Are people really against trade liberalization? Cross-country evidence *

Are people really against trade liberalization? Cross-country evidence * By Channary Khun, Sajal Lahiri and Sokchea Lim Department of Economics, Southern Illinois University Carbondale, IL, USA Abstract We investigate the effect of trade restrictions on the perception of well-being by individuals, and examine whether the extent of the impact varies across groups in a society. Using cross-sectional micro data from surveys in 89 countries over the period 1981-2009, our findings suggest that citizens of a country with a lower degree of trade restrictions are possibly more satisfied with their life. Furthermore, the extent of satisfaction depends on individuals and countries relative endowment of human capital, as predicted by the well-known Stolper-Samuelson theorem. However, there is no robust evidence of diverging effects across ages and income levels.. Keywords: Trade policy, subjective perception of well-being, trade restriction, trade liberalization, human capital, Stolper-Samuelson theorem. JEL Classification: F13, F14, I31 Corresponding author: Channary Khun, Department of Economics, Southern Illinois University Carbondale, MC 4515, 1000 Faner Drive, Carbondale IL 62901, USA; E-mail: nary@siu.edu * The authors are grateful to the participants at the 12 th Annual Missouri Economics Conference and SIUC Brown Bag Seminar for helpful comments. We are also extremely thankful to an anonymous referee and the Joint Editor for very constructive suggestions. Usual disclaimer applies.

1. Introduction Are citizens of a country with a lower degree of trade restrictions happier than those in a more closed country? What are the characteristics of such people? These are some of the research questions that we address in this paper. Some events, and results of direct surveys, seem to suggest a substantial opposition to trade liberalization in many countries. Political battles over free trade agreements and protests disrupting the World Trade Organization meetings are widely reported. According to World Values Survey over the period 1994-1999 and 2005-2007, approximately two-thirds of 69,312 respondents in 52 countries expressed their preferences for trade restrictions when asked about their opinion on trade policy. Similar proportion of voters in the United States and in other Western economies where governments advocate greater trade openness seems to share the same anti-trade sentiment (Hiscox, 2006). However, drawing conclusions from such protests and surveys can be misleading for a number of reasons. First of all, as we have learned from the work of Olson (1965), the public good characteristics of lobbying means that if a policy imposes large losses on a small group of people and small gains for each in a large group, it will only face vociferous opposition from the losers even though the net gain may be significantly positive. Second, many of the opinion surveys of attitudes toward trade typically incorporate anti-trade wordings, which can significantly influence responses (Hiscox, 2006). 1 To avoid such framing, we simply correlate the perception of life satisfaction of individuals with changes in measures of trade restrictions, controlling for several individual and country characteristics. Based on a cross-sectional micro data set for 89 countries over the period 1981-2009, we in fact find support for a negative association between trade restrictions and the perception of life satisfaction by individuals. 1

In terms of the second research question mentioned at the outset, we find evidence in support of the well-known Stolper Samuelson theorem (Stolper and Samuelson, 1941), which predicts that individuals owning the factors in which the economy is relatively abundant, benefit from trade liberalization in the sector which is intensive in that factor, while those owning the other factors are adversely affected. Dividing the sample into two groups of countries based on their level of endowments in skilled labor, the results suggest that trade liberalization of skilled intensive goods is supported by highly skilled individuals in countries that are well endowed with higher level of human capital. On the other hand, in countries endowed with low skilled labor, the less educated individuals are more satisfied with their life as a result of trade liberalization on unskilled intensive goods, while the satisfaction of the highly educated falls. Furthermore, the estimation results with the sub-sample of highly educated U.S. citizens, Canadians, and several European countries are also in line with the prediction of the theorem. This paper makes a contribution in the emerging empirical literature on the effect of globalization on perceived wellbeing. Using responses from opinion surveys in OECD countries, Di Tella and MacCulloch (2008) show that there is a marginal negative effect of economic openness on subjective well-being. Using the same dataset, Xin and Smyth (2010) also find evidence of significantly low levels of perceived well-being for people living in cities which are presumed to be more open to international trade than small towns and villages. Hessami (2011) uses a data for fifteen EU countries and measures trade openness with the KOF globalization index which combines political, social, and economic aspects of globalization. 2 The KOF globalization index is found to be positively associated with life satisfaction. Using German household survey, Frijters and Geishecker (2008) find no compelling evidence that international outsourcing affects perceived job insecurity. Geishecker (2012) and Geishecker et al. (2012) 2

