Policy Research Working Paper 10091 Discrimination Toward Migrants During Crises Marisol Rodriguez Chatruc Sandra V. Rozo Development Economics Development Research Group June 2022 Policy Research Working Paper 10091 Abstract How do crises shape native attitudes towards migrants? their attitudes towards Venezuelan migrants. The findings A common threat could pro-duce an empathy channel suggest that native attitudes towards migrants are substan- among natives, but the perception of competition for tially more suggestive of the resentment channel in the scarce economic resources could just as easily spark prej- treatment group. However, respondents in the so-called udice through a resentment channel. 3,400 Colombian impressionable years—ages 18 to 25—showed more altruism citizens were surveyed and randomly primed to consider towards migrants after priming. Interestingly, both effects the economic consequences of COVID-19 before eliciting disappear in response to positive news. This paper is a product of the Development Research Group, Development Economics. 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 sandrarozo@worldbank.org. 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 Discrimination Toward Migrants During Crises* ıguez Chatruc† Marisol Rodr´ Sandra V. Rozo‡ JEL Classification: D72, F2, O15, R23 Keywords: migration, COVID-19, attitudes, priming, altruism * We thank the editor, Chris Parsons, the two anonymous reviewers, and Berk Ozler for useful suggestions. We are also grateful to Andr´es Barinas, Tatiana Hiller Zapata, Camila Cort´ es, and Mar´ e Urbina for excellent research assistance. USC IRB approval UP-20-00479. This ıa Jos´ study is registered in the AEA RCT Registry with the unique identifying number: AEARCTR-0006182. The views in this paper are strictly those of the authors and should not be attributed to the Inter-American Development Bank or the World Bank, its executive directors, or its member countries. Rodr´ ıguez acknowledges funding from the Inter-American Development Bank. Rozo acknowledges financial support from the Research Support Budget at the World Bank. † Inter-American Development Bank, email: marisolro@iadb.org ‡ Development Research Group, The World Bank, IZA, CEGA - University of California Berkeley, CESR - University of Southern California. email: sandrarozo@worldbank.org I INTRODUCTION Migration is a divisive issue. Anti-immigrant sentiments are widespread and recent opinion surveys suggest that natives worldwide have become less accepting of migrants.1 Prejudice against migrants can preclude integration and lead to their social exclusion, inducing large economic and social costs in hosting societies. It is necessary to understand how natives alter their attitudes towards migrants in response to different stimuli in order to design policies and programs more effectively. We examine how crises affect native altruism and attitudes towards migrants. When facing a common threat, natives may feel solidarity or even kinship with migrants in the same situation, or with society as a whole. We call these positive responses the empathy channel. Alternatively, a crisis could increase competition for scarce resources and trigger native resentment towards migrants, a response we call the resentment channel. We also study whether individuals between 18 and 25 years old respond differently to migrants during crises. According to a vast literature in social psychology and economics, core attitudes, beliefs, and values are formed mostly during early adulthood and change only slowly after this critical period. The theory of the impressionable years hypothesis posits that people 18 to 25 years old have superior mental plasticity. For instance, several studies have shown that the historical and economic contexts of this time shape basic attitudes, values, and world views (e.g., Greenstein, 1965; Easton et al., 1969; Cutler, 1974; Sears, 1983; Giuliano and Spilimbergo, 2014; Torney-Purta, 2017;). Glenn (1980) and Spear (2000) have proposed scientific explanations for this phenomenon. Glenn (1980) suggests that people are flexible in reacting to social circumstances when they are young but become gradually less so as they age. This decrease in flexibility is due to a drop in energy, loss of brain tissue, disengagement, and less interest in events distant from one’s immediate life, as well as to the accumulation of friends with similar world views. Spear (2000) states that a younger brain in a transitional period differs anatomically and neurochemically from an adult brain. In a young brain, the grey matter in the cortex gradually increases until about the age of adolescence, then sharply declines as the brain sheds neuronal connections superfluous to adult needs. We examine the effects of the COVID-19 pandemic in Colombia on native altruism and attitudes towards migrants. It is challenging to identify the effect of COVID-19 on attitudes because a simple comparison of individual views beforehand and afterward could confound the effects of the pandemic with unrelated but concurrent events. For instance, in addition to the pandemic and its economic consequences in 2020, many nations implemented policies that same year to restrict immigration, which could arguably also have affected attitudes towards migrants. To isolate the effects of COVID-19, we conducted an online survey experiment with 3,400 Colombian nationals. 1 See the evolution of the Migrant Acceptance Index from the Gallup report collected in 140 countries between 2016 and 2019 at: https: //news.gallup.com/poll/320678/world-grows-less-accepting-migrants.aspx 2 We randomly assigned half our sample (the treatment group) to receive priming about COVID-19 before they took a survey that evaluated their attitudes towards Venezuelan migrants. Priming is a psychological technique that exposes people to stimuli. The prime typically consists of meanings (e.g., words, images, and sounds) that activate associated memories. This process may influence performance on a subsequent task. We recruited respondents through Facebook advertisements and stratified our experiment by age and gender. Colombia was an appropriate setting for our experiment since it is the primary destination for the exodus of migrants from the humanitarian crisis in Venezuela. By 2020, more than five million people had fled Venezuela; of those, approximately two million settled in Colombia.2 Thus we examine the impacts of a profound economic crisis (the COVID-19 pandemic) in a country hosting a massive inflow of migrants that was equivalent to a shock of approximately four percent of the total population in the last five years.3 We study the effects of the priming on five main outcomes: self-reported altruism, a measure of policy altruism that corresponds to support for public policies to help migrants, opinions about the work efforts of migrants, opinions about the effects of Venezuelan migration on Colombia’s economy, and views on whether Venezuelan migrants pay more or fewer taxes than Colombians. We find that our treatment made the COVID-19 crisis more salient for treated individuals than for the control group. At the end of the survey, we asked respondents in each group two questions. The first was an open-ended question in which they reported the worst crisis in Colombia during the last 50 years. Our idea was to elicit their views without specifically referring to certain crises. The second question asked them to order from first to third the worst crises in Colombia during the last 10 years, with the following options: illegal drug trafficking, internal armed conflict, and COVID-19.4 Responses to both questions confirmed that our treatment increased the salience of the pandemic. Individuals in the treatment group raised the COVID-19 crisis in the open-ended question more often than did the control group. They also ranked COVID-19 as a worse crisis than illegal drug trafficking and internal armed conflict. We document four main findings. First, our results strongly support the validity of the resentment channel whereby treated natives resent migrants during economic crises. In particular, treated individuals consistently reported more negative attitudes towards migrants than did the control group. Natives primed with the COVID treatment were less likely to think an immigrant was poor due to circumstances beyond their control (0.07 standard deviations (sd) lower than the control group). These same natives were more likely to hold negative opinions about both the economic impacts of migrants (0.07 sd lower than the control group) and about their tax contributions (0.19 sd lower than the control group). The effects are small for the first two outcomes but larger for the latter case. These responses may 2 In practice, migration from Venezuela to Colombia is likely higher still as many migrants may never formally register due to fear of deportation. 