A vibrant literature has recently blossomed on public attitudes toward international economic integration. Most frequently, this research involves survey evidence,Footnote 1 sometimes with an experimental component.Footnote 2 While these methods can test many theories, they face nagging questions of external validity.Footnote 3 This problem is especially acute for low-salience issues, which international economic flows, like many issues in international relations, often are: questions about trade or international investment frequently require respondents to formulate an off-the-cuff view about measures that they have not previously contemplated and about which they do not have much information.Footnote 4 Alongside the broader problems of surveys—the artificiality of the setting, the low stakes and hence low credibility of response, and potential reluctance to share politically sensitive views honestly with a stranger—this leaves uncertainty about how well survey-based findings apply to circumstances where people really do act on their beliefs about the international economy.Footnote 5 Complementing these studies with an examination of behavior from less self-conscious, more salient settings is consequently of intense theoretical and practical interest.
Unfortunately, opportunities to observe meaningful mass (as opposed to legislative or activist) reactions specific to international economic policy are rare.Footnote 6 Fortunately, there are exceptions. This research note considers one such exception: the 2007 Costa Rican plebiscite on accession to the agreement between the United States and the Central American Free Trade Area (CAFTA). While referendum outcomes have the disadvantage of producing ecological rather than individual results, thus only indirectly testing models of individual preference formation,Footnote 7 many studies have successfully exploited referendum results to study questions of public opinion and individual behavior.Footnote 8 Especially where referendum votes are uncommon, issue-specific plebiscites helpfully prime the public to think about issues—even issues typically having low salience—and greatly reduce participants' cost of obtaining information relevant to decision making.Footnote 9 Moreover, aggregated data's drawbacks diminish when considering attitudes toward international integration, where opinions may reflect aggregate, not directly personal, expectationsFootnote 10 and outcomes.Footnote 11
After briefly outlining previous analyses of policy preferences over the international economy, I compare those earlier theories with voting patterns in the 2007 Costa Rican plebiscite. Some outcomes, especially those highlighting the importance of education and culture, recur in the aggregate voting results. Other theories, though, particularly economic predictions such as those based on the locally dominant sector, prove less predictive in this instance. Economic variables' lack of effect echoes many recent findings,Footnote 12 suggesting that the empirical weakness of traditional economic explanations may not be an artifact of the survey context. Instead, it provides further, albeit suggestive, evidence that public attitudes toward international economic policy primarily stem from factors other than real returns.
The Costa Rican Plebiscite
On 7 October 2007, Costa Ricans voted on CAFTA, a trade treaty that would incorporate the United States, Guatemala, El Salvador, Honduras, Nicaragua, and the Dominican Republic.Footnote 13 Economically, the potential agreement was highly important for Costa Rica as a regional arrangement that would enhance market access.Footnote 14 While the country had pursued customs unions with its Central American neighbors for decades, the United States represented Costa Rica's most important partner for imports, exports, foreign tourism, and international investment, accounting for a third to a half of Costa Rica's international exchange. (The treaty also promised to expand the trade agreement with Panama, a far smaller player but still among Costa Rica's ten largest trading partners.) At the same time, the plebiscite came after years of increasing reliance on trade. Imports and exports as a share of gross domestic product had risen from 86 percent in 2001 to 104 percent in 2006, while anticipation of a trade deal had also helped raise net foreign direct investment inflow from 2.8 to 6.5 percent of the country's gross domestic product (GDP) over the same period. At the same time as this economic interconnectedness increased, the country as a whole prospered. Income per capita had risen by more than 6 percent in 2006 and was on pace nearly to match that figure in 2007. Unemployment was also at historically low figures, with the overall unemployment rate falling below 6 percent for the first time in years—and for 2007 as a whole, the figure would be a mere 4.6 percent.Footnote 15 Accession to the agreement therefore had potential effects for a substantial share of the Costa Rican economy, and voters were asked to make their decisions in a relatively friendly macroeconomic climate.
