¹ Translation from Spanish is as follows: “Over the last few years the country has had more and more business and trade with other countries. This tendency is called ‘free trade.' Do you think that free trade is good or bad for the country? Very or somewhat?”

² All results that are reported may overestimate slightly the degree of aggregate support for free trade because the sampling frame in each country was urban dwellers only. Urban dwellers and free trade proponents tend to be of a higher socioeconomic status, making them, according to results reported below, less protectionist.

³ Because citizens may not always understand a potentially complex issue such as free trade, a comment on nonresponse levels is merited. High aggregate rates of “don't know” and other “no opinion” responses are typically an indication that citizens, including many who do respond with an opinion, are uncertain or unfamiliar with the policy or issue being queried. See Bartels, Larry M., “Issue Voting under Uncertainty: An Empirical Test,” American Journal of Political Science 30 (1986)CrossRefGoogle Scholar. The degree of nonresponse in Tables 1, 2, and 3 varies substantially by question wording but, overall, indicates widespread knowledge and certainty about free trade. Nonresponse rates to a question about the past impact of free trade (Table 1) are very low, in single digits in almost all countries. They are only slightly higher, although still in the moderately low 7–17 percent range, when respondents were asked about FTAA, a seemingly more complex question about a hypothetical policy that had not yet been enacted (Table 2). (The nonresponse rate in the 1996 Latinbarometer question is 18 percent.) In comparative perspective, these no-opinion rates are about average or even slightly below average, so there is no reason to think that Latin Americans are more uncertain or less knowledgeable about trade than are other democracies' citizens about typical political issues.

At the same time, a potential problem with this assessment of aggregate nonresponse levels is the existence of nonattitudes, or people who do not have real attitudes but offer an opinion nonetheless. See Converse, Philip, “The Nature of Belief Systems in Mass Publics,” in Apter, David, ed., Ideology and Discontent (New York: Free Press, 1964Google Scholar). To address this, the four-city Brazil survey (Table 3) offered respondents an “easy out” by including “no opinion” as an explicitly read option. Among those who did offer an attitude, the “realness” or validity of reported opinions could also then be double-checked against open-ended justifications for the response, discussed in more detail in Section IV of this article. Not surprisingly, nonresponse increased (compare Table 1 with Table 3) to 30 percent in Brazil. The remaining 70 percent, therefore, are those with definite attitudes toward free trade, and almost all of their open-ended responses indicated familiarity with the issue. That said, research indicates that offering the “easy out” increases the occurrence of “false negatives,” or “no opinion” responses that mask real attitudes. In sum, although nonresponse varies with the issue and question wording, it seems safe to assume that uncertainty or lack of familiarity with free trade varies from about 10 percent to at most 25 percent in most of these countries. See Mikael Gilljam and Donald Granberg, “Should We Take Don't Know for an Answer?” Public Opinion Quarterly 57 (Autumn 1993).

⁴ “An Alarm Call for Latin America's Democrats,” Economist (July 26, 2001).

⁵ The presentation in “An Alarm Call” (fn. 4) aggregates results for six Central American countries (El Salvador, Guatemala, El Salvador, Costa Rica, Honduras, and Nicaragua) into one data point. The eleven countries besides the Central American ones are Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Paraguay, Peru, Uruguay, and Venezuela.

⁶ See “An Alarm Call” (fn. 4).

⁷ Hojman, David E., “The Political Economy of Recent Conversions to Market Economies in Latin America,” Journal of Latin American Studies 26 (February 1994), 210CrossRefGoogle Scholar; Weyland, Kurt, “Risk Taking in Latin American Economic Restructuring: Lessons from Prospect Theory,” International Studies Quarterly 40 (June 1996)CrossRefGoogle Scholar; idem, , “Swallowing the Bitter Pill: Sources of Popular Support for Neoliberal Reform in Latin America,” Comparative Political Studies 31 (October 1998)Google Scholar.

⁸ Berry, Albert, ed., Poverty, Economic Reform and Income Distribution in Latin America (Boulder, Colo.: Lynne Rienner, 1998)Google Scholar; and Sheahan, John, “Effects of Liberalization Programs on Poverty and Inequality: Chile, Mexico and Peru,” Latin American Research Review 32, no. 3 (1997)Google Scholar.

⁹ Inter-American Development Bank, Economic and Social Progress in Latin America: 1997 Report (Baltimore: Johns Hopkins University Press, 1997), 58Google Scholar. See page 90 for the econometric results backing this assertion. Also concurring are Stallings, Barbara and Peres, Wilson, Growth, Employment and Equity: The Impact of the Economic Reforms in Latin America and the Caribbean (Washington, D.C.: Brookings Institution/ECLAC, 2000)Google Scholar.