examine the possible association between subjective well-being and job loss fears. They show that the effects depend on whether the host country is a low- or high- wage one. There are two papers which are more to the point of the present paper and use a similar dataset, and these are Bjørnskov et al. (2008) and a recent working paper Dluhosch and Horgos (2012). Bjørnskov et al. (2008) finds a positive association between trade openness and life satisfaction. They use the 1997-2000 World Values Survey to examine the determinants of life satisfaction. Among other explanatory variables, three measures of trade policy (the ratio of total trade to GDP, average import tariff and KOF globalization index) are included in the regressions. They find that only the trade-gdp ratio has a significant effect. In contrast, we put more focus on trade variables and dig deeper into the factor-endowment model by considering several subsamples. We also use a larger sample size by merging World Values Survey (WVS) and European Values Survey (EVS). Like us Dluhosch and Horgos (2012) use a larger sample, but they investigate the effect of globalization (as measured by trade-gdp ratio and trade freedom published by the Heritage Foundation) on happiness and not on life satisfaction; we do both. Furthermore, like Bjørnskov et al. (2008) they also do not related their results to the factorendowment model of international trade. There are studies that pay a closer look at the impact of trade liberalization on wellbeing, as predicted by the factor-endowment (Heckscher-Ohlin) model, or namely the Stolper Samuelson theorem (see Balistreri, 1997; O Rourke and Sinnott, 2001; Scheve and Slaughter, 2001; Mayda and Rodrik, 2005). However, these studies mainly use trade-opinion surveys. The plan of the paper is as follows: Section 2 describes the data and section 3 outlines the empirical framework. Section 4 presents the estimation results followed by robustness check in section 5. Finally, some concluding remarks are made in section 6. 3

2. Data description and summary statistics We base our analysis on the micro data of reported life satisfaction taken from World Values Survey (WVS) and European Values Survey (EVS) for 89 countries over the period 1981-2009. 3 The dataset differentiates between life satisfaction and happiness. The former is associated with the following question: All things considered, how satisfied are you with your life as a whole these days? The answer to the question is ordered from 1 (dissatisfied) to 10 (satisfied). The question on happiness asks: Taking all things together, would you say you are very happy, quite happy, not very happy, or not at all happy? The response is scaled from 1 to 4. 4 The correlation between the two variables is 0.49. Although it is not significantly high, the estimated happiness and life-satisfaction equations from earlier work are found to have almost identical structural form (Blanchflower and Oswald, 2000). For the purpose of the current study, we base our main analysis on reported life-satisfaction, leaving happiness for robustness check. This is because life satisfaction data have been extensively used in psychological research where it is claimed to have passed validation exercises (Di Tella and MacCulloch, 2008). Furthermore, the question on life satisfaction is included in WVS and EVS in the first place partly due to the imprecise translation of the word happy across languages (Di Tella et al., 2001). The data show that the levels of life-satisfaction generally appear to be skewed towards the top of the response distribution. That is, individuals seem to respond optimistically: almost 60 percent of the respondents report above average life satisfaction (i.e., average life-satisfaction score is greater than 6) while less than 18 percent of them evaluate their satisfaction levels of less than 5. The average life satisfaction scores at the country level show that Columbia, a middle income country, has an average life satisfaction level of more than 8, the highest among the countries in the sample, followed by Denmark, Chile, Iceland, Malta, Guatemala, and so on. On 4

the other end, Tanzania and Zimbabwe are at the bottom of the list with average level of less than 4, followed by Iraq, Ukraine, and Pakistan. This shows that there is significant variation in life satisfaction scores across countries. WVS and EVS also contain a broad range of indicators on personal characteristics such as age, gender, marital status, employment status, number of children, education level and personal income, which are used as controls in the regressions. Income is scaled from 1 (low) to 10 (high); education level is measured on 1 (incomplete secondary school and below), 2 (from complete secondary school to university preparatory classes) and 3 (attending university with/without degree); and number of children is scaled from 0 (no child) to 8 (8 or more children). Marital status includes single, married and other (divorced, separated and widowed). Employment status includes employed, unemployed, self-employed, retired, housewife, students and other. The number of countries included in our study is essentially dictated by the availability of data on income. After excluding all missing information on other micro controls, we are left with 204,488 observations for micro-level regressions. In addition to trade openness (trade-gdp ratio), trade liberalization/restriction is measured by weighted average tariff rate applied on manufactured and on all products, customs and other import duties as a percent of tax revenue, and taxes on international trade as a percent of tax revenue. Other macro control variables are carbon dioxide (CO2) over GDP and mean income. The trade variables and CO2-GDP ratio are obtained from World Development Indicators (WDI) while the data on imports and exports is taken from Barbieri et al. (2012). Finally, for Taiwan we extract macro-economic data from Taiwan National Statistics website. 3. Empirical strategy and specification 5

