Information and Gender Differences in Individual Preferences for Protection. Alexandra Guisinger

Transcription

1 Information and Gender Differences in Individual Preferences for Protection Alexandra Guisinger Conflict Processes Panel 21-30, Public Opinion and IR September 1, 2011 APSA Men and women consistently differ in their preference for trade protection, even after controlling for individual economic characteristics. Differences in information about trade could explain the gap in at least two ways. First, men and women possess different political and economic knowledge particularly knowledge about trading partners leading to divergent expectations about the benefits of trade. Second, men and women process and evaluate information about trade differently. To explore the role of information, I conducted a survey of 1,500 potential voters in the 2010 United States Congressional Election. The survey contained an experiment that isolated the effects of participants perceptions of the United States primary trading partners, and a second experiment that randomly varied exposure to a positive frame on the benefits of such trade. I find evidence on behalf of both roles for information in the formation of trade preferences. Men and women possess different priors on the trade behavior of the United States and reacted differently to the experimental stimuli. I find that women are more likely to erroneously identify the top US trading partner as China and be more protectionist, but in contrast to men s behavior - women's beliefs about the top trading partner do not appear to cause their preference for protection. In the second experiment, positive information about the role of trade in the US economy increased the gender gap: while both men and women hearing a positive message similarly updated on the benefits of trade for US employment, men were far more influenced by the positive message than women in terms of individual employment benefits. 1

2 Over the past decade, international political economy scholars have focused great attention on the determinants of individual preferences for trade protection. The use of survey data, both from the United States and internationally, has become commonplace, particularly as researchers try to close the continuing theoretical divide on whether individual preferences are determined by a Stolper-Samuelson based factor classification or Ricardo-Viner based sectoral (industrial) identification 1. A positive externality of such work has been growing knowledge about nonindividual and non-economic sources of trade protection such as socio-tropic factors (Mansfield and Mutz, 2009), neighborhood attachment (Ceccoli, Gelleny, and Kaltenthaler, 2004; Rodrik and Mayda. 2005), national pride (Rodrik and Mayda. 2005), nationalism, and chauvinism (O Rourke and Sinnot, 2001); and values and ideology (Wolfe and Mendlesohn, 2005). Across such surveys, a single, theoretically unexplained finding persists: ceteris paribus, women are more trade protectionist than men (see for example, Hiscox and Burgoon, m.s.; Ceccoli, Gelleny, Kaltenthaler 2004; Scheve and Slaughter, 2001; O Rourke, Kevin and Richard Sinnott, 2001). Given that gender-clustered differences remain even after controlling for observed economic circumstances, this paper explores the gender gap via differences in information possession and processing, specifically in gendered differences in economic and political knowledge and in the response to economic evidence. The literature on American politics has long noted the political knowledge gap between men and women and debated its effects (see Mondak and Anderson 2004 for a review). Trade policy is no exception. In multiple surveys, women are more likely to express no opinion when asked about trade policy. When asked knowledge questions about specific policies, women are also more 1 Baker 2005; Beaulieu et al. 2005; Ceccoli et al. 2004; Mayda and Rodrik 2005; Mayda et al. 2006; O Rourke and Sinnott 2001; Scheve and Slaughter

3 likely to state Don t Know or provide incorrect answers (see for example, Guisinger 2009). This gap in knowledge may be one reason women differ from men in their preferences. If this is the case, ensuring both groups possess similar information to both groups should lead to similar preferences, all else equal. In contrast, if due to unobservable criteria, men and women assess the applicability of trade-related information differently, then dissemination of similar information may serve to exacerbate gender-based disparities in stated preferences for trade protection. To assess the validity of these hypotheses, I undertake a two-part survey experiment in which 1,500 subjects receive treatments to alter their beliefs about their country s main trading partner and the economic effects of trade on the country. Review of Literature Although the finding of a gender gap in trade preferences is widely noted, only two published papers to date directly address the gap. In a twist on typical production-based models of trade policy preferences, Hall, Kao, and Nelson (1998) argue that women are more likely to be defined by their role as a consumer and thus more likely to support lower tariffs on imported goods. They find a temporal correlation between increased women s suffrage and lower US tariff rates. However, this macro relationship is at odds with micro-level empirical studies which have time and time again found individual women less likely to favor of policies which increase international trade. Why this is so remains up for debate. Breaking from consumer and producer typologies, Gidengil (1995) instead characterizes the gender divide through differences in values and beliefs: economic men are assumed to be more concerned with individual material concerns and less concerned with egalitarianism and social 3

