Skip to main content

Individual Attitudes Towards Trade: Stolper-Samuelson Revisited


Using the 2007 wave of the Pew Global Attitudes Project, this paper finds statistically significant and economically large Stolper-Samuelson effects in individuals’ preference formation towards trade policy. High-skilled individuals are substantially more pro-trade than low-skilled individuals in high-skilled labor abundant countries, and vice versa in a considerable share of low-skilled labor abundant countries. Our novel international survey data combine a number of desirable features which allow us to paint a more distinct and thus more convincing picture of the role of the Heckscher-Ohlin model in shaping free trade attitudes, relative to existing literature.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2


  1. 1.

    In the year prior to the Global Financial Crisis, China’s economy has grown in real terms by 13.0 %, followed by those of India (9.1 %), Russia (8.1 %), and Brazil (5.3 %); the degrees of trade openness (measured as the sum of the value of imports and exports over total output) for these economies range between 25 % (Brazil) and 75 % (China) in 2007; all four countries together comprise nearly 2.8 billion people in 2007, which was then as much as about 42 % of total world population; all data come from the World Development Indicators (2007).

  2. 2.

    In fact, there is not a single OECD economy with a majority of people having attained tertiary education, the level of education which is typically seen as essential in qualifying for a high-skilled job. The OECD average of people with tertiary education in 2007 is 28 % for the population aged 25–64; see OECD (2009, p. 29f.).

  3. 3.

    Gallup’s annual World Affairs poll reports that, in January 2000, 35 % of the American adult population believed that trade is a “threat to the economy from foreign imports”. This number has almost steadily increased over the years, reaching a critical level of 52 % in February 2008, an all time high since September 1992; see A similar trend can be found in Western Europe, but not in China or India; see Pew GAP (2007, p. 1).

  4. 4.

    Recent years have seen a surge in empirical research on individual trade policy preferences. For example, evidence from purely national surveys comes from Scheve and Slaughter (2001), Hoffman (2009), Ehrlich and Maestas (2010), and Blonigen (2011) for the United States and from Wolfe and Mendelsohn (2005) for Canada. Beaulieu et al. (2005) document cross-country evidence from Latin America.

  5. 5.

    For example, Hainmueller and Hiscox (2006) and Mansfield and Mutz (2009) argue that high-skilled individuals are more likely to favor free trade due to a general “enlightenment” that comes with a better educational background. In this paper, we explicitly capture individuals’ economic awareness and their inclinations towards nationalist ideas and carry out baseline estimations of the effects of various aspects of individual enlightenment.

  6. 6.

    O’Rourke and Sinnott (2001) and Mayda and Rodrik (2005) use GDP per capita as a proxy variable for a country’s relative endowment with high-skilled labor. Mayda and Rodrik (2001) pursue a similar strategy as we do here.

  7. 7.

    Arguably, the blind spot in O’Rourke and Sinnott (2001) and Mayda and Rodrik (2005) is that labor-abundant, low-income countries are scarce in the ISSP survey data. Mayda and Rodrik (2005) also employ the World Values Survey (WVS) data collected between 1995 and 1997 and covering 40 countries from all stages of development. However, the WVS does not allow to control for individual income, which is paramount in the estimation. The same holds true for Hainmueller and Hiscox (2006) who additionally use the 2003 wave of the Pew GAP data including 44 countries.

  8. 8.

    Our argument is related, but not identical, to that raised by Ai and Norton (2003) who stress the by now well-known difficulties of computing and interpreting interaction effects in non-linear models such as the Probit or the Logit model.

  9. 9.

    The notion of a “general” version of the theorem was introduced by Bhagwati (1959); see also Deardorff (1993).

  10. 10.

    In what follows, the terms ‘high-skilled labor’ and ‘human capital’ are used interchangeably. Analogously for ‘low-skilled labor’ and ‘labor’.

  11. 11.

    Metzler (1949) shows that the imposition of an import tariff raises the domestic relative price of the imported good only if the elasticity of foreign demand for domestic exports is greater than the domestic marginal propensity to consume the exported good. The restriction to the small economy case precludes any terms-of-trade effects and is therefore sufficient to obtain this result.

  12. 12.

    In fact, many authors have interpreted the marginal effect of the interaction term as the interaction effect; see Ai and Norton (2003).

  13. 13.

    For summary statistics and coding information of these variables, see Table 5 in the Appendix.

  14. 14.

    For further information on the GAP survey data, see also

  15. 15.

    We have also applied alternative dummy definitions. In particular, we have assigned non-respondents to either the pro-trade or the anti-trade group of people. All qualitative results reported in this paper are insensitive to this type of recoding.

  16. 16.

