Baird and Kavanamur2018), health (Emerson Reference Emerson2018) and anti-gender-based violence (Raftery et al. Reference Raftery, Howard, Palmer and Hossain2022). It is reasonable to assume, therefore, that collaborative governance might also be used to design these particular policies. Additionally, we specifically chose policies that were, broadly speaking, largely neutral or, at the very least, not highly polarizing. An experiment that measures support (or not) for appropriately equipped hospitals, tutors in public schools and safe spaces for victims of violence is less likely to capture and subsume preferences based on higher-order religious or social values, as might be the case with abortion policy, or highly politicized issues, such as climate change. It is therefore useful to understand whether preferences for collaboratively designed policies obtain across different sectors. We hypothesize it will:
Hypothesis 2: People’s preference for collaboratively designed policies over government-only produced policies will be consistent across different policy domains.
Different groups elicit different levels of trust. In Figure 1, more citizens had less trust in banks and NGOs than in local government.Footnote 6 Citizens should not equally value collaboration with all types of actors. Individual preferences when it comes to collaborative policy design are likely to correspond with other, individually held political and policy preferences (Bechtel and Scheve Reference Bechtel and Scheve2013; Burgoon et al. Reference Burgoon, Nicoli and Vandenbroucke2022; Huber and Wicki Reference Huber and Wicki2021; Rhodes et al. Reference Rhodes, Axsen and Jaccard2014).
Our last set of hypotheses test these claims. Citizens who express more trust in one type of non-national-government actor are likely to prefer policies designed in collaboration with that actor. This means that citizens with higher levels of trust in the private sector will prefer collaboration with the private sector over collaboration with civil society or the local government. The same will be likely, we posit, for citizens who express higher levels of trust in civil society and local government.Footnote 7
Hypothesis 3a: People who exhibit more trust in the private sector are more likely to prefer policies that have been designed in collaboration with the private sector than with civil society organizations or local government.
Hypothesis 3b: People who exhibit more trust in social movements are more likely to prefer policies that have been designed in collaboration with civil society organizations than with the private sector or local government.
Hypothesis 3c: People who exhibit more trust in local government are more likely to prefer policies that have been designed in collaboration with local government than with the private sector or civil society organizations.
Despite their simplicity, these hypotheses are not obvious. Even if citizens find certain actors to be more trustworthy, their role in policy creation may not be welcome. Non-state actors are unelected. Few mechanisms exist to hold them accountable (Halachmi Reference Halachmi, Bovens, Goodin and Schillemans2014; Nabatchi Reference Nabatchi2010). Additionally, early research suggests that citizens can reject collaborative policy design with certain stakeholders (Cain et al. Reference Cain, Gerber and Hui2021).Footnote 8 Collaborative governance can exacerbate the accountability gap that plagues government, especially in newer democracies (Mayer et al. Reference Mayer, Edelenbos and Monnikhof2005).
Finally, policymakers may exploit the complexity of collaborative governance models to shift policy failure to other actors (Bache et al. Reference Bache, Bartle, Flinders and Marsden2015). Citizens, for their part, may perceive that collaboratively designed policies are created to benefit only one group or sector (Huber and Wicki Reference Huber and Wicki2021). In both cases, CG contributes to the politicization of governance, a phenomenon that is problematic in Latin America, where public administration is ‘characterized by patronage appointments, patrimonialism, and a weak capacity to execute public policies’ (Polga-Hecimovich Reference Polga-Hecimovich2019).
Research design
We designed conjoint experiments to measure citizen preferences regarding different types of policies – some designed by the government alone, others designed in collaboration with different actors – in Argentina and Chile. These two countries exhibit many of the structural problems that affect Latin America as a whole, including poverty and inequality. They have also had to endure different types of crises – hyperinflation in Argentina (Perelman Reference Perelman2024), extended social protests in the case of Chile (Somma et al. Reference Somma, Bargsted, Disi Pavlic and Medel2021) – that have complicated policymaking over the past years. Nevertheless, they both also have relatively high levels of democracy and state capacity (Varieties of Democracy Report, Papada et al. Reference Papada2023). These shared traits make it reasonable to expect that both governments would be capable of proposing the policies included in the conjoint experiments we designed.
