《国际政治经济学》文献资料（Political Economy of International Trade：Domestic Politics）Learning to Love Globalization. Education and Individual Attitudes Toward International Trade

Education and International Trade 473 ically,related studies show that higher levels of economic knowledge among sur- veyed individuals have large positive effects on support for free trade.3 This type of ideational argument does have precedents in both the international political economy and broader international relations literatures.Ideas,viewed as beliefs about cause-and-effect relationships,have been assigned key roles in accounts of policymaking in a variety of areas,including international cooperation on envi- ronmental issues and arms control.14 On the trade issue specifically,scholars have linked the removal of mercantilist restrictions on trade in Europe in the nineteenth century to the ideas of Smith and Ricardo and the birth of classical trade theory.5 The multilateral trade liberalization pursued among Western nations after 1945 has been connected to neoclassical economic theories and the spread of Keynesian ideas in particular.16 These types of accounts have traditionally focused on the impor- tance of particular ideas among policymakers,and the transmission of ideas to gov- ernment officials via transnational networks of experts and activists.7 Examining the distribution of economic ideas among voters,and how this might be connected to policy preferences,might be an interesting and important extension. Another plausible explanation for the relationship between education and atti- tudes toward trade focuses on differences in cultural values:highly educated indi- viduals are less prone than others to nationalist and antiforeigner sentiments that are often linked with protectionism in political debates.There is a large scholarly literature showing that education,at least in the United States,tends to socialize students to have more tolerant,pro-outsider views of the world.18 Education can foster tolerance,not just by increasing students'knowledge of foreign cultures and raising levels of critical thinking,but also by generating more diverse and cosmopolitan social networks,especially at the college level.Indeed,Betts has argued that one aspect of class identity that emerges among the college-educated in Western nations is a commitment to cosmopolitanism and an appreciation for diverse cultures.20 Studies of survey data also show that various measures of nation- alism and national pride are significant predictors of support for trade protection.2 13.See Walstad 1997;and Walstad and Rebeck 2002.Walstad and Rebeck 2002 make the larger point that scholarly analysis of public opinion on economic issues makes the erroneous implicit assump- tion that survey respondents are undifferentiated in terms of their economic knowledge.In fact,indi- viduals differ dramatically in their levels of economic knowledge,as measured by their scores on a set of test questions covering basic economic concepts and facts,and such knowledge scores are signifi- cant predictors of attitudes about a range of economic policy issues 14.For a general discussion,see Goldstein and Keohane 1993;on environmental negotiations,see Haas 1992;on arms control,see Adler 1992;and Price 1998. 15.See Kindleberger 1975;and Bhagwati 1988. 16.See Ruggie 1998;and Goldstein 1988.See also Hall 1989 for essays on the particular impact of Keynesian macroeconomic theory among policymakers in the 1930s and 1940s. 17.For example,Haas 1992;and Keck and Sikkink 1998. 18.For example,Campbell et al.1960,475-81;Erikson,Luttbeg,and Tedin 1991,155-56;and McClosky and Brill 1983. 19.See Case,Greeley,and Fuchs 1989;and Allport 1954. 20.Bets1988. 21.See Mayda and Rodrik 2005,1414-16;and O'Rourke and Sinnott 2002,173.On these points, see also Holsti 1996,87-88

Education and International Trade 473 ically, related studies show that higher levels of economic knowledge among sur￾veyed individuals have large positive effects on support for free trade."3 This type of ideational argument does have precedents in both the international political economy and broader international relations literatures. Ideas, viewed as beliefs about cause-and-effect relationships, have been assigned key roles in accounts of policymaking in a variety of areas, including international cooperation on envi￾ronmental issues and arms control.14 On the trade issue specifically, scholars have linked the removal of mercantilist restrictions on trade in Europe in the nineteenth century to the ideas of Smith and Ricardo and the birth of classical trade theory.15 The multilateral trade liberalization pursued among Western nations after 1945 has been connected to neoclassical economic theories and the spread of Keynesian ideas in particular.16 These types of accounts have traditionally focused on the impor￾tance of particular ideas among policymakers, and the transmission of ideas to gov￾ernment officials via transnational networks of experts and activists.I Examining the distribution of economic ideas among voters, and how this might be connected to policy preferences, might be an interesting and important extension. Another plausible explanation for the relationship between education and atti￾tudes toward trade focuses on differences in cultural values: highly educated