The only topic formally under consideration in the vote was international economic policy. Though the agreement had chapters concerning labor and environmental protections—and provisions in its transparency chapter relating to uprooting corruption—these were explicitly tied to how such policies could affect cross-border flows of goods, services, and investment. Meanwhile, no other major vote took place at the time of the referendum: neither further referendum measures nor general election races were on the ballot to bring citizens to the polls.Footnote 16
The vote on CAFTA garnered public and media attention as the first referendum in Costa Rican history, and the free trade agreement was itself politically contentious. The opposition had forced the government to submit the treaty to popular vote after months of constitutional maneuvering and street protests.Footnote 17 Moreover, polling showed a close and tightening contest throughout the months leading up to the vote. In the event, the referendum passed with 51.6 percent of the vote; 59.2 percent of the electorate cast ballots. This compares with 65.4 percent turnout in the closely contested 2006 presidential election, typically the biggest event in the Costa Rican electoral cycle.
Hence the 2007 Costa Rican vote was, unusually, a vote focused squarely on high-salience international economic policy issues.Footnote 18 It occurred, helpfully, in Latin America, an especially fruitful area for research in mass attitudes toward markets and international cooperation.Footnote 19 Moreover, unlike in popular votes on the European Union, the question at issue had few confounding issues of political unification or regulatory change. The referendum accordingly offers rare insight into the willingness of the public to accept or reject international economic liberalization in a practical setting. The very rarity of this sort of vote of course raises questions about generalizability: the circumstances that allowed such a vote are not likely to have been random, and the results obviously do not speak to how trade policy arises in most circumstances. The referendum nevertheless provides an internally consistent way of looking at economic preferences. Where its results correspond to or diverge from prior findings, it can help to establish their robustness to nonsurvey methodologies.
Using the Referendum to Test Theories of Trade-Policy Preference
Ballot secrecy precludes individual-level data on voting behavior for the Costa Rican referendum. However, information is available at relatively low levels of aggregation. The unit of analysis here is generally the administrative district (distrito), the third-order administrative division of Costa Rica. There were 470 of these at the time of the referendum, with a mean population of about 8,300 people according to the 2000 census (making for an average eligible electorate of about 5,600 voters). Even the most populous district had a population of about 80,000. Election and referendum results, along with a small number of control variables, are also available at the smaller electoral-district level, where the average population size is roughly one quarter that of an administrative district. This alternative unit of analysis will be explored in the next section.
For each administrative district, various statistics about the referendum are available. The dependent variable for the reported analyses is the fraction of valid ballots cast that were in favor of the free trade agreement.Footnote 20 Alternative measures using the fraction of all ballots cast (including blank or spoiled ballots) produce very similar results.
The foremost causal mechanisms among models of trade policy preferences have been economic affiliations, typified by the Heckscher-Ohlin (factoral) and Ricardo-Viner (sectoral) models. A wealth of theory argues that trade's effects on individuals' material welfare and job security will strongly influence attitudes toward international economic exchange: those whose livelihoods are threatened by imports have incentive to resist openness and favor protectionism. Conversely, efficient producers see increased profits with openness, as export markets expand and importing necessary inputs to production becomes cheaper. In the long run, these potential costs and benefits may accrue to all suppliers of a particular factor of production, such as capital or labor. Over shorter time horizons, however, laborers and capitalists tend to have their assets invested in specific sectors/industries from which it is costly to change.Footnote 21 This association with a sector may arise indirectly, as among service workers in a region heavily dominated by a particular industry: even if the service workers face no immediate threats from foreign competition, they may still be sensitive to trade-induced downturns in the local economy. If their neighbors suffer layoffs, their own prospects deteriorate as well.