¹⁰ Stallings and Peres (fn. 9).

¹¹ Greenway, David, Morgan, Wyn, and Wright, Peter, “Trade Liberalization and Growth in Developing Countries: Some New Evidence,” World Development 25, no. 11 (1997)CrossRefGoogle Scholar. See also Rodriguez, Francisco and Rodrik, Dani, “Trade Policy and Economic Growth: A Skeptic's Guide to the Cross-National Evidence,” in Bernanke, Ben and Rogoff, Kenneth S., eds., Macroeconomics Annual 2000 (Cambridge: MIT Press, 2001)Google Scholar.

¹² “An Alarm Call” (fn. 4).

¹³ Another leading earning power theory is based on the Ricardo-Viner, or specific factors, model, which claims that individuals who are employed in or who own capital in internationally competitive industries should favor free trade while individuals with assets and employment in uncompetitive industries should favor protection. See Frieden, Jeffry A., Debt, Development and Democracy: Modern Political Economy and Latin America (Princeton: Princeton University Press, 1991)Google Scholar; and Alt, James, Frieden, Jeffry, Gilligan, Michael J., Rodrik, Dani, and Rogowski, Ronald, “The Political Economy of International Trade,” Comparative Political Studies 29 (January 1996)CrossRefGoogle Scholar. I disregard Ricardo-Viner because the data to test it rigorously (surveys must record respondents' sector of employment) are not available in any cross-national Latin American survey projects and because it has so far received limited empirical support as a theory of mass trade preferences. See Scheve, Kenneth F. and Slaughter, Matthew J., Globalization and the Perceptions of American Workers (Washington, D.C.: Institute for International Economics, 2001)Google Scholar; and Mayda, Anne Maria and Rodrik, Dani, “Why Are Some People More Protectionist Than Others?” Working Paper no. 8461 (Cambridge: National Bureau of Economic Research, 2001)CrossRefGoogle Scholar. The four-city Brazil survey does allow for a test of the Ricardo-Viner model, but it receives no empirical support. (These results are available from the author upon request.) Finally, the Ricardo-Viner model clearly does not provide an answer to this paper's central question about high aggregate support for free trade since only a slim minority of the Latin American population is employed in an internationally competitive sector. In fact, most citizens are not employed in a tradable goods sector at all.

¹⁴ Heckscher, Eli, “The Effect of Foreign Trade on the Distribution of Income,” Readings in the Theory of International Trade, vol. 4 (Philadelphia: Blakiston, 1949)Google Scholar; and Samuelson, Paul, “International Trade and the Equalisation of Factor Prices,” Economic Journal 58 (June 1948)CrossRefGoogle Scholar.

¹⁵ Rogowski, Ronald, Commerce and Coalitions (Princeton: Princeton University Press, 1989)Google Scholar.

¹⁶ Mayda and Rodrik (fn. 13).

¹⁷ Gabel, Matthew, Interests and Integration: Market Liberalization, Public Opinion, and European Union (Ann Arbor: University of Michigan Press, 1998)CrossRefGoogle Scholar.

¹⁸ Iversen, Torben and Soskice, David, “An Asset Theory of Social Policy Preferences,” American Political Science Review 95 (December 2001)Google Scholar. “See Scheve, and , Slaughter (fn. 13); Mayda and Rodrik (fn. 13); Gabel (fn. 17); Alt et al. (fn. 13); Miguel Basanez, Ronald Inglehart, and Neil Nevitte, North American Trajectory (Ann Arbor: University of Michigan Press, 1996)Google Scholar; and Rogowski (fn. 15).

²⁰ Tullock, Gordon, “The Welfare Costs of Tariffs, Monopolies, and Theft,” Western Economic Journal 5 (June 1967)Google Scholar; and Krueger, Anne O., “The Political Economy of the Rent-Seeking Society,” American Economic Review 64 (June 1974)Google Scholar.

²¹ Popkin, Samuel, The Reasoning Voter: Communication and Persuasion in Presidential Campaigns (Chicago: University of Chicago Press, 1994)Google Scholar; Gamson, William, Talking Politics (Cambridge: Cambridge University Press, 1992)Google Scholar; and idem, , “Media Discourse as a Framing Resource,” in Crigler, Ann N., ed., The Psychology of Political Communication (Ann Arbor: University of Michigan Press, 1996)Google Scholar.

²² Popkin (fn. 21); and Sniderman, Paul, Brody, Richard, and Tetlock, Philip, Reasoning and Choice: Explorations in Political Psychology (Cambridge: Cambridge University Press, 1991)CrossRefGoogle Scholar.