    Strictly hierarchical classes are (0) no formal education or incomplete primary education, (1) complete primary education, (2) incomplete secondary education (technical/vocational), (3) complete secondary education (technical/vocational) / incomplete secondary education (university-preparatory) / complete secondary education (university-preparatory), (4) some university education (without degree), (5) university education (with degree). There is some cross-country heterogeneity in the survey categories of educational attainment. More information on how we map country-specific groups of educational attainment into the above hierarchical structure is available upon request.

  17. 17.

    An alternative, more flexible model specification uses dummy variables for the different educational categories. This pays attention to the fact that differences in educational attainment reflect an ordinal instead of a cardinal scale. The estimates (not reported) do not alter any of the conclusions drawn in this paper.

  18. 18.

    Employing the median instead of the mean yields virtually identical results.

  19. 19.

    Sampling weights correct for deviations from random sampling.

  20. 20.

    GTA is a recently established academic initiative for monitoring state policies that may detrimentally affect global trade integration in one way or the other. It is coordinated by the Centre for Economic Policy Research (CEPR), London, UK. See Table 7 in the Appendix for a definition of protectionist policy measures and for more information on this data source.

  21. 21.

    The evidence also suggests that this relationship becomes tighter, the more democratic a political regime is. To see this, we regress the count of protectionist policy measures on the main and interaction terms of the average trade opinion and the country’s democracy index (from the Economist Intelligence Unit). Results show that the link between voter attitudes and policies is strongest in the most democratic countries such as Sweden, and practically non-existent in authoritarian regimes such as China.

  22. 22.

    Throughout our analysis, we estimate heteroskedastic robust standard errors to immunize inference against misspecification; see White (1980). Given our assumptions in Section 2.2, stochastic and non-stochastic country effects (σ c and b c ) induce correlation among individual observations within country clusters. Whenever we introduce country fixed effects, however, the γ c ’s capture any such type of within-country correlation. At any rate, inference based on cluster robust standard errors may be misleading if the number of clusters is small as in our case (< 50); see Cameron and Miller (2010).

  23. 23.

    A drawback is that the variable is binary and does not distinguish among different religious groups; see Daniels (2005) for the effect of religious affiliation on people’s attitudes towards a number of international policy issues. Lewer and Van den Berg (2007) estimate a gravity equation in order to study the role of religion for bilateral trade flows.

  24. 24.

    Throughout most of our regression analysis, the linear models predict probabilities of being pro-trade outside the closed unit interval for about half a percent of all estimation sample observations. Whenever outside the unit interval, predictions exceed one, but only by a marginal amount.

  25. 25.

    Results from Table 1 seem to suggest that the impact of income on free trade preferences varies across countries as well. However, in a robustness check to Table 2 where we include an interaction term of Income with GDP per capita, we find contrary evidence in the sense that this latter interaction effect is insignificant.

  26. 26.

    If we exclude individuals that are self-employed from the estimation sample, Hypothesis 1 is reinforced further. This is reassuring because for self-employed individuals we would expect the impact of trade to be driven by other factors than individual skill.

  27. 27.

    These variables refer to a country’s stage of development (GDP per capita, Country Mean of Skill), its institutions (Electoral Process, Political Pluralism and Participation, Functioning of Government, Freedom of Speech and Belief, Associational and Organizational Rights, Rule of Law, Personal Autonomy and Individual Rights), and other aspects (Trade Openness, Labor Force Share). See Table 7 in the Appendix for coding and data sources.

  28. 28.

    To facilitate comparison to columns (1) to (4), in columns (5) to (8) the marginal effect of Skill is evaluated at h c  = 0 and at estimation sample averages of all other covariates. Similarly, reported marginal effects of Country Mean of Skill are evaluated at h ic  = 0 when interacted with Skill.

  29. 29.

    Regression results for the full set of country-level control variables are available on request from the authors.

  30. 30.

    The first aspect is a particularly serious concern in our application, because the question on trade preferences does not address the distributional consequences of international trade within the respondent’s country, but rather the implications for the country at large.

  31. 31.

    To exclude the possibility that changes in estimated coefficients reflect mere changes in sample composition, we employ exactly the same estimation sample in all specifications.

  32. 32.

    The questionnaire allows for four different answer categories, from “completely disagree” (0) to “completely agree” (3).

  33. 33.

    Moreover, the question does not refer to issues such as international trade, trade liberalization, or globalization, at least not explicitly. Answers to this question are thus not subject to what has been dubbed justification bias in the literature on opinion polls. This type of bias would arise if individuals were partly using their answers as a means of ex post justification for their (positive or negative) preferences towards trade; see Malchow-Møller et al. (2009).

  34. 34.

    This variable is based on the following survey question: “Which of the following two statements best describes you: ‘I follow INTERNATIONAL news closely ONLY when something important is happening.’ OR ‘I follow INTERNATIONAL news closely most of the time, whether or not something important is happening’?”. The indicator Informed is coded (1) for individuals who choose the second statement.