To measure these potential sectoral dependencies, the models below use the fraction of the labor force in the administrative district that works in a sector. Data are available for several sectors, but reported results here include only sectors that employ at least 5 percent of the average district's labor force. These are agriculture (representing 19 percent of employment nationwide), manufacturing (17 percent), and three service sectors: commerce (that is, wholesale and retail trade: 17 percent of employment), construction (6 percent), and education (6 percent).Footnote 22 Costa Rica was and is a net exporter of agricultural goods but a net importer of manufactures, particularly machinery. Hence, Costa Rica appears to have a comparative advantage in agriculture but a disadvantage in manufacturing. The standard sectoral framework then suggests that more agricultural districts should see more support for the treaty and its attendant trade liberalization, while manufacturing-oriented districts should generally be less supportive. Trade in services (with the important exception of tourism) was a relatively small part of the Costa Rican economy, although the World Trade Organization suggests that the country was a net exporter of nontravel services (and with net service exports growing at more than 10 percent a year). This suggests that Costa Rican districts specializing in services should, all else equal, be neutral or in favor of the treaty.
Other economic factors also are conventionally thought to correlate with preferences over trade policy. High unemployment rates generally increase demands for government intervention in the economy, including protection of domestic jobs from foreign export competition.Footnote 23 However, the potential for export jobs in a free trade area may reverse this effect: lower foreign tariffs can create constituencies that reap concentrated benefits from exporting and thereby have a positive interest in trade. This may be of especially keen interest when unemployment is widespread. To consider these possibilities, the models below include measures of the fraction of adults in the workforce that were out of work at the 2000 census, the temporally closest available source of district-level economic statistics.
Alongside its consequences for factor markets, trade-barrier reduction also affects product markets. In particular, free trade tends to reduce prices and allow more consumption. Those who are not active labor market participants thereby have cause to prefer free trade. Accordingly, those districts with a higher proportion of their population that are pensioners, or with a higher dependency ratio (the number of young, old, or otherwise out-of-the-workforce persons per member of the labor force), would have more widespread impetus to vote in favor of CAFTA.
Voting patterns are also likely to reflect partisan preferences, especially on an issue as contentious as the free trade agreement was in Costa Rica. This is both because positions on specific issues may derive from general ideological proclivities and because political leaders' guidance can serve as an important heuristic source of information about controversial issues.Footnote 24 There are three main political parties to consider. (No other party received as much as 5 percent of the vote nationally in votes around the time of the referendum; including smaller parties' vote shares does not noticeably enhance the predictive power of the models.) Two of the major parties, the Partido Liberación Nacional (PLN; National Liberation Party) and Partido Acción Ciudadana (PAC; Citizens' Action Party), are generally left-leaning. Yet they diverged in their attitudes toward the referendum. The social-democratic PLN, the traditional party of the Costa Rican left, supported approval of CAFTA, while the newer PAC led the opposition.Footnote 25 The third major party, the Movimiento Libertario (ML; Libertarian Movement), followed its classically liberal roots and joined the PLN in support of the free trade area. Thus, if leadership by political elites drives political preferences, communities that vote for and are presumably influenced by the PAC should vote more heavily against the referendum, while those with more PLN or ML supporters would be more likely to cast ballots in support of the free trade agreement.
