The Effect of Education on Support for International Trade: Evidence from Compulsory-Education Reforms

Abstract

Across countries and over time, support for economic globalization is strongest among individuals with the highest levels of education. Yet despite long-lasting debates on the sources of this correlation, reliable evidence that isolates the causal effect of education from the nonrandom selection of individuals into education is lacking. To address this fundamental issue, I exploit compulsory-schooling reforms that increased the minimum school-leaving age in eighteen countries. Employing a fuzzy regression discontinuity design, I find that the reform-induced added years of education substantially and durably increased support for trade liberalization. And using new data on the content of school curricula, I find that the effect of schooling largely stems from instilling tolerance and pluralism in citizens and reducing the perceived cultural threat of globalization. In contrast, there is little evidence that the effect of schooling reflects the distributive consequences of international trade, separating globalization winners and losers.

The re-emergence of international trade and investment after World War II produced remarkable economic growth in many regions of the globe and facilitated political cooperation between past rival states.¹ Yet recent decades have seen a rising backlash against globalization. Voters have increasingly voted for protectionist political candidates who promised to limit foreign trade.² These campaign promises were eventually translated into significant policy changes, bringing about a steep increase in worldwide protectionism that was most clearly manifested by, but went beyond, the US–China trade war.³

Much of the electoral force driving the rise of antiglobalization parties and candidates and the recent return to protectionism is composed of voters with less formal education.⁴ Education has also emerged as one of the most important predictors of public support for international trade.⁵ Yet no study to date has provided reliable evidence on the causal impact of education on attitudes toward economic globalization. Is education a causal factor in trade policy preferences or a sorting variable that divides essentially different individuals into different levels of education based on other determinants of trade preferences?

Initially, the robust correlation between education and trade attitudes across various developed countries and time periods was interpreted as supporting prominent models of political economy which suggest that individuals’ preferences reflect their material self-interest in light of the distributional outcomes of free trade.⁶ The personal-interest-based perspective has been challenged by subsequent research, which showed that the effect of education likely stems from two potential sources of influence: an ideational mechanism, by which exposure to economic ideas and information in college increases support for trade; and a cultural mechanism, according to which education socializes students to have more tolerant values and a cosmopolitan worldview.⁷

However, another possibility, which has not been directly tested, is that the correlation between education and trade preferences is spurious: individuals who acquire more education are also more likely to support economic globalization to begin with. Education might be the outcome of various unobserved characteristics, such as cognitive skills, labor-market potential, personality traits, or family background.⁸ And these characteristics might also explain variation in trade preferences. For example, higher cognitive skills are associated with both educational attainment and labor-market outcomes among high-school dropouts.⁹ Similarly, individuals with greater openness to experience are more intellectually curious and motivated to acquire knowledge and education, but they also tend to be more open to different cultures and lifestyles and more socially liberal.¹⁰ Given the centrality and predictive power of formal education in explaining trade attitudes, addressing this puzzle has important implications for understanding the broader causes of the rise of public opposition to economic globalization.

To assess the causal effect of education, I leverage the implementation of compulsory-schooling reforms since World War II. I collect information from multiple sources to identify such reforms in countries that participated in at least two of the three National Identity modules of the International Social Survey Programme (ISSP): 1995, 2003, and 2013. Using these data, I compile a list that encompasses twenty compulsory-schooling reforms implemented over four decades (the 1940s to the 1970s) in eighteen countries spanning Western Europe, Eastern Europe, and East Asia. Those reforms increased the minimum school-leaving age by one to three years (or 1.6 years, on average).

Using a fuzzy regression discontinuity (RD) design in which respondents from the first cohorts affected by the reforms are compared to those from slightly older, unaffected cohorts, I show that these reforms successfully increased the treatment take-up (compulsory-schooling completion) by approximately ten percentage points. The central finding of the analysis is that the added years of education substantially and durably increased support for trade liberalization. An additional year of schooling, induced by the reforms, increases support for overseas commerce by three percentage points in the overall population and by ten points for compliers in the first affected cohorts. This set of results shows that the relationship between education and popular support for trade is causal and not entirely or even mostly driven by other characteristics that affect both education and trade preferences.

