Civic engagement in the Americas

Scartascini, Carlos G.; Vlaicu, Razvan Working Paper Civic engagement in the Americas IDB Working Paper Series, No. IDB-WP-883 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Scartascini, Carlos G.; Vlaicu, Razvan (2018) : Civic engagement in the Americas, IDB Working Paper Series, No. IDB-WP-883, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0001042 This Version is available at: https://hdl.handle.net/10419/208108 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Civic Engagement in the Americas Carlos Scartascini Razvan Vlaicu IDB WORKING PAPER SERIES Nº IDB-WP-883 March 2018 Department of Research and Chief Economist Inter-American Development Bank March 2018 Civic Engagement in the Americas Carlos Scartascini Razvan Vlaicu Inter-American Development Bank Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Scartascini, Carlos G., 1971Civic engagement in the Americas / Carlos Scartascini, Razvan Vlaicu. p. cm. — (IDB Working Paper Series ; 883) Includes bibliographic references. 1. Political participation-America. 2. Voting-America. 3. Elections-America. I. Vlaicu, Razvan. II. Inter-American Development Bank. Department of Research and Chief Economist. III. Title. IV. Series. IDB-WP-883 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. http://www.iadb.org 2018 Abstract This paper estimates the effect of voting eligibility on civic engagement measured along three dimensions: political motivation, political activities, and political knowledge. These outcomes originate in the AmericasBarometer 2004-2016 surveys of eligible voters. To identify the effects the paper exploits variation in field survey dates relative to election dates, given country-specific voting age laws. It is found that voter enfranchisement increases self-reported interest in politics, political socialization, and attendance of political meetings; however, consumption of political news is not statistically different between eligible and non-eligible citizens. Some evidence indicates that the political activities voters engage in translate into increased political knowledge, in contrast with the view that voters are rationally ignorant. The effects are larger in countries with enforced mandatory voting. JEL classifications: D12, D72, D83, O12, O17 Keywords: Civic engagement, Enfranchisement, Elections, Voters 1 Introduction The success of public policies depends critically on the level of citizen engagement in the policymaking process, e.g., through expressing preferences, or through selecting and monitoring policymakers. The research based on random assignment of information to voters in developing countries …nds that the governance of public programs can be improved when voters are well-informed and actively participate in the policymaking process (Pande, 2011). To what extent do citizens choose to exert e¤ort to engage in the policymaking process? And, what factors can stimulate civic engagement? The theoretical literature on whether voters have incentives to acquire information, or whether instead they choose to remain "rationally ignorant" (Downs, 1957), has a long tradition in political economy. The empirical literature addressing this question has disproportionately focused on information consumption through the media. Less is known about the extent of citizen engagement through other modalities, such as talking about politics with family and friends, or attending meetings of political parties. Similarly, while the literature has focused on whether turning out to vote is associated with increased engagement, less attention has been given to whether enfranchisement itself provides a source of motivation for citizens to become politically engaged.1 In this paper we provide evidence on the level of civic engagement in three dimensions: political motivation, political activities, and political knowledge. The main empirical challenge to estimating the civic engagement of eligible voters is …nding an adequate couterfactural that captures how these individuals would have behaved if ineligible to vote. To test how enfranchisement motivates voters to engage in political activities, and whether this translates into political knowledege, we focus on newly enfranchised voters surveyed after general elections. Some were eligible to vote in the most recent election, while others were not. We employ data from the AmericasBarometer surveys covering 34 countries biannually between 2004 and 2016. We focus on the 62,005 young voters in the surveys, ages 16-25, and use countries’minimum voting age requirements and election dates to determine who was eligible to vote in the most recent presidential election. Minimum voting age requirements provide a credibly exogenous source of variation that allows us to identify the e¤ects of voter enfranchisement on a variety of outcomes. The civic outcomes include political interest, self-reported understanding of political issues, talking about politics, news consumption, 1It is not clear that citizens who abstain from voting are always less engaged politically, as some may become alienated once they understand that none of the candidates are representing their preferences. 2 attendance at political meetings, and basic knowledge of the political process. We employ three empirical designs to estimate civic engagement. First, we estimate regressions where individual-level civic outcomes depend on voting eligibility status, controlling for age at interview, income, and country and year …xed e¤ects. Second, we restrict the sample to the youngest eligible and the oldest non-eligible, i.e., the age groups closest to the age eligibility cuto¤; we call this subsample the cuto¤ sample. Third, we approximate age on election day using age at interview and the time between the election date and the interview date. This is an approximation because the surveys only elicit age in years, so actual birthdays are not reported. Then we use this approximate age at election as the running variable in a regression discontinuity design. The results show that voter enfranchisement increases individuals’self-reported interest in politics, an e¤ect that appears in all three empirical strategies and holds up