4 women are assumed to be more influenced by social concerns, particularly a commitment to welfare state. 2 Analyzing survey data concerning a proposed Canada-U.S. Free Trade Agreement, Gilengil finds that differences in material circumstances compose only a small portion of the 16 percentage point difference between men s support (60%) and women s support (44%) for the agreement. Using a regression decomposition technique to separate the contribution of differences in levels and in intensity (saliency) of effect, Gilengil finds empirical support for the contention that women were less likely to perceive the economic benefits of free trade but also that such economic considerations were less likely to impact their assessment of the agreement. However, the observational nature of the study prevents exploration of the source of the differences in saliency and in particular the role of that knowledge may play these differences. A pair of unpublished studies, one by Burgoon and Hiscox (2008) and one by Lee (2011), focus specifically on the role of economic literacy and exposure to orthodox economic theory as the cause of the preference gap. This work is a gendered outgrowth of previous work from the economics field which assesses the impact of economic knowledge on public opinions on economic policy issues in general (see for example, Walstad and Rebeck 2002) and international trade specifically (Hainmueller and Hiscox 2006). Burgoon and Hiscox draw upon observational survey data of 1500 respondents to show that the gender gap in trade preferences cannot be explained away by controlling for skills and job characteristics or risk sensitivity linked to 2 Gidengil directly acknowledges Gilligran s In a Different Voice (1982) for providing this characterization. Sapiro (2002) notes weakly supportive evidence of such a characterization from analysis of the most import problems facing this country question on the American National Election Studies (NES) surveys in 1992, 1994, 1996, and 1998: men were slightly, but statistically significantly, more likely to name economic issues and women were more likely to name social issues, although in general there was wide agreement across both genders and the size of the gap varied across survey years. 4

5 maternity, or gender differences in political values. Rather, they determine that much of the gap can be explained by gender differences in economic knowledge as proxied by knowledge of NAFTA participants and the name of the Secretary of State. What Burgoon and Hiscox can t tell us is whether the knowledge itself is important or whether knowledge is serving as a proxy for other differences between genders. To bypass the limitations of an observational study, Lee (m.s.) undertakes a survey experiment of 190 undergraduates. Among the population studied, Lee finds little empirical support for a link between trade preferences and economic literacy (as measured by economics courses taken) but instead, through a series of survey experiments, finds support for the argument that women s greater concern for those facing hardships causes their relatively greater support for protectionist policies. A goal of this paper is to draw on both designs by expanding a survey design technique to a larger, more nationally representational sample. Gender and trade policy Discussion of the gender divide in policy preferences has a longer history in other - non-trade related - policy realms (see review in Wolbrecht 2005). As described in Shapiro and Mahajan (1986), gender differences in responses on public opinion surveys first began to be widely recognized in the 1970s and early 1980s by Hazel Erkstine, William Schneider, Tom Smith, and the editors of Public Opinion. Early scholarship found distinct differences in preferences over policies concerning force and violence (Smith 1984), social welfare issues (Cook 1979), but surprisingly not on women s rights issues (editors of Public Opinion). In their own analysis of public opinion surveys from the 1960s to 1980s, Shaprio and Mahajan found persistent differences in opinions on the use of force and violence, regulation and protective policies, what 5

6 they term compassion issues to do with social welfare, but again not on women s issues other than abortion. Explanations for gender-related attitudinal differences initially fell into two broad camps: essentialist arguments derived from assumptions about biological differences and socioconstructivist arguments focused on the culturally created expectations about gender-specific behavior. (Andersen 1997). More recently, the debate has focused primarily within the socioconstructivist realm, with numerous over-lapping conceptual explanations for differences (see for example, Carroll 1988; Schlesinger and Heldmen 2001; Howell and Day 2000) divided between structural and situational explanations and socio-psychological differences (Gidengil et al m.s.) and including socialization of different norms and expectations, the impact of differences in education, and differences in work and family roles and responsibilities (Sapiro 2002). This paper focuses on three proposed explanations which interact extant but non-gendered theories about trade policy preferences with concepts from the gender gap literature: the political and economic knowledge gap between men and women, the observed and unobserved differences in objective and subjective individual economic roles and conditions (see for example Eagly and Diekman 2006), and gendered differences in the saliency of individual versus societal conditions, what has been traditionally deemed women s greater compassion for the vulnerable (see for example, Shaprio and Mahajan 1986; Welch and Hibbing 1992). Trade policy preferences and knowledge Non-trade related studies demonstrate that fully-informed respondents have different stated policy preferences than the less informed respondents and that disparities in policy-specific 6