3 There is also evidence that attitudes towards migrants in Colombia have deteriorated: according to Gallup’s Migrant Acceptance Index, it ranks third among countries where acceptance dropped the most between 2016 and 2019. 4 Of course, there have been other crises in Colombia during this period; this list did not intend to be comprehensive but only to offer relevant options besides COVID-19). 3 stem from the growing significance of the refugees’ economic situation amid the pandemic in Colombia. Second, we cannot distinguish statistically negative effects of the COVID priming treatment on the self-reported measures of altruism and policy altruism of treated respondents compared to the control group. The point estimates are negative but their standard errors are large. Third, we observe that people in their impressionable years—that is, ages 18 to 25—report substantial improve- ments in altruism after priming, relative to the control group. As a result, this age bracket could be an excellent target for programs to improve prosocial behaviors and reduce prejudice. Interestingly, the effects are not significant for older populations.5 In fact, Giuliano and Spilimbergo (2014) show that individuals who are exposed to economic crises when young tend to be more prosocial later in life. Finally, we repeated our survey experiment with 2,915 different Colombians after news of a successful trial of a COVID-19 vaccine there. Our objective was to test whether our findings would hold after natives received good news about the economic outlook. Notably, we did not observe significant effects of the priming on any of our five outcomes after this event. This suggests that although crises increase antipathy towards migrants, it is reversible in response to favorable developments. These findings suggest that the effects of crises on attitudes toward migrants are temporary and tend to disappear as economic conditions improve. Another explanation for the results of our second experiment could be that the temporary shock caused by COVID-19 induced only temporary effects on attitudes, and the shift in our second experiment was unrelated to the vaccine news. Relation to the literature: Our research relates closely to four branches of the literature. The first one studies how attitudes and behaviors towards migrants can change in experimental settings. One subset of this work has focused on the effects of information provision and has documented high levels of misinformation among respondents in developed countries regarding the size and characteristics of the migrant population (Alesina et al., 2022; Grigorieff et al., 2020). While providing information about the true size and characteristics of the migrant population can improve attitudes, it does not improve behaviors or policy preferences regarding undocumented migration (Grigorieff et al., 2020). Yet, giving subjects information about research that shows no adverse labor market impacts of migration can shift both attitudes and behaviors towards low-skilled immigrants more positively (Haaland and Roth, 2020).6 A second subset in this area promotes perspective taking through mental exercises, videos, and games. For example, US inhabitants who were asked to imagine themselves as refugees were more likely to write a letter in support of refugees to the president (Adida et al., 2018). Furthermore, an online game and video increased altruism and improved the ıguez Chatruc and Rozo, 2021). attitudes of Colombians towards Venezuelan migrants (Rodr´ 5 Examples of this type of program were implemented by Rodr´ ıguez Chatruc and Rozo (2021), where natives were randomly assigned either to play an online game that immersed them in the life of a Venezuelan migrant or to watch a documentary about Venezuelans crossing the border on foot. 6 Information provision that is done in an anecdotal manner seems to be more effective in shifting attitudes than that conducted in a factual way (Alesina et al., 2022). 4 The second branch of work analyzes the determinants of individual attitudes and preferences regarding income re- distribution and altruism. These studies have shown that context, culture, and history—both individual and collective— shape attitudes about redistribution (Luttmer and Singhal, 2011), as do individual characteristics such as income, race, sex, and education (Alesina and Giuliano, 2010; Facchini and Mayda, 2006). Most of the evidence regarding the impacts of migration on prosocial behaviors has centered on preferences for redistribution and trust. These stud- ies conclude that typical misconceptions about the characteristics of immigrants in receiving countries affect native preferences regarding redistribution and prosocial behaviors (Alesina et al., 2019). Analyses of trust and reciprocity between natives and immigrants have yielded mixed results. For example, in the Dutch context, Cettolin and Suetens (2018) find that natives trust and reciprocate less to immigrants who come from non-Western countries. In contrast, when immigrants have similar languages, cultures, and religions, natives might be more trusting and altruistic towards immigrants than towards other natives (Hassan et al., 2019). The third branch of research is grounded in psychology and examines the effects of priming. This method has recently become popular in economic research. Some of its most common uses highlight individual characteristics (Benjamin et al., 2010; Cohn et al., 2015; Benjamin et al., 2016) or recall past traumatic events to examine causal effects on economic behavior (Lerner et al., 2003; Callen et al., 2014). Finally, we also contribute to a recent and growing literature that uses survey experiments to study the impacts of the COVID-19 pandemic on people’s attitudes. Some studies have used modules to prime respondents and measure outcomes of interest such as interpersonal trust, values, and policy preferences (Daniele et al., 2020) or solidarity and fairness (Cappelen et al., 2020). Others have used information provision to examine changes in personal views about the trade-offs between public health conditions and civil liberties (Alsan et al., 2020), as well as changes in economic anxiety (Fetzer et al., 2020). We contribute to the literature by examining the effects of an economic crisis on the self-reported prosocial be- haviors and attitudes of natives towards migrants. We also examine whether natives exhibit heterogeneous responses according to age and gender. II CONTEXT: VENEZUELAN MIGRATION TO COLOMBIA Venezuela, once known as the “economic gem” of Latin America, was an extremely prosperous country in the late 1980s and early 1990s. This began to change in 1998 following the successive elections