Policy Research Working Paper 10055 Policy Preferences in Response to Large Migration Inflows William Allen Isabel Ruiz Carlos Vargas-Silva Social Sustainability and Inclusion Global Practice May 2022 Policy Research Working Paper 10055 Abstract What are the preferred policy responses of host country conditional access to the labour market (i.e., only in certain residents to large migration inflows, and to what extent are occupations) or full free access to the alternative of no access. these preferences driven by contact with migrants as well There is support for unrestricted location choices and access as values such as humanitarianism? This paper addresses to public services, as well as conditional rights to family these questions using new data on preferences of Colombi- reunification (i.e., only if able to support dependants). ans for responding to the large inflow of Venezuelans into Respondents also support the use of numerical limits and their country. In a conjoint survey experiment, respondents limiting the length of the residency permit. The results show selected and rated different policy packages comprising that those who have less contact with Venezuelans, those variations in six policy dimensions: (1) labour market who put more weight on economic priorities, and those access, (2) location restrictions, (3) public service access, who see the situation in Venezuela as mainly an economic (4) family reunification, (5) numerical limits, and (6) length problem, tend to support policies that are more restrictive.. of residency. The results suggest support for the options of This paper is a product of the Social Sustainability and Inclusion Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at isabel.ruiz@bsg.ox.ac.uk . The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Policy Preferences in Response to Large Migration Inflows* William Allen University of Oxford william.allen@politics.ox.ac.uk Isabel Ruiz University of Oxford isabel.ruiz@bsg.ox.ac.uk Carlos Vargas-Silva University of Oxford & Peace Research Institute of Oslo (PRIO) carlos.vargas-silva@compas.ox.ac.uk Keywords: Policy preferences, migration, humanitarianism, conjoint experiment, Venezuela, Colombia. JEL Codes: F22, F51, J61, J68 * This paper was commissioned by the World Bank Social Sustainability and Inclusion Global Practice as part of the activity “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts.” The activity is task managed by Audrey Sacks and Susan Wong with assistance from Stephen Winkler. This work is part of the program “Building the Evidence on Protracted Forced Displacement: A Multi-Stakeholder Partnership''. The program is funded by UK aid from the United Kingdom's Foreign, Commonwealth and Development Office (FCDO), it is managed by the World Bank Group (WBG) and was established in partnership with the United Nations High Commissioner for Refugees (UNHCR). The scope of the program is to expand the global knowledge on forced displacement by funding quality research and disseminating results for the use of practitioners and policy makers. This work does not necessarily reflect the views of FCDO, the WBG or UNHCR. 1. Introduction What are the preferred policy responses of host country residents to large migration inflows, and to what extent are these preferences driven by contact with migrants as well as values such as humanitarianism? Forced migration has been central to recent electoral shifts in many host countries (Altındağ and Kaushal 2021; Campo et al. 2021; Dinas et al. 2019; Dustmann et al. 2019; Marbach and Ropers 2019; Rozo and Vargas 2021; Steinmayr 2021). It has also arguably contributed to the erosion of public trust in governments who are perceived to be ineffective in dealing with these inflows and raises questions about the role for policymaking in addressing public concern (Connor 2018; Jeannet et al. 2021; McLaren 2015). A large body of work examines the patterns and determinants of immigration attitudes (Dinesen and Hjorth 2020), as well as public preferences for immigration levels (Citrin et al. 1997; Facchini and Mayda 2008; Scheve and Slaughter 2001). However, the control of inward flows is just one of several kinds of policies available to policymakers in response to migration. In reality, migration policies—either as proposed by political parties or implemented by governments—are multidimensional and comprise several elements. Existing observational surveys, typically measuring public support for single policy proposals in isolation from other topics that are also relevant for migration, provide limited insights on how host country citizens perceive these bundles of policies and distinguish among their constituent parts (Helbling et al. 2017; Helbling and Leblang 2019). Yet expressed public preferences potentially contribute to the ways that national and regional policymakers think about and act upon this issue, as implementing migration policies relies on a degree of public consent. We address this problem by considering the case of Venezuelan migration to Colombia. Over the last decade, Venezuela has experienced a multifaceted humanitarian, economic and social crisis that has led to a large outflow of its citizens. Official figures from the R4V platform indicates that, by the end of October 2021, 5.91 million Venezuelans had left the country (about 16% of the country’s population), with over 1.75 million migrating to neighbouring Colombia1 This case is important for empirical and theory development because both populations share ethnolinguistic, social and cultural features. These similarities potentially present divergent possibilities: on the one hand, if Colombians and Venezuelans possess similar occupational skills—then they may be close substitutes in the labour market, giving rise to more restrictive policy preferences on the basis of greater economic 1 Statistics obtained from the R4V website – The platform for interagency coordination for refugees and migrants from Venezuela: http://www.r4v.info 2 competition (Borjas 2014). On the other hand, cultural similarity and shared history between the two groups (such as by being part of the same country in the early 19th century) could lead to greater solidarity—and less restrictive preferences—in light of the serious humanitarian situation experienced by Venezuela. Using a pre-registered conjoint survey experiment fielded online during March 2021 (N=2,508 adult Colombian nationals), we first address how Colombians perceive different packages of policy options towards Venezuelan migrants, and which dimensions they are more willing to support in the aggregate. Conjoint designs, as they have been generally used across political and economic settings, involve asking respondents to evaluate a profile or object that contains randomly varied features (Bansak et al. 2021). In our experiment, we presented respondents with sets of hypothetical policy packages (the ‘treatments’) comprising six dimensions available to policymakers and which are relevant to public debate in both Colombia and in many host countries globally: (1) labour market access, (2) location restrictions, (3) public service access, (4) family reunification, (5) numerical limits, and (6) length of residency. This design reflects how people may encounter policy proposals as several options bundled together by legislators or parties, while also reduces social desirability bias by allowing respondents to justify their choices to themselves along any of the given dimensions. It is also resource efficient because it enables researchers to simultaneously test many more dimensions than would typically be possible using conventional designs that only vary a few elements in each treatment without resorting to unfeasibly large sample sizes. Next, we look for heterogeneous effects involving three factors: the role of economic and humanitarian priorities, perceptions on the Venezuelan crisis, and personal contact with Venezuelans. The political economy literature on preference formation indicates that economic priorities more strongly relate to policies that are more likely to be perceived as protecting the labour market from foreign worker competition. Meanwhile, humanitarian priorities relate to policies that are more likely to be perceived as protecting the wellbeing of migrants (Bansak et al. 2016; Giray-Aksoy and Ginn 2021; Hager and Veit 2019). In forced migration situations, the humanitarian context and past conflict experiences may foster higher levels of empathy and altruism (Adida et al. 2018; Hartman and Morse 2017). Finally, we explore whether heterogeneity also exists with respect to social contact. Increased meaningful contact with migrants can foster more positive attitudes and increase understanding of the migrant community and their challenges and therefore further support for policies seen as humanitarian (Betts et al. 2021; Ghosn et al. 2019; McLaren 2015). By 3 contrast, indirect contact or simple exposure, typically measured by the proportion of migrants in a local area, can increase hostility towards the migrant group and thus support for more restrictive and exclusionary policies (Enos 2014; Hangartner et al. 2019). Therefore, in this study, we emphasize and test for the role of contact using a battery of questions focusing on meaningful forms of contact. Our findings indicate that there is more support for conditional access to the labour market (i.e., only in certain occupations) or full free access compared to the alternative of no access. There is support for unrestricted location choices and access to public health services. Respondents also prefer giving conditional rights to family reunification (i.e., only if able to support dependants), applying numerical limits on inflows, and limiting the length of residency permits. We show that those who put more emphasis on economic priorities relative to humanitarian ones and those who see the situation in Venezuela mainly as an economic problem, tend to support policies that are more restrictive. Finally, respondents reporting higher levels of meaningful social contact with Venezuelans show less restrictive policy preferences, particularly in terms of labour market access and family reunification, compared to those reporting lower levels of such contact. This study makes three novel contributions to scholarly understanding of the politics and economics of migration. First, it moves beyond immigration attitudes to consider which policies host communities are willing to support in response to large-scale migration. Much prior conjoint experimental work has focused on identifying patterns in attitudes towards immigrants along different dimensions such as nationality, reason for entering, occupational skill, and gender (Bansak et al. 2016; Hainmueller and Hopkins 2015; Ward 2019). While attitudes are important antecedents for policy preferences, we present a more direct test of specific policy options bundled together as voters are more likely to encounter them. Here, the conjoint experimental design allows us to measure public support for a variety of policy options as they appear alongside each other, in a manner that has been shown to closely match actual preferences in naturalistic settings (e.g., Hainmueller et al. 2015). Second, it provides evidence of public preferences in a region that, in terms of gross numbers, is far more consequential for the host country. Most existing work on the topic has focused on European or North American contexts, particularly around Europe’s “migration crisis” of 2015 (Bansak et al. 2016; Jeannet et al. 2021).2 By contrast, current forced 2 One exception is the study by Alrababa’h et al. (2020). They consider attitudes and policy preferences towards Syrian refugees in Jordan. However, the emphasis of their experimental study is on attitudes and not on policy preferences. 