7 information can be particularly influential in preference formation (Gilens 2001). Trade policy preference formation experiences knowledge gaps at both the general and specific level. Thus, one potential source for the gender gap in trade policy preferences is the gender gap in trade policy knowledge. Numerous studies have found that men test better than women on questions about politics. For example, analysis of results from a ten question knowledge battery incorporated in the 1990 Citizens Participation Study (CPS) found the gap between men and women scores equaled almost a full question or.93 on the 10 point scale (Verba, Burns, and Schlozman 1997). While similar initial estimates of a considerable political knowledge gap between men and women have been halved by more recent analysis correcting for the greater propensity of men to guess, the new estimates still find the divide between genders to be on par with other constructs such as interest in politics and political efficacy (Mondak and Anderson 2004). Specific to economic knowledge, Walstad and Rebeck (2002) calculated that men scored 7.6 to 11.9 percentage points higher on tests of macroeconomic knowledge, after controlling for age, education, income, party identification, and even economic education. In terms of specific knowledge, survey results have shown that women as a group tend to be less knowledgeable or less certain about trade-related policies. Burgoon and Hiscox (m.s.) found women 27-30% less likely than men to name all three signatories of NAFTA correctly. Guisinger (2009) using data from the CCES 2008 election survey found that compared to women, men were significantly less likely to answer Don t Know and more likely to correctly identify their Senator s policy position on trade (operationalized by knowledge of their Senator s 7

8 roll call vote on the ratification of a new free trade agreement between the US and countries in Central America, CAFTA). Respondents of both genders were far less familiar with trade policy specifics than other policies: when asked about their Senator s behavior on a series of roll call votes, respondents were 48% more likely to provide a wrong answer and 51% more likely to select Don t Know on CAFTA compared to the average for other roll call votes during the same Congressional session (banning late-term abortion, federal funding for stem cell research, a timetable to withdraw from Iraq, citizenship for illegal immigrants, increasing the federal minimum wage, and extending capital gains tax cuts). To the extent that preferences over trade-related policies depend on knowledge about the effects of trade and thus the types of trade, confusion about US trading partners and the role of trade in the economy could manifest in different preferences. If so, correcting and providing information about trade-related policies should remove a source of the gender gap in trade preferences. Individual and societal concerns and trade policy preferences A second explanation for the gender gaps may lie not in knowledge about trade but in beliefs about the effect of trade on one s employment, beliefs which are in turn dependent on how individuals internalize information about trade and its effects. Men and women s differing expectations about individually benefiting from the upside of trade or potentially paying the costs of trade-related economic volatility may create different preferences even among economically similar individuals. 8

9 A standard explanation for policy preference differences between men and women is the gender differences in the distribution of economic roles. Scholars have noted that in the US during the 20 th century, gender was a stronger determinant of employment allocation than race (Padavic and Reskin, 2002). For trade policy, the effect of gendered distributions of economic roles could be particularly acute since the two primary theoretical explanations for an individual s preference depend on a characterization of the individual by class and by industry sector. Both explanations start with Heckscher-Ohlin based assumptions of international trade and thus assume that a country will export goods that intensively use a country s abundant resources and import those goods that intensively use a country s scare resources. In the process of determining individuals responses to such trade, the two major models differ in their assumptions about the mobility of the factors of production (often characterized as land, labor, and capital but now more frequently described as skilled and unskilled labor). Trade preference models referencing Stolper- Samuelson assume perfect mobility within factors of production and thus characterize individuals very generally according to their ownership of factors of production. By changing the relative supply of each factor, trade is assumed to change the prices each factor receives: increasing prices for the abundant factors and decreasing prices for the scarce factor. In developed countries like the US, such models predict that due to direct individual economic concerns, holders of capital and high-skilled labor will be more likely to support international trade while labor, especially low-skilled labor, will be more likely to favor limits on international trade. Trade preference models referencing Ricardo-Viner assume limited factor mobility across different industries and thus characterize individuals primarily according to their industry rather than economic class. Again trade can affect the price each individual receives for their contribution to production: increasing prices in export competing industries and decreasing 9