  35. 35.

    The variable Sociotropic Views takes on integer values from 0 (“completely disagree”) to 3 (“completely agree”).

  36. 36.

    The dichotomous variable Fears of Cultural Spill-Overs is coded (1) if respondents take a positive stance on spreading American ideas and customs, and (0) for negative views. Obviously, answers to this question are heavily loaded by the explicit reference to the United States. Our data show that anti-American sentiments are popular in both developing and developed countries. That said, we argue that our indicator variable also captures fears of the cultural impact of globalization in general, and we expect the purely American-specific element to be independent of individual trade policy preferences.

  37. 37.

    Again, the survey allows for four different answers, ranging from “completely disagree” (0) to “completely agree” (3).

  38. 38.

    In the survey, each of the answers is assigned to one of the following categories: “Economic/financial problems”, “Health”, “Education and children”, “Housing”, “Social relations”, “Work”, “Transportation”, “Crime”, “Problems related to government”, “Terrorism and war”, “Other”. Each category comprises two to six pre-specified subcategories plus a “residual” group for answers which do not fit into any one of the given subcategories. We categorize the following subcategories as indicating economic or financial concerns: “Low wages”, “Unemployment”, “Poverty”, “Other economic/financial problems”, and “Lack of good jobs”. Answers to the survey question are again not applicable for a relevant subset of countries in the GAP. We are left with roughly 17,000 individual observations with economic and/or financial concerns and 12,000 without.


  1. Ai C, Norton EC (2003) Interaction terms in logit and probit models. Econ Lett 80(1):123–29

    Article  Google Scholar 

  2. Balistreri EJ (1997) The performance of the Heckscher-Ohlin-Vanek model in predicting endogenous policy forces at the individual level. Can J Econ 30(1):1–17

    Article  Google Scholar 

  3. Beaulieu E, Benarroch M, Gaisford J (2011) Intra-industry trade liberalization: why skilled workers are more likely to support free trade. Rev Int Econ 19(3):579–94

    Article  Google Scholar 

  4. Beaulieu E, Yatawara RA, Wang WG (2005) Who supports free trade in Latin America? World Economy 28(7):941–59

    Article  Google Scholar 

  5. Bhagwati J (1959) Protection, real wages and real incomes. Econ J 69(276):733–48

    Article  Google Scholar 

  6. Blonigen B (2011) Revisiting the evidence on trade policy preferences. J Int Econ 85(1):129–35

    Article  Google Scholar 

  7. Cameron AC, Miller DL (2010) Robust inference with clustered data. In: Ullah A, Giles DE (eds) Handbook of empirical economics and finance, chap 1. Chapman & Hall/CRC Press, pp 1–28

  8. Chong D, Citrin J, Conley P (2001) When self-interest matters. Polit Psychol 22(3):541–70

    Article  Google Scholar 

  9. Daniels J (2005) Religious affiliation and individual international-policy preferences in the United States. Int Interact 31(4):273–301

    Article  Google Scholar 

  10. Deardorff AV (1993) Overview of the Stolper-Samuelson theorem. In: Deardorff AV, Stern RM (eds) The Stolper-Samuelson theorem: a golden jubilee, chap 2, pp 7–34

  11. Deardorff AV (2004) Local comparative advantage: trade costs and the pattern of trade. Working Paper 500, Research Seminar in International Economics, University of Michigan

  12. Ehrlich SD (2007) Access to protection: domestic institutions and trade policy in democracies. Int Organ 61(3):571–605

    Article  Google Scholar 

  13. Ehrlich SD, Maestas C (2010) Risk orientation, risk exposure, and policy opinions: the case of free trade. Polit Psychol 31(5):657–84

    Article  Google Scholar 

  14. Ehrlich SD, Maestas C, Hearn E, Urbanski P (2010) Trade policy preferences and concern about your job. Paper presented at 2010 Midwest Political Science Association Conference

  15. Feenstra RC, Hanson GH (2003) Global production sharing and rising inequality: a survey of trade and wages. In: Choi EK, Harrigan J (eds) Handbook of international trade, vol 1, chap 6. Blackwell, pp 146–85

  16. Freeman RB (2009) Comments. In: Friedman BM (ed) Offshoring of American Jobs—what response from US economic policy? MIT Press, pp 61–71

  17. Hainmueller J, Hiscox MJ (2006) Learning to love globalization: education and individual attitudes toward international trade. Int Organ 60(2):469–98

    Article  Google Scholar 

  18. Hoffman ME (2009) What explains attitudes across US trade policies? Public Choice 138(3–4):447–60

    Article  Google Scholar 

  19. Hunt CV, Kim A, Borgida E, Chaiken S (2010) Revisiting the self-interest versus values debate: the role of temporal perspective. J Exp Soc Psychol 46(6):1155–58