to a battery of robustness tests. Additionally, we …nd consistent evidence that newly enfranchised voters also report talking about politics more frequently and attending political meetings more regularly than their non-eligible peers, while there is also some evidence that this translates into increased political knowledge. As the surveys were conducted at varying lengths of time after a general election, we exploit this feature of the data to study the dynamic element of becoming eligibile. As expected, the gap in civic engagement between eligible and noneligible is greatest at the time of the most recent election. The gap shrinks as the previously ineligible respondents become eligible, and closes about four-…ve years on average after the election has passed. This convergence is due to the previously ineligible group increasing their engagement levels to match the eligible group by the time the next election takes place. In an ideal democracy, a citizen’s right to vote motivates political interest, which leads the citizen to acquire information, which is translated into knowledge and used in making decisions at the ballot box. Yet it is not clear why citizens take this responsibility seriously, given the seemingly low payo¤ to being a good citizen, and the resulting collective action problem. Exploiting the dual voluntary/compulsory voting rules in Brazil’s electoral system where voting eligibility starts at age 16 but voting becomes mandatory at age 18, Lopez de Leon and Rizzi (2014) …nd that compulsory voting laws increase turnout among the eligible. However, those who vote have the same political knowledge as those who do not. The same is true for the intention-to-treat estimate, which measures the e¤ect of mandatory voting on those whose decision to turn out is changed by the mandatory voting law. Using a similar 3 empirical strategy but di¤erent outcomes, Bruce and Costa Lima (2017) show that Brazil’s mandatory voting rules make citizens on average 13.6 percentage points more likely to watch the country’s main television newscast. This e¤ect, which is particularly strong among lowincome individuals, implies that mandatory voting encourages people to acquire information on issues that could be relevant to their voting decisions, though the study is unable to evaluate whether they ultimately become more knowledgeable. Our paper complements the aforementioned work that exploits compulsory voting laws by focusing on the e¤ect of voter enfranchisement rather than the e¤ect of actually voting. While we …nd no evidence supporting the …nding that enfranchisement increases news consumption, our results do point to a number of other potential outlets for increased political interest, and they provide some of the …rst evidence that enfranchisement may lead to a di¤erence in political knowledge between eligible and non-eligible citizens. This stands in contrast to a study in Japan, which used birthdates and an online survey taken shortly after the 2016 House of Councillors election, that found no di¤erence between the attitudes of respondents just below and just above the voting age cuto¤ of 18 in terms of their political interest and trust in government (Horiuchi, Katsumata and Woodard, 2017). Our results also suggest that while voter enfranchisement in voluntary voting systems has a motivating e¤ect on voters, stronger e¤ects may be seen in countries with enforced mandatory voting laws. This provides support for the Lijphart (1997) hypothesis that compulsory voting improves a country’s civic culture. In the next section, we discuss our data and empirical strategy. In Section 3, we present our main results. Section 4 shows the results of a series of robustness tests, and Section 5 explores the mechanisms behind the main e¤ects, using interactions between eligibility and country-level variables such as mandatory voting laws and freedom of the press. We conclude in Section 6. 2 Data and Empirical Strategy The main data for this study comes from the AmericasBarometer, a survey administered by the Latin American Public Opinion Project (LAPOP) at Vanderbilt University in countries across North, Central, and South America, and much of the Caribbean. The survey has been conducted in waves every two years from 2004 to 2016, and coverage has grown from 11 4 countries in 2004 to 29 in 2016; in total, the survey has been applied in 34 countries over this period. Table 1 shows the country and wave coverage of our data. For each wave, the sample is designed to be representative of each country and uses a multi-stage probability design, stratifying by region within the country, sub-stratifying each region by municipality size and urban/rural areas within the municipality, and selecting households in clusters. LAPOP then uses frequency matching to obtain a sample with similar age and gender distributions to the national census, interviewing only respondents who are of legal age to vote. With few exceptions, respondents are interviewed in their homes and data are collected with handheld electronic devices. Restricting our sample to respondents between 16 and 25 years old, our …nal dataset contains 62,005 observations from 34 countries, organized as a repeated cross-section of a group of countries that grows with each subsequent wave. For each respondent, the data include the date of interview, as well as socioeconomic information such as years of schooling, ethnicity, sex, age, marital status, and information on the area where they live. Most LAPOP questions focus on public opinion topics, such as trust in institutions, perceptions of crime and the economy, and political participation. For our project we use questions related to political motivation, political activities, and political knowledge. Table 1 contains a description of the variables in our dataset. Our research question focuses on whether citizens become more engaged in politics when they become eligible to vote. To determine voting eligibility, we need to know not just how old respondents are at the time of interview, but also how old they were on the date of the most recent election. LAPOP does not include respondents’dates of birth, requiring a strategy to infer a respondent’s