The second part of the analysis tests the observable implications of prominent explanations for the effect of education on support for free trade. Since the reforms under study here affect trade attitudes by increasing secondary education, the potential influence of the ideational mechanism—according to which exposure to economic ideas and knowledge in college produces the effect of education—is largely neutralized.¹¹ Instead, the analysis reveals that the additional years of secondary education made individuals more open to foreign cultures and influences more broadly, even when such openness is not directly associated with international trade. The increase in education due to the reforms did not sort individuals into occupations that are less vulnerable to foreign economic competition. Moreover, the effect of education on support for trade liberalization remains similar across starkly different economic circumstances and degrees of exposure to import competition or offshoring. Whether or not individuals actively participate in the labor force or work in the manufacturing sector or in offshorable occupations, more reform-based schooling leads to less opposition to overseas commerce.

Finally, to further assess the mechanism underlying the effect of secondary education, I use a new cross-national data set on the politicization of education and the content of teaching curricula. Constructing cohort-specific measures of pluralism, patriotism, and nationalism in the curricula, I find that the effect of schooling is strongly moderated by the extent to which cosmopolitan worldviews are embedded in education. Individuals who acquire more pluralistic education grow significantly more supportive of trade, but the treatment effect is significantly weaker in cohorts exposed to more patriotic or nationalist education. The strength and consistency of these empirical results suggest that individuals with more education tend to assess international trade more positively, primarily due to their stronger consmopolitanism, even when there is no significant change in the distributional consequences of trade for their own economic standing.

The findings contribute to research on the political economy of international integration. This literature has long been occupied with the interpretation of the robust and strong association between formal education and support for trade and its implications for the broader question of what explains variation in mass opinion on trade. Yet both data and design limitations have left important uncertainties about the question of causality. By leveraging compulsory-schooling reforms in a fuzzy RD design, the analysis presented here makes headway in identifying the causal effect of education on support for trade. Moreover, the study advances this body of research by employing new data on the content of school curricula individuals were exposed to as youngsters. Coupled with the exogenous variation in schooling based on the implementation of compulsory-education reforms, these novel data and measures offer a means to assess how the quantity and quality of formal education interact in shaping public attitudes. Consequently, the analysis provides direct evidence on the causal mechanism underlying the effect of education, elucidating specific components in education that are particularly important in driving popular support for trade liberalization.

Beyond international political economy, the study's findings resonate with a growing literature that has conceptualized or employed a normative approach to explain or criticize cosmopolitanism as an educational response to the challenges of globalization.¹² I provide empirical evidence that this educational strategy is not only a response to globalization but also a source of popular support for economic integration. The research note also connects the international political economy and political socialization literatures by focusing on the effect of secondary, pre-adult education on attitudes toward international economic policy.¹³ Previous studies mainly focused on college education and portrayed it as the key line dividing losers and winners, as well as opponents and proponents, of globalization. The new findings demonstrate that secondary education also affects public support for trade liberalization, and they shed light on the enduring consequences of adolescent socialization in school for economic policy preferences and how citizens think about globalization more broadly.

Why Do the Educated Support Economic Globalization?

Theories based on material self-interest suggest that those who benefit from international trade support trade liberalization, and those who lose from its economic implications support protectionism. In this view, education promotes pro-trade attitudes primarily by improving individuals’ labor-market standing or by changing the trade-related circumstances workers are exposed to in their specific industry or occupation.

The economic-self-interest approach can be divided into three more specific models. The Heckscher–Ohlin or factor proportions model assumes that factors of production are mobile across sectors, and predicts that trade liberalization will benefit owners of factors of production that are abundant relative to the rest of the world. In developed economies such as the ones included in the sample of countries I study here, highly skilled workers will therefore be the primary distributional winners from trade liberalization.

The Ricardo–Viner or specific factors model assumes that workers cannot easily move from one sector to another. This assumption implies that, at least in the short term, the primary losers from overseas commerce in developed economies would be workers in import-competing industries—primarily in manufacturing.¹⁴ A third, material-based model is centered on the extent to which workers’ occupations are offshorable (can be outsourced to another country).¹⁵ In developed countries, low-to-medium-skilled workers in routine-task-intensive occupations are vulnerable to offshoring and job displacement from international economic integration.¹⁶

The observable implications of these material-based models are that education might affect trade preferences by reducing employment in low-skilled jobs, import-competing sectors, or offshorable and routine-task-intensive occupations, respectively. Also, education is expected to increase support for trade among individuals currently in the active labor force, and primarily among those in offshorable occupations.