For each of these parties, the vote share in the district from the valid votes cast in the 2006 presidential election measures local support. These variables are of special interest in this case because the 2006 election was largely, though not exclusively, fought over the free trade agreement. Hence divergences between votes in the two elections offer a rough approximation of the limits of international economic affairs' effect on vote choice in general elections: if foreign affairs indeed are a low-salience issue that voters do not engage with, then the association between preferences over trade and votes in the election will be weak despite the seeming centrality of CAFTA to the presidential campaign. These partisan variables also speak to the idea that support for free trade may be contingent: free trade may be attractive only in the presence of a left-leaning government that will provide for trade-affected workers through a strong welfare state.Footnote 26 Yet voters who support the PLN may not be confident that future governments would maintain the commitment to the welfare state, especially in light of the extremely narrow victory that the PLN won in the 2006 election.Footnote 27 Thus it is quite likely that even voters who were willing to support free trade under the auspices of the PLN government would hesitate when faced with a more permanent commitment codified in a treaty.Footnote 28
A final political variable of note is district turnout. Higher levels of participation may reflect a local culture that is more engaged with (or informed about)Footnote 29 political issues, which might influence opinions about trade. Greater interest in the referendum may lead to a larger number of conversations about the potential consequences of the free trade area, for example, which might alleviate fears that CAFTA threatened the jobs and livelihoods of many acquaintances and increase the perceived benefits of trade for the nation at large. Or, equally, it could increase the perceived cost of trade if any acquaintances are leery of their employment prospects in the free trade area. Turnout also matters as an indicator of the comparability of populations across the units of analysis. If, like most protectionist barriers, Costa Rican trade policy features concentrated benefits and diffuse costs (or if voters perceive policy to feature such an asymmetrical incidence),Footnote 30 then random events such as rain that reduce electoral participation in a locale may also be likely to lead to disproportionately large shares of voters opposing the referendum in places with low turnout.Footnote 31 This would suggest that the referendum results might represent not differences in actual opinion toward CAFTA, but only differences in expression of those preferences.
Recent studies of policy preference have considered cultural and sociological influences beyond those of economic return and political party. Education, for instance, may not simply involve an increase of human capital, but also a greater exposure to ideas and various forms of cosmopolitanism that may make people more willing to accept the costs of trade exposure.Footnote 32 This is an especially interesting possibility in a developing-world context such as Costa Rica. Here, the society is likely to be relatively scarce in human capital and skilled labor, and so the classic factor-model prediction is for those with higher skills (along with their families and neighbors) to be more protectionist.Footnote 33 But the cosmopolitanism argument suggests the opposite: educated populations—along with those who interact with the educated and are exposed to their ideas—will be more likely to embrace free trade, especially at the top end of the education spectrum. To capture these possibilities, the models below include two measures of education: the proportion of the district population with a university degree, along with the proportion of the population that is illiterate. Insofar as the Heckscher-Ohlin factoral model holds, voters in districts with university-educated populations would be expected to favor protectionist policies, while areas with more illiteracy would gain more from trade and hence favor the agreement. To the extent that the cosmopolitanism model applies, however, lower levels of illiteracy and (especially) higher levels of university education should associate with votes supporting the referendum.
Other sociological factors may also relate to acceptance of trade. Those living in the diversity of an urban environment, for example, may know more people and hence be more likely to be exposed to some whose job is at risk from foreign competition. (In addition, urban areas may provide a denser network of local job options and wider consumption options, thus changing the relative costs and benefits that imports or job displacement might have.) Thus the models control for the (natural logarithm of) population density in a district.Footnote 34 Acquaintance with the foreign-born may also change views on the value of international exchange. Although exposure to immigrants and their attendant job competition can under some circumstances reduce contentment with open economic policies,Footnote 35 it simultaneously exposes people to the benefits of international openness and to natural cheerleaders for international engagement. Living near foreigners may also affect nationalist sentiments, another frequently observed predictor of economic protectionism.Footnote 36 To look for these effects, the models include the proportion of each district's population that was born abroad.Footnote 37 Another cultural group that may have distinctive attitudes to transnational exchange is the indigenous community, whose ancestors experienced the invasion of foreign cultures in past centuries. To account for this possibility, the models include an indicator of districts that, according to the electoral authority, include indigenous zones. Table 1 provides summary information about these variables, dependent and independent, for Costa Rica's administrative districts.
Administrative-district summary statistics
Finally, to account for simple sectional differences in opinion, there is also a battery of cantonal fixed effects. Cantons (cantones) are the second-order administrative unit of Costa Rica, one level in the hierarchy above administrative districts. The eighty-one cantons generally class the districts into small regional groupings with historical and bureaucratic ties to one another and thereby account for many unobserved factors.