To explore the effect of trade restrictions on perceived life satisfaction, the regression model is specified as follows: = + + + + +, (1) where subscript i is for individual and j for country. W is life satisfaction scores, a measure of perceived well-being, and X is a set of personal characteristics including age, age-squared, gender dummy, dummies for marital status, dummies for employment status, number of children, education level and personal income. We include age-squared to capture the non-linearity of perceived well-being across ages. T is the variable of interest that measures how restrictive a country is in international trade. Z is a vector of country-level control variables including average income and CO2 to GDP ratio. The former is calculated as the average of all respondents incomes in a country, and is a proxy for the overall economic condition in the surveyed areas; the latter is a measure of environmental damage. 5 These variables are included to control for possible effect of trade policy on countries economic and environmental conditions that in turn influence the level of life satisfaction. We expect that improvement in overall economic condition leads to higher reported life satisfaction (see Di Tella et al., 2003) while environmental pollution reduces the subjective well-being (see Di Tella and MacCulloch, 2008). The variable represents individuals level of education, age, or income scales. Its interaction with the trade restrictions variable is to capture the differential impact of trade liberalization across individuals with different attributes. Using education as a proxy for individuals skill level, its interaction with the variable T will allow us to test the well-known Stolper-Samuelson theorem. While trade as a proportion of GDP has been extensively used in the literature, we use other measures of trade openness/restrictions as there are three main problems with the traditional openness variable. First of all, there is an issue of measurement: a country s high 6

measure of openness does not necessarily reflect a more open economy; this is because GDP is calculated as value added in production while export measure is gross, i.e., not net of intermediate inputs which can be partly imported. This could result in an overestimate of the level of trade openness. 6 Second, a high level of openness may not be due to polices but because of the small size of the economy, its resource endowments which are in high demand, or its inherent dislike for imported goods. Third, the endogeneity of openness is a real concern: the resource endowments and policies which determine trade flows may also influence reported life satisfaction. More importantly, trade openness is not itself a policy variable. Recall that the aim of the paper is to determine the effect of restrictive trade policies on the perceived well-being; thus, we focus our analysis mainly on one of the policy variables, the tariff rate applied on manufactured products, while leaving others for sensitivity analysis. Another concern is that the regression results could be biased due to the fact that the reported quality of life across countries may systematically differ as individuals face different policy environments and economic contexts. To the extent that these factors are country specific, including macro controls such as CO2-GDP ratio and mean income may address this concern. Finally, the dependent variables life satisfaction and happiness are ordinal in nature. Yet, some studies often cardinalize them and apply a linear regression model (see, for example, Di Tella et al., 2001). This may provide misleading results if the actual distance between two consecutive thresholds of life satisfaction scores are not uniform. Furthermore, the errors are heteroscedastic and not standard normal, violating the assumptions of OLS. To deal with these issues, we use Ordered Probit regression and additionally employ other estimation methods such as Ordinary Least Squares (OLS), Ordered Logit, and Tobit model for sensitivity analysis. 4. Results 7