age at the last election. Using dates from presidential or general elections (for parliamentary systems) from the Institute for Democracy and Electoral Assistance (IDEA) and the Inter-Parliamentary Union, we created a database on elections between 1999-2016 in our sample of countries, merged this with the LAPOP data, and determined the most recent election at the time of each interview. Then we generated a new variable called Scoreict as follows: Scoreict =InterviewAgeict Gapict MinAgec(1) where InterviewAgeict is the age at interview, for respondent i, living in country c, inter5 (Lopez de Leon and Rizzi, 2014) and Japan (Horiuchi, Katsumata and Woodard, 2017). The results suggest a story where political interest increases in newly enfranchised voters compared to their ineligible peers, perhaps because for the …rst time they have both a stake and a voice in national politics. This increased interest may translate into a higher degree of socialization, as newly enfranchised voters share their increased interest and excitement about politics with those around them in politically-tinged conversations. This interest may also lead them to attend meetings of political movements or parties for the …rst time as they attempt to learn more about the political process and gain a greater voice. Importantly, the results also suggest that the e¤ect is strongest as of the time of the last election. As the next election approaches, previously ineligible voters become engaged, closing the gap with eligible voters. 4 Robustness In this section we present a series of checks to show the robustness of our results to di¤erent sample compositions, control variables, model speci…cations, eligibility de…nitions, and eligibility cuto¤ thresholds. 4.1 Country Composition Recall that our sample includes 34 countries, spanning a diverse group that includes large Latin American countries like Mexico, Brazil, and Colombia, smaller countries in Central America, English-speaking Caribbean islands, the United States, and Canada. Countries were also added progressively, with the original 2004 wave of the survey including Mexico, Central America and some of South America, while most South American countries were included in 2006, and the United States, Canada, and Caribbean islands were included progressively starting in 2006. To show that our results are not driven by any one speci…c country and robustness to di¤erent samples, we provide evidence from two robustness tests. First, we run our baseline regressions separately for each country and plot the coe¢ cients to show that outliers are not driving the results. Second, we exclude the United States, Canada, and English-speaking Caribbean from the sample to show robustness to a di¤erent composition of countries. Figure 1 shows the coe¢ cient distributions from country-by-country regressions of the 12 baseline speci…cation with the full sample. As shown, no one country appears to be driving the results. The distribution for political interest is centered slightly to the right of zero, as expected from the positive coe¢ cient seen in the baseline regressions. To the extent that extreme outliers are seen, these tend to fall on the negative end of the distribution and represent small Caribbean islands that appear in only one wave (2016) and have relatively few observations. Table 8 presents the results for the restricted sample regressions. Odd-numbered columns present the results using the full sample, but excluding the United States and Canada, while even-numbered columns show the results excluding the United States, Canada, and Englishspeaking Caribbean countries. Few observations are lost from the baseline when we exclude the USA and Canada, re‡ecting the fact that while these are large countries, LAPOP does not interview more of their citizens than it does for other countries, and neither country was included in the the original 2004 wave. A greater proportion of observations is dropped when we also exclude the English-speaking Caribbean, resulting in a 10-20% reduction in observations from the full sample speci…cation, depending on the outcome variable. Results do not vary signi…cantly between speci…cations and are similar to the dynamic results in Table 5; while the estimate is somewhat smaller for political interest in the restricted sample, the estimate for socialization is marginally greater in magnitude, and the estimate for knowledge increases slightly and becomes marginally signi…cant in the second restriction. The interaction between eligibility and interview gap continues to go in the expected direction and is little changed in terms of magnitude or signi…cance. 4.2 Control Variables Table 9 shows the results for the full sample dynamic regressions, with variations on the set of included control variables. In odd-numbered columns, we include three additional controls - a dummy for sex, equal to one if the respondent is male, years of schooling,2and a dummy equal to one if the respondent has never been married and is not cohabitating with a domestic partner. While the estimates for eligibility decline slightly compared to the baseline regressions, they remain comparable in magnitude and retain signi…cance. Estimates for the dynamic elements are also qualitatively similar to those in our baseline regressions. With 2Education was excluded from our baseline regressions because of concerns about endogeneity, i.e., education may respond to becoming eligible to vote. 13 respect to the new control variables, we observe a positive, statistically signi…cant correlation between males and all of our outcomes of interest, with the exception of political knowledge. Similarly, respondents with more education, as expected, appear more politically engaged, interested in politics, and knowledgeable, while the relationship between being single and the outcomes of interest is less consistent. In the even-numbered columns, we present results analagous to the original full sample dynamic regressions, where income percentile has been replaced by an alternative measure of …nancial wellbeing. The measure is derived from a question asking respondents about their family’s economic situation, and takes four possible values ranging from "Not enough income, and having a hard time" (1) to "Good enough and can save" (4). In general, the results are qualitatively similar to those in the odd-numbered columns of the table and to those in the full sample dynamic regressions in Table 5, with the most obvious di¤erence being the marginally signi…cant positive result for the e¤ect of eligibility on political knowledge seen in Column 12. On the whole, we believe the estimates in Table 9 con…rm the robustness of our main results and assuage concerns that the exclusion of education in our main set of control variables impacts the estimates on our coe¢ cients of interest. 