Additionally, both countries have fairly stable party systems (Piñeiro Rodríguez and Rosenblatt Reference Piñeiro Rodríguez and Rosenblatt2020) anchored by left and right options.Footnote 9 In Argentina, at the time of data collection, the centre-left coalition, called Unión por la Patria (UxP) included the long-running Peronist party. The centre-right coalition, by contrast, was Juntos por el Cambio (JxC). In Chile, the right option was Partido Republicano (PRCH), and the left choice was the Frente Amplio (FA).Footnote 10 Given the ideological distinctions between these two parties/coalitions, citizens are likely to have opinions about them – a feature of party politics that is far from common in the region (Kitschelt et al. Reference Kitschelt2010).
Finally, Argentina and Chile display interesting variations when it comes trust in different institutions and actors, especially vis-à-vis the regional average (see Table 1). For example, Argentines and Chileans have lower levels of trust in the presidency, Congress and political parties, when compared with the regional average. By contrast, Argentines and Chileans had a lot or some trust in NGOs, while fewer said they had a lot or some trust in banks. Argentines with a lot or some trust in local government nearly aligned with the regional average, whereas Chileans were slightly lower. These variations allow us to assess whether trust in non-national-government stakeholders plays a role in preferring collaboratively designed policies.
Table 1. Levels of Citizen Trust in Latin America (Average), Argentina, and Chile (%)
Source: Latinobarómetro (Reference Latinobarómetro2020); data on local government came from LAPOP (Reference LAPOP2021).
After Jens Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2014), conjoint experimental designs became a prominent method to assess: immigration preferences (Hainmueller and Hopkins Reference Hainmueller and Hopkins2015), bureaucratic behaviour (Oliveros and Schuster Reference Oliveros and Schuster2018), corruption (Klasnja et al. Reference Klasnja, Lupu and Tucker2021; Mares and Visconti Reference Mares and Visconti2020), vote choice (Franchino and Zucchini Reference Franchino and Zucchini2015; Kirkland and Coppock Reference Kirkland and Coppock2018) and perceived polarization in social media (Aruguete et al. Reference Aruguete, Calvo and Ventura2023). Conjoint experiments provide treated individuals with two competing profiles – in our case, policies – with randomized traits. After exposure, the participants must select the profile they prefer or, in our case, the type of policy they prefer.
Our experiment adapts the traditional conjoint design to compare policy characteristics and evaluate levels of support for the different frame elements. We include three conjoint experiments, each of which presents two vignettes describing the same policy crafted in different ways. In typical conjoint survey experiments, the task is repeated multiple times, exposing a respondent to different combinations of attribute values (levels). In our case, each vignette pair addressed a different policy area, allowing us to test for preferences across policy sectors. Respondents evaluated two vignettes for three different policies: one providing tutors in all public schools; a second seeking to improve medical equipment supply in public hospitals; and a third providing safehouses for women victims of domestic violence.Footnote 11
These exact policies were not contemplated in Chile or Argentina. However, all three were relevant in the aftermath of the COVID-19 pandemic, when data collection occurred. In each country, concerns about children being left behind academically in the public school system, under-equipped hospitals and increasingly vulnerable victims of domestic violence were salient during the pandemic. Therefore, policies on each of these issues might emerge. The presentation of each pair of vignettes was randomly assigned to avoid ordering effects.