of populist regimes that expropriated private property and changed the country’s constitution, creating a profound institutional and economic crisis. In recent years, the economic crisis has worsened as oil prices fell, the United States imposed drastic sanctions, 5 and the private sector virtually disappeared (Bahar et al., 2021). Since 2015, social protests and political violence have become rampant, and Venezuelans have been forced to migrate on a scale never seen before in Latin America (Ib´ an˜ ez et al., 2021). By 2020, the Venezuelan exodus had become one of the world’s largest humanitarian crises with more than five million people having fled the country. Colombia is the primary host of these forced migrants. That same year, the Colombian government reported that more than 1.72 million Venezuelans had registered with the United Nations Refugee Agency to stay in the country. Nonetheless, the real number of Venezuelan migrants in Colombia is likely larger due to irregular migration. Colombia is also one of the most generous countries in terms of the support it has offered these migrants. In 2018, the government regularized approximately half a million Venezuelans and in 2020, it granted all irregular migrants access to complete health services. Moreover, in 2021, it also offered a temporal regularization (for up to 10 years) to all irregular migrants living there.7 As in other countries that have experienced large migration inflows in a short period of time, the Venezuelan migration has sparked strong reactions from Colombian natives (Rozo and Vargas, 2021). Some Colombians have welcomed the migrants, but others resent them and blame them for current socioeconomic problems. III THE SURVEY EXPERIMENT a with 3,400 Colombian Facebook users. Bogot´ We conducted an online survey experiment in Bogot´ a has a population of 10.7 million, of which 7.8 million use Facebook. As noted above, we recruited respondents through Facebook ads and stratified the experiment by age group and gender. Table A.1 of the Appendix compares sociodemographic a population (as collected by the population census characteristics of the survey respondents with those of the Bogot´ conducted by the Colombian Statistics Agency in 2018). As the table illustrates, our survey respondents were com- parable to the rest of the population in terms of age distribution and income (measured as economic strata). Yet, our a. This aligns with our expectations as Facebook respondents were more educated than the average inhabitant of Bogot´ access requires respondents to read, write, and proficiently use digital apps and surveys. Figure A.1 illustrates our recruiting ads. Without mentioning migrants or the COVID-19 pandemic, these ads a. In total, 47,376 individuals in Bogot´ invited users to answer questions about the current situation in Bogot´ a clicked on the ads and 34,034 went to the first page of the survey. Of these, 5,908 began the survey and 4,333 finished it. These numbers imply a success rate of 12.7 percent of the population exposed to the ads. The actual number 7 See Ib´ an˜ ez et al., 2022 for details on the impacts of the amnesty offered in 2018. 6 of completed surveys was 3,413 because some respondents were not Colombians, did not consent to participate in a, or answered the survey more than once. Figure A.2 illustrates the actual the experiment, were not living in Bogot´ location of respondents. All respondents answered 31 questions divided into five modules: (i) basic sociodemographic characteristics, (ii) COVID-19, (iii) crises, (iv) attitudes towards migrants, and (v) social desirability.8 Table A.2 presents the order in which the treatment and control groups answered each module. We chose the order carefully to avoid priming the control group before asking them about prosocial behaviors and attitudes towards migrants. The treatment group answered the modules in the order listed above, whereas the control group answered them in the following order: (i) basic sociodemographic characteristics, (ii) attitudes towards migrants, (iii) crises, (iv) COVID-19, and (v) social desirability. Respondents were not able to identify the different modules as we applied a continuous questionnaire. Specifically, each module collected the following information, respectively: 1. Basic sociodemographic characteristics: included gender, age, education, economic strata (measured in Colom- bia according to the area of residence), religion, and political orientation (measured in scale of one to 10 from left to right). 2. COVID-19: this module asked individuals to think about their situation in March 2020 (before the pandemic) and compare it with their current situation to assess whether someone in their family had lost their job or expe- rienced a reduction in working hours. They were also asked how many people they knew who had contracted COVID-19, and to report their perceptions on poverty trends in Colombia as a consequence of the pandemic. 3. Crises: this module’s objective was to verify whether the COVID-19 module made the crisis more salient to the treatment group. For this purpose, we asked individuals two questions. The first was an open-ended question about the worst crisis faced by Colombia in the last 50 years. Using the answers to this question, we created an indicator variable equal to one for any answer that included the words “pandemic,” “COVID-19,” or “Coro- navirus.” The second question asked individuals to rank the following three crises in Colombia from worst to least bad: illegal drug trafficking, internal armed conflict, and COVID-19. We then created a variable that gave a score of three to anyone who ranked COVID-19 as the worst crisis, two if COVID-19 was the second-worst crisis, and one if it was the third-worst crisis. 8 We used a reduced scale since Facebook surveys must be short to increase response rates. Since the original index counts each question as one or zero depending on the answer, summing the number of questions still yields changes in the same direction. Dahr et al., 2022 also applied a reduced-form version of the scale. 7 4. Prosocial behaviors and attitudes towards migrants: this module collected information on our five main out- comes of analysis and additional secondary outcomes for exploratory analysis. Our five main outcomes include two measures of altruism: one was a self-reported measure and the other measured policy altruism related to voter support for public aid to migrants. The other three outcomes measured native attitudes towards migrants (described in detail in the next section). We also collected information on respondents’ perceptions of Venezue- lan migrants (the share of Venezuelans in the Colombian population and their levels of education) and elicited native perceptions about migrants’ impacts on Colombian labor markets, culture, and crime. 5. Social desirability: this module included four questions to construct a social desirability scale for each individ- ual. We measure social desirability bias by using four questions from Crowne and Marlowe’s social desirability scale (see Crowne and Marlowe, 1964 for details). The questions assess whether or not respondents are con- cerned with social approval. A high number of socially desirable responses suggests the respondent is concerned with social approval. Particularly, we used four of the 33 questions of the scale to construct an index from one (no social desirability) to four (maximum social desirability) and standardize it for ease of interpretation. We used only four questions in order to minimize the duration of the survey and increase response rates. Each an- swer was assigned a score of one or zero depending on whether the scale identified the respondent as someone who wanted social approval. The four questions—with possible answers of true or false—were: • It is sometimes hard for me to go on with my work if I am not encouraged. (False was associated with social desirability). • There have been times when I was quite jealous of the good fortune of others. (False was associated with social desirability). • I am always courteous, even to people who are disagreeable. (True was associated with social desirability). • I’m always willing to admit it when I make a mistake. (True was associated with social desirability). Descriptive statistics for all survey variables are in Table I. Figure I offers a timeline illustration of the different pandemic shocks and survey rounds. 