indi￾viduals are less prone than others to nationalist and antiforeigner sentiments that are often linked with protectionism in political debates. There is a large scholarly literature showing that education, at least in the United States, tends to socialize students to have more tolerant, pro-outsider views of the world.'" Education can foster tolerance, not just by increasing students' knowledge of foreign cultures and raising levels of critical thinking, but also by generating more diverse and cosmopolitan social networks, especially at the college level.19 Indeed, Betts has argued that one aspect of class identity that emerges among the college-educated in Western nations is a commitment to cosmopolitanism and an appreciation for diverse cultures.20 Studies of survey data also show that various measures of nation￾alism and national pride are significant predictors of support for trade protection.21 13. See Walstad 1997; and Walstad and Rebeck 2002. Walstad and Rebeck 2002 make the larger point that scholarly analysis of public opinion on economic issues makes the erroneous implicit assump￾tion that survey respondents are undifferentiated in terms of their economic knowledge. In fact, indi￾viduals differ dramatically in their levels of economic knowledge, as measured by their scores on a set of test questions covering basic economic concepts and facts, and such knowledge scores are signifi￾cant predictors of attitudes about a range of economic policy issues. 14. For a general discussion, see Goldstein and Keohane 1993; on environmental negotiations, see Haas 1992; on arms control, see Adler 1992; and Price 1998. 15. See Kindleberger 1975; and Bhagwati 1988. 16. See Ruggie 1998; and Goldstein 1988. See also Hall 1989 for essays on the particular impact of Keynesian macroeconomic theory among policymakers in the 1930s and 1940s. 17. For example, Haas 1992; and Keck and Sikkink 1998. 18. For example, Campbell et al. 1960, 475-81; Erikson, Luttbeg, and Tedin 1991, 155-56; and McClosky and Brill 1983. 19. See Case, Greeley, and Fuchs 1989; and Allport 1954. 20. Betts 1988. 21. See Mayda and Rodrik 2005, 1414-16; and O'Rourke and Sinnott 2002, 173. On these points, see also Holsti 1996, 87-88

474 International Organization This type of argument resonates with a growing body of research in inter- national relations that focuses on the importance of cultural values and concep- tions of identity and how they shape the interests pursued by policymakers in international affairs-in areas including military strategy,humanitarian interven- tion,and support for human rights.22 It is more difficult to find cultural accounts of trade politics or policymaking,specifically,in the political economy literature, though references to the popular appeal of protectionism when linked with nation- alism and xenophobia are common.23 How then should we interpret the observed connection between education and support for trade?Besides the standard account,which focuses on Stolper- Samuelson-style distributional concerns,alternative accounts that stress the impor- tance of economic ideas and related values also seem highly plausible.We suggest a simple test to establish whether the education connection is primarily reflecting concerns about the effects of trade on respondents'personal incomes or whether it is a manifestation of broader differences in ideas and/or values among sur- veyed individuals.We examine the impact of education levels on attitudes toward trade among respondents currently in the active labor force and among those who are not.If the Stolper-Samuelson interpretation of the education effect is accu- rate,this effect should be contingent on whether or not individuals are actually being paid for the employment of their skills in the labor market.24 The Effects of Education on Trade Preferences Our new empirical tests draw from two key sources of data on individual trade preferences:the NES and the International Social Survey Program (ISSP).These are the same data sets employed by the two most prominent studies of individual attitudes toward trade in recent years:the analyses by Scheve and Slaughter and Mayda and Rodrik.5 In the following section we briefly describe each dataset and present the results for our split-sample tests. Tests Using the NES Data The NES surveys are fielded in the United States around the time of presidential elections and designed to gather data on Americans'social backgrounds,political predispositions,opinions on questions of public policy,and participation in polit- 22.For general discussions,see Wendt 1999;Ruggie 1998;and Finnemore and Sikkink 1998.For studies of strategy,see Katzenstein 1996;on human rights,see Sikkink 1993. 23.For example,Bauer,Pool,and Dexter 1972,103. 24.Similar split-sample tests have been used in studies of anti-immigrant sentiments to help dis- cern whether greater opposition to immigration among less educated survey respondents(compared with more educated counterparts)reflects different degrees of concern about having to compete with immigrants in the job market:see Scheve and Slaughter 2001c;Mayda 2004;and also Hainmueller and Hiscox 2004. 25.Scheve and Slaughter 2001a and 2001b;and Mayda and Rodrik 2005