4.3 Naive Eligibility De…nition As discussed in the Data and Empirical Strategy section, the nature of the data creates some ambiguity as to a respondent’s true age at the time of the last election, resulting in a group of observations for whom eligibility status at the last election cannot be determined with certainty, constituting about 11.5% of the sample. Here, we present two alternative, "naive" eligibility de…nitions, which allows us to show the robustness of our main results to the inclusion of this group of observations. For our …rst naive strategy, we de…ne a new eligibility variable as follows: Eligibility0 ict =1if InterviewAgeict GapY earict MinAgec 0if InterviewAgeict GapY earict < MinAgec (6) where GapY earict =InterviewY earict ElectionY earct:To return to our hypothetical Mexican respondent examples from the Data and Empirical Strategy section, respondents i= 1;2;3would continue to be classi…ed as before. However, respondent i= 4, previously assigned to the ambiguous-eligibility group and assigned a missing value for eligibility, is 14 now assigned an eligibility value. Their age at interview, 20, is subtracted by 2 (2008 minus 2006), which is equal to the minimum voting age of 18; this naive strategy assigns them to the eligibile group. In our second strategy, we employ the Scoreict de…ned previously. Recall that this variable takes an individual’s age at interview and compares it to the minimum voting age, such that aScore of zero means the individual was exactly the voting age, while values above (below) zero refer to years above (below) the minimum voting age at the time of the last election. Recall also that this variable can take decimal values, re‡ecting the degree of uncertainty about the exact age of the respondent at the last election, and that the ambiguous group always falls between 1< Scoreict <0(i.e., there is some degree of uncertainty whether they were exactly the cuto¤ age for voting or one year below the cuto¤). Hypothetical Mexican respondent i= 4 had a Scoreict =0:06, meaning there is a 6% chance they were 17 at the last election (i.e., ineligible) and a 94% chance they were 18 (i.e., eligible), assuming a uniform distribution of birthdays throughout the year. Thus, we de…ne this alternative eligibilty measure as: Eligibility00 ict =1if Scoreict  0:5 0if Scoreict <0:5(7) Results are presented in Table 10, with oddand even-numbered columns showing results from the …rst and second de…nitions, respectively. The results are qualitatively similar to those from the full sample dynamic regressions in Table 5, though somewhat smaller in magnitude. This suggests an attenuation bias due to measurement error, which would be expected due to the fact that these strategies erroneously assign some non-eligible observations to the eligible group, and vice versa. 4.4 Ordered Probit As described previously, our outcome variables are measured on ordinal scales, with the exception of political knowledge, which is a dummy. Here we present the results of ordered probit regressions, showing the robustness of the OLS results to a logistic regression model. Results are shown in Table 11 for the baseline and dynamic speci…cations with the full sample. The strategy con…rms the results from Table 5, and we note the addition of a positive, statistically signi…cant result for the e¤ect on knowledge in the dynamic version of the ordered probit regression. 15 4.5 Eligibility Threshold Falsi…cation Test We conduct a falsi…cation by manipulating the minimum voting age threshold. Speci…cally, we re-run the dynamic regressions for the full sample, de…ning eligibility at two years below the true cuto¤, one year below, at the true cuto¤, one year above the true cuto¤, and two years above. Thus, at the true cuto¤ the regression results would be the same as those presented in the dynamic regressions in Table 5, while the manipulated cuto¤s provide placebo treatment and control groups. If the estimates for eligibility truly re‡ect the e¤ect of being eligible to vote on political engagement, we would expect to …nd two patterns: i) the e¤ect of eligibility should only be signi…cant at the true cuto¤, and ii) plotting the estimated coe¢ cients on a graph where the x-axis ranges from two years below to two years above the true cuto¤ and the y-axis measures the estimated coe¢ cient, we should see an inverted U shape, representing a greater e¤ect at the true cuto¤ and declining estimated e¤ects as the manipulated cuto¤ gets further from the true cuto¤. The results from this test are plotted in Figure 2. Overall, the results suggest the use of the correct cuto¤. For political interest, the plot shows the predicted inverted U shape, though the estimates are only signi…cant at the 95% con…dence interval both one year above and below the true cuto¤. For attendance at political meetings, the e¤ect is only signi…cant at the 95% con…dence interval for the estimate at the true cuto¤ and the plot somewhat resembles an inverted U, though several of the placebo estimates are signi…cant at the 90% con…dence interval. However, socialization performs somewhat poorly on the falsi…cation test, with a shape that does not resemble an inverted U and estimates signi…cant at the 95% con…dence interval for placebo treatments two years below and one year below the true cuto¤. An important limitation to these tests is that the treatment-control split becomes even more heavily skewed when we de…ne the cuto¤ below its true value. This is because LAPOP only interviews voting-age respondents. As the voting age is arti…cially set below the true threshold, more 16-25 year-old respondents at the time of survey end up in the eligible group for the most recent election; in the case of political interest, for example, 95.3% of the sample is eligible when we de…ne the threshold two years below the true cuto¤, and 88.7% is eligible when the threshold is set one year below the true cuto¤, compared to 77.6% when the threshold is set at the true cuto¤. This contributes to the wider con…dence intervals on the left-hand side of the true threshold in Figure 2, and may explain in some measure the 16 poorer performance of the falsi…cation tests on this side of the cuto¤. 