Table 1 reports the estimated regressions. Since the regressions are estimated using Ordered Probit model, the coefficients do not represent the marginal effects of trade protection, but they can be used to infer the qualitative nature of association between the two variables. Columns (1)-(3) report the results in which weighted average tariff rate applied on manufactured products is used to proxy for trade restrictions. As expected, the coefficients for trade policy are negative and statistically significant at the one percent level in all three specifications. This implies that a drop in the tariff rate to liberalize international trade raises individuals perceived life satisfaction. 7 In column (1), the interaction term between trade restrictions and education level that represents the skill level of the respondents is significantly positive at the one percent level. The result implies that individuals with lower educational attainment are relatively more satisfied with less trade restrictions. The results, to certain extent, support the prediction of Stolper-Samuelson theorem since approximately 60 percent of the countries in our sample are middle and low income countries, which are relatively abundant in low-skilled labor. 8 We will examine the hypothesis further by grouping countries according to their endowments in human capital. In column (2), the coefficient of the interaction term between the tariff rate and age is positive and statistically significant at the one percent level, suggesting a U-shaped relationship between trade liberalization and life satisfaction across ages. That is, older individuals could be negatively affected by trade liberalization while younger ones could benefit from it. This can be due to the fact that older-age people are relatively more risk-averse, less flexible in the job markets and less capable of taking advantage of what a more open economy offers such as greater product varieties. 8

Column (3) presents the result in which trade policy is interacted with individual income. The coefficient of the interaction term is significantly positive at the one percent level, suggesting that trade liberalization is more satisfying for low-income individuals than the rich ones. This is perhaps due to the fact that our sample mostly consists of developing countries. Besides the variables of main interest, most other control variables at both individual and country levels are also significant and carry the expected signs as predicted in the literature. The macro variables show that greater environmental degradation reduces individual well-being (Di Tella and MacCulloch, 2008) and the improvement of the overall living standard raises life satisfaction (see Sacks et al., 2010). The results for individual characteristics show that income makes people more satisfied with their lives. The results are consistent with studies in the happiness literature (see Di Tella and MacCulloch, 2008; Sacks et al., 2010). The highlyeducated individuals are more satisfied with their lives than the lower educated. More kids make people happier. Single and married individuals are happier than those who are divorced, separated or widowed. As mentioned before, other than the average tariff rate on manufactured products we also use other proxies for trade restrictions. 9 When customs and duties as a percentage of tax revenue and taxes on international trade are used, the results are consistent except that their interactions with age turn negative. On the other hand, when trade to GDP ratio is used as a measure of trade liberalization, the results are similar to benchmark results except when trade openness is interacted with personal income. The results seem to support the evidence provided by Mayda and Rodrik (2005). Overall, regardless of the measures of trade liberalization it is evident that people are generally more satisfied with their lives in a more open economy. Moreover, the evidence seems to support the factor-endowment model or Stolper-Samuelson theorem. 9

To test the prediction of Stolper-Samuelson theorem as well as to examine further the different effects of trade restrictions across different groups of respondents, we selectively sample two groups of countries based on their levels of endowments in skilled labor proxied by the percentage of tertiary enrollment averaging over the period 2000-2009. The samples include top and bottom 30 countries in skilled labor endowment. 10 The results are presented in Table 2. Interestingly, the estimated results (columns 1 and 4) shows that the interaction term between trade protection and education is negative and statistically significant at the one percent level for countries which are relatively more skilled labor abundant (Top 30 countries), but it turns positive for countries well endowed with unskilled labor (Bottom 30 countries). Although the coefficient of trade policy is statistically insignificant for the former group, it is negative and significant at the 5 percent level for the latter. The evidence suggests that highly skilled individuals benefit more than unskilled labor from trade liberalization in skilled labor abundant countries while trade openness in unskilled labor abundant countries raises the well-being of unskilled individuals, but lowers that of the highly skilled. These results provide evidence in support of the Stolper-Samuelson theorem. Columns (2) and (5) show the results from the inclusion of an interaction between trade policy and age. Though the interaction term is significantly negative at the 10 percent level for the top 30 countries, the coefficient is very small and the coefficient for the trade policy variable is not significant. The results do not seem to provide compelling evidence of any differential effect of trade liberalization across ages even in countries with different level of human capital. The results in columns (3) and (6) show that high-income earners are more satisfied with their lives than low-income earners when trade in developed countries becomes more open. The evidence now seems to be in line with Mayda and Rodrik (2005). Another interesting result from 10