8 IV EMPIRICAL STRATEGY ıguez Chatruc and Rozo (2019), We estimate the average treatment effect as specified in the pre-analysis plan of Rodr´ using the following model: Yi = α + λTi + i (1) where the dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group for ease of interpretation, and T ∈ {0, 1} is the assigned treatment status to COVID priming. Finally, i represents the error term. We examine the effects of COVID priming on five main outcomes, including: (i) Altruism, measured as the self- reported willingness to donate to good causes (ranges from one to 10, where 10 corresponds to “very willing to donate”);9 (ii) Policy altruism, measured as agreement with “the Colombian government should support Venezuelan migrants” (Likert scale from one to four, where four corresponds to “strongly agree”); (iii) Opinion on effort, an indicator variable equal to one if the respondent answered that Venezuelan migrants were poor due to circumstances beyond their control and zero if the respondent answered that Venezuelan migrants were poor due to lack of self-effort; (iv) Opinion on economy, which asked whether migrants were good for the economy (Likert scale from one to four, where four corresponds to “strongly agree”); (v) Opinion on taxes, which asked whether Venezuelan migrants paid more or fewer taxes than Colombians (a scale from one to five, where five represents a lot more).10 The first two variables attempt to capture natives’ prosocial behaviors in general, and variables (iii) through (v) attempt to capture natives’ attitudes towards migrants. We examine the effects for the entire sample and broken down by gender and age group (18–25, 35–44, 45–54, and 55+), as mentioned in the pre-analysis plan. For that purpose, we interact an indicator variable for males and an indicator variable for the impressionable years (18 to 25) with the main explanatory variables in Equation 1. 9 The self-reported question is an assessment of each participant’s willingness to give to good causes in general. It was adapted from Falk et al. (2018), who validated the question experimentally in Colombia, among many other countries. The authors selected this question as the one that best approximated experimental variation in altruism. The question asks: “How willing are you to give to good causes without expecting anything in return?” Respondents answer by choosing a value on a Likert scale from zero to 10, where zero means “completely unwilling to do so” and 10 means “very willing to do so.” 10 It corresponds to the following statement: “Consider two individuals, Carlos and Diego, who currently live in Colombia with their families. Carlos was born in Colombia and Diego was born in Venezuela and moved five years ago to Colombia. They are both 35 years of age, have three children, and earn low incomes. In your opinion, does Diego the Venezuelan pay less, the same, or more taxes than Carlos the Colombian?” Respondents could select a response from a five-item Likert scale where one represented “a lot more” and five represented “a lot less.” The question was adapted from Alesina et al. (2022). 9 V RESULTS V.1 Experiment success First, we demonstrate a successful randomization of the priming treatment by testing whether there were statistically significant differences between participants in the treatment and control groups on the sociodemographic variables we collected. The results are in Table II and are reassuring because both groups were generally balanced in all observed variables. Moreover, the variables are not jointly statistically different between groups, as confirmed by the joint orthogonality test. We also tested the efficacy of our priming treatment by asking both groups to answer two questions. As explained above, we constructed an indicator variable that took the value of one if words like “COVID-19,” “Coronavirus,” or “pandemic” appeared in answers to the open-ended question: “What is the worst crisis that Colombia has faced in the last 50 years?” When coding this variable, we included corrections for phonetic approximations of these words to account for possible spelling mistakes. The second question asked individuals to rank three of Colombia’s crises in the last 10 years, from worst to least intense: illegal drug trafficking, internal armed conflict, and the COVID-19 pandemic. We then created a score that took one of three values: “3” for respondents who ranked COVID-19 as the most serious crisis, “2” for those who ranked it second, and “1” for those who ranked it third. Although both groups answered these questions, the treatment group answered them after the COVID-19 module whereas the control group answered them before the COVID-19 module and after the module on attitudes towards migrants. Thus, the treatment group was primed to think about COVID-19 before answering the crisis questions while the control group was not. Also, since respondents in the control group answered these questions after we assessed their attitudes towards migrants, they were not primed to think about any crisis before we elicited their opinions. The formal test of the efficacy of our treatment is in Table B.1 and is illustrated in Figure II. The table shows a regression of the two variables that measure the salience of the pandemic on a treatment indicator for priming. Our results strongly support the notion that our treatment succeeded in making the COVID-19 crisis salient for treatment recipients. In particular, treatment recipients more often ranked the pandemic as the worst problem in the last 10 years, compared to the control group (column (1)). They were also 7.3 percentage points more likely than the control group to report that COVID-19 was the worst crisis Colombia had experienced in the last 50 years (column (2)). 10 V.2 Main results Figure II depicts the main results of our experiment, which are also presented in Table III. The figure illustrates the point estimates of Equation (1) and its 95 percent confidence intervals. The coefficients for all our primary outcomes suggest the COVID-19 crisis worsened Colombian attitudes towards Venezuelan migrants. Specifically, we identify negative effects of the treatment on opinions regarding migrant effort, opinions concerning the effect of migrants on the Colombian economy, and opinions about migrant tax payments. The effects are small for the first two outcomes but larger for the last one. Notably, the results in columns (3), (4), and (5) show that primed individuals experienced reductions of 0.071 sd in their opinions regarding the effort of migrants, 0.067 sd in their opinions of the effects that migrants had on the Colombian economy, and 0.19 sd in their opinions regarding whether Venezuelan migrants paid more taxes than Colombians. The last outcome can be interpreted as the result that after priming, respondents believed that Venezuelans paid less taxes than Colombians. These responses may stem from the increased salience of the Venezuelan refugee situation amid the pandemic in Colombia. The estimates are robust to multiple hypotheses testing, as shown in Table B.2. We cannot, however, distinguish any statistically significant effects of the priming treatment