4.6 Regression Discontinuity with Varying Slopes Next, we modify the regression discontinuity in equation (4) to allow the slope on the Scoreict term to vary on either end of the eligibility cuto¤ threshold. The modi…ed equation is as follows: Yict =0+1Eligibleict +2Scoreict +3Eligibleict Scoreict +0Xict +c+t+"ict (8) Results for both static and dynamic versions of this speci…cation are presented in Table 12. The interaction between Score and eligibility, determining the di¤erence in slope between the eligible and ineligible groups, tends to be close to zero and is consistently insigni…cant, with the exception of a marginally signi…cant, negative di¤erence in the dynamic version of the regression for knowledge. Overall, this suggests little di¤erence between the slope of the RD running variable for the two groups. For the estimates on eligibility, results are similar to the regression discontinuity results in Table 7, with a positive, signi…cant result for political interest and socialization in the static version, as well as for interest, socialization, and meetings in the dynamic version. However, the marginally signi…cant result for meetings in the static version of the original RD speci…cation is no longer signi…cant in the varying-slope version, though the estimate is moderately larger. 5 Channels In this section we use interactions with country-level characteristics/institutions to test hypotheses regarding the channels through which acquiring the right to vote may lead to increased political engagement. First, we examine whether voters become increasingly engaged in politics because they want to make an informed, optimal decision with their vote. If this is the case, the e¤ect should be stronger in countries with mandatory voting laws, where newly enfranchised voters not only have the opportunity to vote for the …rst time, but also face some form of obligation to exercise their voting rights. Second, we expect that the presence of a free and independent press provides additional incentives for newly enfranchised voters to become informed by o¤ering them information of higher quality; we 17 examine this by interacting a measure of freedom of the press with voting eligibility. 5.1 Mandatory Voting Although a number of countries in Latin America and the Caribbean have mandatory voting laws, these laws are only enforced and punished through some form of sanction in Ecuador, Peru, Uruguay, Brazil, and Argentina, as well as in Chile until 2012. Such policies have been found to be e¤ective at increasing turnout in elections, particularly among citizens with lower levels of education (Jaitman, 2013). It is possible that, by inducing registered voters to get to the polls, the e¤ect of voter enfranchisement on political engagement will be stronger in these countries. This would provide evidence that it is not simply being eligible to vote that results in increased engagement, but rather that engagement is an active part of voters taking their voting obligations seriously, and preparing to vote by informing themselves on candidates and policy issues. On the other hand, increased engagement could result from the act of voting, as voters feel a greater sense of civic duty after casting their vote and pay closer attention to how politicians they voted for (or against) are performing (Lijphart, 1997). To test this, we take data on the enforcement of mandatory voting laws for the countries in our sample from IDEA. We de…ne a dummy variable called V oluntary, equal to one if no mandatory law is enforced, and equal to zero otherwise (i.e., the aforementioned group of six countries)3. We then interact eligibility with the V oluntary dummy, expecting a negative coe¢ cient re‡ecting a more moderate e¤ect for eligibility in countries without enforced mandatory voting laws. The results from the static model are shown in Table 13 in even and odd-numbered columns for the full and cuto¤ samples, respectively. The coe¢ cient for the e¤ect of voting eligibility on political interest is substantially larger than in the baseline speci…cation, suggesting the e¤ect is being driven by newly enfranchised voters in countries with enforced mandatory voting. Consequently, the interaction term consistently shows a negative coef- …cient, suggesting an attenuated e¤ect in countries without enforced mandatory voting; in the cases of political interest and socialization, the magnitude of this e¤ect is such that it essentially cancels out the e¤ect of voting eligibility. We also note that the coe¢ cient for the 3Note, however, that while all of these countries have enforced mandatory voting, in Ecuador, Brazil, and Argentina it is mandatory starting at age 18, but not at the minimum voting age in those countries (16). 18 e¤ect on knowledge remains positive and signi…cant, as in the baseline regressions, and that the estimate for the e¤ect on understanding is positive and now marginally signi…cant. While the estimates for the e¤ect of V oluntary are inconsistent in direction, it is worth pointing out that Chile is the only country with variation in its mandatory voting laws over the period of study; this is also why the coe¢ cient cannot be estimated in the political socialization regression, as the 2006-2010 period for which this item was included in the survey does not cover Chile’s 2012 transition to a voluntary voting regime. 