    Article  Google Scholar 

  20. Krugman PR (2008) Trade and wages, reconsidered. Brookings Pap Econ Act 1:103–54

    Article  Google Scholar 

  21. Lawrence RZ (2008) Is trade to blame for rising US income inequality? Peterson Institute for International Economics

  22. Lewer J, Van den Berg H (2007) Religion and international trade—does the sharing of a religious culture facilitate the formation of trade networks? Am J Econ Sociol 66(4):765–94

    Article  Google Scholar 

  23. Malchow-Møller N, Munch J, Schroll S, Skaksen J (2009) Explaining cross-country differences in attitudes towards immigration in the EU-15. Soc Indic Res 91(3):371–90

    Article  Google Scholar 

  24. Mankiw NG (2008) Principles of economics, 5th edn. Cengage Learning

  25. Mansfield ED, Mutz DC (2009) Support for free trade: self-interest, sociotropic politics, and out-group anxiety. Int Organ 63(3):425–57

    Article  Google Scholar 

  26. Mayda AM, O’Rourke KH, Sinnott R (2007) Risk, government and globalization: international survey evidence. Working Paper 13037, National Bureau of Economic Research

  27. Mayda AM, Rodrik D (2001) Why are some people (and countries) more protectionist than others? Working Paper 8461, National Bureau of Economic Research

  28. Mayda AM, Rodrik D (2005) Why are some people (and countries) more protectionist than others? Eur Econ Rev 49(6):1393–430

    Article  Google Scholar 

  29. Metzler LA (1949) Tariffs, the terms of trade, and the distribution of national income. J Polit Econ 57(1):1–29

    Article  Google Scholar 

  30. OECD (2009) Education at a Glance 2009: OECD Indicators. OECD Publishing

  31. O’Rourke KH (2006) Heckscher-Ohlin theory and individual attitudes towards globalization. In: Findlay R, Henriksson R, Lindgren H, Lundahl M (eds) Eli Heckscher, international trade, and economic history. MIT Press, pp 107–38

  32. O’Rourke KH, Sinnott R (2001) The determinants of individual trade policy preferences: international survey evidence. In: Collins SM, Rodrik D (eds) Brookings trade forum 2001. Brookings Institute Press, pp 157–96

  33. Pew GAP (2007) World publics welcome global trade—but not immigration. Survey Report of the Pew Global Attitudes Project, October 4th, 2007. Washington DC

  34. Rodrik D (1995) Political economy of trade policy. In: Grossman GM, Rogoff K (eds) Handbook of international economics, vol 3, chap 28. Elsevier, pp 1457–94

  35. Scheve KF, Slaughter MJ (2001) What determines individual trade-policy preferences? J Int Econ 54(2):267–92

    Article  Google Scholar 

  36. Scheve KF, Slaughter MJ (2006) Public opinion, international economic integration, and the welfare state. In: Bardhan PK, Bowles S, Wallerstein M (eds) Globalization and egalitarian redistribution, chap 9. Princeton: Princeton University Press, pp 217–60

    Google Scholar 

  37. Scheve KF, Slaughter MJ (2007) A new deal for globalization. Foreign Affairs 86(4):34–47

    Google Scholar 

  38. Stolper WF, Samuelson PA (1941) Protection and real wages. Rev Econ Stud 9(1):58–73

    Article  Google Scholar 

  39. White H (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48(4):817–38

    Article  Google Scholar 

  40. Wolfe R, Mendelsohn M (2005) Values and interests in attitudes toward trade and globalization: the continuing compromise of embedded liberalism. Can J Polit Sci 38(1):45–68

    Article  Google Scholar 

Download references


We thank conference and seminar participants at universities in Aarhus, Duisburg-Essen, Göttingen, Lausanne, Milan, Stuttgart-Hohenheim, and Tübingen for useful comments. We are heavily indebted to Wilhelm Kohler for numerous valuable suggestions. Thanks are also due to Christian Bjørnskov, Rainer Grundmann, Benjamin Jung, Onur Koska, Philipp Schröder, and an anonymous referee for helpful comments. Inga Heiland and Marie-Christin Scholl have provided excellent research assistance. Ina Jäkel acknowledges financial support from the Danish Agency for Science, Technology and Innovation.

Author information



Corresponding author

Correspondence to Marcel Smolka.



Table 5 Summary statistics for individual-level variablesa
Table 6 Country-level informationa
Table 7 Coding information and data sources for country-level data

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Jäkel, I.C., Smolka, M. Individual Attitudes Towards Trade: Stolper-Samuelson Revisited. Open Econ Rev 24, 731–761 (2013).

Download citation


  • Trade policy
  • Voter preferences
  • Political economy
  • Heckscher-Ohlin trade theory

JEL Classification

  • F11
  • F13