²³ Katzenstein, Peter, Small States in World Markets: Industrial Policy in Europe (Ithaca, N.Y.: Cornell University Press, 1985)Google Scholar.

²⁴ This includes financial services such as savings and investment. According to free trade advocates, liberalization increases the demand for, and therefore the prices of, nontraded goods and services because it leads to greater economic growth. For reasons discussed above, I do not make this assumption; however, this does not negate the fact that prices M_P and M_L will decrease relative to those NT_L.

²⁵ Household budget surveys require family expenditures to be tracked rigorously for at least a week.

²⁶ Houthakker, H. S., “An International Comparison of Household Expenditure Patterns: Commemorating the Centenary of Engel's Law,” Econometrica 25 (October 1957)CrossRefGoogle Scholar; and Ray, Debraj, Development Economics (Princeton: Princeton University Press, 1998)Google Scholar.

²⁷ Ogaki, Masao, “Engel's Law and Cointegration,” Journal of Political Economy 100 (October 1992)CrossRefGoogle Scholar; and Ogaki, Masao and Reinhart, Carmen M., “Measuring Intertemporal Substitution: The Role of Durable Goods,” Journal of Political Economy 106 (October 1998)CrossRefGoogle Scholar.

²⁸ Inter-American Development Bank, Economic and Social Progress in Latin America: 1998–1999 Report (Baltimore: Johns Hopkins University Press, 1998)Google Scholar.

²⁹ Ventura-Dias, Vivianne, Cabezas, Mabel, and Contador, Jaime, Trade Reform and Trade Patterns in Latin America (Santiago, Chile: Economic Commission on Latin America, 1999)Google Scholar.

³⁰ Although the government agencies that record CPIs do not report prices for a “durable goods” category, they do typically report price changes for “home equipment and maintenance.” This category consists largely of tradable consumer durables like electronic appliances and furniture.

³¹ To determine the first year of trade liberalization, I consulted the commercial policy reform index in Morley, Samuel A., Machado, Roberto, and Pettinato, Stefano, Indexes of Structural Reform in Latin America (Santiago, Chile: ECLAC, 1999)Google Scholar.

³² Results were weighted by city size to portray a representative sample of the aggregated four cities.

³³ Because individuals who had never heard of their country's regional trade bloc were not asked these questions, there was a substantial amount (about 40 percent) of missing data.

³⁴ These questions were not asked in the Dominican Republic.

³⁵ By using a linear combination of these two variables instead of just one of them, I tap an underlying dimension of free trade orientations that is less subject to either item's idiosyncrasies of question wording or measurement error. It is also crucial to note that hypothesized relationships are the same for both questions in the dependent variable. Both the Heckscher-Ohlin and consumption-based theories rely on the notions of relative abundance and comparative advantage. Therefore, a comparison of resource endowments between trading partners is necessary to understand economic consequences and to predict individual preferences. Throughout this paper, I have developed predictions based on Latin American countries' comparative advantage as developing countries vis-a-vis the global economy. Both questions are similar in this regard, then, since one refers to unilateral trade liberalization and the other to the FTAA, which would be dominated by the skill-intensive and capital-intensive economy of the U.S.

³⁶ The estimation procedure adjusts for the multilevel nature of the data by correcting for noninde-pendence or clustering within countries. See Eltinge, J. and Sribney, W., “Svy4: Linear, Logistic, and Probit Regressions for Survey Data,” Stata Technical Bulletin Reprints 6 (1997)Google Scholar; and Steen-bergen, Marco R. and Jones, Bradford S., “Modeling Multilevel Data Structures,” American Journal of Political Science 46 (January 2002)Google Scholar.

³⁷ To avoid a severe loss of cases, the reported estimates are averages of model results from twenty data sets with different imputations for missing values on income (10 percent missing), education (1 percent missing), and the variables used to construct the presidential approval instrument. King, Gary, Honaker, James, Joseph, Anne, and Scheve, Kenneth, “Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation,” American Political Science Review 95 (March 2001)Google Scholar. Imputations were made by country using the EMis algorithm in Amelia, 2.0; Honaker, James, Joseph, Anne, King, Gary, Scheve, Kenneth, and Singh, Naunihal, Amelia: A Program for Missing Data (Cambridge: Harvard University, 1999)Google Scholar (http://Gking.Harvard.edu, accessed January 2002).

³⁸ Data are 1999 figures from Robert J. Barro and Jong-Wha Lee, “International Data on Educational Attainment” (http://www.cid.harvard.edu/ciddata/ciddata.html, accessed January 2002). Education is expressed as deviation from the overall weighted sample mean while supply of skilled labor is expressed as deviation from the fourteen countries' mean, and the interaction term is therefore the product of these two variables. Centering these variables in this manner provides a more intuitive and substantive interpretation of the coefficients. The education coefficient is the relationship between education and free trade support in a country with the region's mean supply of skilled labor, while the supply of skilled labor coefficient expresses the relationship between skilled labor and free support for a citizen with the region's mean education level.