Rather than changing economic circumstances and interests, an alternative, cultural explanation sees education primarily as a socializing agent that shapes attitudes, norms, and perceptions. According to this perspective, education fosters cultural openness and a cosmopolitan worldview, promotes humanitarian values, and gives people the capacity to override anti-foreigner sentiments and ethnocentrism.¹⁷ Since international economic integration exposes countries to various foreign cultures, arts, people, goods, and firms, such cultural dispositions play a key role in shaping individuals’ responses to economic globalization. Many people do not disentangle the broad and complex effects of globalization and separately assess the material effects of trade. Instead, they often view the material effects of trade liberalization as part of an “openness package,” which also includes the nonmaterial aspects of globalization.¹⁸ In surveys across the world, people say they are worried about the proliferation of foreign fast-food chains, music, and fashion trends, and about exposure to foreign goods eroding traditional values, cultural customs, and national uniqueness.¹⁹ To the extent that a cosmopolitan view of the world reduces such cultural concerns, it should also increase popular support for international trade.

Moreover, according to theories of pre-adult socialization, exposure to civic values and democratic norms in secondary school, when individuals are more malleable, should have a particularly profound and long-lasting impact on political attitudes and policy preferences.²⁰ After this critical period of openness, political attitudes remain relatively stable and resistant to change even in the face of meaningful events.

Beyond its effect on trade preferences, the first observable implication of this cultural mechanism is that education will also increase tolerance of various foreign influences, such as films, books, and artworks. The second implication is that the effect of education will be moderated by the content of the curricula students are exposed to. Specifically, the effect of education on pro-trade attitudes should be stronger when curricula more strongly emphasize liberal-democratic norms such as tolerance and pluralism, and weaker when they more strongly promote the values of patriotism and nationalism.

Finally, economic knowledge of the virtues of trade openness might also be a key influence.²¹ However, since most students are not exposed to such economic knowledge before college—and even then, only when taking an economics class or related courses in the social sciences—this mechanism should be largely muted in this study, which focuses on the effect of secondary education.²² Instead, secondary education can develop students’ professional skills through either vocational or general programs. Vocational programs prepare students to work in specific occupations by providing specific job-related skills. General programs develop broader knowledge and cognitive skills like numeracy, literacy, and abstract thinking, which are also useful for further skill acquisition through on-the-job learning and training.²³ Secondary schools can also develop an orientation of openness to foreign cultures through the study of foreign languages, civics, history, and social studies. Indeed, since World War II, cosmopolitan perspectives have been developed in the philosophy of education, effectively promoting a vision of the world that sees all humanity as belonging to the same community.²⁴ Table E-1 (in the online supplement) summarizes these competing theories of the effect of education on trade preferences.

Research Design

To identify the causal effect of education on public support for international trade, I leverage compulsory-education reforms in eighteen countries as a source of exogenous variation. This setup creates a rare opportunity to isolate the effect of education from the nonrandom selection of individuals into education.

Data

I use repeated cross-sectional surveys from the ISSP's National Identity modules of 1995, 2003, and 2013.²⁵ Using various single- and multi-country reports and studies on compulsory-education reforms, I assemble a data set on eighteen countries that implemented twenty such reforms since World War II and participated in at least two of the three relevant waves of the ISSP (Table 1). For the main sources used to code the reforms, and more details on each reform, see Section A of the online supplement.

Table 1. Compulsory-education reforms

Notes: For respondents living in Scotland, the pivotal cohort in UK's 1972 reform is 1959. In Germany, 1954 is the pivotal cohort in four states: Nordrhein-Westphalia, Hessen, Rheinland-Pfalz, and Baden-Wurtenberg. Respondents living elsewhere are treated based on their state-specific year of reform. In Sweden, the reform was implemented gradually across municipalities, I follow Mocan and Pogorelova 2017 and code respondents born in 1952 as the first fully affected cohort.