4 migration patterns demonstrate how neighbouring countries in the Global South actually host the majority of forced migrants worldwide (UNHCR 2020). The relatively few studies involving host communities in the Global South observe how, despite obvious differences in the scale of forced migratory flows, most published work focuses on the experiences of higher-income destination countries (Alrababa’h et al. 2020). Third, we extend and improve upon prior work that considers key moderating factors for migration policy preferences. Specifically, we consider the role of humanitarian and economic values in impacting preferences for specific policies. This is relevant in a context in which the saliency of the economic and humanitarian crisis is particularly strong in explaining the driving forces of the migration inflows. We also revisit the role of social contact with migrants by moving beyond approaches used in existing observational studies (e.g., measures of distance to the border or migrant density as an approximation of exposure), instead using a detailed battery of questions regarding self-reported encounters with Venezuelans. The paper is organised as follows: section 2 of the paper presents the conceptualisation of the different dimensions of policy; section 3 reviews the context of Venezuelan migration into Colombia; section 4 discusses the research design for the conjoint survey and the analysis and estimation strategies; section 5 presents the results; section 6 presents a series of robustness tests; and section 7 contains the conclusion and policy implications. 2. Policy Preferences and Dimensions of Policy Responses 2.1 Dimensions of Policy Responses to Large Migration Inflows Policymakers in host countries must decide among several options when setting migration policies. In this section, we describe the six dimensions that we included in our hypothetical policy packages: (1) labour market access, (2) location restrictions, (3) public service access, (4) family reunification, (5) numerical limits, and (6) length of residency. Then, we show how they theoretically relate to economic and humanitarian priorities. Labour market access Restricting labour market access is a popular policy in the forced migration context. This policy intends to reduce concerns about labour market competition between hosts and the 5 newcomers (e.g., labour market replacement of the local population, reduction in wages, worsening of employment conditions). However, labour market restrictions can have negative consequences for migrant well-being. These policies can push forced migrants into informal employment where there is greater vulnerability and higher likelihood of exploitation. Moreover, labour market restrictions can have other effects such as psychological discouragement and the potential for mental health problems (Hainmueller et al. 2016; Hvidtfeldt et al. 2019; Laban et al. 2004). The global evidence on labour market access suggests that concerns regarding job competition dominate other concerns about the well-being of migrants, particularly in economies with weak labour markets (Zetter and Ruadel 2016). Location restrictions Restricting the location of incoming migrants is also common around the world and it is particularly popular in the forced migration context. This includes restricting mobility to a “camp” location or temporary restrictions on moving across administrative divisions (e.g., municipalities, provinces). This policy aims to create a balance regarding the “burden” of immigration on regional labour markets and to protect domestic workers in certain regions from foreign competition (Auer 2018; Bratsberg et al. 2020; Ruiz and Vargas-Silva 2016). Public service access Restricting access to public services such as health access is a common policy in the case of migrants without formal legal status (Juanmarti Mestres et al. 2021). In the cases of other migrants, including those with refugee status, there is typically some level of access to public services, which varies across countries. In general, these restrictions aim to reduce concerns about a decrease in the quality of services received by the host population, for example, waiting times for health services (Giuntella et al. 2018). However, the evidence suggests that public services restrictions have substantial negative long-term effects for migrant well-being (Bozorgmehr and Razum 2015) and there is strong public support for some types of access (e.g., maternity services). Family reunification Restrictions on family reunification are also common around the world. These restrictions are typically imposed to reduce the possible burden on public finances of having additional dependants entering the country (Cholewinski 2004; Jeannet et al. 2021; Sumption and 6 Vargas-Silva 2019). Across countries, there is substantial evidence that family migrants have low employment rates compared to other types of migrants (Gillespie et al. 2020; Ruiz and Vargas-Silva 2018). However, there is evidence that policies restricting family reunification have major negative consequences for the well-being of migrants (Bragg and Wong 2016). This separation can increase financial vulnerability for some groups and create substantial mental distress. Numerical limits Using numerical restrictions to limit immigration is a common practice around the world, taking form in policies such as limiting the number of visas or caps on work permits. These restrictions can be seen as a way to protect the economic situation of workers that compete with migrants in the labour market (Borjas 2014; Boubtane et al. 2016). Restricting numbers is a more controversial practice in the context of forced migration as there is an internationally recognised right to claim asylum. However, in many cases of large-scale forced displacement, most of those who are in need of protection are not recognised as refugees, hence different restrictions on numbers could be imposed. Yet from a humanitarian perspective, limiting the number of individuals who are allowed to enter the country can be seen as providing fewer numbers of people with essential protection. Length of residency Finally, countries often only offer protection for the period of the crisis or conflict and require forced migrants to return home after this period (Black and Koser 1999; Ruiz and Vargas- Silva 2021). Requiring the return of migrants is a way of avoiding the abuse of humanitarian protection as way to gain long-term access to the labour market. However, there is a strong humanitarian concern that temporariness related to the status of forced migrants affects their well-being and leads to mental distress (Tize 2020). 2.2 Treatment Design and Hypotheses In our experimental design, we chose the possible levels within each dimension to correspond with the reality of current Colombian policy towards Venezuelans, as well as the range of possibilities available to policymakers. These options are expressed in Table 1. 7 Table 1. Policy dimensions and levels Policy Levels Venezuelans can work in Colombia without restrictions. Access to Venezuelans can work in Colombia only in in selected occupations. labour markets Venezuelans cannot work in Colombia. Venezuelans are allowed to locate in their city of preference in Geographic Colombia. Location Venezuelans are allowed to locate in certain designated cities in Colombia. Venezuelans can access the subsidised public healthcare system on Access to an equal basis to Colombians. health Venezuelans cannot access the subsidised public health care system in Colombia. Venezuelans are allowed to bring their spouse and children. Family Venezuelans are allowed to bring their spouse and children if they reunification can pay for their cost of living. Venezuelans are not allowed to bring their spouse and children. Venezuelans are allowed into Colombia without numerical limits. Numerical limit Venezuelans are allowed into Colombia until an annual limit is reached. Venezuelans are allowed into Colombia for an indefinite period. Protection Venezuelans are allowed into Colombia for a period of ten years, period which can be renewed. Venezuelans are allowed into Colombia for a period of ten years, which cannot be renewed. Based on the above discussion, we set out four pre-registered hypotheses about aggregate Colombian policy preferences, as well as the extent to which contact and either 8 economic or humanitarian priorities are driving those preferences (Table A1 presents a summary of the hypotheses and expected results). First, we expect that economic concerns will cause respondents to express more restrictive preferences with respect to Venezuelans’ access to the labour market and the extent to which they are free to settle across the country. Meanwhile, we expect that humanitarian concerns will cause respondents to express less restrictive preferences with respect to Venezuelans’ abilities to access public health services and be reunited with family members. H1 (economic concerns). On average, respondents will prefer more restrictive policies related to labour market access and location restrictions. H2 (humanitarian concerns). On average, respondents will prefer less restrictive policies related to access to health services and allowing for family reunification. We also consider possible variation in preferences between respondents who hold either economic or humanitarian priorities. In the methods section, we explain how we measured these priorities. H3 (economic versus humanitarian concerns). Respondents who prioritize economic values or see the Venezuelan crisis as a primarily economic problem will prefer more restrictive policies related to numerical limits, location, labour market access and length of residence, compared to those who prioritize humanitarian values or see the Venezuelan crisis as a primarily humanitarian problem. Finally, we consider the role of personal contact with Venezuelan migrants. Existing work on contact suggests that there is theoretical scope for contact to foster more favourable attitudes towards outgroups, particularly when the type of contact in question is more personal (Ghosn et al. 2019; McLaren 2003). Additionally, recent experimental work from Europe suggests that contact specifically matters by shaping how receiving communities differentiate between Western and non-Western migrants (Clayton et al. 2021). Given the relative lack of research specifically examining the moderating effects of contact in the Global South—and on outcomes beyond attitudes—we test whether contact produces differences in policy preferences. We move away from the conventional idea of contact as mere exposure, instead focusing on meaningful contact as reflected in longer standing or more substantial forms of interaction. 9 H4 (social contact). Respondents who report having higher levels of contact with Venezuelans will prefer less restrictive policies across all dimensions. 