Each vignette pair varied randomly in three ways. First, they varied in terms of whether the government collaborated or not. This first attribute had four levels in each country (no collaboration – just the national government; collaboration with civil society; collaboration with the private sector; collaboration with municipal/local government). For operationalization purposes, we treated each level dichotomously. Therefore, we tested preferences over a policy designed by Government-only (1) or in collaboration with other actors (0). We tested preferences over a policy designed by national government in collaboration with NGOs (Organizations), versus all other options (i.e. by government alone or in collaboration with companies or municipalities). The same occurred with Companies and Municipalities (collaboration with a certain actor versus others or no collaboration at all).Footnote 12
Second, the vignettes varied in terms of the party in power when the policy was (hypothetically) designed. In Argentina, either the centre-right Juntos por el Cambio or the centre-left Unión por la Patria was in power. In Chile, it was the radical right Partido Republicano or the left Frente Amplio option. Third, they varied in terms of cost. This third attribute had two levels (0.5% or 1% of the national budget in each country).
For each policy, participants received two vignettes, which varied the three attributes in Table 2 randomly, uniformly (i.e. with equal probabilities for all levels in a given attribute), and independently from one another. Figure 2 presents examples of the rotation of vignettes in the Argentine survey (the only difference with the Chilean survey is that the government parties rotated between Partido Republicano and Frente Amplio).Footnote 13
Figure 2. Conjoint Experiment, Examples of Rotations in the Argentine Survey
Table 2. Conjoint Experiment, Policy Attributes
The three conjoint designs were included in nationally representative surveys that were carried out online in June and July 2023, through panels conducted by Netquest. Each included approximately 2,500 respondents. The entire sample was profiled with demographic filters applied for gender and geographic distribution. The number of participants met national representative samples for each country and held enough statistical power for the different experimental treatments in the survey.Footnote 14
Regarding the number of attributes in a conjoint experiment, there is a trade-off between masking and satisficing (Kirk et al. Reference Kirk, Hainmueller, Hopkins and Yamamoto2018). With too few attributes, respondents may associate those provided with others omitted from the design. To avoid this, we include the additional attributes of party in power and policy cost in our vignettes. Too many attributes in a conjoint design can increase the cognitive burden for participants (Krosnick Reference Krosnick1999). We sought to mitigate this burden, because our vignettes contained a substantial amount of text.
After reading each of the policy pairs, participants were asked to identify the one they preferred (see Table 3 below), allowing us to directly test our hypotheses. We also introduced a second measure of support for the individual policies in each pair. Participants rated each policy description on a scale of 0 to 10, where 0 denoted ‘no support’ for a particular policy, and 10 denoted ‘total support’ for that policy. This second measure added nuance to the respondent’s expressed preference. One may prefer one policy over the other but still exhibit quite low levels of support altogether. For ease of presentation below, we only present models for the question on policy preference, since these directly address the hypotheses.
Table 3. Questions Included after Each Pair of Vignettes
To test Hypothesis 2, we included a variable with three categories to test preferences on collaboration in the design of the different types of policies (education – our model’s baseline – as well as health and anti-gender-based violence). This allowed us to examine whether preferences for collaboratively designed policies obtained across three policy domains. Finally, to test Hypotheses 3a–3c, we measured individual preferences for policies designed in collaboration with the private sector, civil society organizations or local government (here we exclude the Government-only dummy). These three actors collaborated with different Latin American governments in the design of policies to address the COVID-19 pandemic (Cyr et al. Reference Cyr, Coda, Santamarina and Bianchiforthcoming). It is therefore reasonable to expect that they would collaborate with the national government on the policies included in the vignettes. We measured levels of trust in Congress, judges, police, political parties, social movements, the private sector, national government and local government using the following question, ‘On a scale of 1 to 7, where 1 denotes “least trustworthy” and 7 denotes “most trustworthy”, how trustworthy do you think these institutions are?’ To avoid ordering effects, the institutions’ order was randomly assigned across respondents. Based on these results, we include the variables Trust, Private Sector; Trust, Social Movements; and Trust, Local Government in different estimations.