on altruism (in gen- eral) or on our measure of policy altruism as illustrated in columns (1) and (2). Our results accord with previous studies on how economic downturns affect the attitudes of white people towards African Americans in the United States (Bianchi et al., 2018). In interpreting the results, a valid concern is whether these changes in attitudes reflect more discrimination towards migrants during the COVID-19 pandemic or updated beliefs about how the pandemic affected the migrant commu- nity. Venezuelan refugees typically work in sectors such as construction and services. Since these sectors suffered heavily due to the Colombian lockdowns, an increase in the salience of the pandemic’s economic consequences could potentially make the negative employment shocks to Venezuelans more salient as well. A negative treatment effect on beliefs concerning how much tax Venezuelans pay would thus be a rational response and less likely due to resentment. These confounding effects are only problematic for the respondents’ opinions on migrant impacts on the economy and their tax contributions, but not—for example—for respondents’ opinions on migrant effort. Our findings, however, show negative changes in all three outcomes. This suggests that even if the results partly arise from updated beliefs about the crisis’s effect on Venezuelan migrants, there is also significant change in the perceptions of how responsible migrants are for their own poverty due to lack of effort. 11 V.3 Heterogeneous effects We examine heterogeneous effects of our treatment by gender and age as specified in our pre-analysis plan. The results are in panels B and C of Table III. Gender We cannot estimate any statistically significant heterogeneous effects of the program by gender, as evidenced by the first row of panel B. Yet, the marginal effects of our main outcomes for males are statistically significant for the variables of opinions regarding the economy and taxes. They suggest, in general, that males have worse attitudes towards migrants, relative to females. In fact, the coefficient of the marginal effects for males is negative for all our main outcomes. Impressionable Years Two main results emerge in this regard. First, respondents in their impressionable years (ages 18 to 25) showed remarkably higher altruism after priming relative to other adults (see the first row of panel C in Table III). The results on attitudes towards migrants (as measured by opinions on migrant effort, the economy, and taxes) of individuals in this age group are mixed, exhibiting both negative and positive coefficients. Second, we observe that participants in this age group, in both the treatment and control groups, had more positive views about migrants (as illustrated by the estimated coefficients for the indicator variable of population ages 18–25). Overall, these two findings translate into positive marginal effects of the impressionable years on altruism.11 V.4 Assessing social desirability bias Another relevant concern about the validity of our results is that individuals may have responded to our questions in a more socially desirable way rather than one that accurately reflected their true thoughts or feelings. These behaviors could be problematic if the treatment and control groups showed different social desirability biases. We were especially concerned that treated individuals might exhibit more social desirability bias and thus be afraid to honestly report their views about migrants. To assess this threat, we used the social desirability index (described in section III) to determine any heterogeneous effects of the priming treatment according to each individual’s index. These results are in Table B.3 and suggest a disproportionate response among treatment recipients, who had a higher social desirability index for two of our main outcomes (opinions on migrant effort and the economy). However, the same people also showed more negative attitudes towards migrants—proving that a higher social desirability bias did not prevent honesty. 11 We also examined the treatment effects in the five age groups we proposed in our pre-analysis plan but did not see any clear patterns in the heterogeneous effects of the treatment by smaller age groups. 12 V.5 Exploratory analysis Impacts on beliefs about migrants We also studied if changes in attitudes towards migrants after priming could be explained by or relate to beliefs about the impacts migrants have on the economy: (i) migrants increase competition for national jobs, (ii) migrants increase crime, or (iii) migrants bring new ideas. These results are in Table B.4. We cannot identify significant changes in any of these three outcomes. Yet, the signs of the coefficients suggest that respondents in the treatment group have more negative views than those in the control group. Impacts on misinformation Moreover, we explored whether the treatment could impact misconceptions about Venezuelan migration (see Alesina et al., 2022; Grigorieff et al., 2020). To do this, we analyzed treatment effects on beliefs about the Venezuelan share of the Colombian population and the migrants’ average years of education. Results show the treatment did affect these variables, increasing misconceptions among the general public regarding the size of the migration shock and the average years of education (see Table B.5). In particular, after priming, respondents reported substantially larger migration shocks relative to the control group. They also said migrants had lower levels of education. VI DOES POSITIVE NEWS AFFECT ATTITUDES TOWARDS MIGRANTS? We repeated our survey with 2,915 individuals after news of a successful trial of the first COVID-19 vaccine in Colombia. Our objective was to test if our findings held after respondents received positive information related to the crisis. The results are in Appendix C (Tables C.1 and C.2). We could no longer distinguish significant effects of the priming on any of our five outcomes. Our results suggests that altruism and attitudes towards migrants depend dramatically on the general economic context. Although crises increase antipathy towards migrants, these effects seem to be temporary and reversible in reaction to good news. VII CONCLUSION We explore how crises, such as the one caused by the COVID-19 pandemic, affect altruism and attitudes towards migrants in contexts of large migration inflows. For this purpose, we conducted a survey experiment that primed Colombians to think about the COVID-19 pandemic before eliciting self-reported behaviors and attitudes towards migrants. We find that priming negatively affects these attitudes; this result supports the validity of the resentment channel. Respondents in their impressionable years reacted otherwise, showing more altruism after priming. 