474 International Organization This type of argument resonates with a growing body of research in inter￾national relations that focuses on the importance of cultural values and concep￾tions of identity and how they shape the interests pursued by policymakers in international affairs-in areas including military strategy, humanitarian interven￾tion, and support for human rights.22 It is more difficult to find cultural accounts of trade politics or policymaking, specifically, in the political economy literature, though references to the popular appeal of protectionism when linked with nation￾alism and xenophobia are common.23 How then should we interpret the observed connection between education and support for trade? Besides the standard account, which focuses on Stolper￾Samuelson-style distributional concerns, alternative accounts that stress the impor￾tance of economic ideas and related values also seem highly plausible. We suggest a simple test to establish whether the education connection is primarily reflecting concerns about the effects of trade on respondents' personal incomes or whether it is a manifestation of broader differences in ideas and/or values among sur￾veyed individuals. We examine the impact of education levels on attitudes toward trade among respondents currently in the active labor force and among those who are not. If the Stolper-Samuelson interpretation of the education effect is accu￾rate, this effect should be contingent on whether or not individuals are actually being paid for the employment of their skills in the labor market.24 The Effects of Education on Trade Preferences Our new empirical tests draw from two key sources of data on individual trade preferences: the NES and the International Social Survey Program (ISSP). These are the same data sets employed by the two most prominent studies of individual attitudes toward trade in recent years: the analyses by Scheve and Slaughter and Mayda and Rodrik.25 In the following section we briefly describe each dataset and present the results for our split-sample tests. Tests Using the NES Data The NES surveys are fielded in the United States around the time of presidential elections and designed to gather data on Americans' social backgrounds, political predispositions, opinions on questions of public policy, and participation in polit- 22. For general discussions, see Wendt 1999; Ruggie 1998; and Finnemore and Sikkink 1998. For studies of strategy, see Katzenstein 1996; on human rights, see Sikkink 1993. 23. For example, Bauer, Pool, and Dexter 1972, 103. 24. Similar split-sample tests have been used in studies of anti-immigrant sentiments to help dis￾cern whether greater opposition to immigration among less educated survey respondents (compared with more educated counterparts) reflects different degrees of concern about having to compete with immigrants in the job market: see Scheve and Slaughter 2001c; Mayda 2004; and also Hainmueller and Hiscox 2004. 25. Scheve and Slaughter 2001a and 2001b; and Mayda and Rodrik 2005

Education and International Trade 475 ical life.26 In 1992,for the first time,the NES included a question that asked respondents about their attitudes toward international trade.The question was the following:"Some people have suggested placing new limits on foreign imports in order to protect American jobs.Others say that such limits would raise con- sumer prices and hurt American exports.Do you favor or oppose placing new limits on imports,or haven't you thought much about this?" Scheve and Slaughter used responses to this question from the 1992 NES sur- vey as their measure of individual trade policy preferences.27 We have replicated their approach here to conduct the split-sample test,while also examining data from the 1996 NES survey that included the same trade question.28 We created the dichotomous dependent variable TRADE OPINION,coded as 1 for responses that favored protection("new limits on foreign imports")and 0 for those opposed.The "haven't thought much about this"answers are coded as missing,as in the Scheve and Slaughter analysis. The principal measure of education is sCHOOLING,which simply records the years of full-time education completed by each respondent (a cap is set by the NES at seventeen years).This is the measure employed by Scheve and Slaugh- ter.To allow for nonlinear education effects,we have also constructed a set of dummy variables indicating each respondent's highest level of educational at- tainment:JUNIOR HIGH (1 =8 years of schooling;0 otherwise);HIGH SCHOOL (1 high school degree;0=otherwise);HIGHER EDUCATION (1 some years of post-high school education,including junior or community college;0=other- wise);CoLLEGE (1 four-year bachelor's degree;0 otherwise);and GRADU- ATE(1=postgraduate degree;0=otherwise).29 Assuming that education has linear effects on trade preferences seems appropriate if one assumes that each additional year of education (of any type)improves an individual's skills(and changes his or her attitudes)by a roughly constant amount.It is less appropriate if education has other,nonlinear types of effects associated with the ideas and information individuals possess about the way the economy works or the kinds of values and preferences they develop-as noted above,college education appears to play an overwhelming role in terms of its impact on ideas and cultural values among individuals. For different subsamples of respondents,we estimated binary probit models using TRADE OPINION as the dependent variable,testing for the effects of SCHOOLING or the different education dummy variables while controlling for a variety of other individual characteristics that might plausibly affect trade preferences.We esti- mated each model with two sets of covariates:a limited set of standard sociode- 26.For details,see Miller et al.1992;and Rosenstone et al.1996.For more on NES see (http:// www.umich.edu/-nes/).Accessed 10 January 2006. 