I follow Cavaille and Marshall and measure schooling as the number of completed years of education, with a limit of thirteen, since the reforms did not affect postsecondary education.²⁶ In some specifications, I also use the variable compulsory education, which is coded 1 if the respondent had completed the years of schooling eventually required by law following the reform, and 0 otherwise.

The key dependent variable measures support for international trade. Respondents were asked whether they agreed or disagreed that their country should limit the import of foreign products to protect its national economy (using a scale from 1 to 5). In the main analysis, I use a pro-trade dummy variable that is coded 1 for individuals who disagree with the statement, and 0 otherwise. In section D of the online supplement, I show that using the original five-point measure does not change the results.

Another set of variables I use to test the observable implications of competing explanations of the effect of education are related to respondents’ occupation and cultural protectionism. To sort respondents by their professional skills, I rely on Oesch's classification of four-digit International Standard Classification of Occupations (ISCO-08) codes.²⁷ High-skilled is a dummy variable for respondents in highly skilled occupations, including higher-grade managers, skilled service or manual workers, sociocultural professionals, and technical experts. Occupation-specific measures of working in manufacturing, job routineness, and offshorability are taken from Owen and Johnston and are based on their classification of industry occupation shares, a routine-task-intensity index, and an index of offshorability, respectively.²⁸ Owen and Johnston provide binary indicators for occupational offshorability and manufacturing occupations. I also construct a variable for offshorable routine-task-intensive occupation by interacting the offshorability indicator and the continuous routine-task-intensity index. Respondents in non-offshorable occupations get a score of 0 on this measure, while individuals working in offshorable occupations are rated according to their occupation's routine-task-intensity index, where higher values indicate that occupations are more intensive in routine tasks.

As for cultural openness, the ISSP collected data on respondents’ agreement or disagreement (on a scale from 1 to 5) with two items: cultural protectionism, measured by “[Country's] television should give preference to [Country] films and programmes” and available in all three waves; and cultural threat, measured by “Increased exposure to foreign films, music and books is damaging our national and local cultures” and available only in the 2003 wave. These measures capture an important part of the theoretical concept of cultural openness without explicitly mentioning the economic aspects of globalization.

Finally, I use a new global data set on the politicization of education: Varieties of Indoctrination (V-Indoc).²⁹ It covers 160 countries since World War II (1945–2021) and draws on information provided by 760 country experts about indoctrination in education and the content of school curricula. Neundorf and colleagues show that V-Indoc measures have high face, convergent, and construct validity.³⁰ For each item, the experts were asked a question about the content of school curricula, followed by a clarification with specific examples. Responses to each country-year observation were provided by multiple country experts on an ordinal scale.³¹ V-Indoc uses a measurement model that aggregates the multiple ratings, takes disagreement and measurement error into account, and produces a probability distribution over country-year scores on a standardized interval scale that typically varies between –5 and 5, with 0 approximately representing the sample mean.³² The point estimates are the median values of these distributions for each country-year.

I use three V-Indoc variables. Pluralism in the curriculum measures the extent to which students are exposed to diverse views and interpretations when historical events are taught. Patriotic education in the curriculum measures the extent to which the language-studies curriculum promotes feelings of love, pride, loyalty, and commitment to one's country. And Nationalist ideology character in the curriculum captures the degree to which nationalism is promoted by the history curriculum as the dominant societal model or ideology. Detailed information on these measures is provided in section B of the online supplement.

I use these country-year point estimates to calculate a series of birth-cohort-level measures, each of which is a twelve-year (ages six to eighteen in most countries) moving average of the relevant interval variable for each cohort. For example, for ISSP respondents who were born in 1945 in France, the cohort-level measure is the (standardized) mean value of the original V-Indoc variable from 1951 to 1963. These measures are meant to capture potential cohort-level exposure to values or knowledge ingrained in school curricula. Summary statistics for each measure are presented in Table A-2 in the online supplement.

Identification Strategy

To capture the causal effect of education on public support for international trade, I use a fuzzy RD design in which respondents just young enough to be included in the reforms are compared to respondents who are just too old to be included. Accordingly, respondents’ birth cohort relative to the year when the reform came into force is used as the running variable that determines assignment to the treatment. The cutoff point represents the first cohort to be affected.