10 prices in import-competing industries. Preference formation thus crosses factor lines: capitalists, high-skilled labor, and low-skilled labor all share similar interests when the ability to move to a different industry is limited. The centrality of job types, industry, and skill-levels as the source of differences in individual trade preferences means that few studies have failed to control for these types of observable socio-economic indicators. Additionally, the expectations provided by such models do not neatly fit with the empirical finding on high support for protection among women. To the extent that international trade allows for growth in US services industries, traditional gender segregation trends in the US disadvantage American men not women (Anker 1998) and would suggest greater support for protection by low-skilled men in particular. Focusing on individuals current economic roles fails to capture individuals perception of vulnerability or insecurity in their economic status. A corollary to the orthodox trade theory s expectation that increased liberalization generates increased growth is that increased liberalization additionally increases uncertainty, volatility, and risk. If women as a group perceive themselves to be less likely to benefit from the potential upside of increased trade and more likely to pay the cost of any downside because of disadvantages in recruitment processes (Padavic and Reskin 2002) and retention practices (Hall 1972; Diebold et al 1998) 3 ; if women are less supportive of the idea that people should move to regions where employment is more available (Gidengil et al. m.s.); or if women as primary caregivers even in dual-income families (Presser 1994) are more concerned with jobs compatible to family life (Darian 1975; Glass and 3 Hall (1982) estimated that in 1978, men were 30 percent more likely than women to reach five or more years of employment tenure, a much greater retention divide than he calculated between whites and blacks. Ureta (1992, cited in Dielbold 1998, corrected Hall s calculation to arrive at the lower but still substantial figure of 22% difference between men and women. Sheeran ( ) noted that while the application of Title VII in the United States increased employment opportunities for women and minorities, it did nothing to adjust workers seniority status which had been built on a foundation of discrimination and thus left women and minorities more vulnerable during economic downturns. 10

11 Camarigg 1992), then their responses may differ significantly from men s even when provided the same information. Perceptions of vulnerability rather than current circumstances have been found to effect attitudes towards general domestic policies (Rusciano 1992) as well economic policies (Gidengil 1995; Welch and Hibbing 1992); thus they may similarly explain differences in trade preferences. In contrast with the knowledge gap hypothesis, the provision of information will do little to alleviate the gender gap and may perhaps enlarge the gap if men and women update their perceptions differently even when presented the same factual information. A third source of the gender gap may arise from gender-related differences in the salience of societal economic concerns relatively individual self-interests. Mansfield and Mutz (2009) have found that an individual s perception of whether trade benefits or hurts the country has a stronger effect on an individual s trade preferences than his or her own economic circumstances or perceptions about trade s effects on those circumstances. This new model of trade preferences meshes with long-standing arguments that women are more likely to support altruistic policies which benefit society in general and disadvantaged groups specifically. Some scholars have argued that women s relative vulnerability increases identification with disadvantaged groups, enhancing women s altruistic instincts (for review, see Schlesinger and Heldman 2001, 61-62). Previous research has found women more supportive towards public assistance in general (Edlund 1999; Blekesaune, Morten, and Jill Quadagno 2003) and to disadvantaged groups specifically (Shapiro and Mahajan 1986). Trade protection offers just this type of assistance to groups that would be disadvantaged by competition from imports. If altruism explains a portion 11

12 of trade preferences and women tend to be more altruistic, gendered differences in sociotropic concerns then serve as a source or magnifier of gender differences in trade preferences. Research Strategy To test the information-based explanations for the gender divide in trade protection preferences, I designed a pair of survey experiments: one that isolated the effect of participants perceptions of the United States primary trading partners, and a second experiment that randomly varied exposure to a positive frame on the benefits of such trade. Unlike observational studies, survey experiments can independently vary individual circumstances, allowing for more careful testing of causal mechanisms: here, the effect of information on trade preferences. In both experiments, only truthful albeit carefully chosen and selectively applied information was provided in order to see the changes that could be created by information correction and provision. The survey also collected data on standard individual characteristics, such as age, gender, income, education, and political views. As such, the analysis combines the broad sample of traditional observational data with the inferential power of experimental data. A few works on international trade and related policies have undertaken an experimental survey framework (Hermann et al 2001; Hiscox and Hainmueller 2010). In these studies, the researchers provided treatment descriptions with very generic nomenclature and hypothetical scenarios. Even if one is willing to make the assumptions that survey respondents have meaningful opinions about abstract scenarios, such a research design is inappropriate for a study focused on knowledge. Thus, I use real names of countries and current data from government sources to describe the US s relationship with these actual countries. 12