3. Venezuelan Migration to Colombia and Institutional Context The Venezuelan humanitarian, economic and social crisis over the last decade, has resulted in resulted in the outflow of over 5.91 million Venezuelan citizens (around 16% of the country’s population) from the country. Over 4.5 million have stayed in the Latin American region, with Colombian being the main host. Colombia shares a large border with Venezuela spanning 1,378 miles (see Figure A1). In addition, Venezuela and Colombia were part of the same country in the early 19th century (Gran Colombia). As a result, both countries have a shared history and share important cultural similarities. It is therefore not surprising that Colombia has been a major recipient of Venezuelans with an estimated 1.75 million now living in the country.3 The official crossing points at the Colombia-Venezuela border have remained open for much of the Venezuelan crisis. The border crossing points were temporarily closed in 2015 but re-opened in 2016, after which a large number of migrants crossed (see Figure 1). It has always been relatively easy for Venezuelans to cross the border at unofficial border-crossing points. Figure 1 - Number of Venezuelan migrants in Colombia 2000000 1500000 Number of Venezuelan migrants 1000000 500000 0 2014 2016 2018 2020 Year Source : Authors' calculations using data published by the Ministry of External Relations 2020. The vertical solid line indicates the border closure in 2015 and the dashed line the re-opening of the border in 2016. 3 This represents about 3.6% of Colombia’s population and 7.2% of the labour force (Santamaria 2020). 10 The situation of Venezuelan migrants displays many similarities to the situation of refugees (i.e., those deserving the legal status). Mindful of Venezuelans’ conditions of vulnerability, in 2017 the Colombian government created a new migratory status called the ‘Permiso Especial de Permanencia’ (Special Permanency Permit), known as PEP. This permit has had different versions: the 2018 iteration was initially valid for up to two years and renewable every 90 days. Moreover, Venezuelans who held that version of the PEP were able to join the labour market and access public services.4 While the permit regularised a significant proportion of Venezuelans, the number of unauthorised Venezuelans in the country remained high. In fact, until late 2020, nearly 50% of Venezuelans did not have regular status as shown in Figure 2 (Migración Colombia 2020). However, given the large level of informality in the Colombian labour market—estimated at over a third of all employment in Colombia (Departamento Nacional de Estadística 2021)—the lack of regularisation and formal work authorisation might not have represented a major impediment for Venezuelans to find employment.5 Figure 2 - Number of Regular and Irregular Venezuelan migrants since 2018 1000000 Number of Venezuelan migrants 800000 600000 Regular Irregular 400000 2018m7 2019m1 2019m7 2020m1 2020m7 2021m1 Year Source : Authors' calculations using data published by the Ministry of External Relations 2020. 4 The PEP was initially rolled out in 2017 and came to be known as the PEP-I. Three resolutions were enacted in 2018 with three different forms of the PEP. Current evidence shows negligible effects of these amnesties on the Colombian labour markets (Bahar et al. 2021). 5 There is evidence that, even after being eligible for regularisation, some migrants chose to keep working in the informal sector (Bahar et al. 2021). 11 Comparing demographic and labour market characteristics of Colombians and Venezuelans during the first quarter of 2021 (when our survey fieldwork occurred) reveals patterns that largely accord with prior studies. As shown in Table 2, while both labour market populations have similar sex ratios, Venezuelans are younger and have slightly higher levels of education (Bahar et al. 2021; Santamaria 2020) compared to Colombians. Venezuelan migrants also work longer weekly hours and report lower wages. They also present a much higher level of informality. Table 2: Demographic and labour market characteristics of Colombian and Venezuelans – 1st quarter of 2021. Colombians Venezuelans Average Std. Dev Average Std. Dev Gender [1 = Male] 0.48 0.49 0.48 0.49 Age 35 22.0 25.7 16.5 Married 0.15 0.36 0.08 0.26 Years of education 5.6 3.9 6.1 4.4 Education None 0.05 0.22 0.06 0.24 Primary 0.25 0.43 0.24 0.43 Secondary 0.40 0.48 0.52 0.50 Tertiary 0.23 0.42 0.14 0.35 Labour market Labour force 0.84 0.36 0.76 0.42 Employment 0.46 0.49 0.56 0.49 Informal 0.34 0.47 0.76 0.42 Weekly hours worked 44.4 14.3 47.8 15.3 Monthly wages 1,392.3 1,522.8 814.3 571.7 Note: Data comes from the Colombian Labour Market Survey (GEIH), Q1 2021. Monthly wages expressed in thousands of Colombian pesos. In February 2021, immediately prior to our survey fieldwork, the Colombian government announced a mass regularisation for Venezuelan migrants currently living in the country, called Estatuto Temporal de Protección para Migrantes Venezolanos (ETPV). This granted Venezuelans a temporary regular migratory status for a non-renewable 10-year period.6 It also granted permit holders access to formal labour markets and public services 6 https://www.unhcr.org/news/press/2021/2/60214cf74/unhcr-iom-welcome-colombias-decision-regularize- venezuelan-refugees-migrants.html 12 such as health and education.7 This scheme, which requires Venezuelans to apply for it, covers migrants already living in Colombia with either regular or irregular status. However, we know very little about the public’s preferences for specific policies towards Venezuelan migrants. To be sure, Proyecto Migración Venezuela has collected data on public opinion towards Venezuelans in Colombia. Using data collected in September 2020, Figure 3 shows the percentage of individuals agreeing to different statements. At the time of that survey, a large percentage of Colombians perceived Venezuelans as a threat to employment (i.e., taking jobs away) (57%) while the majority also perceived them as a fiscal burden (58%). Yet a majority of respondents also agreed that Venezuelan migrants should be given access to healthcare (72%) and employment (54%), while a strong plurality thought they should be granted legal documents to access health and labour markers under the same conditions as Colombians (48%).8 However, these observational results have several potential limitations, including difficulty in interpreting whether Colombians prioritise some policy aspects over others. A survey experiment is therefore well-suited to elicit Colombians’ views towards realistic policy options and understand their likely drivers. Figure 3 - Attitudes towards Venezuelan migrants, September 2020 Threat to employment Fiscal Burden Should be given access to employment Should be granted access to health Should be given legal documents 0 20 40 60 80 Percent Note: Authors's calculation using data from Proyecto Migracion Venezuela, Revista Semana 2020. Percentage of respondents agreeing with the statements.N(2020-9)= 2,289 7 https://www.cancilleria.gov.co/newsroom/news/presidente-duque-anuncia-decision-historica-crear-estatuto- proteccion-temporal 8 Proyecto Migración also collected data in April 2019 and March 2020. The proportions of respondents agreeing to these statements have remained stable between these waves. We thank Adriana Sabogal and Proyecto Migracíon Venezuela for granting us access to the data. https://migravenezuela.com/ 13 4. Data and Research Design 4.1 Conjoint Experimental Survey In order to measure Colombian policy preferences, we conducted a fully randomized choice- based conjoint experiment. Specifically, the conjoint design allows estimation of causal effects of different treatment components simultaneously in a multidimensional manner (Bansak et al. 2021; Hainmueller et al. 2014; Rodon and Sanjaume-Calvet 2020). As noted in Bansak et al. (2021) and Horiuchi, Markovich, and Yamamoto (2021), evaluating several attributes simultaneously helps to obscure the intent of researchers as well as allows respondents to justify their choices along any number of the dimensions. Indeed, evidence shows that fully randomised conjoint designs do mitigate social desirability bias (Dahl 2018; Horiuchi et al. 2021). We presented respondents with hypothetical pairs of packages comprising policies towards Venezuelans that reflect current Colombian policy as well as options that policymakers in the country could plausibly implement. Each package contained randomly selected levels from within each of the six policy dimensions shown in Table 1. Then, each respondent was asked to choose which of the two policy packages they preferred more, as well as to rate each package on a 1-7 scale, where 1 indicated “absolutely dislike” and 7 indicated “absolutely like.”9 Each respondent completed five trials (Figure A2 provides an example of the task within the survey interface). In line with good practice, we randomised the order of the dimensions between respondents to reduce ordering effects (where respondents disproportionately focus on items appearing near the beginning of a list), but kept it constant across the five trials to minimize survey fatigue. We also ask a series of questions that allow us to explore the level of concern regarding economic vs humanitarian aspects, as well as respondents’ levels of social contact. Table A2 and Table A3 in the Appendix explain each of these variables in detail. 4.2 Survey Sample 9 The introductory paragraph and exact wording of the questions were: “Imagine that the Colombian government is considering various packages of measures to address migration from Venezuela. Next we will show you two possible options of packages, A and B. Read their descriptions carefully.” Then, in each conjoint task they are asked: (1) Which package is closest to what you would prefer?, and (2) On a scale from 1 to 7, where 1 means “you absolutely dislike” and 7 means “you absolutely like”, how would you rate each of these packages? 14 We fielded the survey experiment between March 1-13, 2021 to a sample of 2,508 Colombian adults who were at least 18 years old. 10 These respondents came from a larger online panel of approximately 140,500 respondents.11 Our sampling aimed to be representative across different dimensions of the country’s population based on the 2018 Colombian census, including key demographic variables including age, gender, region, and socioeconomic strata (a measure of income used in Colombia). Table A4, which displays the distributions of these variables in our sample against those from the 2018 census, indicates how our sample shows close similarity with the census estimates.12 We also recorded additional socio-demographic and economic variables for the respondents. Description and descriptive of other additional features of our sample appear in the Appendix (Table A5). Since each respondent saw five pairs of policy packages, our analysis comprises 25,080 observations. 