these regressions is that children raise individual well-being in developed countries, but do not significantly affect the well-being in the developing world. Finally, we examine the impact of trade liberalization on the highly educated in some selective countries or sub-groups. The regressions are performed on three countries, the U.S., Canada, France and a sub-group of OECD European countries. The samples consist of only individuals who attended universities or higher education, or received tertiary certificates. The results are presented in columns 7-10 of Table 2. The mean income and CO2 to GDP ratio are dropped from the equations for the U.S., Canada and France due to lower variation across the periods and high correlation with the tariff rate. The regressions are estimated using Ordinary Least Squares (OLS); however, the results from the Order Probit estimation gave similar results. The results show that the coefficients of average tariff rate are all statistically significant and different from zero. The evidence indicates that the highly educated in selected developed countries are more satisfied with their lives when the level of protection is lower. 5. Robustness checks As a robustness check over the choice of the variable to represent individual well-being, we also use happiness instead of life satisfaction. The results are not reported for the sake of brevity. However, the findings are similar to those based on life satisfaction as the dependent variable. However, we find no robust evidence indicating any differential impacts of trade liberalization across age groups. Moreover, to ensure that our results are not driven by any particular regression models, we additionally estimate equation (1) using Ordinary Least Squares (OLS), Ordered Logit, and Tobit model. Once again these results are not reported, but results do not change qualitatively. Cardinalizing life satisfaction scores and using OLS might not be appropriate as it violates OLS 11

assumptions regarding the error terms. However, Ferrer-i-Carbonell and Frijters (2004) find that assuming ordinality or cardinality of the well-being data makes little difference. Moreover, one advantage with OLS estimates is that the coefficients directly represent the marginal effects of independent variables on the dependent variable. With OLS regressions, we consider the impact of a one standard deviation increase in the tariff rate, which is equal to 12.88 percentage points. According to our OLS estimate, the rise in tariff rate by one standard deviation diminishes the life satisfaction of unskilled individuals (education = 1) by 0.15 points, implying a shift of 15 percent of the population of unskilled labor downwards from one life satisfaction category to another. Other estimation methods also provide qualitatively similar results. 6. Conclusion In this paper we have analyzed the effect of reducing trade restrictions on subjective wellbeing. Based on surveyed data of 89 countries over the period 1981-2009, we find significant negative impact of protectionist trade policy on the quality of life. Furthermore, using subsamples of top and bottom 30 countries in terms of human capital endowment, we are able to provide evidence confirming the prediction of the well-know Stolper Samuelson theorem. That is, the highly skilled individuals are more satisfied with their lives than the unskilled when countries with high human capital are more open to trade. Similarly, less trade restriction raises the life satisfaction of the unskilled, but hurts the highly skilled in countries with low level of human capital. 12

Notes 1 For instance, WVS on trade attitudes reads: Do you think it is better if: (1) Goods made in other countries can be imported and sold here if people want to buy them; or that: (2) There should be stricter limits on selling foreign goods here, to protect the jobs of people in this country; or: (9) Don t Know. A similar question from International Social Survey Program reads: Now we would like to ask a few questions about relations between (respondent s country) and other countries. How much do you agree or disagree with the following statement: (Respondent s country) should limit the import of foreign products in order to protect its national economy. Note that the former links trade to the possibility of job losses, while the latter incorporates a similar anti-trade wording. These are technically referred to as the issue of framing. 2 See Dreher et al. (2008) for details on the index and with the respective data available for download at http://globalization.kof.ethz.ch/ (last accessed February 8, 2013). 3 The two datasets are largely compatible; we merge the data following the instruction provided on the websites. A list of the countries included in the dataset is provided in Appendix A 4 For comparative purposes, we reverse the order of happiness answers with not at all happy taking value 1 and very happy taking value 4. 5 We prefer mean income to GDP per capita as it represents more precisely the living standards in the surveyed community. 6 Singapore s export, for example, is 230% of its GDP. 7 Though not reported, this result also holds without the interaction terms. 8 There are 35 high income and 54 middle and low income countries. 9 The results for the tariff rate on all products are very similar to those from the tariff rate manufactured products; hence, they are not presented. 10 We also considered top 20 and bottom 20 countries and obtained similar results; but these are not reported. 13