We also included controls for other individual-level traits that have been shown to impact citizen preferences for different policies. We included age (18–25; 26–35; 36–45; 46–55; 56–65; and older than 65); gender (baseline woman, including other genders different from man); level of education (ranging incomplete primary education to complete, postgraduate education); subjective income (ranging from 1 to 10, with 1 as the lowest income); self-reported ideology (ranging from left to right on a scale of 1 to 7); self-perception of skin colour (ranging from 1 to 10, with 1 as the lightest); political knowledge (we ask how long the Senate term in each country is); and whether the respondent works, is partisan, has participated in demonstrations in the last year, has attended public school for elementary and secondary levels (public elementary and public secondary, respectively) and has ever received treatment at a public hospital. To control for order effects, we included a dummy indicating whether the vignette with the preferred policy was shown on the right side (PolicySide).
In half of the surveys, the three conjoint experiments were preceded by a battery of questions on participant attitudes towards government and democracy. In the other half, the same battery of questions came immediately after the conjoint experiments. We include a dummy variable named treatment that indicates the former with the value of 1 and 0 for the latter. We segmented some models to separately address intended vote in an eventual runoff between Juntos por el Cambio (JxC) and Unión por la Patria (UxP) in Argentina, and Partido Republicano (PRCH) and Frente Amplio (FA) in Chile.Footnote 15
Below, we report logit models and marginal means to detect subgroup frame effects (Leeper et al. Reference Leeper, Hobolt and Tilley2020). The estimated coefficient represents the difference in the likelihood of an attribute being selected (relative to the baseline) – that is, the marginal effect of one level relative to another. In Figure B in the Supplementary Material, we also present the plot for average marginal component effects (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014).
Results
Table 4 shows the results for Argentina. Model 3 considers Argentine respondents whose declared vote was for Juntos por el Cambio (JxC), while Model 4 considers those whose declared vote was for Unión por la Patria (UxP). All other models in Table 4 include all respondents.
Table 4. Estimates of the Likelihood of Support for Collaboratively Produced Policies in Argentina
Notes: Controls were calculated but omitted from the table. In models 1 to 6, collaboration with municipalities is omitted from the table. In model 7, collaboration with companies is omitted from the table. Government-only is excluded from the analysis in models 5 to 7. Education category is the baseline in type of policy. Standard errors in parentheses.
+ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 5 shows the results for Chile. Model 10 considers Chilean respondents whose declared vote was for Partido Republicano (PRCH), and Model 11 includes those who said they would vote for Frente Amplio (FA). All other models in Table 5 include all respondents. Descriptive statistics are available in Table A in the Supplementary Material, including control variables that, although calculated, were omitted from the tables.
Table 5. Estimates of the Likelihood of Support for Collaboratively Produced Policies in Chile
Notes: Controls were calculated but omitted from the table. In models 8 to 13, collaboration with municipalities is omitted from the table. In model 14, collaboration with companies is omitted from the table. Government-only is excluded from the analysis in models 8 to 11. Education category is the baseline in type of policy. Standard errors in parentheses.
+ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.
Hypothesis 1 expresses our general expectation that respondents in both countries will prefer policies that were designed collaboratively over those designed by government alone. We find support for this argument. When respondents in each country had to choose between two policies, they were more likely to prefer the one designed collaboratively – 4% more likely in Argentina and 5% more likely in Chile.Footnote 16 The Government-only dummy is negative and statistically significant across the models that consider all respondents in Tables 4 and 5. Figure 3 shows all coefficients based on Model 1 for Argentina and Model 8 for Chile. This preference obtains if we disaggregate by individuals whose intended vote was for the left (UxP in Argentina, Model 4, and FA in Chile, Model 11). By contrast, there is no statistically significant difference regarding the preferences of those who would vote for right options (JxC in Argentina, Model 3, and PRCH in Chile, Model 10).
Figure 3. Conjoint Experiment Results (Likelihood of Support)
Table 4 shows us that Argentines in general prefer collaboratively designed policy. Beyond this general outcome, however, patterns about preferences are harder to discern. For example, Argentines have a statistically significant preference against collaborations between the government and organizations, on one hand, and for government and companies, on the other. They also do not have clear preferences when it comes to policy cost.
In Chile, the results are slightly different. There is no statistically significant relationship between collaboration with organizations or companies, except with those whose intended vote is for the left coalition, Frente Amplio. These individuals prefer policies that are not designed in collaboration with companies. Chileans are also more likely to prefer policies that are slightly more expensive, representing 1% versus 0.5% of the national budget.