Importantly, receiving the treatment—education—is not perfectly determined by the compulsory-schooling reforms. For various unobserved reasons, students can drop out early even after such laws are enacted, and they can decide to remain in school even when it is not required. Thus, compulsory-education reforms produce a fuzzy discontinuity in the treatment take-up: though the decision whether to remain in school is not perfectly determined by the reform, the probability of remaining in school still jumps due to the reform (Figure 1).³³ Accordingly, exposure to compulsory-education reforms is used as an instrumental variable (IV) that discontinuously increases the probability of receiving additional years of education.

Figure 1. Compulsory-schooling completion around education reforms

Like the sharp RD design, the fuzzy RD design relies on the assumption that individuals’ potential outcomes are continuous at the threshold.³⁴ This assumption would be violated if individuals (or their parents) were able to foresee the timing of the reform and precisely manipulate respondents’ year of birth accordingly, but this is highly implausible because it requires parents to know about those reforms years before they are even proposed. Around the cutoff, therefore, the variation in the treatment is as good as randomized: the pre-reform and post-reform cohorts should be comparable to each other in all relevant aspects, except for their exposure to the new legal requirements regarding compulsory education.³⁵

To validate the continuity assumption, I use density and balance tests. The density test indicates that respondents were not able to sort around the threshold (Figure A-4 in the online supplement). If respondents or their parents had precise control over their birth cohort, then an unusually large proportion of them should have been observed either below or above the threshold (depending on whether they perceive compulsory education positively or negatively). Instead, the manipulation test fails to reject the hypothesis that the density of the score changes discontinuously at the cutoff point. In addition, the control and treatment groups are similar across all pretreatment characteristics available in the ISSP data, consistent with the proposition that treatment assignment is as good as randomly assigned for sequential birth cohorts around the cutoff (Figure A-5).³⁶

I estimate local linear regressions using the optimal bandwidth recommended by Calonico, Cattaneo, and Titiunik.³⁷ This continuity-based RD framework focuses on the treatment effect at the cutoff. In other parts of the analysis, I use an IV design that estimates the treatment effect not at the cutoff but rather in a small window around it. To determine the size of this window, I use Cattaneo, Idrobo, and Titiunik's procedure, which considers a sequence of nested windows, starting with the smallest, and in each one conducts balance tests for each of the pretreatment covariates. The optimal window is the smallest window with covariate balance.³⁸

The IV estimation relies solely on the exogenous variation in education that comes from whether individuals’ birth cohort is included in or excluded from the schooling reform. Compulsory-education reforms can be viewed as an exogenous encouragement, where individuals just young enough to be included in the reform are “encouraged” to remain in school for several more years, while individuals just too old to be under the mandate of the new law are not.

The causal identification of the local average treatment effect—that is, the effect of education for compliers who remain in school only if required to do so by compulsory-education laws—requires two additional assumptions: monotonicity, which rules out the possibility that some students drop out if and only if they are subject to compulsory-education reforms; and exclusion restriction, by which the effect of the reforms on trade attitudes runs exclusively through the education treatment.³⁹ Both assumptions are plausible in this context: students are unlikely to respond to the reforms by completing less education, and the proximity of the reforms to students’ choice to stay in school or drop out early makes potential violations of the exclusion restrictions highly unlikely.⁴⁰

Education Increases Support for International Trade

Column 1 of Table 2 provides the first-stage estimate of the effect of compulsory-schooling reforms on completing compulsory education: a substantial 9.7 percentage-point increase. Similarly, compulsory-education reforms increase educational attainment by 0.31 schooling years, on average (column 2). The F-statistic for the relevance of the reform instrument for completing compulsory education (schooling years) is 270 (201), far past the conventional threshold of 10 and indicating a strong first stage.

Table 2. Effect of Secondary Education on Support for International Trade

Notes: Entries are local linear regression estimates with a triangular kernel using the Calonico, Cattaneo, and Titiunik 2014 optimal bandwidth (±12 birth cohort years). The F-statistic is drawn from a baseline first-stage regression. Pre-treatment covariates include respondents’ sex, ethnicity, and parents’ immigrant status. Robust standard errors in parentheses, clustered by subnational region. ITT = intention to treat. LATE = local average treatment effect. *p < .05; **p < .01.