Appendix Table A1. 89 countries in the dataset High income Middle income Low income Australia AUS Albania ALB Armenia ARM Austria AUT Algeria DZA Bangladesh BGD Belgium BEL Argentina ARG Burkina Faso BFA Canada CAN Azerbaijan AZE Egypt EGY Croatia HRV Belarus BLR El Salvador SLV Cyprus CYP Bosnia and Herzegovina BIH Ethiopia ETH Czech Republic CZE Brazil BRA Georgia GEO Denmark DNK Bulgaria BGR Ghana GHA Estonia EST Chile CHL Guatemala GTM Finland FIN China CHN India IND France FRA Colombia COL Indonesia IDN Germany DEU Dominican Republic DOM Iraq IRQ Greece GRC Iran IRN Kyrgyzstan KGZ Hungary HUN Jordan JOR Mali MLI Iceland ISL Latvia LVA Moldova MDA Ireland IRL Lithuania LTU Morocco MAR Italy ITA Macedonia MKD Nigeria NGA Japan JPN Malaysia MYS Pakistan PAK Luxembourg LUX Mexico MEX Philippines PHL Malta MLT Peru PER Rwanda RWA Netherlands NLD Romania ROM Tanzania TZA New Zealand NZL Russian Federation RUS Uganda UGA Norway NOR Serbia and Montenegro YUG Ukraine UKR Poland POL South Africa ZAF Viet Nam VNM Saudi Arabia SAU Thailand THA Zambia ZMB Slovakia SVK Turkey TUR Zimbabwe ZWE Slovenia SVN Uruguay URY South Korea KOR Venezuela VEN Spain ESP Sweden SWE Switzerland CHE Taiwan TWN Trinidad and Tobago TTO United Kingdom GBR United States USA Note: The three-letter abbreviations represent ISO country code 14

References Balistreri, E.J., 1997. The performance of the Heckscher Ohlin Vanek model in predicting endogenous policy forces at the individual level. Canadian Journal of Economics 30(1), 1 17. Barbieri, K., Keshk, O., Pollins, B., 2012. Correlates of war project trade data set codebook, Version 3.0. Online: http://correlatesofwar.org. Bjørnskov C., Dreher A., Fischer, A.V.J., 2008. Cross-country determinants of life satisfaction: exploring different determinants across groups in society. Social Choice and Welfare 30, 119 173. Blanchflower, D., Oswald, A., 2000. Well-being over time in Britain and the USA, mimeo. Di Tella, R., Macculloch, R.J., 2008. Gross national happiness as an answer to the Easterlin Paradox? Journal of Development Economics 86, 22-42. Di Tella, R., Macculloch, R.J., Oswald, A.J., 2001. Preferences over inflation and unemploynnent: Evidence from surveys of happiness. American Economic Review 91(1), 335-341. Di Tella, R., MacCulloch, R.J., Oswald, A.J., 2003. The macroeconomics of happiness. Review of Economics and Statistics 85(4), 809-827. Dluhosch B., Horgos, D., 2012. Trading up the happiness ladder. Working Paper No. 115. Helmut Schmidt University Hamburg. Dreher, Axel; Gaston, Noel; Martens, Pim (2008): Measuring Globalization Gauging its Consequences, Springer, New York and Heidelberg. European and World Values Surveys four-wave integrated data file, 1981-2004, v.20060423, 2006. Surveys designed and executed by the European Values Study Group and World Values Survey Association. File producers: ASEP/JDS, Madrid, Spain and Tilburg University, Tilburg, the Netherlands. File distributors: ASEP/JDS and GESIS, Cologne, Germany. Ferrer-i-Carbonell, A., Frijters, P., 2004. How important is methodology for the estimates of the determinants of happiness? The Economic Journal 114, 641 659. Frijters, P., Geishecker, I., 2008. International outsourcing and job loss fears: An econometric analysis of individual perceptions. Mimeo. Geishecker, I., 2012. Simultaneity bias in the analysis of perceived job insecurity and subjective well-being. Economics Letters 116, 319-321. Geishecker, I., Riedl, M., Frijters, P., 2012. Offshoring and Job Loss Fears: An Econometric Analysis of Individual Perceptions. Labor Economics 19, 738 747. 15