13 Our results highlight the importance of support for migrants during crises as these vulnerable populations could experience more prejudice and receive less assistance from hosting communities. Our findings also suggest that the impressionable years are a period in which treatments to improve prosocial behaviors could be most effective. Our study highlights multiple avenues for future research. First, it is likely that users of Facebook—our survey platform—have specific characteristics that may not compare directly to the rest of the population. Facebook users, for example, have more exposure to media content in Facebook platforms. They also read and are proficient users of technology, and as such, have higher education levels than non-users. Moreover, our respondents reported an unusually high likelihood of job loss due to COVID-19. This may explain why they had time to take part in our experiment. If these characteristics are comparable across treatment and control groups, it should not affect our results. However, future research initiatives should examine whether our results can be generalized to other groups such as rural residents with limited internet access. Second, Colombia is a country that is relatively new to international migration inflows. As such, future research could also study the effects of the COVID-19 pandemic in countries where international migration is more common, such as those in the Global North. Third, our analysis only addresses the effects of worldwide crises that affect sending and receiving countries equally. It is possible, therefore, that localized crises can trigger different effects that are more directly associated with migrants, as shown by Rozo and Vargas (2021) concerning the voting effects of Venezuelan migration in Colombia. The authors show that voters responded to higher immigration inflows by supporting more right-wing ideologies because they associated migrants with the leftist regime in Venezuela. Fourth, future studies should assess the validity of our results in situations where respondents truly sacrifice something when they report being more altruistic. References Adida, C. L., A. Lo, and M. R. Platas (2018). Perspective taking can promote short-term inclusionary behavior toward syrian refugees. Proceedings of the National Academy of Sciences 115(38), 9521–9526. Alesina, A., C. Favero, and F. Giavazzi (2019). Effects of austerity: Expenditure- and tax-based approaches. Journal of Economic Perspectives 33(2), 141–62. Alesina, A. and P. Giuliano (2010). Preferences for redistribution. In A. Bisin and J. Benhabib (Eds.), Handbook of Social Economics, pp. 93–132. North Holland. Alesina, A., A. Miano, and S. Stantcheva (2022). Immigration and redistribution. Technical report, Review of Eco- nomic Studies, Forthcoming. 14 Alsan, M., L. Braghieri, S. Eichmeyer, M. J. Kim, S. Stantcheva, D. Y. Yang, et al. (2020). Civil liberties in times of crisis. NBER Working Paper (27972). Bahar, D., A. M. Ib´ an˜ ez, and S. 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. Benjamin, D. J., J. J. Choi, and G. Fisher (2016). Religious identity and economic behavior. Review of Economics and Statistics 98(4), 617–637. Benjamin, D. J., J. J. Choi, and A. J. Strickland (2010). Social identity and preferences. American Economic Re- view 100, 1913–28. Bianchi, E. C., E. V. Hall, and S. Lee (2018). Reexamining the link between economic downturns and racial antipathy: Evidence that prejudice against blacks rises during recessions. Psychological Science 29(10), 1584–1597. Callen, M., M. Isaqzadeh, J. D. Long, and C. Sprenger (2014). Violence and risk preference: Experimental evidence from afghanistan. American Economic Review 104(1), 123–48. Cappelen, A. W., R. Falch, E. Ø. Sørensen, and B. Tungodden (2020). Solidarity and fairness in times of crisis. NHH Dept. of Economics Discussion Paper (06). Cettolin, E. and S. Suetens (2018). Return on trust is lower for immigrants. The Economic Journal 129(621), 1992–2009. echal (2015). Evidence for countercyclical risk aversion: An experi- Cohn, A., J. Engelmann, E. Fehr, and M. A. Mar´ ment with financial professionals. American Economic Review 105(2), 860–85. Crowne, D. P. and D. Marlowe (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology 24(4), 349. Crowne, D. P. and D. Marlowe (1964). The approval motive: Studies in evaluative dependence. New York: Wiley. Cutler, N. E. (1974). Aging and generations in politics: The conflict of explanations and inference. Public Opinion and Political Attitudes. Dahr, D., J. Tarun, and S. Jayachandran (2022). Reshaping adolescents’ gender attitudes: Evidence from a school- based experiment in India. American Economic Review 112(3), 899–927. Daniele, G., A. F. Martinangeli, F. Passarelli, W. Sas, and L. Windsteiger (2020). Wind of change? Experimental survey evidence on the COVID-19 shock and socio-political attitudes in Europe. Working Paper 2020–10 of the Max Planck Institute for Tax Law and Public Finance. 15 Easton, D., J. Dennis, and S. Easton (1969). Children in the political system: Origins of political legitimacy. McGraw– Hill–New York. Facchini, G. and A. M. Mayda (2006). Individual attitudes towards immigrants: Welfare-state determinants across countries. Review of Economics and Statistics 91(2), 295–314. Falk, A., A. Becker, T. Dohmen, B. Enke, D. Huffman, and U. Sunde (2018). Global evidence on economic prefer- ences. The Quarterly Journal of Economics 133(4), 1645–1692. Fetzer, T., L. Hensel, J. Hermle, and C. Roth (2020). Coronavirus perceptions and economic anxiety. Review of Economics and Statistics 103(5), 1–36. Giuliano, P. and A. Spilimbergo (2014). Growing up in a recession. Review of Economic Studies 81(2), 787–817. Glenn, N. D. (1980). Values, attitudes, and beliefs. In O. Brim and J. Kagan (Eds.), Constancy and Change in Human Development. Harvard Univesity Press. Greenstein, F. (1965). Children and politics. Yale University Press. Grigorieff, A., C. Roth, and D. Ubfal (2020). Does information change attitudes toward immigrants? Demogra- phy 57 (3), 1117–1143. Haaland, I. and C. Roth (2020). Labor market concerns and support for immigration. Journal of Public Economics 191, 104256. Hassan, M., S. Mansour, S. Voigt, and M. Gadallah (2019). When Syria was in Egypt’s land: Egyptians cooperate with Syrians, but less with each other. Public Choice 191, 337–362. Ib´ an˜ ez, A. M., A. Moya, M. A. Ortega, S. V. Rozo, and M. J. Urbina (2022). Life out of the Shadows: Impacts of Amnesties in the Lives of Migrants. IZA – Institute of Economic Labor. Ib´ an˜ ez, A. M., S. V. Rozo, and M. J. Urbina (2021). Forced migration and the spread of infectious diseases. Journal of Health Economics 79, 102491. Lerner, J., R. Gonzalez, D. Small, and B. Fischhoff (2003). Effects of fear and anger on perceived risks of terrorism: A national field experiment. Psychological Science 14, 144–50. Luttmer, E. F. P. and M. Singhal (2011). Culture, context, and the taste for redistribution. American Economic Journal: Economic Policy 3(1), 157–79. ıguez Chatruc, M. and S. V. Rozo (2021). In someone else’s shoes: Promoting prosocial behavior through Rodr´ perspective taking. Technical report, The World Bank. 16 Rozo, S. V. and J. F. Vargas (2021). Brothers or invaders? How crisis-driven migrants shape voting behavior. Journal of Development Economics 150, 102636. Sears, D. O. (1983). The persistence of early political predispositions: The roles of attitude object and life stage. Review of Personality and Social Psychology 4(1), 79–116. Spear, L. P. (2000). Neurobehavioral changes in adolescence. Current Directions in Psychological Science 9(4), 111–114. Torney-Purta, J. V. (2017). The development of political attitudes in children. New York: Routledge. 