We included an additional attribute regarding which party was in government when the (hypothetical) policy was designed. Policies that were designed by the left coalition – UxP in Argentina and FA in Chile – were less likely to be preferred over those that were produced by the other, opposition party, JxC and PRCH, respectively.Footnote 17 Notably, at the time of data collection, the president in each country was from the left coalition – Alberto Fernández (UxP) and Gabriel Boric (FA) – and both registered low levels of support. We randomly rotated the party in government in the conjoint experiment. Nevertheless, it may have been difficult for respondents to isolate their evaluation of hypothetical policies from their (generally negative) evaluation of the current government.
To have a better sense of policy design preferences in each population, we follow (Leeper et al. Reference Leeper, Hobolt and Tilley2020) recommendations and use marginal means to detect subgroup frame effects in our conjoint analysis. We disaggregate the population by vote intention – specifically, whether an individual said they would vote for the left (UxP or FA), the right (JxC or PRCH) or no one (blank) in a hypothetical runoff election. This filter acknowledges the likelihood that if a policy is designed by one’s preferred party it may impact the evaluation of other aspects of the policy design. The marginal means are presented in Figure 4 and include point estimates and confidence intervals. The y-axis contains the attributes and levels from the conjoint. The marginal mean effect is the average probability of a policy being selected given an attribute level, which does not depend on the choice of base or reference level.
Figure 4. Marginal Means
When we disaggregate our surveyed population by declared vote intention, we see more statistically significant variation in the stated preferences towards differently designed policies. In Chile, for example, all subgroups were more likely to prefer collaboratively produced policies over those produced by government alone. In Argentina, this preference was statistically significant for those who would vote blank and for those who said they would vote for the centre-right JxC (see Figure 4).
Individuals whose declared vote was for right parties, that is, Juntos por el Cambio and Partido Republicano, tended to prefer policies that were created by their party of choice. They were also more likely to prefer collaboration with the private sector (Companies) versus civil society or local government (Municipalities). The opposite, however, was preferred by declared voters for the (centre-)left Unión por la Patria and Frente Amplio, who were more likely to prefer collaboration with Municipalities over that with Companies or civil society Organizations. Finally, Figure 4 shows that, generally speaking, more expensive policies are very slightly preferred over those that are less costly (1% versus 0.5% of the national budget). This relationship holds for all subgroups except blank voters in Argentina, for whom the difference is not statistically significant.
Hypothesis 2 posits that the preferences for collaboratively produced policies will be consistent across distinct policy areas. In both countries, in all three policy areas (health, education, violence against women), collaboratively produced policies were more likely to be preferred than those that were created exclusively by government, as Figure 5 shows. However, education policy in Argentina and gender policy in Chile exhibited the smallest differences between collaboratively produced preferences and those generated solely by the government, averaging around three percentage points, while the largest differences were in health and gender policies in Argentina, and education in Chile, with an average of four and six percentage points, respectively.
Figure 5. Conjoint Experiment Results (Likelihood of Support)
Our hypotheses suggest that citizens will prefer policies produced in collaboration with other actors when governmental actors on their own elicit low levels of trust. These non-national-government actors may be perceived as more trustworthy. Hypotheses 3a–3c allow us to test this claim more directly. Specifically, we test the notion that citizens who exhibit higher levels of trust in one actor (e.g. private sector) versus other actors with whom the government might collaborate (e.g. social movements or local government) will tend to prefer policies that were produced collaboratively with that trusted actor.
We find evidence for many of these claims. Models 5 to 7 (Argentina) and 12 to 14 (Chile) exclude the dummy variable Government-only, permitting us to focus on preferences for collaboration with different actors. For example, we find that, as trust in the private sector grows, policies designed in collaboration with Companies (Hypothesis 3a) were more likely to be supported than policies designed with civil society Organizations or Municipalities (see the upper images in Figure 6). The predicted probability of support increases by about 10 percentage points when we consider respondents with the lowest and the highest levels of confidence in the private sector.