27.Scheve and Slaughter 2001a. 28.Also examined in Scheve and Slaughter 2001b. 29.Note that there are too few respondents who failed to finish elementary school to allow us to estimate the separate effect of elementary-level education here:the excluded category in our analysis of education effects when using these dummy variables is all those with less than eight years of schooling

Education and International Trade 475 ical life.26 In 1992, for the first time, the NES included a question that asked respondents about their attitudes toward international trade. The question was the following: "Some people have suggested placing new limits on foreign imports in order to protect American jobs. Others say that such limits would raise con￾sumer prices and hurt American exports. Do you favor or oppose placing new limits on imports, or haven't you thought much about this?" Scheve and Slaughter used responses to this question from the 1992 NES sur￾vey as their measure of individual trade policy preferences.27 We have replicated their approach here to conduct the split-sample test, while also examining data from the 1996 NES survey that included the same trade question.28 We created the dichotomous dependent variable TRADE OPINION, coded as 1 for responses that favored protection ("new limits on foreign imports") and 0 for those opposed. The "haven't thought much about this" answers are coded as missing, as in the Scheve and Slaughter analysis. The principal measure of education is SCHOOLING, which simply records the years of full-time education completed by each respondent (a cap is set by the NES at seventeen years). This is the measure employed by Scheve and Slaugh￾ter. To allow for nonlinear education effects, we have also constructed a set of dummy variables indicating each respondent's highest level of educational at￾tainment: JUNIOR HIGH (1 = 8 years of schooling; 0 = otherwise); HIGH SCHOOL (1 = high school degree; 0 = otherwise); HIGHER EDUCATION (1 = some years of post-high school education, including junior or community college; 0 = other￾wise); COLLEGE (1 = four-year bachelor's degree; 0 = otherwise); and GRADU￾ATE (1 = postgraduate degree; 0 = otherwise).29 Assuming that education has linear effects on trade preferences seems appropriate if one assumes that each additional year of education (of any type) improves an individual's skills (and changes his or her attitudes) by a roughly constant amount. It is less appropriate if education has other, nonlinear types of effects associated with the ideas and information individuals possess about the way the economy works or the kinds of values and preferences they develop-as noted above, college education appears to play an overwhelming role in terms of its impact on ideas and cultural values among individuals. For different subsamples of respondents, we estimated binary probit models using TRADE OPINION as the dependent variable, testing for the effects of SCHOOLING or the different education dummy variables while controlling for a variety of other individual characteristics that might plausibly affect trade preferences. We esti￾mated each model with two sets of covariates: a limited set of standard sociode- 26. For details, see Miller et al. 1992; and Rosenstone et al. 1996. For more on NES see (http:// www.umich.edu/-nes/). Accessed 10 January 2006. 27. Scheve and Slaughter 2001a. 28. Also examined in Scheve and Slaughter 2001b. 29. Note that there are too few respondents who failed to finish elementary school to allow us to estimate the separate effect of elementary-level education here: the excluded category in our analysis of education effects when using these dummy variables is all those with less than eight years of schooling

476 International Organization mographic controls (age,in years,gender,and race),which preserved the maximum number of observations across the subsamples;and a more extensive set of con- trols (the standard controls plus indicators of UNION MEMBERSHIP,PARTY IDENTI- FICATION,and IDEOLOGY),which closely matched the more extensive specifications used by Scheve and Slaughter.30(See the Appendix for a description of all vari- ables used along with summary statistics). We expect the measures of education to be negatively associated with support for trade protection(as measured by TRADE OPINION),either because highly skilled individuals expect trade to increase their real wages and poorly skilled respon- dents expect trade to decrease their real wages (a la Stolper-Samuelson),because more-educated respondents know more about the overall economic benefits asso- ciated with trade openness,and/or because the more-educated are less likely to nurture antiforeigner sentiments.If this link between education and individual atti- tudes toward trade is primarily due to