Results
Table 2 presents the results of the regression using the above variables with and without cantonal dummy variables. For comparison, columns (1) through (4) present regressions using only subsets of the independent variables, very loosely grouped into economic, political, and sociological theories of trade policy preference formation. These groupings are certainly not definitive, as many variables here are open to multiple interpretations, but they provide an organizational structure for considering the various causal factors. Columns (5) and (6) give the full models, with column (6) including cantonal dummies.
OLS models of percent of Costa Rican voters favoring free trade, by administrative district
Note: Robust standard errors in parentheses. OLS = ordinary least squares.
Both with and without the cantonal dummies, the sectoral distribution of a district's labor force has consistent effects on the local referendum vote. The greater the share of workers in the manufacturing sector, the more supportive a district was of the referendum. By contrast, more workers in the education sector consistently reduced the support for the referendum. The result for education may partially stem from data anomalies: one district (Mercedes, in Guácimo canton) has about 34 percent of its workforce in education (likely because of the district's proximity to EARTH University), almost twice the proportion of the next-most education-centric district. Still, dropping this observation preserves a negative relationship between districts' education workers and referendum support: the coefficient in column (5) becomes −0.47 (standard error of 0.25) and that in column (6) −0.62 (standard error of 0.23). Regardless, this does not support the standard idea that employees in nontraded sectors should tend to be most concerned with the price effects of international economic integration.Footnote 38 The effect of manufacturing employment may fit better with traditional models of sectoral effects on policy preferences, although Costa Rica's net imports of manufactures raise questions about whether the preferences reflect sectoral comparative advantage. Other sectors—agriculture, commerce, construction—prove not to have substantial effects one way or the other when other variables are taken into account, at least compared to the baseline represented by the sectors excluded from the model. Table 2's models suggest that each 5-percentage-point increase in the share of workers in manufacturing resulted in around a 1-point increase in the share of votes for the referendum. The estimated effect of education-sector employment was larger but more variable in size, with estimated effects ranging from somewhat under 0.5 (implying that having 2 percentage points more workers in education resulted in a percentage-point fall in the share of pro-CAFTA votes) to 0.9 (so that there was nearly a one-to-one relationship between more education workers and lower shares of voters favoring the plebiscite).
Other variables relating to real consumption also offer mixed support for traditional models of policy preferences. As would be expected from populations that tend to have a fixed income and so are price-sensitive, districts dominated by pensioners tend to have higher proportions of voters supporting CAFTA; in the fuller models of columns (5) and (6), a 1-percentage-point increase in citizens on pensions translates into an increase of just under two-thirds of a percentage point in the share of voters supporting the referendum. However, once this effect of pensioners is taken into account, higher district dependency ratios tend to associate with lower local support for free trade. This may suggest that having to provide for dependents (other than pensioned retirees) increases aversion to risk of job or income loss, which a shift in trade rules could change in unforeseeable ways. This is especially plausible given the link between protectionist sentiment and risk aversion.Footnote 39
Several of the political variables have significant effects. Support for the PAC, the only major party to actively campaign against the referendum, has a strong correlation with protectionist voting on the CAFTA referendum, with a 3-point increase in PAC support associating with a 1- to 2-point decline in CAFTA support. The other political variables show somewhat less consistent results. Higher levels of general political engagement, as proxied by presidential election turnout rates, associate with greater district-level support for the free trade agreement—but this effect vanishes when the cantonal indicator variables are included. A similar pattern occurs with the vote for the libertarian ML party: ML-supporting districts also tend to be pro-CAFTA in their votes until the canton effects enter the model.Footnote 40 Support for the PLN, the largest party of all, has much weaker association with the referendum results, although it consistently takes a positive coefficient in line with the party's supportive stance on CAFTA. The weakness of this association may stem from the big-tent nature of the party, or from the ideological tension between the party's general left-wing beliefs and its embrace of international markets in this instance (although leftists who rejected CAFTA could have shifted their support to the PAC).