Figure 6. Likelihood of Support to Policies Produced in Collaboration with Companies/Municipalities According to the Confidence in the Private Sector/Local Government
Hypothesis 3c posits that citizens who expressed higher trust in local government will tend to prefer policies produced in collaboration with Municipalities compared to those produced with Companies or civil society Organizations. The lower images in Figure 6 demonstrate support for this hypothesis. In both countries, the predicted probability of support increases by around five percentage points when we compare respondents with the lowest levels of trust in the local government versus those with the highest.
Alternatively, we do not find support for Hypothesis 3b. Argentine and Chilean respondents with higher levels of trust in social movements were not more likely to prefer policies designed in collaboration with civil society (Organizations). We likely conflated theoretically civil society organizations with social movements – a common but not necessarily justifiable decision, given recent work on the Global South (Daniel and Neubert Reference Daniel and Neubert2019; see also Della Porta Reference Della Porta, In Baert, Koniordos, Procacci and Ruzza2010). Evidence suggests that these two phenomena should be conceptually differentiated. Indeed, in Argentina, social movements tend to be associated with the centre-left, Peronist party, whereas civil society organizations, and NGOs in particular, tend to be associated with the centre-right (Sorj Reference Sorj2007). In Chile, they were fairly integrated into institutionalized politics following the first constitutional assembly (Rozas-Bugueño Reference Rozas-Bugueño2024).
Contextual particularities aside, it is important to understand these conceptual differentiations, so that future models can isolate the effect of the theoretically appropriate phenomenon rather than conflating them. Specifically, we follow Antje Daniel and Dieter Neubert (Reference Daniel and Neubert2019: 176) in recommending that future research use the term ‘civil society organization’ when the theoretical focus is on existing associations in the public sphere arena, and the term ‘social movement’ when the focus is on ‘mobilization and action’.
Finally, the order effects control is negative and statistically significant in almost all the estimations. The vignette order of appearance had an effect on the likelihood of support: those viewed on the right side of the survey were less likely to be preferred than those on the left.Footnote 18
A concluding discussion
What, then, seems to be the relationship between collaboratively designed policy and citizen support? Our findings offer initial support for the notion that citizens, in general, prefer policies that the national government crafts in collaboration with other actors. This was the case in Argentina and Chile, and we found support for these preferences across different policy areas, including education, health and gender-based violence. The probability of preferring collaboratively produced policies was, on average, five percentage points higher than compared with policies designed exclusively by government.
When we drill down and disaggregate from this broad, general finding, preferences become more nuanced. Take, for example, the question of with whom the national government collaborates. Citizens in both countries, with different degrees of significance, preferred policies designed in collaboration with municipal governments. The likelihood of preferring collaborations with civil society organizations and also the private sector was much more variable and often negative (that is, citizens tended to not prefer collaboration with these actors). At least in Argentina and Chile, then, the collaboration of local governments in national policy design may be a boon for policy support. Research shows that local governments elicit different levels of trust than national government. Additionally, the source of said trust is often different (Fitzgerald and Wolak Reference Fitzgerald and Wolak2016). Our findings may tap into a similar phenomenon. They also cohere with recent qualitative case-study work on collaborative responses to the COVID-19 pandemic which suggest that collaboration with local governments in Argentina and Chile, when it occurred, produced responsive policies (Cyr et al. Reference Cyr, Coda, Santamarina and Bianchiforthcoming).
Partisanship also helps to distinguish preferences over collaboration in both countries. For example, individuals with intentions to vote for a particular party tended to prefer policies designed by that party. Additionally, right-leaning voters were more likely to prefer policies designed in collaboration with the private sector. Left-leaning voters, by contrast, preferred policies designed in collaboration with municipal governments. These findings reinforce the conclusion that the in-group loyalty that partisanship tends to foster can be fundamentally important for collaborative processes (Liu Reference Liu2024).