Column 3 shows the reduced-form or intention-to-treat estimate: exposure to the compulsory-schooling reforms caused an average increase in support for international trade of three percentage points, even before accounting for imperfect compliance. Since 21 percent of the pre-reform cohorts take a pro-trade stance, this estimate implies that affected cohorts are over 14 percent more supportive of trade due to the implementation of compulsory-education reforms.

Yet the intention-to-treat effect underestimates the actual effect of education. Many students are always-takers who acquire the compulsory level of education regardless of whether they are required by law to do so. And a few are never-takers who drop out early either way. In column 4, I correct for this noncompliance by scaling the intention-to-treat effect by the compliance ratio. The local average treatment effect is substantial, increasing support for overseas commerce by thirty-two percentage points.

Marshall shows that coarsening a multivalued treatment (for example, treating a continuous endogenous variable as binary) can upwardly bias IV estimates. Conversely, when the treatment is coded as a multivalued variable, the IV estimation produces a consistent causal estimate.⁴¹ Column 5 addresses this concern by using the multivalued schooling variable. The effect of education remains large: one additional reform-induced year of education increases support for overseas commerce by ten percentage points.

Here and in section D of the online supplement, I present additional checks that underscore the robustness of these findings. First, the results are statistically similar when I use a local-randomization RD approach—that is, estimating the effect of the reform-induced schooling not at the cutoff but in a neighborhood around the cutoff (Figure 2). This analysis also further demonstrates that pre-adult education has a long-lasting impact on how people think of economic globalization, as the three percentage-point treatment effect (or 14 percent above the baseline rate of support) is stable across all three survey waves, which span over two decades.

Figure 2. Effect of education across survey waves

A second set of tests shows the results from IV models on either a subset of near-cutoff observations within the optimal window of ±11 birth cohort years or any other window between ±1 and ±20 cohorts (Figure A-7 in the online supplement). Not only are the point estimates remarkably stable across bandwidths, their precision remains high even at the smaller bandwidths, with fewer observations than optimal. Third, the results (in Figure A-7) remain similar in a minimal specification without controlling for covariates, which corroborates the identification strategy and suggests that, as in a randomized experiment, the covariates are controlled for by design. Fourth, using a higher-order polynomial yields similar estimates of the effect of reform-induced education (Table A-6). Fifth, the results are robust to leaving out of the analysis any of the eighteen countries originally included (Figure A-8). Finally, I conduct placebo cutoff tests and find no significant change around placebo reforms either before or after the timing of real reforms (Table A-5).

Overall, in contrast to the view that education is merely a proxy for various personality traits and socioeconomic or cultural characteristics, we see that education has a sizable causal effect on support for international economic integration. Compared to individuals from cohorts unaffected by the new compulsory-education reform, otherwise similar individuals from affected cohorts became considerably more supportive of economic globalization. Not only is this attitudinal difference significant and substantively meaningful, it lasts decades after students have finished their school years and reached adulthood.

Education, Economic Circumstances, and Cultural Threat

Why does secondary education increase public support for trade openness? As outlined in the theory section, two competing but not mutually exclusive explanations are material self-interest and cultural openness. According to the former, education shapes individuals’ economic interests by affecting their position in the labor market and improving their employment and wage prospects in the global economy—making them more highly skilled workers, less likely to work in import-competing industries, and less vulnerable to offshoring. According to the latter, education reduces the perceived cultural threat of globalization by fostering a cosmopolitan view of the world and tolerance of foreign cultures and lifestyles.

The results provide little evidence for the material-self-interest account (Table 3). Specifically, the added, reform-induced years of education did not significantly change the probability that respondents would find employment in high-skilled jobs, in the manufacturing sector, or in offshorable routine-task-intensive occupations. Of course, exposure to offshoring or import competition might generally account for some of the variation in public support for trade more broadly,⁴² but I find no evidence that they could have more specifically generated the effect of education on support for trade.⁴³

Table 3. Compulsory education, economic outcomes, and cultural protectionism

Notes: Entries are IV estimates with robust standard errors in parentheses, using the ±11 optimal bandwidth. Data on cultural threat are available only in the 2003 ISSP wave. RTI = routine-task-intensity. *p < .05; **p < .01.