Hessami, Z., 2011. Globalization s Winners and Losers Evidence from Life Satisfaction Data, 1975-2001. Economics Letters 112, 250 253. Hiscox, J.M., 2006. Through a glass and darkly: Attitudes toward international trade and the curious effects of issue framing. International Organization 60, 755-780. Mayda, M.A., Rodrik, D., 2005. Why are some people (and countries) more protectionist than others? European Economic Review 49, 1393-1430. Olsen, M., 1965, The Logic of Collective Action, Cambridge: Harvard University Press. O Rourke, K.H., Sinnott, R., 2001. What determines attitudes towards protection? Some crosscountry evidence. In: Collins, S.M., Rodrik, D. (Eds.), Brookings Trade Forum 2001. Brookings Institute Press, Washington DC, 157 206. Sacks, D.W.; Stevenson, B.; Wolfers, J., 2010. Subjective well-being, income, economic development and growth. NBER Working Paper No. 16441. Scheve, K.F., Slaughter, M.J., 2001. What determines individual trade policy preferences? Journal of International Economics 54(2), 267 292. Stolper, W.F., Samuelson, P.A., 1941. Protection and Real Wages. Review of Economic Studies 9, 58-73. World Bank World Development Indicator Online. World Values Survey (1981-2008). Official aggregate v.20090901, 2009. World Values Survey Association (www.worldvaluessurvey.org). Aggregate file producer: ASEP/JDS, Madrid. Xin, W., Smyth, R., 2010. Economic openness and subjective well-being in China. China & World Economy 18 (2), 22 40. 16

Table 1: Micro equations for 89 countries, 1981-2009: Life satisfaction as subjective well-being Trade restriction variable: Independent Tariff rate on manufactures Customs and duties Taxes on international trade Trade-GDP ratio variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Trade restriction -0.008*** -0.005*** -0.003*** -0.027*** -0.005*** -0.019*** -0.036*** -0.008*** -0.022*** 0.088*** 0.109*** -0.204*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.021) (0.020) (0.019) Trade policy*edu 0.003*** 0.008*** 0.012*** -0.058*** (0.003) (0.0001) (0.001) (0.010) Trade policy*age 0.0001*** -0.0002*** -0.0001*** -0.003*** (0.003) (0.0001) (0.0001) (0.0001) Trade Policy*Income 0.0004*** 0.002*** 0.002*** 0.038*** (0.0001) (0.0001) (0.0001) (0.0030) CO2/GDP -0.181*** -0.182*** -0.181*** -0.137*** -0.135*** -0.137*** -0.140*** -0.138*** -0.140*** -0.124*** -0.124*** -0.123*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.001) Mean income 0.025*** 0.026*** 0.025*** -0.009** -0.013*** -0.011*** 0.005 0.005 0.004 0.046*** 0.046*** 0.048*** (0.003) (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.003) (0.003) (0.003) Income 0.080*** 0.080*** 0.077*** 0.087*** 0.086*** 0.075*** 0.084*** 0.084*** 0.075*** 0.075*** 0.075*** 0.054*** (0.001) (0.001) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.001) (0.001) (0.002) Education Level 0.029*** 0.060*** 0.059*** -0.026*** 0.039*** 0.039*** -0.023*** 0.041*** 0.041*** 0.097*** 0.064*** 0.065*** (0.005) (0.004) (0.004) (0.005) (0.004) (0.004) (0.005) (0.004) (0.004) (0.007) (0.003) (0.003) Number of children 0.011*** 0.009*** 0.010*** 0.018*** 0.018*** 0.016*** 0.017*** 0.016*** 0.015*** 0.003 0.002 0.002 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Age -0.029*** -0.030*** -0.029*** -0.030*** -0.028*** -0.030*** -0.030*** -0.029*** -0.030*** -0.026*** -0.025*** -0.026*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Age-squared 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** 0.0003*** (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) Single 0.105*** 0.099*** 0.103*** 0.096*** 0.101*** 0.094*** 0.085*** 0.084*** 0.083*** 0.102*** 0.101*** 0.100*** (0.012) (0.012) (0.012) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.010) (0.010) (0.010) Married 0.195*** 0.193*** 0.195*** 0.192*** 0.192*** 0.196*** 0.190*** 0.187*** 0.192*** 0.192*** 0.192*** 0.189*** (0.009) (0.009) (0.009) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.008) (0.008) (0.008) Number of obs. 142,097 142,097 142,097 120,394 120,394 120,394 125,666 125,666 125,666 200,787 200,787 200,787 Pseudo R 2 0.029 0.029 0.029 0.030 0.029 0.029 0.030 0.029 0.029 0.029 0.029 0.029 Note: Though not reported, dummies for gender and employment status are also included in all specifications. Income ranges from 1 (low) to 10 (high); education level is measured on 1 (incomplete secondary school and below), 2 (from complete secondary school to university preparatory classes) and 3 (attending university with/without degree). The regressions are estimated using Ordered Probit method. Robust standard errors are in parenthesis. Tariff rate on all products is also regressed as a measure of trade policy; however, its results resemble those of tariff rate on manufactured products thus are not reported here. * denotes significant at 10% level; ** significant at 5% level; and *** significant at 1% level. 17