expectations about wages,however,the results from our estimations of trade preferences among respondents not actively engaged in the labor market should differ substantially when compared with those from our estimations of preferences among respondents who are currently employed.To test for this difference,we created subsamples of the full NES survey sample,separat- ing those who were in paid work from those not in paid work.31 Since those not 30.We have used age in years,because this is the most straightforward approach and is the mea- sure provided in the NES data;Scheve and Slaughter 2001b instead included a range of dummy vari- ables covering separate age brackets.We have reestimated all our results using the respective age dummies and the results are virtually identical.We have also used dummy variables for multiple racial categories here,while Scheve and Slaughter appear to have used a single-race dummy variable.Again, the results are almost identical regardless of how the race variable is entered in the models.The main variables we exclude are the ones constructed by Scheve and Slaughter using non-NES data:sEcToR NET EXPORT SHARE(net exports as a share of output for the industry in which the respondent is employed),and sECToR TARIFF(customs duties as a share of the total value of imports for the industry in which the respondent is employed).Neither of these variables have significant effects on trade pref- erences,according to the results reported by Scheve and Slaughter(2001b,59-60),and their inclusion in the estimations makes no difference at all to the estimated effects of education on trade opinion.We also excluded the other variables Scheve and Slaughter derive from non-NES data-the measure of oCCUPATION WAGE (the national average weekly wage for the respondent's occupation,which they use as an alternative to years of education as an indicator of skill level),and measures of couNTY EXPOSURE to trade liberalization (the shares of total employment in the respondent's home county accounted for by industries with above average tariffs or net imports).Scheve and Slaughter find that county exposure,when interacted with a dichotomous indicator of home ownership,does have a sig- nificant negative impact on support for trade openness,but the inclusion of these added controls in their models has a minuscule and statistically insignificant effect on the estimated impact of education on trade preferences(2001a,285;2001b,64),so we have not attempted to replicate the construction of these added controls here.Note that,because there is very little missing data for variables measured in the NES (the schooling measure has only 6 percent of observations missing in 1992,and 0 percent missing in 1996),and there is less missing data for all the other variables),we do not have to resort to imputation of missing data-Scheve and Slaughter(2001a,278)report having to impute up to 73.4 percent of the observations for other(non-NES)variables they included in their analysis. 31.The subsamples are determined by answers to a NES question asking respondents about their current employment status.Answer choices included the following:working now,temporarily laid-off, unemployed,retired,student,homemaker,and permanently disabled.Our"currently in paid work" subsample includes those that answered"in work now";those answering otherwise are classified as "currently not in paid work."Those with missing employment status are coded as missing.We also

476 International Organization mographic controls (age, in years, gender, and race), which preserved the maximum number of observations across the subsamples; and a more extensive set of con￾trols (the standard controls plus indicators of UNION MEMBERSHIP, PARTY IDENTI￾FICATION, and IDEOLOGY), which closely matched the more extensive specifications used by Scheve and Slaughter.30 (See the Appendix for a description of all vari￾ables used along with summary statistics). We expect the measures of education to be negatively associated with support for trade protection (as measured by TRADE OPINION), either because highly skilled individuals expect trade to increase their real wages and poorly skilled respon￾dents expect trade to decrease their real wages (at la Stolper-Samuelson), because more-educated respondents know more about the overall economic benefits asso￾ciated with trade openness, and/or because the more-educated are less likely to nurture antiforeigner sentiments. If this link between education and individual atti￾tudes toward trade is primarily due to expectations about wages, however, the results from our estimations of trade preferences among respondents not actively engaged in the labor market should differ substantially when compared with those from our estimations of preferences among respondents who are currently employed. To test for this difference, we created subsamples of the full NES survey sample, separat￾ing those who were in paid work from those not in paid work.31 Since those not 30. We have used age in years, because this is the most straightforward approach and is the mea￾sure provided in the NES data; Scheve and Slaughter 2001b instead included a range of dummy vari￾ables covering separate age brackets. We have reestimated all our results using the respective age dummies and the results are virtually