Importantly, the limited overall support for the material-based explanations does not seem to be the result of a weak treatment in terms of labor-market outcomes and income mobility. Treated individuals were able to take advantage of their extra schooling years and find better-paying jobs (column 5). One additional year of schooling increased respondents’ position on their countries’ income distribution by more than half a decile. However, these better-paying jobs seem to be mostly within respondents’ broader level of skill, and they did not reduce individuals’ exposure to import competition or offshoring. In other words, while education significantly affected workers’ economic fortunes, it did not dramatically change their potential benefits from, and material self-interest in regard to, international trade and globalization.

In contrast, education had a large effect on cultural protectionism (last two columns of Table 3). Specifically, one additional year of schooling decreased individuals’ support for giving preference to national films and programs in their country's television by ten percentage points. It also decreased individuals’ belief that exposure to foreign films, music, and books is damaging their national and local cultures by nine points. Taken together, these findings suggest that in the eighteen countries surveyed here, education strongly reduced perceived cultural threat from globalization, while its potential to change economic interests was limited.

An important limitation in exploring the mechanisms of the effect of education is that most individual characteristics, such as employment status, income, occupation, or partisanship, are post-treatment variables that may themselves be affected by pre-adult exposure to education. Controlling for these post-treatment variables or subsetting the data based on them may introduce severe bias, because estimating the treatment effect “holding all other variables constant” becomes impossible.⁴⁴ That said, it is worth subsetting the data based on employment status to test one important observable implication of the material-based account. If education increases support for trade because it shapes workers’ expectations about the impact of international trade on their real wages, then the pro-trade effect of education should be large and significant among respondents who either actively participate in the labor market or work in offshorable occupations, but much weaker among their counterparts.⁴⁵

I find that more years of education, induced by the compulsory-schooling reforms, increase public support for international trade regardless of whether respondents are currently in paid work, working in the manufacturing sector, or in offshorable occupations (Figure 3). Despite the stark differences in exposure to trade shocks or vulnerability to offshoring across these groups of respondents, and in their material self-interests derived from these different circumstances, the impact of education on how individuals think about international economic integration remains statistically and substantively similar across all subsamples. While these results should be taken with caution given the mentioned potential post-treatment bias, they provide no evidence that the effect of education on trade is driven by material circumstances or self-interest.

Figure 3. Effect of years of schooling, by labor-market characteristics

The Effect of Education by Curriculum Content

Another way to assess the mechanisms of the effect of education on pro-trade preferences is to examine how the treatment effect interacts with the curriculum content individuals were exposed to as students. Previous research was unable to do this due to data availability limitations. Here, I use V-Indoc, a new cross-national data set on indoctrination in education and school curricula since 1945.⁴⁶

In Figure 4, I create subsamples of the full ISSP sample, dividing respondents by the median value of three curriculum measures: pluralism, nationalism, and patriotism. Comparing the results across the three subsamples, I find a significant, systematic difference in the estimated effects of education on pro-trade preferences. One more year of schooling, induced by the compulsory-schooling reforms, increases support for trade by nearly six percentage points (p < 0.001) for respondents from cohorts exposed to more pluralism in education, while this effect is only 0.5 points (p < 0.59) among those exposed to less pluralism. The difference between the two effects is statistically significant (p < 0.01). Similarly, another year of schooling for respondents from cohorts exposed to less nationalism as a dominant societal model in the school curriculum brings about a 4.6-point (p < 0.001) increase in support for trade, while the effect is smaller than 0.1 points for those exposed to more nationalism; and the difference between these two effects is also statistically significant (p < 0.01). The differential effect of schooling across exposure to different levels of patriotism in the curriculum is more subtle, but is also statistically significant (p < 0.01).