5.2 Free Press Free, independent, quality journalism is necessary in a well-functioning democracy. Voters rely on the media to highlight important policy issues, frame the political debate, and provide basic information on when events like elections are taking place and identify candidates to watch. The presence of local Spanish-language news programming has been found to increase voter turnout among Latinos in the United States, suggesting the mere presence of relevant media in a format viewers can understand increases engagement and participation in politics (Oberholzer-Gee and Waldfogel, 2009). In a randomized experiment, Gerber, Karlan and Bergan (2009) …nd that exposure to media in‡uences voters’ support for particular candidates, though they …nd no e¤ect on knowledge of political events, political opinions, or voter turnout. How the quality and indepence of news media might impact political engagement is less obvious. In a number of countries in Latin America, including Mexico and Brazil, journalists face violence by criminal groups (Comittee to Protect Journalists, 2016), resulting in a di¢ cult media environment and likely leading to self-censorship for fear of reprisal. In such cases, voters may become disengaged with politics as the media focuses on softer stories, and motivated voters become frustrated by the lack of quality information and their inability to identify the perpetrators of media opression. In other countries, such as Venezuela, increasing invervention from the government into the media market, coupled with obstruction from security forces and prosecutions for defamation have led to the deterioration of the freedom of the press in recent years (Freedom House, 2017); yet the resulting political polarization has arguably led to increased political engagement as citizens mobilize in support or opposition to the government’s actions. To test whether voter enfranchisement has an increased impact on political engagement 19 in countries with a freer press, we employ data from the Freedom House Freedom of the Press Index (FPI). We take the median FPI for each year, and de…ne a CapturedPress dummy variable equal to one if the respondent lived in a country equal to or above the median (i.e., less free) and equal to zero if the country had an FPI value below the median (i.e., more free) at the time of the interview. If living in a country where the media faces more hostility provides disincentives to becoming politically engaged, because information is lower-quality and less useful for making informed voting decisions, we would expect the coe¢ cient on the interaction term to be negative. As in the case of mandatory voting laws, we estimate the regression for the full and cuto¤ samples. Results are presented in Table 14 in even and odd-numbered columns for the full and cuto¤ samples, respectively. The estimates for eligibility are similar to the results from our baseline speci…cation presented in Tables 5 and 6. The direction of the estimate for the interaction between eligibility and captured press is not always consistent and shows little statistical signi…cance. On the other hand, the estimate for the e¤ect of living in a more oppressive media environment (CapturedPress = 1) is positive and marginally signi…cant for the e¤ect on political interest and socialization in the cuto¤ sample regressions. Perhaps unexpectedly, the estimate is positive and marginally signi…cant for news consumption in both samples (p=:101 for the cuto¤ sample). CapturedPress appears to have no impact on political knowledge or attendance at political meetings, and may have a negative e¤ect on self-reported understanding of political issues. This suggests not a correlation between oppression of the media and self-reported political interest, but rather an environment where it becomes more di¢ cult to convert that interest into action. Consequently, while the average citizen might be more interested in politics in a more oppressive media environment, no disproportionate increase in interest occurs in such countries when citizens become newly enfranchised voters, perhaps again because they feel it is di¢ cult to convert interest into actionable change. 6 Conclusions This paper provides evidence on the level of civic engagement of young voters in the Americas using variation in eligibility provided by minimum voting age rules. Civic engagement is measured along three dimensions: political motivation, political activities, and political 20 knowledge. These measures are based on surveys taken between elections. The three empirical strategies employed have similar implications: enfranchisement is associated with more political interest; this manifests itself in more political socialization and more frequent attendance of political meetings; these activities translate into better political knowledge, although this latter e¤ect is somewhat small. Noticeably, increased political interest is not accompanied by more consumption of political news. We also …nd that non-eligible voters close the engagement gap by the time they become eligible in the next election. Finally, we …nd some evidence that civic engagement is stronger under mandatory voting laws. 21 Table 5: Full Sample Results Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.061** 0.188*** 0.070 0.128 -0.045 0.270** -0.001 0.037 0.018 0.085** 0.046*** 0.038 (0.024) (0.06) (0.043) (0.105) (0.054) (0.100) (0.034) (0.072) (0.018) (0.039) (0.012) (0.027) Eligibility Gap -0.046*** -0.033 -0.094*** -0.021 -0.022* -0.011 (0.017) (0.029) (0.030) (0.019) (0.013) (0.007) Gap 0.031 0.004 0.105*** 0.003 0.018 -0.012 (0.019) (0.028) (0.031) (0.019) (0.013) (0.009) Age -0.009** -0.007** -0.008 -0.004 0.021*** -0.019** 0.024*** 0.027*** 0.004* 0.005** 0.001 0.005*** (0.004) (0.004) (0.007) (0.006) (0.008) (0.008) (0.005) (0.005) (0.002) (0.002) (0.002) (0.001) Income Percentile 0.266*** 0.266*** 0.538*** 0.538*** 0.598*** 0.597*** 0.342*** 0.343*** -0.012 -0.013 0.077*** 0.077*** (0.023) (0.023) (0.039) (0.039) (0.047) (0.047) (0.030) (0.030) (0.013) (0.013) (0.010) (0.010) Observations 40136 40136 33059 33059 13281 13281 26995 26995 40069 40069 30714 30714 Eligible (%) 77.6 77.6 76.9 76.9 77.8 77.8 77.4 77.4 77.0 77.0 78.4 78.4 Countries 34 34 34 34 23 23 34 34 26 26 27 27 Survey Period 2006-2016 2006-2016 2008-2016 2008-2016 2006-2010 2006-2010 2010-2016 2010-2016 2004-2016 2004-2016 2004-2016 2004-2016 Note: Regressions with country and year …xed e¤ects. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 28 Table 6: Cuto¤ Sample Results Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.056** 0.438*** 0.086 0.243 -0.042 0.346 -0.046 -0.028 0.003 0.215*** 0.027** 0.025 (0.025) (0.111) (0.058) (0.255) (0.051) (0.235) (0.045) (0.164) (0.022) (0.072) (0.011) (0.055) Eligibility Gap -0.046** -0.048 -0.032 0.000 -0.011 -0.010 (0.020) (0.033) (0.030) (0.029) (0.015) (0.008) Gap 0.161*** 0.047 0.170 0.009 0.096*** -0.007 (0.052) (0.112) (0.114) (0.082) (0.033) (0.024) Age -0.008 -0.136** -0.026 -0.038 -0.081*** -0.075 0.027 0.018 0.015 -0.076** -0.011** 0.005 (0.010) (0.052) (0.025) (0.116) (0.019) (0.125) (0.018) (0.080) (0.009) (0.034) (0.005) (0.024) Income Percentile 0.232*** 0.233*** 0.555*** 0.556*** 0.475*** 0.477*** 0.315*** 0.315*** -0.035* -0.035* 0.074*** 0.074*** (0.041) (0.041) (0.073) (0.073) (0.089) (0.089) (0.050) (0.050) (0.020) (0.020) (0.016) (0.016) Observations 9294 9294 7835 7835 2939 2939 6372 6372 9301 9301 6959 6959 Eligible (%) 58.3 58.3 57.5 57.5 58.6 58.6 58.2 58.2 57.4 57.4 58.3 58.3 Countries 34 34 34 34 23 23 34 34 26 26 27 27 Survey Period 2006-2016 2006-2016 2008-2016 2008-2016 2006-2010 2006-2010 2010-2016 2010-2016 2004-2016 2004-2016 2004-2016 2004-2016 Note: Regressions with country and year …xed e¤ects. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 29 Table 7: Modi…ed Regression Discontinuity Results Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.090*** 0.226*** 0.043 0.125 0.177*** 0.327*** -0.018 0.045 0.036* 0.083* -0.006 0.031 (0.033) (0.060) (0.063) (0.112) (0.053) (0.093) (0.053) (0.083) (0.021) (0.044) (0.014) (0.026) Eligibility Gap -0.048*** -0.030 -0.060** -0.026 -0.017 -0.014* (0.017) (0.030) (0.026) (0.022) (0.013) (0.007) Gap 0.169** -0.058 -0.228 -0.399 0.010 -0.030 (0.067) (0.171) (0.354) (0.250) (0.081) (0.040) Score -0.013 0.117* 0.029 -0.052 -0.089*** -0.360 0.002 -0.415* -0.009 -0.012 0.022*** -0.018 (0.013) (0.069) (0.022) (0.178) (0.021) (0.349) (0.018) (0.250) (0.010) (0.084) (0.006) (0.041) Age -0.004 -0.132* -0.037** 0.046 0.067*** 0.343 0.025* 0.446* 0.012 0.016 -0.014*** 0.027 (0.010) (0.069) (0.017) (0.175) (0.017) (0.347) (0.013) (0.247) (0.008) (0.083) (0.005) (0.041) Income Percentile 0.222*** 0.222*** 0.465*** 0.465*** 0.578*** 0.581*** 0.336*** 0.336*** -0.020 -0.020 0.079*** 0.079*** (0.029) (0.029) (0.050) (0.050) (0.060) (0.060) (0.036) (0.036) (0.016) (0.016) (0.012) (0.012) Observations 23051 23051 19240 19240 7438 7438 15692 15692 22990 22990 17210 17210 Eligible (%) 69.5 69.5 68.9 68.9 68.5 68.5 69.9 69.9 68.5 68.5 70.1 70.1 Countries 34 34 34 34 23 23 34 34 26 26 27 27 Survey Period 2006-2016 2006-2016 2008-2016 2008-2016 2006-2010 2006-2010 2010-2016 2010-2016 2004-2016 2004-2016 2004-2016 2004-2016 Note: Regressions with country and year …xed e¤ects. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 30 Table 8: Restricted Sample Regressions Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.186*** 0.147** 0.134 0.085 0.270** 0.277** 0.045 0.045 0.086** 0.075* 0.037 0.049* (0.060) (0.065) (0.106) (0.120) (0.100) (0.105) (0.073) (0.080) (0.040) (0.042) (0.027) (0.029) Eligibility Gap -0.047*** -0.034* -0.036 -0.018 -0.094*** -0.093*** -0.023 -0.024 -0.022* -0.018 -0.011 -0.014* (0.018) (0.019) (0.029) (0.031) (0.030) (0.030) (0.020) (0.022) (0.013) (0.013) (0.007) (0.008) Gap 0.030 0.026 0.004 -0.001 0.105*** 0.111*** 0.005 0.010 0.018 0.018 -0.012 -0.009 (0.019) (0.021) (0.028) (0.031) (0.031) (0.031) (0.019) (0.020) (0.013) (0.014) (0.009) (0.009) Age -0.007* -0.008* -0.004 -0.003 0.019** 0.017* 0.027*** 0.027*** 0.005** 0.004* 0.006*** 0.005*** (0.004) (0.004) (0.006) (0.007) (0.008) (0.009) (0.005) (0.005) (0.002) (0.002) (0.001) (0.001) Income Percentile 0.268*** 0.281*** 0.546*** 0.591*** 0.597*** 0.642*** 0.341*** 0.346*** -0.012 -0.009 0.077*** 0.083*** (0.024) (0.026) (0.039) (0.042) (0.047) (0.048) (0.031) (0.035) (0.013) (0.014) (0.010) (0.010) Observations 39150 32992 32312 26752 13281 11824 26142 21184 39718 35607 30630 26403 Excluded US,Can English US,Can English US,Can English US,Can English US,Can English US,Can English Note: Regressions with country and year …xed e¤ects. Restriction 1 refers to the full sample, excluding USA and Canada; Restriction 2 excludes USA, Canada, and English-speaking Caribbean. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 31 Table 9: Alternative Controls Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.174*** 0.176*** 0.109 0.136 0.224** 0.253** 0.031 0.035 0.086** 0.080** 0.034 0.048* (0.058) (0.057) (0.109) (0.097) (0.100) (0.099) (0.076) (0.073) (0.039) (0.037) (0.027) (0.025) Eligibility Gap -0.046*** -0.042** -0.036 -0.040 -0.082*** -0.088*** -0.024 -0.016 -0.022* -0.020* -0.011 -0.014** (0.017) (0.017) (0.030) (0.028) (0.029) (0.028) (0.020) (0.020) (0.013) (0.012) (0.007) (0.007) Gap 0.031* 0.028 0.004 0.007 0.103*** 0.099*** 0.005 -0.002 0.019 0.013 -0.011 -0.008 (0.018) (0.018) (0.028) (0.026) (0.029) (0.026) (0.020) (0.021) (0.013) (0.012) (0.008) (0.008) Age -0.007* -0.004 -0.001 0.005 0.020** 0.022** 0.019*** 0.029*** 0.005** 0.006** 0.006*** 0.006*** (0.004) (0.004) (0.006) (0.006) (0.008) (0.009) (0.005) (0.005) (0.002) (0.002) (0.001) (0.001) Income Percentile 0.131*** 0.310*** 0.347*** 0.226*** -0.048*** 0.025** (0.022) (0.039) (0.042) (0.030) (0.013) (0.010) Male 0.098*** 0.218*** 0.225*** 0.042*** 0.059*** 0.003 (0.011) (0.021) (0.026) (0.015) (0.009) (0.004) Education 0.034*** 0.054*** 0.051*** 0.039*** 0.007*** 0.013*** (0.002) (0.004) (0.005) (0.003) (0.001) (0.001) Single - Never married 0.012 0.042* 0.010 -0.120*** 0.001 0.017*** (0.012) (0.023) (0.024) (0.018) (0.008) (0.005) Alt. Income Measure 0.068*** 0.162*** 0.126*** 0.093*** -0.001 0.020*** (0.007) (0.013) (0.013) (0.010) (0.005) (0.004) Observations 38745 43718 31956 37562 13152 13104 25867 30802 39282 41207 30337 32303 Note: Regressions with country and year …xed e¤ects. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1 32 Table 10: Naive Eligibility De…nitions Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.139*** 0.128*** 0.076 0.052 0.149** 0.185** 0.035 -0.019 0.046* 0.073** 0.013 0.042** (0.037) (0.044) (0.068) (0.075) (0.072) (0.078) (0.043) (0.053) (0.024) (0.031) (0.016) (0.018) Eligibility Gap -0.038*** -0.040*** -0.018 -0.012 -0.058** -0.072*** -0.018 -0.005 -0.013 -0.021* -0.006 -0.013** (0.012) (0.014) (0.019) (0.021) (0.023) (0.025) (0.012) (0.016) (0.009) (0.011) (0.005) (0.006) Gap 0.020 0.022 -0.011 -0.016 0.071** 0.084*** 0.000 -0.011 0.010 0.017 -0.018*** -0.011* (0.014) (0.016) (0.019) (0.021) (0.027) (0.028) (0.016) (0.016) (0.010) (0.011) (0.006) (0.007) Age -0.006* -0.002 -0.002 -0.000 0.017** 0.018*** 0.027*** 0.028*** 0.005*** 0.005*** 0.006*** 0.005*** (0.003) (0.003) (0.006) (0.006) (0.007) (0.007) (0.005) (0.004) (0.002) (0.002) (0.001) (0.001) Income Percentile 0.270*** 0.270*** 0.530*** 0.530*** 0.600*** 0.600*** 0.353*** 0.353*** -0.012 -0.012 0.076*** 0.076*** (0.023) (0.023) (0.038) (0.038) (0.045) (0.045) (0.031) (0.031) (0.013) (0.013) (0.010) (0.010) Observations 45339 45339 37245 37245 14991 14991 30481 30481 45310 45310 34636 34636 Strategy 1 2 1 2 1 2 1 2 1 2 1 2 Note: Regressions with country and year …xed e¤ects. Strategy 1 de…nes eligibility by subtracting the di¤erence between interview year and election year from age at interview and comparing to the voting age. Strategy 2 considers observations with Score greater than or equal to -0.5 as eligible. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 33 Table 11: Ordered Probit Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.072*** 0.230*** 0.041 0.078 -0.023 0.309*** -0.011 0.034 0.075 0.245** 0.184*** 0.206** (0.028) (0.070) (0.026) (0.064) (0.056) (0.104) (0.036) (0.073) (0.046) (0.100) (0.050) (0.104) Eligibility Gap -0.056*** -0.020 -0.098*** -0.023 -0.061* -0.067** (0.020) (0.018) (0.029) (0.019) (0.031) (0.028) Gap 0.039* 0.003 0.112*** 0.005 0.041 -0.037 (0.022) (0.017) (0.028) (0.020) (0.031) (0.031) Age -0.010** -0.008** -0.005 -0.002 0.021*** 0.018** 0.030*** 0.032*** 0.010 0.013** 0.007 0.026*** (0.004) (0.004) (0.004) (0.004) (0.008) (0.008) (0.005) (0.005) (0.006) (0.006) (0.007) (0.006) Income Percentile 0.312*** 0.312*** 0.329*** 0.329*** 0.607*** 0.607*** 0.371*** 0.371*** -0.042 -0.042 0.364*** 0.369*** (0.027) (0.027) (0.025) (0.025) (0.045) (0.045) (0.035) (0.034) (0.032) (0.032) (0.048) (0.048) Observations 40136 40136 33059 33059 13281 13281 26995 26995 40069 40069 30714 30714 Note: Regressions with country and year …xed e¤ects, using the full sample. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1 34 Table 12: RD Results With Varying Slopes Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.102** 0.211*** 0.029 0.100 0.228** 0.324*** -0.031 0.019 0.045 0.081* -0.021 0.013 (0.049) (0.062) (0.089) (0.119) (0.086) (0.104) (0.082) (0.093) (0.030) (0.045) (0.021) (0.028) Eligibility Gap -0.051*** -0.034 -0.061** -0.030 -0.018 -0.017** (0.018) (0.030) (0.026) (0.023) (0.014) (0.007) Gap 0.171** -0.055 -0.227 -0.397 0.011 -0.027 (0.067) (0.171) (0.357) (0.251) (0.081) (0.040) Score -0.020 0.131* 0.037 -0.030 -0.119** -0.357 0.010 -0.394 -0.015 -0.010 0.031** -0.001 (0.024) (0.071) (0.045) (0.178) (0.047) (0.361) (0.037) (0.256) (0.016) (0.084) (0.012) (0.041) Eligibility Score 0.009 -0.017 -0.010 -0.027 0.037 -0.003 -0.009 -0.026 0.007 -0.003 -0.010 -0.020* (0.023) (0.023) (0.045) (0.047) (0.048) (0.048) (0.035) (0.033) (0.016) (0.017) (0.012) (0.012) Age -0.004 -0.132* -0.036** 0.047 0.066*** 0.342 0.025* 0.448* 0.012 0.016 -0.014*** 0.028 (0.010) (0.069) (0.017) (0.175) (0.016) (0.347) (0.013) (0.247) (0.008) (0.083) (0.005) (0.041) Income Percentile 0.222*** 0.222*** 0.465*** 0.465*** 0.578*** 0.581*** 0.336*** 0.336*** -0.020 -0.020 0.079*** 0.079*** (0.029) (0.029) (0.050) (0.050) (0.060) (0.060) (0.036) (0.036) (0.016) (0.016) (0.012) (0.012) Observations 23051 23051 19240 19240 7438 7438 15692 15692 22990 22990 17210 17210 Note: Regressions with country and year …xed e¤ects. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1 35 Table 13: Institutional Interactions - Mandatory Voting Laws Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.125** 0.172*** 0.160*** 0.129* 0.132*** -0.020 0.030 -0.018 0.001 0.005 0.046* 0.037* (0.054) (0.046) (0.056) (0.070) (0.044) (0.048) (0.042) (0.065) (0.026) (0.023) (0.023) (0.021) Eligibility Voluntary -0.074 -0.140*** -0.106* -0.054 -0.201*** -0.027 -0.035 -0.030 0.019 -0.003 -0.000 -0.012 (0.056) (0.049) (0.057) (0.076) (0.054) (0.069) (0.041) (0.068) (0.028) (0.025) (0.023) (0.022) Voluntary 0.119 0.232* -0.068 -0.190 0.000 0.000 0.129** 0.283** -0.103** -0.069 0.076** 0.063 (0.108) (0.123) (0.137) (0.316) (.) (.) (0.064) (0.129) (0.043) (0.048) (0.035) (0.045) Observations 40136 9294 33059 7835 13281 2939 26995 6372 40069 9301 30714 6959 Sample Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Note: Regressions with country and year …xed e¤ects, controlling for age at interview and income percentile. All speci…cations use the static model. Voluntary de…ned as the country not having an enforced mandatory voting law in place at the time of the most recent election. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 36 Table 14: Institutional Interactions - Freedom of the Press Interest Understanding Socialization News Meetings Knowledge (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Eligibility 0.066** 0.073* 0.057 0.085 -0.067 -0.087 0.020 -0.005 0.017 0.020 0.039*** 0.032** (0.031) (0.038) (0.056) (0.070) (0.054) (0.088) (0.037) (0.053) (0.021) (0.029) (0.014) (0.014) Eligibility Captured -0.009 -0.029 0.021 0.001 0.034 0.042 -0.034 -0.076 0.002 -0.028 0.011 -0.006 Press (0.034) (0.043) (0.060) (0.080) (0.070) (0.094) (0.041) (0.063) (0.023) (0.028) (0.016) (0.016) Captured Press 0.023 0.102* -0.083 -0.145 0.015 0.145* 0.129* 0.139 -0.029 0.015 -0.019 -0.024 (0.044) (0.058) (0.097) (0.106) (0.062) (0.078) (0.076) (0.084) (0.041) (0.041) (0.024) (0.023) Observations 40136 9294 33059 7835 13281 2939 26995 6372 40069 9301 30714 6959 Sample Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Full Cuto¤ Note: Regressions with country and year …xed e¤ects, controlling for age at interview and income percentile. All speci…cations use the static model. Standard errors clustered at the country-wave level. *** p<0.01,**p<0.05,*p<0.1. 37