In line with past results linking education to support for trade, prevalence of university education in a district associates with greater referendum support,Footnote 41 while illiteracy, though associated with reduced support for market openness, has a substantively smaller, statistically insignificant effect. Other cultural factors also appear to matter for district votes just as past, survey-based results have suggested they matter for individuals. Areas with higher levels of immigration, which may suggest a more cosmopolitan local culture, tend to see greater support for the free trade area: a 10-percentage-point increase in the share of residents that are foreign born associates with a 2- to 3-percentage-point increase in referendum support. Conversely, districts containing indigenous zones see rates of support for the referendum a few percentage points below those of districts without indigenous zones, albeit at substandard levels of statistical significance.
The independent variables measuring urbanness and—except in column (2)—unemployment do not show statistically significant effects at standard levels. This is perhaps most interesting in the case of unemployment, given the traditional association between periods of unemployment and increased protectionism. Of course, the dynamics of effects across time such as those seen comparing recessions and booms need not carry over to cross-jurisdiction or cross-individual comparisons. It is nonetheless noteworthy that Costa Rican districts with higher unemployment rates show greater levels of support for the free trade area in every regression. While statistically insignificant effects are indicative at best, a positive point estimate certainly does not match prior findings that the unemployed are significantly more protectionist.Footnote 42 The finding could be a function of ecological inference, or it may result from Costa Rica being a relatively labor-rich, capital-poor society, where international flows are likely to benefit laborers.
Electoral District Results
As noted, the privacy of the ballot box typically forces the use of aggregate levels of analysis when considering referendum results, with the risks of ecological inference that poses. Aggregated results also pose another trade-off, in that larger geographic areas tend to have a broader range of available statistics but smaller geographic areas produce finer-grained results and a larger sample size. Considering results at smaller areas allows for a check of the robustness of the results in a context closer to the theoretically optimal level of individual response.
Along these lines, for some variables—including the essential dependent variable of share voting in favor of the free trade area in the plebiscite—information is available at the level of the electoral district (distrito electoral). These units are subdivisions of the (administrative) districts used in Table 2, and accordingly allow for observations of patterns within somewhat smaller populations. Whereas the average administrative district had about 5,600 eligible voters at the time of the plebiscite, the average electoral district had slightly fewer than 1,400. While this still does not allow for individual-level inference, it focuses on communities closer to the size of typical social networks (if further from the breadth of local economic interactions).Footnote 43
As the name “electoral districts” suggests, these units' primary purpose relates to tabulation of votes, and most available data for these districts relates to politics and elections. In particular, figures are available for vote shares for each of the political parties in the 2010 presidential elections, as well as the presence or absence of indigenous zones. Direct controls are not obtainable for the other measures used in Table 2. However, it is possible to control for the administrative district and thereby take into account all of the previously controlled-for factors at this higher level of aggregation, observing the effect of the political and indigenousness variables within those districts. Table 3 presents regression results both with and without these controls for administrative district.
OLS models of percent of Costa Rican voters favoring free trade, by electoral district
Note: Robust standard errors in parentheses. N = 1944. OLS = ordinary least squares.
The electoral-district models generally parallel those of the administrative-district models of Table 2. The only party whose support correlates in a statistically significant way with plebiscite votes is that of the anti-CAFTA PAC; greater support for that party consistently associates with lower CAFTA support. Even when taking administrative-district level fixed effects into account, each percentage-point increase in support for the PAC predicts slightly more than a one-half percentage-point fall in the proportion of voters supporting the referendum. The association with district vote shares for other parties is smaller, though it is notable that votes for the pro-CAFTA ML may associate with slightly lower support for the referendum.
The indigenous zones produce even more striking results. Areas associated with native populations tend to see support for joining the free trade area more than 10 percentage points lower than do otherwise equivalent nonindigenous areas, even when controlling for the electoral district population. This effect is highly statistically significant, and indeed far larger (14 to 18 percentage points rather than 2 to 7) and more precisely estimated than are the corresponding figures in Table 2.Footnote 44 This enhanced precision and larger effect could stem from the electoral districts' actually aligning perfectly with the definition of indigenous zones used.