Figure 4. Effect of years of schooling, by the content of education

Thus, the extent to which the curriculum promotes and celebrates feelings of love, pride, and loyalty toward the nation, and the extent to which it exposes students to diverse views of the world and alternative interpretations of history, clearly and durably matter to the formation of citizens’ attitudes toward international trade and globalization. The findings also highlight the significance of secondary education and the role of the state in shaping citizens’ political attitudes. Compared to universities, secondary schools offer governments not only broader access to the population but also greater influence over the content of education, which may be used for either patriotic or democratic indoctrination purposes.

In section D.8 of the online supplement, I present robustness tests that use a different coding of the treatment or a different threshold for splitting the sample based on the school curriculum measures. The results are robust to these different coding decisions and alternative measures. Furthermore, interacting schooling and continuous measures of school curricula using the full sample yields similar results (Table A-9). This analysis also demonstrates that a more intense focus on subjects related to the development of labor-market skills does little to change the pro-trade effect of schooling.⁴⁷

Taken together, the findings show that education—even at the secondary level—durably increases public support for international trade. Material circumstances and interests play little role in driving this effect. Instead, the evidence is very much consistent with the argument that education heightens public support for international trade by instilling tolerance and pluralism in students and reducing perceived cultural threat from economic globalization.⁴⁸

Discussion

Education has emerged as one of the clearest lines dividing supporters and opponents of globalization and international trade.⁴⁹ However, whether this robust correlation stems from nonrandom selection of individuals into education or represents a causal effect of education itself has yet to be established. This research note presents the first evidence of a causal effect of education on public support for trade liberalization. I find that more years of education, instrumented by compulsory-schooling reforms, brings about a sizable and long-lasting rise in public support for international trade. Several decades after these reforms came into force, treated individuals were still not only significantly more supportive of free trade but also markedly more tolerant of foreign cultures and influences. Furthermore, the effect is most pronounced when individuals are exposed to more pluralistic and less nationalist education. In contrast, I find that neither the extent to which school curricula focus on professional skill-building subjects nor the vulnerability of individuals to the distributional consequences of trade later in life moderate the effect. These patterns are most consistent with a cultural channel through which education fosters a cosmopolitan outlook and reduces the perceived cultural threat of economic globalization.

In evaluating these findings, several issues must be considered. First, although I find that the effect of secondary education primarily operates through a cultural channel, other explanations I have mentioned, including material self-interest and economic knowledge, might still account for the effect of college education, as well as other determinants of public opposition to trade.⁵⁰ Second, while the compulsory-education reforms studied here successfully increased levels of education, they have not transformed globalization losers into winners. Instead, the economic profile of treated individuals remained similar in terms of employment in low-skilled jobs, import-competing sectors, and offshorable occupations. The balance in these labor-market circumstances provides a rare opportunity to isolate the often-overlapping economic and cultural mechanisms underlying the effect of education on public support for international trade. However, this also leaves an open question about the role of economic factors—and their interaction with non-material factors⁵¹—in cases where education substantially improves labor-market circumstances, a question future research should seek to answer.

Although the evidence is mixed, several studies find that education correlates with support for trade in skill-abundant countries, and opposition to trade in skill-scarce ones, in line with the factor-endowment model.⁵² While relative factor endowment might be important, this study raises an additional possibility: low-income, skill-scarce countries, often governed by autocratic regimes, may use patriotic indoctrination in education to a much larger degree than skill-abundant countries, which are more likely to be consolidated democracies.⁵³ Understanding these dynamics is a promising avenue for future research.

From a policy perspective, these findings speak to a growing trend in cases where right-wing populist parties and leaders have come to power. After World War II, nation-states increasingly incorporated a global perspective into their educational curricula.⁵⁴ More recently, however, right-wing populist parties, as well as conservative lawmakers, have increasingly attempted to change teaching curricula, textbooks, and education practices in several countries, with varying success.⁵⁵ The evidence presented here suggests that making school curricula more nationalist and less pluralistic might plant the seeds for an even stronger protectionist shift by the next generation of political leaders and supported by a broader coalition of antiglobalization voters. While voters sometimes punish incumbent leaders for promoting protectionist policies,⁵⁶ the gradual, slowly evolving nature of potential protectionist shifts driven by education reforms and generational replacement is likely to be more hidden from the public eye and less susceptible to democratic accountability.