4.3 Estimation Strategy We analyse the data in three steps. First, we estimate both the average marginal component effects (AMCEs) and marginal means (MMs) to measure public preferences towards each policy option within each dimension. AMCEs report a series of linear probability estimations where the dependent variable is a binary (forced choice) variable indicating the preferred policy by the respondent. The dependent variable is regressed on the levels of each policy dimension (with a researcher-chosen reference category as the baseline level) to estimate the probability of respondents choosing that policy option in relation to the reference category. To test for the consistency of the results, we also use a scaled rating version as a dependent variable where the policy ratings are re-scaled to vary between 0-1 (see Table A3 for a description). In all estimations, we report robust standard errors clustered at the respondent level since each respondent saw multiple treatments (Hainmueller et al. 2014). 10 The survey experiment was reviewed and approved by the Central University Research Ethics Committee (CUREC) of the University of Oxford (C1A-20-2) and the pre-analysis plan was registered with As-Predicted (#60418). 11 We worked with Invamer (https://www.invamer.com.co/), a reputable nationally recognised survey firm with one of the largest online panels in the country. They operated through the Netquest platform, and their recruitment strategy is based on incentives that are granted to the respondents at the end of the survey. 12 Nevertheless, we acknowledge that our use of online sampling might have over-represented individuals with online access and sufficient digital literacy to access an online survey—who in turn may have been more educated and urban-dwelling within broad regions. For context, in terms of internet access, Colombia ranks in the middle of other Latin American countries of similar income levels: according to World Bank estimates, 65% of the population in the country has access to use the internet. This is higher than Ecuador (54%), Peru (60%), and Bolivia (44.2%), but lower than Mexico (70%), Argentina (74%), and Brazil (70%) [https://data.worldbank.org/indicator/IT.NET.USER.ZS]. 15 Meanwhile, MMs represent the mean outcome across all appearances of a particular feature, averaging across all other features. They are the differences in the outcome of interest caused by the presence of a specific attribute, all other attributes being equal. We report MMs to reduce concerns about how researchers’ choices of reference categories might impact the substantive interpretation of the results—particularly among subgroups (Leeper et al. 2020; Ratkovic 2021). Second, we estimate conditional marginal means (CMMs) to measure subgroup preferences for the policy options. We assessed the extent to which respondents hold either economic or humanitarian priorities in two ways. On the one hand, prior to the experimental treatment we measured respondents’ levels of agreement with five statements that covered a range of economic and humanitarian stances: for example, “protecting the most vulnerable should be given priority even if that reduces economic growth.” This was measured on a scale from 1 indicating “strongly disagree” to 5 indicating “strongly agree.” The full battery of statements appears in Table A2 and the corresponding summary statistics in Figures A3 and A4. We randomised the order of these statements by respondent. Then, after recoding responses so that higher values indicated humanitarian priorities, we assigned respondents whose mean response across the five statements was greater than 3 as holding more humanitarian priorities. By contrast, we assigned respondents whose mean response were below 3 as holding more economic priorities. On the other hand, we measured respondents’ views as to whether the situation in Venezuela was primarily an economic or humanitarian problem using a single question (see Table A2 Appendix). Third, we also consider the extent to which social contact with Venezuelans matters for Colombians’ policy preferences. Using a series of three questions measuring self-reported meaningful social contact—rather than chance or superficial exposure to a large group—we created an index that allowed us to code respondents as having either “high” or “low” levels contact (Clayton et al. 2021).13 The questions asked about daily interactions, trust in individuals, and social interactions (i.e., gathering for dinner). The precise questions and construction of the index are reported in the Appendix in Table A2 and the summary statistics in Figure A5. The order of these questions was also randomised across respondents. 13 To test whether our results with respect to contact are sensitive to the type of contact in question, we repeated the analysis using a more conventional measure of migrant density. This measure involved calculating the proportion of migrants as a share of the total population of each state (department). Then, using respondents’ reported location at the state-level, we split the sample into two groups: those living in states whose proportion of migrants was below the median of all states (coded as 0), and those living in states with migrant proportions above the median (coded as 1). 16 5. Results In this section, we report the main results of the survey experiment. We focus on the forced choice outcome, but also report the results using the rating outcome in the Appendix. The substantive findings do not change whether looking at the forced choice or rating outcome. 5.1 Aggregate-Level Policy Preferences Figure 4 shows the AMCEs and MMs for each level across the six policy dimensions. Among the AMCEs, the reference category is the least restrictive option within each dimension (except for length of residency). The results provide mixed support for our aggregate-level hypotheses. H1 stated that respondents will favour more restrictive policies with respect to labour market access and geographic location. On the one hand, Colombians prefer restricting Venezuelans’ access to the labour market over no restrictions by 2 percentage points. On the other hand, they clearly do not prefer completely preventing Venezuelans from employment: this option was 12 percentage points less favoured compared to free access. There is also opposition (by 3 percentage points) to location restrictions compared to the option of no restrictions. 17 Figure 4 - Colombian policy preferences towards Venezuelan migration Average Marginal Component and Marginal Means AMCE Marginal Mean Labour Market Access 0.00 0.53 Allowed to work 0.02 0.55 Restricted work -0.12 0.42 Not allowed to work Location Restrictions 0.00 0.52 No restriction on location -0.03 0.48 Restricted Location Health Access 0.00 0.52 Access to subsidised health -0.04 0.48 No access subsidised to health Family Reunification 0.00 0.47 Allowed to bring family 0.11 0.57 Only bring family if can support it -0.01 0.46 Not allowed to bring family Numerical Limits 0.00 0.47 No numerical limits 0.06 0.53 Annual Limit Length of Residency 0.00 0.47 Indefinite period 0.05 0.52 Ten years that can be renewed 0.05 0.52 Ten years that cannot be renewed Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. In the AMCE column, single points without horizontal bars denote the reference category used for that dimension. Meanwhile, H2 stated that respondents should prefer less restrictive policy options with respect to Venezuelans’ access to public health services and ability to reunify with family members. Respondents do indeed prefer allowing Venezuelans to access health services over not allowing access (by 4 percentage points). However, respondents preferred the intermediate option for family reunification that introduced economic criteria over the unrestricted option (by 11 percentage points, respectively). Although we did not pre-register aggregate-level hypotheses with respect to numerical limits and length of residency, our results suggest that respondents prefer some kind of annual limit (by 6 percentage points, compared to no annual limit) and having a time-limit on residency—whether or not it can be renewed (by 5 percentage points over having an 18 unlimited time period). The results using the rating outcome are included in Figures A6 and A7 in the Appendix. As mentioned above, the substantive findings remain unchanged. 5.2 Role of Humanitarian versus Economic Priorities In a second step, we move beyond overall policy preferences to consider the role of humanitarian vs economic priorities. Our pre-registered hypothesis stated that respondents who held economic priorities would prefer more restrictive policy options with respect to numerical limits, location, labour market access and length of residence, compared to respondents who held more humanitarian priorities. Figure 5 displays the conditional marginal means for each subgroup, using a measure of agreement with our battery of questions to determine respondents’ priorities (marginal means and differences in sub-group means are presented in Table A6 and A7 in the Appendix). The results broadly support this hypothesis, although some nuance is needed with respect to intermediate options. For example, respondents holding more humanitarian priorities are less likely to prefer completely preventing Venezuelans from accessing either the labour market or public health services (when looking at the AMCE’s, this is 14 and 7 percentage points less likely compared to the base categories), or bringing family members, compared to those holding more economic priorities. Yet in absolute terms, both groups prefer some degree of restriction when it comes to labour market access, family reunification, and allowed length of stay. We repeat this analysis using a question asking whether respondents viewed the Venezuelan crisis as a primarily economic or humanitarian crisis: the substantive results, reported in Figure 6, are broadly the same. 19 Figure 5 - How economic and humanitarian priorities matter for Colombian policy preferences towards Venezuelans Humanitarian Index Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Humanitarian Economic Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. 20 Figure 6 - How economic and humanitarian priorities matter for Colombian policy preferences towards Venezuelans Humanitarian Question Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Humanitarian Economic Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. 5.3 Social Contact We now turn to social contact. The results, presented in Figure 7, partially support our hypothesis that greater meaningful social contact leads to preferences for less restrictive policies.14 On the one hand, this is clearly the case when it comes to labour market access: respondents reporting higher levels of contact are more likely to prefer policies that allow Venezuelans to have unrestricted access to employment, and less likely to prefer complete restriction, compared to those with lower levels of contact. This is also the case for family reunification, and—to lesser degrees of significance—preferences for numerical limits and 14 The marginal means and test for statistical difference in sub-group means are presented in Table A8 in the Appendix. 