Preferences for collaboratively designed policy were also shaped by prior levels of trust. Citizens were more likely to prefer policies created in collaboration with actors they trusted more, including local government and especially the private sector. One implication may be that governments seeking to shore up support with constituents may benefit from designing policy with more trustworthy non-state and/or non-national actors – especially local government officials and the private sector (see also Bundi and Pattyn Reference Bundi and Pattyn2023).
Our finding that citizens tend to prefer collaboratively produced policies is not trivial. Ours is the first systematic testing of how citizens view collaborative governance in Latin America, a region that confronts several, overlapping challenges, including climate change (Araos et al. Reference Araos, Ford, Berrang-Ford, Biesbroek and Moser2017), forced migration (Castles Reference Castles2006) and global pandemics (Al-Ali Reference Al-Ali2020). Nevertheless, given weak state capacity and national governments that inspire low levels of trust, Latin American countries are likely to struggle to address the policy consequences of the wicked problems they disproportionately face. In short, Latin America could benefit greatly from adopting a more collaborative approach to governance (Cyr et al. Reference Cyr, Coda, Santamarina and Bianchiforthcoming).
Nevertheless, and despite its theorized benefits, collaboration between state and non-state actors is difficult to achieve. Collaborative governance requires investment, time and, yes, trust from all parties involved. It is costly and unwieldy in practice (Lahat and Sher-Hadar Reference Lahat and Sher-Hadar2020). Collaborative arrangements may lose steam over time, as stakeholders bow out or participate less (Heikkila and Gerlak Reference Heikkila and Gerlak2016). Especially in countries like Argentina and Chile, which are vulnerable to economic, social and political crisis, collaboration may be difficult to achieve over time.
Given these obstacles, it is difficult to imagine that collaborative policy design could be feasible. Government actors tend to avoid risks and revert to ‘conventional interpretations of policy problems and solutions’ (Doberstein Reference Doberstein2016: 823). Other stakeholders may need to be convinced of the value of collaborative governance (Hileman and Bodin Reference Hileman and Bodin2019). Towards that end, our article offers a potent argument: Collaboratively produced policy is likely to be preferred by citizens.
This argument is strengthened by the fact that past research (Cyr et al. Reference Cyr, Bianchi, González and Perini2021) provides evidence that collaboratively designed policies are associated with better outcomes in Latin America. When national governments in the region devised COVID-19 policy responses in collaboration with other actors, the rates of contagion and mortality were, on average, lower than in those countries where governments did not collaborate. This was especially the case where national governments collaborated with local government and civil society organizations, two of the actors included in our study.
Popular support for collaborative policymaking may well be crucial for Latin American democracies. Regimes are transitioning away from democracies across the globe (Bermeo Reference Bermeo2016; Lührmann and Lindberg Reference Lührmann and Lindberg2019). If citizen perceptions about governance improve, the increasingly strained relationships between citizens and political institutions may begin to reverse (Knutsen et al. Reference Knutsen, Marquardt, Seim, Coppedge, Edgell, Medzihorsky and Lindberg2024). Engaging other actors in policy design may help to rebuild that trust.
Certainly, we cannot know for sure that our initial findings in Argentina and Chile travel beyond these countries. Despite the numerous similarities observed across Latin American countries regarding low levels of institutional trust, we must exercise caution when attempting to generalize our results. We consider it important to test hypotheses in cases with variation in levels of trust and in real-life party in government. Future research, therefore, should continue to measure citizen preferences on collaborative governance in Latin America.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/gov.2025.10014.
Financial support
This work emerged from a larger project, Colabora.lat, which was carried out with the aid of a grant from the International Development Research Centre, Ottawa, Canada (Grant no. 109,500). The views expressed herein do not necessarily represent those of IDRC or its Board of Governors. Julieta Suarez-Cao also acknowledges the funding of FONDECYT Regular #1240104 and the Millennium Nucleus NCS2024_065.
Ethics statement
The research using human subjects has been approved by the Institutional Review Board (IRB) of the University of Maryland, no. 2054425-1.
Open access