Table 2: Countries endowed with skilled and unskilled labor: Life satisfaction as subjective well-being Sub-sample by level of human capital Sub-sample of people with higher education in: Independent variables Top 30 Bottom 30 USA Canada France OECD Europeans (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Tariff rate on manuf. -0.007-0.007-0.018*** -0.007*** -0.0015* -0.003*** -0.171* -1.741** -0.250** -0.0919*** (0.005) (0.005) (0.004) (0.001) (0.001) (0.001) (0.096) (0.846) (0.115) (0.010) Tariff on manuf.*edu -0.007*** 0.002*** (0.002) (0.000) Tariff on manuf.*age -0.0002** -0.00004* (0.0001) (0.0001) Tariff on manuf.*income 0.0001 0.0001 (0.001) (0.0001) CO2/GDP -0.200*** -0.202*** -0.202*** -0.080*** -0.082*** -0.082*** -0.572*** (0.004) (0.004) (0.004) (0.006) (0.006) (0.006) (0.013) Mean income 0.069*** 0.070*** 0.070*** -0.049*** -0.051*** -0.049*** 0.0365*** (0.006) (0.006) (0.006) (0.007) (0.007) (0.007) (0.013) Income 0.060*** 0.060*** 0.060*** 0.086*** 0.087*** 0.086*** 0.141*** 0.0592*** 0.0934*** 0.116*** (0.002) (0.002) (0.003) (0.003) (0.003) (0.004) (0.024) (0.022) (0.033) (0.004) Education level 0.034*** 0.013* 0.013* 0.069*** 0.107*** 0.107*** (0.010) (0.007) (0.007) (0.010) (0.008) (0.008) Number of children 0.040*** 0.040*** 0.040*** -0.005-0.004-0.005-0.0652 0.081* 0.126 0.0159* (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.045) (0.042) (0.080) (0.009) Age -0.040*** -0.039*** -0.040*** -0.021*** -0.020*** -0.021*** -0.071*** -0.047*** -0.086** -0.067*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.019) (0.017) (0.039) (0.004) Age-squared 0.0004*** 0.0004*** 0.0004*** 0.0003*** 0.0003*** 0.0002*** 0.001*** 0.0004** 0.0005 0.0006*** (0.00002) (0.00002) (0.00002) (0.00003) (0.00003) (0.00003) (0.0002) (0.0002) (0.0004) (0.00004) Single 0.105*** 0.105*** 0.105*** -0.014-0.015-0.017 0.225-0.078 0.520 0.143*** (0.019) (0.019) (0.019) (0.028) (0.028) (0.028) (0.218) (0.188) (0.352) (0.040) Married 0.273*** 0.273*** 0.273*** 0.073*** 0.074*** 0.072*** 0.769*** 0.496*** 0.741** 0.586*** (0.014) (0.014) (0.014) (0.024) (0.024) (0.024) (0.172) (0.161) (0.293) (0.031) Number of obs. 50,167 50,167 50,167 41,160 41,160 41,160 1,120 971 512 42,849 (Pseudo) R 2 0.039 0.039 0.039 0.014 0.014 0.014 0.114 0.09 0.139 0.137 Note: Though not reported, dummies for gender and employment status are also included in all specifications. Income ranges from 1 (low) to 10 (high); education level is measured on 1 (incomplete secondary school and below), 2 (from complete secondary school to university preparatory classes) and 3 (attending university with/without degree). The regressions are estimated using Ordered Probit method for the top and bottom 30 sub-samples and OLS for USA, Canada, France and OECD European countries. Robust standard errors are in parenthesis. The percentage of tertiary enrollment averaging over the period 2000-2009 is used to divide the sample into countries endowed with skilled and unskilled labor. * denotes significant at 10% level; ** significant at 5% level; and *** significant at 1% level. 18