21 access to healthcare. On the other hand, both groups still outrightly prefer intermediate policies on family reunification and labour market access. By contrast, contact does not seem to matter for policy preferences with respect to length of residency: both groups prefer time- limited options, with the prospects of renewal not making a detectable difference. We further conduct the analysis with an alternative version of the social contact index that was less restrictive (segmenting by those who had any index value greater than zero) and with the question of whether the respondent has shared a meal with a Venezuelan and the results remain qualitatively the same (Figure A8). Here, it is worth comparing the results from the self-reported social contact index and measures that rely on migrant density calculations as proxies for contact (included in Figure A9 in the Appendix). As described earlier, we hypothesised that more meaningful forms of contact involving personal interaction would have greater impacts on policy preferences. While the likelihood of contact has often been measured by the proportion of migrants in a local area, it is possible that experiencing groups of migrants at ‘arms-length’ in this way might actually foster greater feelings of threat—although empirically measuring this is beyond the scope of our study. Inspecting the CMMs corresponding to each subgroup labelled as having ‘high’ or ‘low’ contact on either measure reveals that, while the direction of preferences among policy options does not substantively vary, the differences between subgroups do change in magnitude for some dimensions. For example, meaningful contact (compared to mere migrant density) appears to matter more for respondents’ preferences on family reunification, health access, and location restrictions. While more work needs to be done to empirically document the mechanisms at play, our results suggest that the quality of contact likely matters for migration policy preferences. 22 Figure 7 - How social contact with Venezuelans matters for Colombian policy preferences Social Contact Index Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy High Contact Low Contact Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. 5.4 Heterogeneities: Conjoint results by other respondents’ characteristics We also considered other demographic features as potential moderators, which was done in an exploratory manner separate from the pre-registered analyses. These included age, gender, socio-economic status, education, and employment status. The description of all these variables are included in Table A2 in the Appendix. The results are presented in Figure A10. We observe some notable differences among subgroups in some dimensions that correspond with prior work on the drivers of immigration attitudes in the Global North. For example, younger respondents (those aged below 40) are less likely to prefer restricting work 23 compared to those that are older, although the two groups express largely similar responses with respect to the other policy dimensions. 6. Robustness Checks We conduct a series of robustness checks for the validity of our results. First, we report the results of Figure 5 when we include additional control variables (age and education level, political orientation, attitudes towards migration). As shown in Table A6, the ACMEs remain unaffected by additional control variables in terms of magnitude and statistical significance. This in turn shows that the experimental estimates are robust across policy dimensions.15 We also consider the potential for survey fatigue since conjoint experiments can be cognitively demanding. In our design, respondents completed five tasks which is well within the limits of what respondent can typically complete without significant losses in quality (Bansak et al. 2019; Hainmueller et al. 2014). Figure A11 in the Appendix displays all estimates by task number: the results do not significantly differ among tasks. 7. Conclusion As countries around the world continue to experience large-scale migration inflows, governments face choices of which policies to implement in response. To a certain degree, these choices will be informed by public demands and consent for different kinds of interventions. While existing research has tried to ascertain host communities’ preferences for immigration policies, this work often displays shortcomings that limit both theoretical and empirical development—not least of all a lack of focus on cases in the Global South which, by sheer numbers, experience most of the world’s forced migration flows. By harnessing the power of a conjoint experiment fielded in Colombia, we explored the extent to which host country residents prefer a range of policies realistically available to policymakers as they respond to Venezuelan immigration, which is one of the largest migratory flows globally. We have also addressed whether these preferences are driven by different priorities (e.g., economic or humanitarian priorities) and meaningful social contact. Our results reveal support for the options of conditional access to the labour market (i.e., only in certain occupations) or full free access to the alternative of no access, for unrestricted location choices and access to public services, as well as conditional rights to family 15 Acknowledging debates about when adding control variables is needed in experimental settings where researchers control randomisation (Mutz 2011) —and particularly when looking for heterogeneous effects among subgroups using covariates which are not randomly assigned (Kam and Trussler 2017) —we lean more heavily on the treatment effects reported in the main text. 24 reunification (i.e., only if immigrants are able to support their dependants). Respondents also support the use of numerical limits and limiting the length of residency permits. The results also show that those who put more emphasis on economic priorities relative to humanitarian ones, and those who see the situation in Colombia as mainly an economic problem, tend to support policies that are more restrictive. Meanwhile, people who report having higher levels of social contact with Venezuelans also hold preferences for less restrictive policies when it comes to labour market access and family reunification. This is particularly important given our focus on meaningful contact rather than incidental exposure. Our results present several implications for policymaking. First, aggregate public preferences vary among migration policy components, with more support found for those aspects having more direct consequences for migrants’ welfare—including access to work, healthcare, and family (Figure 5). This suggests that, at least in the Colombian context, policymakers are likely already working with broadly positive policy preferences in these domains. What is more, as indicated by the lack of polarization on migration policy in terms of sociodemographic characteristics, this appears to cut across major divisions that prior research tends to show as mattering for immigration attitude formation globally (Dinesen and Hjorth 2020; International Organization for Migration 2015; Rodrigues-Chatruc and Rozo 2021). As a result, there may be more political room for manoeuvre whereby legislators may be able to build public consent for a range of practically consequential policies towards Venezuelans. Second, public messaging that invokes humanitarian messaging and values may be a potentially effective lever for building that consent among citizens, particularly in those migration policy domains that have more directly-felt impacts on migrants’ daily lives. People holding stronger humanitarian values or viewing the Venezuelan crisis as a humanitarian issue expressed more favourable preferences in the domains of family reunification and healthcare access (Figure 6 and Figure 7). While testing the durability and longevity of this effect is beyond the scope of this single experiment, future work should investigate the extent to which migration policy preferences are sensitive to different framings that highlight competing values—and for whom these framings are more or less effective. For example, it could be possible that those most opposed to immigration could have a more favourable reaction to messaging indicating that Venezuelans can boost economic growth in Colombia. Moreover, emphasising values and objectives that are shared across national groups may also impact the extent to which citizens perceive migration as having primarily humanitarian features. Besides values, messaging also involves 25 communicating about the dynamics, drivers, and impacts of migration: we suggest greater attention to the extent to which policy preferences are sensitive to factual information involving migrants and migration—much of which is often filtered through mass media (see e.g., Allen and Blinder 2018; Grigorieff et al. 2020). Third, meaningful contact may be a mechanism for fostering social cohesion, particularly in settings where the sociocultural distance between host and migrant communities may be less. People reporting higher levels of meaningful contact were more likely to favour fewer location restrictions and granting healthcare access. What is more, such levels of contact are not rare in the Colombian context: about 75% of our sample reported having at least some form of meaningful contact with Venezuelans (Figure A5). Our results pave the way for testing interventions that generate substantive interactions among host and migrant communities, suggesting that these kinds of efforts may have consequential effects. 26 8. References Adida, Claire L., Adeline Lo, and Melina R. Platas (2018), “Perspective Taking Can Promote Short-Term Inclusionary Behavior toward Syrian Refugees.” Proceedings of the National Academy of Sciences of the United States of America 115(38): 9521–26. Allen, William, and Scott Blinder (2018), “Media Independence through Routine Press-State Relations: Immigration and Government Statistics in the British Press.” The International Journal of Press 23(2): 202–26. Alrababa’h, Ala’ et al. (2020), “Attitudes Toward Migrants in a Highly Impacted Economy: Evidence From the Syrian Refugee Crisis in Jordan.” Comparative Political Studies 54(1). Altındağ, Onur, and Neeraj Kaushal (2021), “Do Refugees Impact Voting Behavior in the Host Country? Evidence from Syrian Refugee Inflows to Turkey.” Public Choice 186(1–2): 149–78. Auer, Daniel (2018), “Language Roulette–the Effect of Random Placement on Refugees’ Labour Market Integration.” Journal of Ethnic and Migration Studies 44(3): 341–62. Bahar, Dany, Ana María Ibáñez, and Sandra V. Rozo (2021), “Give Me Your Tired and Your Poor: Impact of a Large-Scale Amnesty Program for Undocumented Refugees.” Journal of Development Economics 151: 102652. Bansak, Kirk, Jens Hainmueller, and Dominik Hangartner (2016), “How Economic, Humanitarian, and Religious Concerns Shape European Attitudes toward Asylum Seekers.” Science 354(6309): 217–22. Bansak, Kirk, Jens Hainmueller, Daniel J. Hopkins, and Teppei Yamamoto (2019), “Beyond the Breaking Point? Survey Satisficing in Conjoint Experiments.” Political Science Research and Methods: 1–19. ———. (2021), “Conjoint Survey Experiments.” In Advances in Experimental Political Science, Cambridge University Press, 19–41. Betts, Alexander, Maria Stierna Flinder, Naohiko Omata, and Olivier Sterck (2021), "Social Cohesion and Refugee-Host Interactions: Evidence from East Africa." Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Black, Richard, and Khalid Koser (1999), “The End of the Refugee Cycle?” In The End of the Refugee Cycle? Refugee Repatriation and Reconstruction, , 288. Borjas, George J. (2014), Immigration Economics Immigration Economics. https://www.hup.harvard.edu/catalog.php?isbn=9780674049772 (April 22, 2021). Boubtane, Ekrame, Jean Christophe Dumont, and Christophe Rault (2016), “Immigration and Economic Growth in the OECD Countries 1986-2006.” Oxford Economic Papers 68(2): 340–60. Bozorgmehr, Kayvan, and Oliver Razum (2015), “Effect of Restricting Access to Health Care on Health Expenditures among Asylum-Seekers and Refugees: A Quasi- Experimental Study in Germany, 1994–2013” ed. Joan A Caylà. PLOS ONE 10(7): e0131483. Bragg, Bronwyn, and Lloyd L. Wong (2016), “‘Cancelled Dreams’: Family Reunification 27 and Shifting Canadian Immigration Policy.” Journal of Immigrant & Refugee Studies 14(1): 46–65. Bratsberg, Bernt, Jeremy Ferwerda, Henning Finseraas, and Andreas Kotsadam (2020), “How Settlement Locations and Local Networks Influence Immigrant Political Integration.” American Journal of Political Science: ajps.12532. Campo, Francesco, Sara Giunti, and Mariapia Mendola (2021), The Refugee Crisis and Right-Wing Populism: Evidence from the Italian Dispersal Policy *. https://www.iza.org/publications/dp/14084/the-refugee-crisis-and-right-wing-populism- evidence-from-the-italian-dispersal-policy (May 26, 2021). Cholewinski, Ryszard, (2004), “Family Reunification and Conditions Placed on Family Members: Dismantling a Fundamental Human Right.” European Journal of Migration and Law 4(3): 271–90. Citrin, Jack, Donald P. Green, Christopher Muste, and Cara Wong (1997), “Public Opinion toward Immigration Reform: The Role of Economic Motivations.” Journal of Politics 59(3): 858–81. Clayton, Katherine, Jeremy Ferwerda, and Yusaku Horiuchi (2021), “Exposure to Immigration and Admission Preferences: Evidence from France.” Political Behavior 43(1): 175–200. Connor, Phillip (2018), “A Majority of Europeans Favor Taking in Refugees, but Most Disapprove of EU’s Handling of the Issue.” Pew Research Center. Dahl, Malte (2018), Social Desirability Bias in Candidate Conjoint Experiments: What Is the Optimal Design When Studying Sensitive Topics? Departamento Nacional de Estadística (DANE) (2021), https://www.dane.gov.co/files/investigaciones/boletines/ech/ech_informalidad/bol_ech_i nformalidad_jun20.pdf Boletín Técnico Gran Encuesta Integrada de Hogares (GEIH). Noviembre 2020 - Enero 2021. Medición de Empleo Informal y Seguridad Social. Dinas, Elias, Konstantinos Matakos, Dimitrios Xefteris, and Dominik Hangartner (2019), “Waking Up the Golden Dawn: Does Exposure to the Refugee Crisis Increase Support for Extreme-Right Parties?” Political Analysis 27(2): 244–54. Dinesen, Peter Thisted, and Frederik Hjorth (2020), “Attitudes toward Immigration: Theories, Settings, and Approaches.” In The Oxford Handbook of Behavioral Political Science, Oxford University Press. Dustmann, Christian, Kristine Vasiljeva, and Anna Piil Damm (2019), “Refugee Migration and Electoral Outcomes.” Review of Economic Studies 86(5): 2035–91. Enos, Ryan D. (2014), “Causal Effect of Intergroup Contact on Exclusionary Attitudes.” Proceedings of the National Academy of Sciences of the United States of America 111(10): 3699–3704. Facchini, Giovanni, and Anna Maria Mayda (2008), “From Individual Attitudes towards Migrants to Migration Policy Outcomes: Theory and Evidence.” Economic Policy 23(56): 651–713. Ghosn, Faten, Alex Braithwaite, and Tiffany S. Chu (2019), “Violence, Displacement, Contact, and Attitudes toward Hosting Refugees.” Journal of Peace Research 56(1): 118–33. Gillespie, Brian Joseph, Clara H. Mulder, and Michael J. Thomas (2020), “Migration for 28 Family and Labour Market Outcomes in Sweden.” Population Studies: 1–11. Giray Aksoy, Cevat, and Thomas Ginn (2021), "Host Country Policies and Attitudes toward Refugees: A Global Evaluation." Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Giuntella, Osea, Catia Nicodemo, and Carlos Vargas-Silva (2018), “The Effects of Immigration on NHS Waiting Times.” Journal of Health Economics 58: 123–43. Grigorieff, Alexis, Christopher Roth, and Diego Ubfal (2020), “Does Information Change Attitudes Toward Immigrants?” Demography 57: 1117–43. Hager, Anselm, and Susanne Veit (2019), "Attitudes Toward Asylum Seekers: Evidence from Germany." Public Opinion Quarterly Oxford University Press. Hainmueller, Jens, Dominik Hangartner, and Duncan Lawrence (2016), “When Lives Are Put on Hold: Lengthy Asylum Processes Decrease Employment among Refugees.” Science Advances 2(8): e1600432. Hainmueller, Jens, Dominik Hangartner, and Teppei Yamamoto (2015), “Validating Vignette and Conjoint Survey Experiments against Real-World Behavior.” Proceedings of the National Academy of Sciences 112(8): 2395. Hainmueller, Jens, and Daniel J. Hopkins (2015), “The Hidden American Immigration Consensus: A Conjoint Analysis of Attitudes toward Immigrants.” American Journal of Political Science 59(3): 529–48. Hainmueller, Jens, Daniel J. Hopkins, and Teppei Yamamoto (2014), “Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Experiments.” Political Analysis 22(1): 1–30. Hangartner, Dominik et al. (2019), “Does Exposure to the Refugee Crisis Make Natives More Hostile?” American Political Science Review 113(2): 442–55). Hartman, Alexandra, and Ben Morse (2017), “Violence, Empathy, and Altruism: Evidence from the Ivoiran Refugee Crisis in Liberia.” SSRN Electronic Journal. https://papers.ssrn.com/abstract=2769889 (May 10, 2021). Helbling, Marc, Liv Bjerre, Friederike Römer, and Malisa Zobel (2017), “Measuring Immigration Policies: The IMPIC Database.” European Political Science 16(1): 79–98. Helbling, Marc, and David Leblang (2019), “Controlling Immigration? How Regulations Affect Migration Flows.” European Journal of Political Research 58(1): 248–69. Horiuchi, Yusaku, Zachary D. Markovich, and Teppei Yamamoto (2021), “Can Conjoint Analysis Mitigate Social Desirability Bias?” Political Analysis, forthcoming. Hvidtfeldt, Camilla, Jørgen Holm Petersen, and Marie Norredam (2019), “Prolonged Periods of Waiting for an Asylum Decision and the Risk of Psychiatric Diagnoses: A 22-Year Longitudinal Cohort Study from Denmark.” International Journal of Epidemiology 49(2): 400–409. International Organization for Migration (2015), How the World Views Migration. Geneva. http://publications.iom.int/system/files/pdf/how_the_world_gallup.pdf. Jeannet, Anne-Marie, Tobias Heidland, and Martin Ruhs (2021), “What Asylum and Refugee Policies Do Europeans Want? Evidence from a Cross-National Conjoint Experiment.” European Union Politics. 29 Juanmarti Mestres, Arnau, Guillem López Casasnovas, and Judit Vall Castelló (2021), “The Deadly Effects of Losing Health Insurance.” European Economic Review 131. Kam, Cindy D., and Marc J. Trussler (2017), “At the Nexus of Observational and Experimental Research: Theory, Specification, and Analysis of Experiments with Heterogeneous Treatment Effects.” Political Behavior 39(4): 789–815. Laban, Cornelis J. et al. (2004), “Impact of a Long Asylum Procedure on the Prevalence of Psychiatric Disorders in Iraqi Asylum Seekers in The Netherlands.” Journal of Nervous and Mental Disease 192(12): 843–51. Leeper, Thomas J., Sara B. Hobolt, and James Tilley (2020), “Measuring Subgroup Preferences in Conjoint Experiments.” Political Analysis 28(2): 207–21. Marbach, Moritz, and Guido Ropers (2019), “Not in My Backyard: Do Increases in Immigration Cause Political Violence?” SSRN Electronic Journal. https://papers.ssrn.com/abstract=3310352 (October 20, 2021). McLaren, Lauren (2003), “Anti-Immigrant Prejudice in Europe: Contact, Threat Perception, and Preferences for the Exclusion of Migrants.” Social Forces 81(3): 909–36. ———. (2015), Immigration and Perceptions of National Political Systems in Europe Immigration and Perceptions of National Political Systems in Europe. Oxford University Press. Migración Colombia (2020), “Mas de 1 Millon de Venezolanos Estarian Radicados En Colombia - Migración Colombia.” Viernes, 03 Abril 2020. https://www.migracioncolombia.gov.co/noticias/mas-de-1-millon-825-mil-venezolanos- estarian-radicados-en-colombia (April 22, 2021). Mutz, Diana C (2011), Population-Based Survey Experiments Population-Based Survey Experiments. Princeton: Princeton University Press. Ratkovic, Marc (2021), “Subgroup Analysis: Pitfalls, Promise, and Honesty.” In Advances in Experimental Political Science, eds. James Druckman and Donald P Green. Cambridge University Press, 271–88. Rodon, Toni, and Marc Sanjaume-Calvet (2020), “How Fair Is It? An Experimental Study of Perceived Fairness of Distributive Policies.” Journal of Politics 82(1): 384–91. Rodrigues Chatruc, Marisol, and Sandra Rozo (2021), Attitudes Towards Migrants during Crisis Times. https://www.iza.org/publications/dp/14319/attitudes-towards-migrants- during-crisis-times. Rozo, Sandra V., and Juan F. Vargas (2021), “Brothers or Invaders? How Crisis-Driven Migrants Shape Voting Behavior.” Journal of Development Economics 150: 102636. Ruiz, Isabel, and Carlos Vargas-Silva (2016), “The Labour Market Consequences of Hosting Refugees.” Journal of Economic Geography 16(3): 667–94). ———. (2018), “Differences in Labour Market Outcomes between Natives, Refugees and Other Migrants in the UK.” Journal of Economic Geography 18(4): 855–85). ———. (2021), “The Impacts of Refugee Repatriation on Receiving Communities.” Journal of Economic Geography 21(2): 169–94. Santamaria, Julieth (2020), “When a Stranger Shall Sojourn with Thee”: The Impact of the Venezuelan Exodus on Colombian Labor Markets. Scheve, Kenneth F., and Matthew J. Slaughter (2001), “Labor Market Competition and 30 Individual Preferences over Immigration Policy.” Review of Economics and Statistics 83(1): 133–45. Schmid, Katharina, Ananthi Al Ramiah, and Miles Hewstone (2014), “Neighborhood Ethnic Diversity and Trust: The Role of Intergroup Contact and Perceived Threat.” Psychological Science 25(3): 665–74. Steinmayr, Andreas (2021), “Contact versus Exposure: Refugee Presence and Voting for the Far Right.” The Review of Economics and Statistics: 1–18. Sumption, Madeleine, and Carlos Vargas-Silva (2019), “Love Is Not All You Need: Income Requirement for Visa Sponsorship of Foreign Family Members.” Journal of Economics, Race, and Policy 2(1–2): 62–76. Tize, Carola (2020), “Living in Permanent Temporariness: The Multigenerational Ordeal of Living under Germany’s Toleration Status.” Journal of Refugee Studies. UNHCR (2020), “UNHCR - Figures at a Glance.” https://www.unhcr.org/en-us/figures-at-a- glance.html (April 22, 2021). Ward, Dalston (2019), “Public Attitudes toward Young Immigrant Men.” American Political Science Review 113(1): 264–69. Zetter, Roger, and Heloise Ruadel (2016), Knomad Working Paper Refugees’ Right to Work and Access to Labor Markets – An Assessment (Part 1) | KNOMAD. https://www.knomad.org/publication/refugees-right-work-and-access-labor-markets- assessment-part-1 (April 22, 2021). 31 Appendix Contents Table A1. Summary of pre-registered hypotheses .................................................................. 33 Table A2. Definition of socio-demographic and other covariates ........................................... 34 Table A3. Definition of dependent variables ........................................................................... 36 Table A4. Descriptive statistics of respondents ....................................................................... 38 Table A5. Average marginal component effects (AMCEs) of migration policy features ......... 39 Table A6. Conditional marginal means, CMMs (humanitarian vs economic) – Index............. 