17 Table (I) Descriptive Statistics Variable Observations Average Std. Deviation Min Max Panel A: Sociodemographic characteristics Male [=1] 3,399 0.53 0.50 0.00 1.00 Age (years) 3,413 40.80 14.61 18.00 95.00 Education: Secondary or More [=1] 3,407 0.97 0.16 0.00 1.00 Economic Strata: 1 [=1] 3,397 0.09 0.29 0.00 1.00 Economic Strata: 2 [=1] 3,397 0.35 0.48 0.00 1.00 Economic Strata: 3 [=1] 3,397 0.39 0.49 0.00 1.00 Economic Strata: 4 [=1] 3,397 0.12 0.33 0.00 1.00 Economic Strata: 5 [=1] 3,397 0.04 0.19 0.00 1.00 Economic Strata: 6 [=1] 3,397 0.01 0.11 0.00 1.00 Religious [Yes =1] 3,311 0.87 0.34 0.00 1.00 Ideology (1 Left - 10 Right) 2,446 5.39 1.92 1.00 10.00 Panel B: Primary Outcomes Altruism*: willingness to donate (1= very willing to do so) 3,131 0.70 0.26 0.10 1.00 Policy Altruism*: Col. government should help Venezuelans (1= Strongly Agree) 3,384 0.47 0.24 0.25 1.00 Opinion on Effort*: Ven. are poor due to circumstances beyond their control (1=Yes) 3,346 0.57 0.49 0.00 1.00 Opinion on Economy*: Ven. are good for the economy (1= Strongly Agree) 3,360 0.47 0.24 0.25 1.00 Opinion on Taxes*: Ven. pay more taxes than Colombians (1= A lot more) 3,413 0.42 0.18 0.20 1.00 Panel C: Secondary Outcomes Opinion on Jobs*: Migrants compete for national jobs (1= Strongly Agree) 3,380 0.74 0.26 0.25 1.00 Opinion on Culture*: Migrants bring new ideas and cultures (1= Strongly Agree) 3,383 0.45 0.24 0.25 1.00 Opinion on Crime*: Migrants increase crime (1= Strongly Agree) 3,377 0.80 0.26 0.25 1.00 18 Perception about Size: Venezuelans’ share of Colombian population 3,273 32.38 25.50 1.00 99.00 Perception about Education: Venezuelans’ average years of education 2,901 6.66 3.81 1.00 21.00 Panel C. Covid 19 priming Did you lose your job because of the pandemic [1=Yes] 3,412 0.80 0.40 0.00 1.00 Colombian poverty has increased because of the pandemic* (1=Increased) 3,411 0.95 0.16 0.33 1.00 Pop.you know with COVID: No one 3,409 0.19 0.39 0.00 1.00 Pop.you know with COVID: 1-2 3,409 0.25 0.43 0.00 1.00 Pop.you know with COVID: 3-5 3,409 0.29 0.46 0.00 1.00 Pop.you know with COVID: 6-10 3,409 0.15 0.36 0.00 1.00 Pop.you know with COVID: More than 10 3,409 0.12 0.32 0.00 1.00 Panel D. Listing Experiment and Social Desirability Index COVID-19 worst crisis that Colombia has faced in the last 50 years (1=Yes) 3,413 0.14 0.35 0.00 1.00 COVID-19 worst problem in last 10 years (3=Worst, 2= Second Worst , 1= Third Worst) 3,278 1.45 0.75 1.00 3.00 Social desirability index 3,376 2.70 1.00 0.00 4.00 Notes: *We normalized these variables by dividing them on the maximum value to facilitate interpretation. Altruism is measured as the self-reported willingness to donate to good causes (ranges from one to 10, where 10 corresponds to “very willing to donate”); it was adapted from Falk et al. (2018), who validated the question experimentally in Colombia. The authors selected the question as the one that best approximated experimental variation in altruism. The question asks: “How willing are you to give to good causes without expecting anything in return?” Individuals answer by choosing a value on a Likert scale from zero to 10, where zero means “completely unwilling to do so” and 10 means “very willing to do so.” Policy altruism is measured as agreement with “the Colombian government should support Venezuelan migrants” (Likert scale from one to four, where four corresponds to “strongly agree”); Opinion on effort is an indicator variable equal to one if the respondent answers that Venezuelan migrants are poor due to lack of self-effort; Opinion on economy measures whether migrants are good for the economy (Likert scale from one to four, where four corresponds to “strongly agree”); and Opinion on taxes assesses whether Venezuelan migrants pay more or fewer taxes than Colombians (scale from one to five, where five represents a lot more). It corresponds to the following statement: ”Consider two individuals, Carlos and Diego, who currently live in Colombia with their families. Carlos was born in Colombia and Diego was born in Venezuela and moved five years ago to Colombia. They are both 35 years of age, have three children, and earn low incomes. In your opinion, does Diego the Venezuelan pay less, the same, or more taxes than Carlos the Colombian?” Individuals could select from a five-item Likert scale where one represented “a lot more” and five represented “a lot less.” The question was adapted from Alesina et al. (2022). Table (II) Testing Balance Between Groups Means by Treatment Assignment Priming Group Control Group Overall Sample P-value (1) (2) Mean St. Deviation (1) vs (2) Male [=1] 0.59 0.6 0.6 0.49 0.73 Age 40.7 40.13 40.4 14.92 0.35 Ed: Secondary or More 0.99 0.98 0.99 0.12 0.38 Economic Strata: 1 0.07 0.08 0.08 0.27 0.36 Economic Strata: 2 0.32 0.31 0.31 0.46 0.39 19 Economic Strata: 3 0.41 0.4 0.4 0.49 0.67 Economic Strata: 4 0.14 0.15 0.15 0.35 0.39 Economic Strata: 5 0.04 0.05 0.05 0.21 0.21 Economic Strata: 6 0.02 0.01 0.02 0.12 0.10 Religious [Yes =1] 0.85 0.82 0.84 0.37 0.12 Ideology (1 Left - 10 Right) 5.37 5.41 5.39 1.92 0.58 Joint orthogonality test p-value 0.15 Sample Size 1,160 1,228 2,388 Notes: The joint orthogonality test is a joint significance test of all the covariates from a probit of the treatment status in all the independent covariates listed in the table. Table (III) Priming Effects On Altruism and Attitudes Towards Migrants Panel A: Main Estimates (All outcomes in zscores) (1) (2) (3) (4) (5) Altruism Policy Altruism Opinion on Effort Opinion on Economy Opinion on Taxes Treatment (Priming = 1) -0.036 -0.021 -0.071** -0.067** -0.188*** (0.036) (0.035) (0.035) (0.034) (0.034) R-squared 0.000 0.000 0.001 0.001 0.009 Observations 3131 3384 3346 3360 3413 Panel B: Heterogeneous Effect - Gender Priming = 1 × Male -0.057 0.038 0.087 -0.097 -0.028 (0.072) (0.070) (0.070) (0.068) (0.068) Treatment (Priming = 1) -0.006 -0.040 -0.119** -0.013 -0.172*** (0.053) (0.051) (0.051) (0.050) (0.049) Male -0.051 0.217*** 0.045 0.373*** 0.041 (0.051) (0.049) (0.049) (0.048) (0.048) Marginal Effects [Male = 1] -0.063 -0.002 -0.032 -0.110∗∗ -0.200∗∗∗ ( 0.049) ( 0.048) ( 0.048) ( 0.046) ( 0.046) R-squared 0.002 0.014 0.004 0.028 0.009 20 Observations 3,117 3,370 3,332 3,346 3,399 Panel C: Heterogeneous Effect - Age (Impressionable Years) Priming = 1 × Pop. Aged 18-25 = 1 0.237*** 0.048 -0.065 -0.051 0.140* (0.089) (0.087) (0.087) (0.086) (0.083) Treatment (Priming = 1) -0.084** -0.028 -0.057 -0.055 -0.210*** (0.040) (0.039) (0.039) (0.038) (0.037) Pop. Aged 18-25 = 1 0.011 0.163*** 0.160*** 0.214*** 0.335*** (0.061) (0.061) (0.060) (0.059) (0.058) Marginal Effects [Pop. Aged 18-25 = 1] 0.153∗ 0.020 -0.121 -0.105 -0.070 ( 0.079) ( 0.078) ( 0.078) ( 0.077) ( 0.075) R-squared 0.005 0.006 0.004 0.007 0.036 Observations 3,131 3,384 3,346 3,360 3,413 Notes: The dependent variable represents the outcome for individual as measured in the survey. It was standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to the COVID-19 priming. The table presents the estimates of a simple OLS regression of the dependent variable on the treatment status. For the heterogeneous effects, we interact the variable on Treatment (Priming=1) with an indicator variable for being male and ages 18 to 25 years old, respectively. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Robust standard errors are reported in parentheses. Figure (I) Timeline of Survey Rounds 6th March 2020   At the end of October 2020   First case of COVID-19 in Public announcement of successful trial of one of Colombia the COVID-19 vaccines in Colombia First Wave - Survey Round  Second Wave - Survey Round 14th October 2020 – 20th October 2020  13th November 2020 – 20th November 2020 The actual number of completed surveys was 3,413  The actual number of completed surveys in the second wave was 2,915 21 Figure (II) Summary Results 0.10 0.07 0.05 0.06 -0.02 -0.04 -0.07 -0.07 [Covid Priming =1] -0.15 -0.10 -0.05 0.00 22 -0.19 -0.20 -0.25 e * * m r t y s o* sm s fo o m xe on tw trui trui Ef n Ta g g o in in Al Al on Ec on li st st cy on 9 li li on on -1 9 Po ni n ni D -1 pi o pi D O ni O VI VI pi O O O C C Notes: The figure presents the point coefficient estimates of a simple OLS regression of the dependent variable on the treatment status. The point estimates are written in black and the 95 percent confidence intervals are presented in bars. *COVID-19 listing one ranks three crises in the last 10 years (the COVID-19 pandemic, drug trafficking, and internal conflict). It gives a score of 3 to anyone who ranked COVID-19 as the worst crisis, 2 if COVID-19 was the second-worst crisis, and 1 if COVID-19 was listed as the third-worst crisis. ** COVID-19 listing two codes an open-ended question of what was the worst crisis in the last 50 years. It takes a value of 1 if the individual mentioned “COVID”, the “pandemic”, or “Coronavirus” in their