³⁹ Income is recalculated as deviations from the sample mean, with income squared equivalent to the square of this value. Neter, John, Wasserman, William, and Kutner, Michael H., Applied Linear Regression Models, 2d ed. (Homewood, 111.: Irwin, 1989)Google Scholar. The decision to measure income in 1997 U.S.$1, 000 without adjusting for purchasing power parity is crucial. It retains cross-national differences in purchasing power in international markets even though over- or undervalued exchange rates may contribute to these differences.

⁴⁰ This variable is not a percentage but a measure of the sheer size of this sector. More specifically, this is the natural log of the country's 1997 manufacturing sector output in US$ at purchasing power parity. Data are from World Bank, World Development Indicators 2000 CD-Rom (Washington, D.C.: World Bank, 2000)Google Scholar.

⁴¹ World Bank (fn. 40).

⁴² Stokes, Susan, ed., Public Supportfor Market Reforms in New Democracies (Cambridge: Cambridge University Press, 2001)CrossRefGoogle Scholar; Roberts, Kenneth M. and Arce, Moises, “Neoliberalism and Lower-Class Voting in Peru,” Comparative Political Studies 31 (April 1998)CrossRefGoogle Scholar; Fiorina, Morris, Retrospective Voting in American National Elections (New Haven: Yale University Press, 1981)Google Scholar; and Kinder, Donald and Kiewiet, D. Roderick, “Economic Discontent and Political Behavior: The Role of Personal Grievances and Collective Economic Judgements in Congressional Voting,” American Journal of Political Science 23 (August 1979)CrossRefGoogle Scholar.

⁴³ Kaufman, Robert R. and Zuckerman, Leo, “Attitudes toward Economic Reform in Mexico: The Role of Political Mediations,” American Political Science Review 92 (June 1998)CrossRefGoogle Scholar; and Zaller, John, The Nature and Origins of Mass Opinion (Cambridge: Cambridge University Press, 1992)CrossRefGoogle Scholar. Stokes, Susan C., Mandates andDemocracy: Neoliberalism by Surprise in Latin America (Cambridge: Cambridge University Press, 2001)CrossRefGoogle Scholar; and Corrales, Javier, “Presidents, Ruling Parties, and Party Rules: A Theory on the Politics of Economic Reform in Latin America,” Comparative Politics 32 (January 2001)Google Scholar.

⁴⁵ I categorized each president's economic policy orientation using Stokes's (see fn. 44) classification of the president's government as either “security”-oriented or “efficiency”-oriented (pp. 14–15). For the two security-oriented presidents, I reversed the coding of presidential approval for the respondents in these two countries.

⁴⁶ Kaufman and Zuckerman (fn. 43).

⁴⁷ Hinich, Melvin J. and Munger, Michael C., Analytical Politics (Cambridge: Cambridge University Press, 1997)CrossRefGoogle Scholar.

⁴⁸ Edwards, Sebastian, Crisis and Reform in Latin America: From Despair to Hope (Oxford: Oxford University Press, 1995)Google Scholar; and Williamson, John, ed., Latin American Adjustment: How Much Has Happened? (Washington, D.C.: Institute for International Economics, 1990)Google Scholar.

⁴⁹ Cue-taking theory dictates an interactive effect between awareness and political orientation, since highly aware citizens are both more exposed to the persuasive messages of their favorite politicians and more likely to reject those of their least preferred politicians. See Zaller (fn. 43). Although I estimated a model with this interaction term, I do not present the results because scholars do not yet know how to calculate the correct standard errors for a model with an interaction term that contains an instrumental variable. Besides this, the inclusion of this variable did not significantly enhance the predictive power of the model nor did it noticeably change any of the other coefficients.

⁵⁰ Frieden(fn. 13).

⁵¹ Alvarez, R. Michael and McCaffery, Edward J., “Are there Sex Differences in Fiscal Political Preferences?” Political Research Quarterly 56 (March 2003)CrossRefGoogle Scholar

⁵² This finding on education coincides with the results in Mitchell Seligson, “Popular Support for Regional Economic Integration in Latin America,” Journal of Latin American Studies 31 (February 1999)Google Scholar.

⁵³ Guatemala's value on the supply of skilled labor variable is -8.8, or almost 9 percentage points below the mean of the fourteen nations. Therefore, the education coefficient for Guatemala would be -8.8 x .0011 + .0211 = .011.