40 Table A7. Conditional marginal means, CMMs (humanitarian vs economic) – question ....... 41 Table A8. Conditional marginal means, CMMs (social contact) .............................................. 42 Figure A1. Distribution of Venezuelans in Colombia ............................................................... 43 Figure A2. Examples of policy bundles as presented to respondents ..................................... 44 Figure A3. Distribution of humanitarian index ........................................................................ 45 Figure A4. Distribution of humanitarian vs economic question .............................................. 46 Figure A5. Distribution of social contact measures ................................................................. 47 Figure A6. Comparison of results: AMCEs for forced and scaled-rating dependent variables48 Figure A7. Comparison of results: Marginal means for forced and scaled-rating dependent variables. .................................................................................................................................. 49 Figure A8. Subgroup analysis: Social contact........................................................................... 50 Figure A9. Subgroup analysis: Migrant Density ....................................................................... 51 Figure A10. Subgroup analysis: other moderating factors ...................................................... 52 Figure A11. Testing for survey fatigue ..................................................................................... 55 32 Table A1. Summary of pre-registered hypotheses Prioritise economic over More restrictive (+) Aggregate humanitarian Labour market - - Location - - Public services + Numerical limits - Family reunification + Length of residence - Note: - more restrictive, + less restrictive 33 Table A2. Definition of socio-demographic and other covariates Variable Definition Age Self-reported age in years. Responded coded as “old” if above the median age in the sample or “young” otherwise. Education Self-reported highest level of education. The scale contains the following categories: Primary education, Secondary education, Technical education, University (professional) degree, Postdoctoral degree. Employment status Self-reported employment status. Respondents answer the following question: “Which of the following applies to you?” Answer options included: Currently working as self-employed; Currently working as employed; Full time student, Retired/pension, Unemployed, Not working, Have never worked, Other. Gender Self-reported gender. Humanitarian Index A humanitarian index was created based on the answer to the following statements: “To what extent do you agree with the following statements, where 1 means ‘strongly disagree’ and 5 means ‘strongly agree’?” [Ranking 1 to 5] 1. The government must spend more to help the poor. 2. The government should ensure that everyone can meet basic human needs such as food and shelter. 3. Protecting the most vulnerable should be given priority even if that reduces economic growth. 4. People should take more responsibility to provide for themselves. Statements are randomised at the respondent level. Impacted by COVID The following questions were asked to those that were either employed or unemployed: “Many people have been affected financially by the coronavirus. We’d like to know how you have been affected, and how you and your household are coping.” [If unemployed] – Have you been unemployed since before the pandemic or did you become unemployed during the pandemic? [Yes, No] [If employed] – think about the last week compared to your life in February of 2020, before the pandemic. Due to the coronavirus outbreak, did you [thick yes/no]: (i) work fewer hours than usual, (ii) earn less money than usual, (iii) had trouble paying for your usual bills or expenses? Migrant Density For each state (department), we divided the number of migrants by the state’s population. We split the sample between those above and below the median and created an index of exposure to migrants (0 below median, 1 above median). Political ideology Self-reported ideological leaning. The question used was: “Nowadays, when we speak of political leanings, we talk of those on the left and those on the right. In other words, some people sympathize more with the left and others with the right. According to the meaning that the terms 34 “left” and “right” have for you, and thinking of your own political leanings, where would you place yourself on this scale? [Ranking 1 to 10]”. Three ideological categories are defined: left (0-4), centre (5) and right (6-10). Region of residence of Region 1 – Bogota respondent Region 2 – North/Caribbean Region 3 – East Center Region 4 – Antioquia and coffee growing area Region 5 – West - South Social Contact Index Three questions about contact were asked: 1. Over the past few years, how often if at all do you have everyday relationships with people from Venezuela, such as exchanging a few words, for example, or buying a newspaper at the store, and so on? [never, rarely, sometimes, often, every day]; (Schmid et al. 2014) 2. This question concerns people you trust, for example good friends, those with whom you discuss important subjects, with whom you keep in touch, or who are there for you if you need help. How many people you trust are from Venezuela? [0, 1, 2-5, 6-10, More than 10], (General Social Survey) 3. Over the past few years, have you shared a meal in your home with someone from Venezuela? [Yes, No]. (General Social Survey) We follow (Clayton et al. 2021) and randomised questions per respondent and create an index of proximity to stratify the sample in two groups: those who report high levels of contact, and those who report low levels of contact. These groups are based on an index of contact with Venezuelans (on a scale of 0-3) using the three questions on contact. Respondents will receive one point if they report having everyday interactions with Venezuelans “sometimes,” “often,” or “everyday;” if they report trusting “2-5,” “5-10,” or “more than 10” Venezuelans; or if they report having shared a meal in their home with a Venezuelan in the past year. Those who score either a 2 or 3 on the index will be considered in the “high contact” group, and those scoring either a 0 or 1 will be considered in the “low contact” group. In our robustness checks, we will also re-run the analyses using (a) anyone scoring more than 0 as having “high contact,” (b) anyone answering “yes” to the shared meal question as having “high contact. Socioeconomic strata Self-reported social strata [1 to 6]. Social strata (estrato socioeconomico) is considered low for those living in 1, 2 and 3 – which is where residents receive subsidies for their utilities (water, gas, electricity). 5 and 6 are the highest and pay higher rates to subsided those in lower stratas. Vote in Presidential For those who answer yes in a question that asks: “Did you vote in the election last presidential elections in Colombia?” The choices were Gustavo Petro, Ivan Duque, in blank, rather not respond. Our subgroup analysis is based on Petro vs Duque. 35 Table A3. Definition of dependent variables Variable Description A binary indicator for whether the policy package was the preferred choice from the pair of policy packages presented. The respondent answers based on the following question: “Imagine that the Colombian government is considering various packages of Forced Choice measures to address migration from Venezuela. Next will show you two possible options of packages, A and B Read their descriptions carefully. Which package is closest to what you would prefer?” A dummy variable is created with a value of 1 for the option chosen and 0 otherwise. A dichotomised version of the rating outcome coded 1 for profiles rated higher than 4, 0 otherwise. Based on the following question: On a scale from 1 to 7, where 1 means “you absolutely dislike” Scaled rating and 7 means “you absolutely like”, how would you rate each of these packages? 36 Table A4. Sampling comparison with 2018 Census Sample 2018 Census Gender Female 0.49 0.52 Age Structure 41.31 18-24 0.17 0.17 25-34 0.23 0.22 35-44 0.19 0.19 45-54 0.15 0.16 > 55 0.26 0.26 Economic Strata 1-2 0.68 0.69 3-4 0.26 0.27 5-6 0.06 0.04 Region Bogota 0.17 0.19 Norte/Caribe 0.22 0.21 Centro/Oriental 0.19 0.24 Antioquia/Eje cafetero 0.16 0.21 Sur Occidental 0.25 0.16 37 Table A5. Descriptive statistics of respondent’s characteristics Mean SD Marital Status Single 0.41 0.50 Married 0.29 0.45 Free union 0.21 0.41 Other 0.09 0.28 Economic Strata 1-2 0.69 0.46 3-4 0.26 0.26 5-6 0.06 0.22 Education Less than secondary 0.12 0.32 Less than University 0.38 0.48 University 0.50 0.50 Employment Status Employed 0.65 0.47 Unemployed 0.12 0.33 In education 0.10 0.30 Other (e.g. retired) 0.11 0.32 N = 2,508 38 Table A6. Average marginal component effects (AMCEs) of migration policy features Model 1 Model 2 Model 3 Model 4 Intercept Numerical limits (baseline: no limit) Annual limits 0.065*** 0.065*** 0.065*** 0.065*** (0.007) (0.007) (0.007) (0.007) Location restrictions (baseline: No restrictions) Restrictions -0.031*** -0.031*** -0.031*** -0.031*** (0.007) (0.007) (0.007) (0.007) Labour market access (baseline: allowed to work) Restricted work 0.022** 0.022** 0.022** 0.022** (0.009) (0.009) (0.009) (0.009) Not allowed to work -0.116*** -0.116*** -0.116*** -0.116*** (0.009) (0.009) (0.009) (0.009) Public services access – Health (baseline: Access) No access -0.040*** -0.040*** -0.040*** -0.040*** (0.007) (0.007) (0.007) (0.007) Family reunification (baseline: allowed to bring family) Bring family if can support it 0.107*** 0.107*** 0.107*** 0.107*** (0.008) (0.008) (0.008) (0.008) Not allowed to bring family -0.012 -0.012 -0.012 -0.012 (0.008) (0.008) (0.008) (0.008) Length of residency (baseline: indefinite period) Ten years that can be renewed 0.053*** 0.053*** 0.053*** 0.053*** (0.008) (0.008) (0.008) (0.008) Ten years that cannot be renewed 0.050*** 0.050*** 0.050*** 0.050*** (0.008) (0.008) (0.008) (0.008) Controls for age and education No Yes Yes Yes Control for liberalism No No Yes Yes Control for attitude to migration No No No Yes N 25,080 25,080 25,080 25,080 39 Table A7. Conditional marginal means, CMMs (humanitarian vs economic) – Index MMs MMs Humanitarian Economic Difference p-value Allowed to work 0.542 0.516 0.026 (0.011) 0.024 Restricted work 0.554 0.555 -0.001 (0.011) 0.906 Not allowed to work 0.408 0.43 -0.022 (0.011) 0.052 No restriction on location 0.522 0.505 0.017 (0.009) 0.087 Restricted Location 0.479 0.495 -0.016 (0.009) 0.085 Access to subsidised health 0.531 0.502 0.029 (0.009) 0.002 No access subsidised to health 0.47 0.498 -0.028 (0.009) 0.002 Allowed to bring family 0.482 0.45 0.032 (0.011) 0.005 Only bring family if can support it 0.571 0.58 -0.009(0.011) 0.436 Not allowed to bring family 0.447 0.47 -0.023(0.011) 0.039 No numerical limits 0.468 0.468 0.00(0.009) 0.983 Annual Limit 0.533 0.533 0.00(0.009) 1 Indefinite period 0.471 0.457 0.014(0.011) 0.221 Ten years that can be renewed 0.519 0.521 -0.002(0.011) 0.867 Ten years that cannot be renewed 0.511 0.522 -0.011(0.011) 0.311 N 15190 9890 Note: Marginal means. Standard errors clustered by respondent in parentheses. 