written responses. APPENDIX A: SURVEY EXPERIMENT DETAILS Table (A.1) Comparison between survey respondents and Bogot´ a inhabitants Variable Sample DANE Average Std. Deviation Average Std. Deviation Male[=1] 0.53 0.50 0.47 0.50 Age (Age groups) 18 years - 24 years [=1] 0.18 0.007 0.13 0.000 25 years - 34 years [=1] 0.21 0.007 0.24 0.000 35 years - 44 years [=1] 0.21 0.007 0.20 0.000 45 years - 54 years [=1] 0.20 0.007 0.17 0.000 ≥ 55 years [=1] 0.20 0.007 0.25 0.000 Education (Education Level) Nothing [=1] 0.00 0.001 0.04 0.000 Primary [=1] 0.03 0.003 0.15 0.000 Secondary [=1] 0.28 0.008 0.38 0.000 Technical [=1] 0.33 0.008 0.13 0.000 University [=1] 0.24 0.007 0.22 0.000 Postgraduate [=1] 0.12 0.006 0.08 0.000 Economic Strata Strata 1 [=1] 0.09 0.005 0.09 0.000 Strata 2 [=1] 0.35 0.008 0.41 0.000 Strata 3 [=1] 0.39 0.008 0.36 0.000 Strata 4 [=1] 0.12 0.006 0.09 0.000 Strata 5 [=1] 0.04 0.003 0.03 0.000 Strata 6 [=1] 0.01 0.002 0.02 0.000 Notes: We use the CNVP 2018 (National Population and Housing Census 2018) to construct the following variables: male, age and education a, Colombia. To construct the economic strata data, we use the EM 2017 (Multipurpose survey 2017). Both, CNVP 2018 and EM levels for Bogot´ 2017 were designed and implemented by the National Administrative Department of Statistics (DANE) from Colombia. Table (A.2) Order of survey modules in treatment and control groups Treatment Control 1) Sociodemographics 1) Sociodemographics 2) COVID-19 2) Attitudes towards migrants 3) Crisis questions 3) Crisis questions 4) Attitudes towards migrants 4) COVID-19 5) Social Desirability 5) Social Desirability 23 Figure (A.1) Recruiting Facebook Ads 24 Figure (A.2) Location of Survey Respondents 25 APPENDIX B: EXPLORATORY ANALYSIS Table (B.1) Priming Efficacy Dep. Variable (1) (2) COVID-19 listing one COVID-19 listing two Treatment (Priming=1) 0.060** 0.073*** (0.026) (0.012) R-squared 0.002 0.011 Observations 3,278 3,413 Mean Dep. Variable (Control Group) 1.420 0.102 Notes: The COVID-19 listing one variable ranks three crises in the last 10 years (the COVID-19 pandemic, drug trafficking, and internal conflict). It gives a score of 3 to anyone who ranked COVID-19 as the worst crisis, 2 if COVID-19 was the second-worst crisis, and 1 if COVID-19 was listed as the third-worst crisis. The COVID-19 listing two variable codes an open-answer question of what was the worst crisis in the last 50 years. It takes a value of 1 if the individual mentioned COVID, the pandemic, or Coronavirus in their written responses.∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. Table (B.2) Multiple Hypothesis Testing (q-values) 26 (1) (2) (3) (4) (5) Altruism Policy Altruism Opinion on Effort Opinion on Economy Opinion on Taxes Treatment (Priming = 1) -0.036 -0.021 -0.071** -0.067** -0.188*** (0.036) (0.035) (0.035) (0.034) (0.034) [0.186] [0.286] [0.072] [0.072] [0.002] Mean Dep. Variable (Control Group) 0 0 0 0 0 R-squared 0.000 0.000 0.001 0.001 0.009 Observations 3131 3384 3346 3360 3413 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. We present q-values for multiple hypotheses testing correction in brackets. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. Table (B.3) Heterogeneous Effects by Social Desirability Dep. Variable (1) (2) (3) (4) (5) Altruism Policy Altruism Opinion on Effort Opinion on Economy Opinion on Taxes Priming=1 × STD (Social Desirability Index) -0.028 -0.028 -0.067* -0.105*** -0.038 (0.037) (0.035) (0.035) (0.035) (0.034) Treatment (Priming = 1) -0.043 -0.008 -0.055 -0.055 -0.188*** (0.036) (0.035) (0.035) (0.034) (0.034) STD (Social Desirability Index) 0.093*** -0.037 -0.063*** 0.007 -0.008 (0.025) (0.024) (0.024) (0.024) (0.023) R-squared 0.007 0.003 0.011 0.005 0.010 Observations 3100 3347 3311 3324 3376 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. The variable Social Desirability Index is constructed using 4 of the 33 questions of the Crowne and Marlowe (1960) scale. Each question’s answer was assigned a score of 1 or 0 depending on whether the scale identified the answer with someone who wanted to be socially desirable as explained in Crowne and Marlowe (1960). We add the answer of the 4 questions having a scale from 1 (no social desirability) to 4 (maximum social desirability) and standardize it for ease of interpretation. We interact the variable on Treatment 27 (Priming=1) with the standardized variable of Social Desirability Index. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. Table (B.4) Exploring Secondary Outcomes: Priming Effects on Beliefs on the Impacts of Migrants Dep. Variable (1) (2) (3) Opinion on Jobs Opinion on Culture Opinion on Crime Treatment (Priming=1) 0.002 -0.021 0.031 (0.026) (0.026) (0.025) R-squared 0.000 0.000 0.000 Observations 6,174 6,180 6,181 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. Table (B.5) Exploring Secondary Outcomes: Priming Effects on Information Misperceptions Dep. Variable (1) (2) Perception about Size Perception about Education Treatment (Priming=1) 6.225*** -0.391*** (0.885) (0.142) R-squared 0.015 0.003 Observations 3,273 2,901 Av. dep. var before treatment 29.18 6.84 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey of their personal beliefs about the Venezuelans’ share of Colombian population and the Venezuelans’ average years of education. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. Table (B.6) Heterogeneous Effects: Losing job during the pandemic 28 Dep. Variable (1) (2) (3) (4) (5) Altruism Policy Altruism Opinion on Effort Opinion on Economy Opinion on Taxes Priming = 1 × Lose Job = 1 0.137 -0.002 0.003 0.053 -0.031 (0.090) (0.087) (0.087) (0.086) (0.084) Treatment (Priming = 1) -0.146* -0.018 -0.074 -0.109 -0.162** (0.080) (0.078) (0.077) (0.077) (0.075) Because of the pandemic you lost your job -0.048 -0.111* -0.092 -0.198*** -0.062 (0.063) (0.061) (0.060) (0.060) (0.059) Marginal Effects [Lose Job = 1] -0.010 -0.020 -0.071∗ -0.056 -0.193∗∗∗ ( 0.040) ( 0.039) ( 0.039) ( 0.038) ( 0.038) R-squared 0.001 0.002 0.003 0.006 0.010 Observations 3,130 3,383 3,345 3,359 3,412 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses. APPENDIX C. SECOND WAVE RESULTS Table (C.1) Testing Balance Between Groups (Second Wave) Means by Treatment Assigment Priming Group Control Group Overall Sample P-value (1) (2) Mean St. Deviation (1) vs (2) Male [=1] 0.52 0.55 0.54 0.5 0.16 Age 42.8 43.98 43.41 14.14 0.06 Ed: Secondary or More 0.98 0.98 0.98 0.14 0.22 Economic Strata: 1 0.09 0.08 0.08 0.28 0.18 Economic Strata: 2 0.33 0.31 0.32 0.47 0.56 Economic Strata: 3 0.39 0.39 0.39 0.49 0.71 Economic Strata: 4 0.13 0.15 0.14 0.35 0.26 Economic Strata: 5 0.05 0.04 0.04 0.21 0.18 Economic Strata: 6 0.01 0.03 0.02 0.15 0.02 Religious [Yes =1] 0.82 0.82 0.82 0.39 0.98 Ideology (1 Left - 10 Right) 5.34 5.25 5.3 1.9 0.27 Joint orthogonality test p-value 0.01 29 Sample Size 1,012 1,070 2,082 Notes: The joint orthogonality test is a joint significance test of all the covariates from a multinomial logit of the treatment status in all the independent covariates listed in the table Table (C.2) Priming Effects of Altruism and Attitudes Towards Migrants (Second Wave) Dep. Variable (1) (2) (3) (4) (5) Altruism Policy Altruism Opinion on Effort Opinion on Economy Opinion on Taxes Treatment (Priming = 1) -0.034 0.008 -0.037 -0.007 0.009 (0.038) (0.037) (0.037) (0.037) (0.038) R-squared 0.000 0.000 0.000 0.000 0.000 Observations 2710 2892 2876 2874 2916 Notes: The dependent variable Yi represents the outcome for individual i as measured in the survey and standardized using the mean and standard deviation of the control group. The variable Treatment (Priming=1) is the assigned treatment status to COVID priming of the individual. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Standard errors are reported in parentheses.