Conclusion
Survey-based studies have done an excellent job establishing patterns of trade policy preferences. Less obtrusive measures, even at aggregated levels, can however usefully complement surveys by providing a more direct picture of actual political behavior. The 2007 Costa Rica referendum offers one source with which to compare survey results. These referendum-based data have weaknesses of their own—they typically rule out the causal claims of experimental designs, or even the individual-level correlations of survey studies—but they do provide a sense of how people react to questions of trade when they have given the matter some thought, as befits the higher stakes of a binding policy vote.Footnote 45
The plebiscite results are consistent with several of the major theories of attitudes toward the international economy. As predicted by consumption-based models, communities with more pensioners (who are likely to have a fixed income and so be sensitive to trade's price effects) have higher rates of voting in favor of the free trade agreement, while higher dependency ratios have a negative association with support for CAFTA. Socialization- and culture-based arguments also fit easily with the results here: more highly educated communities tend to show more support for trade liberalization in most models, while districts with fewer immigrants or more associations with indigenous groups tend to be less receptive to trade. (Education, of course, can also affect real returns to trade, though a factor-based view of education might not predict reduced protectionism among the educated in a developing country.) All these results accord with prior findings.
Also interesting are cases where the Costa Rican vote does not wholly confirm prior results. The results for partisanship, for example, belie simple left-right dynamics, hinting at a rich relationship between party and protectionism. Only one party's support showed a consistent connection to the plebiscite results, and, perhaps significantly, that was the PAC, the only major party that opposed the referendum. Other parties' votes had relatively weak relationships with votes on the free trade area, even though the 2006 presidential election used as a baseline measure of partisan support is generally framed as CAFTA-focused. While differences in national institutions and partisan structures make it hard to directly generalize this finding cross-nationally, it does suggest that careful consideration of those structures and institutions may yield insights about trade-policy preferences. The weak relationship with most parties' support further raises the possibility that the occasional attempts to interpret other general electionsFootnote 46 as being essentially about trade or international economic policies may rest on shaky premises.
Another suggestive finding is the generally poor showing of traditional economic theories of protectionism. While sectors of employment still very much correlate with vote choices—districts with more manufacturing tended to be more supportive of the free trade area, districts with larger proportions of the workforce in education and (usually) agriculture were less supportive—these alignments do not correspond neatly to the predictions of standard sectoral theories such as the Ricardo-Viner model. This is especially true of the connection between education-sector workers and protectionist sentiment, even if education is becoming more tradable. Nor do economically depressed circumstances, as measured here, appear to decrease willingness to open to the world economy, as often argued: the vote share in favor of CAFTA in districts with higher unemployment rates was either indistinguishable from or (in less complete models) higher than the vote share in districts with lower unemployment. Naturally, these findings do not rule out any effect of real returns or economic climate on attitudes toward trade, just as the lack of statistical significance does not conclusively indicate the absence of a relevant causal effect. Yet the indifferent influence of sectoral and unemployment variablesFootnote 47 observed here, using a novel methodology not based on surveys, matches a growing body of literature. Although some studies have found that those in more comparatively disadvantaged sectors are more likely to be protectionist,Footnote 48 many others have not.Footnote 49
Dependent on the Costa Rican context as all these results are, they open up another window on mass political attitudes—and, with turnout, behaviors—concerning international economic policy. The plebiscite provides further evidence in line with recent alternatives to the atomistic, materialistic traditional models of political economy. Subsequent analyses of this vote could more deeply explore how the various observable factors interact in their effects on trade preferences. More generally, further theoretical development may focus more on influences other than the direct sources of real income, which appear in this data set to be important in determining community attitudes toward trade. Furthermore, because attitudes toward other economic flows generally seem to have the same precursors as those concerning trade,Footnote 50 these ideas, and a referendum-based approach to studying public opinion, may extend to other issue areas throughout political economy.