40 Table A8. Conditional marginal means, CMMs (humanitarian vs economic) – question MMs MMs Humanitarian Economic Difference p-value Allowed to work 0.545 0.517 0.028 (0.011) 0.01 Restricted work 0.552 0.557 -0.005 (0.011) 0.67 Not allowed to work 0.403 0.431 -0.028(0.011) 0.01 No restriction on location 0.523 0.506 0.017(0.009) 0.06 Restricted Location 0.477 0.494 -0.017(0.009) 0.06 Access to subsidised health 0.533 0.503 0.03(0.009) 0.001 No access subsidised to health 0.466 0.497 -0.031(0.009) 0.001 Allowed to bring family 0.482 0.455 0.027 (0.011) 0.01 Only bring family if can support it 0.575 0.575 0(0.011) 0.97 Not allowed to bring family 0.444 0.469 -0.025(0.011) 0.026 No numerical limits 0.472 0.463 0.009(0.009) 0.319 Annual Limit 0.529 0.537 -0.008(0.009) 0.353 Indefinite period 0.472 0.458 0.014(0.011) 0.244 Ten years that can be renewed 0.507 0.534 -0.027(0.011) 0.018 Ten years that cannot be renewed 0.522 0.508 0.014(0.011) 0.204 N 13300 11780 Note: Marginal means. Standard errors clustered by respondent in parentheses. 41 Table A9. Conditional marginal means, CMMs (social contact) MMs High SC MMs Low SC Difference p-value Allowed to work 0.548 0.52 0.028(0.011) 0.013 Restricted work 0.556 0.554 0.002(0.011) 0.875 Not allowed to work 0.401 0.428 -0.027(0.011) 0.014 No restriction on location 0.529 0.505 0.024(0.009) 0.008 Restricted Location 0.471 0.495 -0.024(0.009) 0.009 Access to subsidised health 0.531 0.511 0.02(0.009) 0.033 No access subsidised to health 0.47 0.489 -0.019(0.009) 0.035 Allowed to bring family 0.489 0.455 0.034(0.011) 0.003 Only bring family if can support it 0.576 0.574 0.002(0.011) 0.795 Not allowed to bring family 0.437 0.471 -0.034(0.011) 0.002 No numerical limits 0.478 0.46 0.018(0.009) 0.054 Annual Limit 0.522 0.54 -0.018(0.009) 0.054 Indefinite period 0.467 0.464 0.003(0.011) 0.763 Ten years that can be renewed 0.512 0.526 -0.014(0.011) 0.228 Ten years that cannot be renewed 0.52 0.511 0.009(0.011) 0.409 N 10760 14320 Note: Marginal means. Standard errors clustered by respondent in parentheses. 42 Figure A1. Distribution of Venezuelans in Colombia (a) Venezuela and Colombia’s shared border (b) Distribution of Venezuelans in Colombia (by Department) (83921.5,432716] (14095.5,83921.5] (4713.5,14095.5] [72,4713.5] No data Note: The Administrative Dataset was obtained from the Colombian Minister of Foreign Affairs. The dataset compiles information drawn from the Colombian register of foreigners (Información de Registro de Extranjeros, SIRE), the permit residency data (Permiso Especial de Permanencia, PEP), Migratories entries (Intención de hospedaje) and the 2018 National Census (DANE). 43 Figure A2. Examples of policy bundles as presented to respondents 44 Figure A3. Distribution of humanitarian index The government should spend more to help the poor The government should ensure everyone can meet Protecting the most vulnerable should be given priority basic human needs such as food and shelter even if that reduces economic growth 41.75 40 50 48.80 40 35.01 40 30.14 30 30 24.80 Proportion 23.60 Proportion Proportion 21.33 30 25.20 20 20 20 18.42 10 8.41 10 4.90 10 5.78 4.23 4.27 3.35 0 0 0 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 People should take more responsibility When jobs are scarce, employers should give priority to Humanitarian Index to provide for themselves. Colombians over foreign residents of the country 60.57 43.66 60 80 40 65.31 60 39.43 30 40 Proportion 26.75 Proportion Proportion 21.09 40 20 20 20 17.07 10 5.10 10.53 3.39 4.27 2.83 0 0 0 1 2 3 4 5 1 2 3 4 5 0 1 Note: The question prior to each statement was: To what extent do you agree with the following statements, where 1 means ‘strongly disagree’ and 5 means ‘strongly agree’? [Ranking 1 to 5]. For the Humanitaria Index, the values are 0 for Economic Priorities and 1 for Humanitarian Priorities. Details about the construction of the index are in Table A3 inthe Appendix 45 Figure A4. Distribution of humanitarian vs economic question Humanitarian vs Economic 53.03 50 46.97 40 Proportion 30 20 10 0 0 1 Note: This is the response to a question: If you had to choose one of these statements, which one is closer to what your view is: The situation in Venezuela is mainly an economic one (=0) or the situation in Venezuela is mainly a humanitarian one (=1) More details about the construction are in Table A3 inthe Appendix 46 Figure A5. Distribution of social contact measures Daily Interactions People you trust 40 60 57.14 33.13 30 24.64 40 Proportion Proportion 22.37 20 21.89 9.65 10.21 20 14.51 10 3.75 2.71 0 0 Never Rarely Sometimes Often Every Day 0 1 2-5 6-10 more than 10 Shared a meal Contact Index 60 57.10 40 36.76 31.22 42.90 30 25.88 40 Proportion Proportion 20 20 10 6.14 0 0 Yes No 0 1 2 3 Note: Details about the construction of the index are in Table A4 47 Figure A6. Comparison of results: AMCEs for forced and scaled-rating dependent variables Forced Scale Labour Market Access 0.00 0.00 Allowed to work 0.02 0.00 Restricted work -0.12 -0.07 Not allowed to work Location Restrictions 0.00 0.00 No restriction on location -0.03 -0.02 Restricted Location Health Access 0.00 0.00 Access to subsidised health -0.04 -0.03 No access subsidised to health Family Reunification 0.00 0.00 Allowed to bring family 0.11 0.05 Only bring family if can support it -0.01 -0.02 Not allowed to bring family Numerical Limits 0.00 0.00 No numerical limits 0.06 0.04 Annual Limit Length of Residency 0.00 0.00 Indefinite period 0.05 0.02 Ten years that can be renewed 0.05 0.03 Ten years that cannot be renewed Change in Pr(Policy Pack) Change in Pr(Policy Pack) Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. Single points without horizontal bars denote the reference category used for that dimension. The scaled variable is a dichotomised version of the rating outcome coded 1 for profiles rated higher than 4, 0 otherwise. Based on the following question: On a scale from 1 to 7, where 1 means “you absolutely dislike” and 7 means “you absolutely like”, how would you rate each of these packages? 48 Figure A7. Comparison of results: Marginal means for forced and scaled-rating dependent variables. 49 Figure A8. Subgroup analysis: Social contact Social Contact 2 Shared a Meal Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. Blue (Red) coefficients are for high (low) social contact and shared (has not) a meal with Venezuelans 50 Figure A9. Subgroup analysis: Migrant Density The role of migrant density as a moderator Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy High Low Note: N = 2,508 respondents. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals. The migrant density measure involved calculating the proportion of migrants as ashare of the total population of each state (department). Then, using respondents’ reported location at the state-level, we split the sample into two groups: those living in states whose proportion of migrants was below the median of all states (coded as 0), and those living in states with migrant propor- tions above the median (coded as 1). 51 Figure A10. Subgroup analysis: other moderating factors Gender Age Labour Market Access Labour Market Access Allowed to work Allowed to work Restricted work Restricted work Not allowed to work Not allowed to work Location Restrictions Location Restrictions No restriction on location No restriction on location Restricted Location Restricted Location Health Access Health Access Access to subsidised health Access to subsidised health No access subsidised to health No access subsidised to health Family Reunification Family Reunification Allowed to bring family Allowed to bring family Only bring family if can support it Only bring family if can support it Not allowed to bring family Not allowed to bring family Numerical Limits Numerical Limits No numerical limits No numerical limits Annual Limit Annual Limit Length of Residency Length of Residency Indefinite period Indefinite period Ten years that can be renewed Ten years that can be renewed Ten years that cannot be renewed Ten years that cannot be renewed 0.40 0.50 0.60 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Marginal Mean: Likelihood of Preference for Policy Female Male Older Younger Region Economic Strata Labour Market Access Labour Market Access Allowed to work Allowed to work Restricted work Restricted work Not allowed to work Not allowed to work Location Restrictions Location Restrictions No restriction on location No restriction on location Restricted Location Restricted Location Health Access Health Access Access to subsidised health Access to subsidised health No access subsidised to health No access subsidised to health Family Reunification Family Reunification Allowed to bring family Allowed to bring family Only bring family if can support it Only bring family if can support it Not allowed to bring family Not allowed to bring family Numerical Limits Numerical Limits No numerical limits No numerical limits Annual Limit Annual Limit Length of Residency Length of Residency Indefinite period Indefinite period Ten years that can be renewed Ten years that can be renewed Ten years that cannot be renewed Ten years that cannot be renewed 0.40 0.50 0.60 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Marginal Mean: Likelihood of Preference for Policy Region1 Region2 Region3 Region4 Region5 Low High 53 Employment Education Labour Market Access Labour Market Access Allowed to work Allowed to work Restricted work Restricted work Not allowed to work Not allowed to work Location Restrictions Location Restrictions No restriction on location No restriction on location Restricted Location Restricted Location Health Access Health Access Access to subsidised health Access to subsidised health No access subsidised to health No access subsidised to health Family Reunification Family Reunification Allowed to bring family Allowed to bring family Only bring family if can support it Only bring family if can support it Not allowed to bring family Not allowed to bring family Numerical Limits Numerical Limits No numerical limits No numerical limits Annual Limit Annual Limit Length of Residency Length of Residency Indefinite period Indefinite period Ten years that can be renewed Ten years that can be renewed Ten years that cannot be renewed Ten years that cannot be renewed 0.40 0.50 0.60 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Marginal Mean: Likelihood of Preference for Policy Employed Unemployed Less than University University or more 54 Figure A11. Testing for survey fatigue Marginal Means by Task Labour Market Access Allowed to work Restricted work Not allowed to work Location Restrictions No restriction on location Restricted Location Health Access Access to subsidised health No access subsidised to health Family Reunification Allowed to bring family Only bring family if can support it Not allowed to bring family Numerical Limits No numerical limits Annual Limit Length of Residency Indefinite period Ten years that can be renewed Ten years that cannot be renewed 0.40 0.50 0.60 Marginal Mean: Likelihood of Preference for Policy Task 1 Task 2 Task 3 Task 4 Task 5 Note: N = 2,508 respondents. Results are from different estimations based on the task or trial number. That is, 2 refers to the second conjoint task. Estimates based on linear probability models with clustered and robust standard errors with 95% confidence intervals.