identical. We have also used dummy variables for multiple racial categories here, while Scheve and Slaughter appear to have used a single-race dummy variable. Again, the results are almost identical regardless of how the race variable is entered in the models. The main variables we exclude are the ones constructed by Scheve and Slaughter using non-NES data: SECTOR NET EXPORT SHARE (net exports as a share of output for the industry in which the respondent is employed), and SECTOR TARIFF (customs duties as a share of the total value of imports for the industry in which the respondent is employed). Neither of these variables have significant effects on trade pref￾erences, according to the results reported by Scheve and Slaughter (2001b, 59-60), and their inclusion in the estimations makes no difference at all to the estimated effects of education on trade opinion. We also excluded the other variables Scheve and Slaughter derive from non-NES data-the measure of OCCUPATION WAGE (the national average weekly wage for the respondent's occupation, which they use as an alternative to years of education as an indicator of skill level), and measures of COUNTY EXPOSURE to trade liberalization (the shares of total employment in the respondent's home county accounted for by industries with above average tariffs or net imports). Scheve and Slaughter find that county exposure, when interacted with a dichotomous indicator of home ownership, does have a sig￾nificant negative impact on support for trade openness, but the inclusion of these added controls in their models has a minuscule and statistically insignificant effect on the estimated impact of education on trade preferences (2001a, 285; 2001b, 64), so we have not attempted to replicate the construction of these added controls here. Note that, because there is very little missing data for variables measured in the NES (the schooling measure has only 6 percent of observations missing in 1992, and 0 percent missing in 1996), and there is less missing data for all the other variables), we do not have to resort to imputation of missing data-Scheve and Slaughter (2001a, 278) report having to impute up to 73.4 percent of the observations for other (non-NES) variables they included in their analysis. 31. The subsamples are determined by answers to a NES question asking respondents about their current employment status. Answer choices included the following: working now, temporarily laid-off, unemployed, retired, student, homemaker, and permanently disabled. Our "currently in paid work" subsample includes those that answered "in work now"; those answering otherwise are classified as "currently not in paid work." Those with missing employment status are coded as missing. We also

Education and International Trade 477 currently in paid work include a varied set of individuals,such as those that are unemployed,students,and homemakers(and may be seeking paid work or plan to seek to work soon),we have also isolated one particular group-those individuals who are retired-who are highly unlikely to reenter paid work in the future and be concerned about how their(potential)wages might be affected by trade.32 The key results from the estimations are reported in Table 1,which displays the estimated effects of education on individual trade preferences in the full NES sam- ple and in each of the different subsamples.To facilitate comparison across sub- samples,rather than showing estimated probit coefficients,we report estimated marginal effects:that is,the change in the probability of favoring protectionism associated with an infinitesimal change in sCHOOLING(for the specific dummy vari- ables for levels of highest educational attainment,the discrete change in the prob- ability is shown). Comparing the results across the subsamples,we find little difference in the estimated effects of education on attitudes toward trade.In all cases,the estimated effects of SCHOOLING are similar,both in terms of magnitude and level of statisti- cal significance,across all models(none of the coefficients is significantly differ- ent from the others across subsamples at conventional levels).This is true for estimations using both the 1992 and the 1996 NES data.For example,in the case of the 1996 survey,using the extensive set of covariates,a change from zero to seventeen years of sCHOOLING(while holding the other covariates at their respec- tive sample means)is associated with an average decrease in the probability of favoring protection of about 0.59(s.e.0.05)for the full sample,0.51 (s.e.0.09) for those currently in paid work,0.48(s.e.0.11)for those currently not in paid work,and 0.53(s.e.0.11)for those who are retired (Models 9 to 12,Panel B).33 The observed relationship between education and trade preferences becomes even more similar across subsamples once we replace the sCHOOLING measure with the separate education dummies.For example,compared to individuals with less than junior high-level educations,completing a CoLLEGE education decreases the probability of being in favor of protection by about 0.28(s.e.0.08)for the full include those few retired/students/disabled/homemakers who also indicated that they are "currently working more than 20 hours per week"in our"currently in paid work"subsample.When the latter are excluded from the"currently in paid work"subsample,the magnitudes of the schooling effect become, if anything,more similar across the in-and out-of-paid-work subsamples than in the results we show here.Full results of these robustness tests are available on request. 32.While pensions for retired workers in some prominent U.S.industries (for example,steel)have been linked to the financial health of their former employers,this is the exception and not the rule. Recent studies of U.S.retirees indicate that less than 17 percent of retirement income in the median household comes from employer-provided pension plans(see Sass 2003,6;and Social Security Admin- istration 2002).The connection between employer-provided pensions and the financial health of the firm also is attenuated by the standards for funding and fiduciary conduct established by the Employee Retirement Income Security Act in 1974(see Sass 1997).On this issue,we might also note that we get identical results when we perform the same tests comparing retirees with workers using the ISSP data (see below),drawn from a variety of countries with a variety of pension and retirement income systems. 33.Predicted effects here,and below,are calculated using the"Clarify"software developed by King, Tomz,and Wittenberg 2001.For each such calculation,all other covariates are set at the sample mean values

Education and International Trade 477 currently in paid work include a varied set of individuals, such as those that are unemployed, students, and homemakers (and may be seeking paid work or plan to seek to work soon), we have also isolated one particular group-those individuals who are retired-who are highly unlikely to reenter paid work in the future and be concerned about how their (potential) wages might be affected by trade.32 The key results from the estimations are reported in Table 1, which displays the estimated effects of education on individual trade preferences in the full NES sam￾ple and in each of the different subsamples. To facilitate comparison across sub￾samples, rather than showing estimated probit coefficients, we report estimated marginal effects: that is, the change in the probability of favoring protectionism associated with an infinitesimal change in SCHOOLING (for the specific dummy vari￾ables for levels of highest educational attainment, the discrete change in the prob￾ability is shown). Comparing the results across the subsamples, we find little difference in the estimated effects of education on attitudes toward trade. In all cases, the estimated effects of SCHOOLING are similar, both in terms of magnitude and level of statisti￾cal significance, across all models (none of the coefficients is significantly differ￾ent from the others across subsamples at conventional levels). This is true for estimations using both the 1992 and the 1996 NES data. For example, in the case of the 1996 survey, using the extensive set of covariates, a change from zero to seventeen years of SCHOOLING (while holding the other covariates at their respec￾tive sample means) is associated with an average decrease in the probability of favoring protection of about 0.59 (s.e. 0.05) for the full sample, 0.51 (s.e. 0.09) for those currently in paid work, 0.48 (s.e. 0.11) for those currently not in paid work, and 0.53 (s.e. 0.11) for those who are retired (Models 9 to 12, Panel B).33 The observed relationship between education and trade preferences becomes even more similar across subsamples once we replace the SCHOOLING measure with the separate education dummies. For example, compared to individuals with less than junior high-level educations, completing a COLLEGE education decreases the probability of being in favor of protection by about 0.28 (s.e. 0.08) for the full include those few retired/students/disabled/homemakers who also indicated that they are "currently working more than 20 hours per week" in our "currently in paid work" subsample. When the latter are excluded from the "currently in paid work" subsample, the magnitudes of the schooling effect become, if anything, more similar across the in- and out-of-paid-work subsamples than in the results we show here. Full results of these robustness tests are available on request. 32. While pensions for retired workers in some prominent U.S. industries (for example, steel) have been linked to the financial health of their former employers, this is the exception and not the rule. Recent studies of U.S. retirees indicate that less than 17 percent of retirement income in the median household comes from employer-provided pension plans (see Sass 2003, 6; and Social Security Admin￾istration 2002). The connection between employer-provided pensions and the financial health of the firm also is attenuated by the standards for funding and fiduciary conduct established by the Employee Retirement Income Security Act in 1974 (see Sass 1997). On this issue, we might also note that we get identical results when we perform the same tests comparing retirees with workers using the ISSP data (see below), drawn from a variety of countries with a variety of pension and retirement income systems. 33. Predicted effects here, and below, are calculated using the "Clarify" software developed by King, Tomz, and Wittenberg 2001. For each such calculation, all other covariates are set at the sample mean values