13 Use of experimental survey data requires some degree of caution. Experimental economists and psychologists claim women s preferences are more situationally specific and their social preferences are more malleable than those of men (Croson and Gnezzy 2009). The initial expectation is that women should be more responsive to cues within the survey than men. Also, without panel data it may be difficult to determine how long treatment effects last. However, these limitations are outweighed by the benefits of the ability to rigorously test the constructs of interest. Data The data comes from the 2010 Cooperative Congressional Election Survey conducted by YouGov/Polimetrix both before and after the 2010 US Midterm election. From an opt-in panel of respondents, YouGov/Polimetrix used a matched random sample technique to select a nationally representative, stratified sample of 30,000 registered and unregistered US adults 4. The full sample received a common set of questions and a randomly selected sub-sample of 1,500 respondents received a smaller module of additional questions. Data collection occurred in two waves, pre- and post- election, with approximately a 15% respondent drop off between the two waves. All respondents answered a battery of questions concerning their individual characteristics (including income, education, gender, and age) and political views (creating a 7 point political 4 The sample included three strata: Registered and Unregistered voters, State Size, and Competitive and Uncompetitive Congressional Districts. Because of the common research interest in voters, YouGov/Polimetrix purposefully oversampled registered voters. 13

14 identification scale). Respondents in the sub-sample received two sets of questions concerning trade and trade-related policies. In one phase of the study, respondents were asked to identify Who do you think ranks as the United States top trading partner? from a randomly rotated list of five countries: Canada, China, the European Union, Japan, and Mexico. The first experiment 5 followed immediately afterwards; a randomly selected half of those who selected China were informed that Canada was in fact the US s largest trading partner and half of those who selected Canada were told that China and Mexico were the next two largest trading partners. All respondents then received a battery of questions concerning trade related policies. Specifically, they were asked to select Agree, Somewhat agree, Somewhat disagree, Disagree, Don't know on whether the government should increase limits on imported goods and services (Limits on Imports), increase subsidies to affected industries (Subsidies), do more to stabilize the dollar (Stabilize Dollar), increase unemployment benefits and job retraining programs (Increase Benefits), and ensure its trading partners protect basic rights of workers (Worker s Rights) 6. Answers were coded to a 5 point scale (-2 to 2) from Disagree to Agree with Don t knows coded as 0s. The treatment highlights two potential shifts: first, whether shared information leads to convergence between gendered groupings and second, whether correcting information leads to convergence within gendered groupings. If a lack of information, uncertainty of information, or poor information is driving differences then shared, correct information should eradicate these differences. In contrast, if genders process the information differently then preference gaps 5 Although the respondents were the same, the two experiments occurred on different dates: the data for experiment 1 was collected after the US Congressional election and the data for experiment 2 was collected in the weeks prior to the election. 6 Due to drop-off between the pre- and post-surveys, approximately 1,240 individuals completed this set of questions. 14

15 between genders should stay stable or even increase whereas differences within gender groups should decline. In a separate wave of the survey, the second experiment started with a random assignment of two different prompts prior to a series of questions about trade s effect on employment. A randomly selected half of the participants received a treatment in the form of factual, but positive, information about the rank of the US as the largest trade country and the link between exports and jobs in both the service and the manufacturing sectors: The US continues to rank as the largest trading country in the world. In the service sector, the US dominates the international market, exporting more than the next two countries combined. US exports in services more than doubled between 1998 and 2008 and service exports are estimated to support at least 3.5 million US jobs. Even in the manufacturing sector, where the US faces stronger competition, more than 1 in 5 jobs are dependent on exports. Then all participants were informed that The US government continues to expand opportunities to trade through bi-lateral and multi-lateral agreements with foreign countries before being asked to answer a set of four questions concerning how international trade effected Your employment, Employment of friends and family, Employment in your region, and Employment in the United States with the choices of Benefit Greatly, Benefit Slightly, No Difference, Hurt Slightly, Hurt Greatly, and Don t know. Again, answers were coded on to a -2 to 2 scale ( Disagree to Agree ) with No Difference and Don t know comprising the center score of 0. Furthermore, respondents were then asked to select why they thought trade affected employment in the way they did, selecting among provided answers or providing their own response. (See Appendix A for full text). By varying the information provided to respondents, the treatment allows for analysis of whether men and women respond differently to the type of positive 15

16 framing often provided by economists, politicians, and others supporting increased trade liberalization. Analysis of Lack of Political Knowledge: Does Wrong Information Lead to a Gender Gap in Trade Policy Preferences The 2010 CCES survey results corroborate prior findings of a divide between men and women in their support for trade-related policies. For a set of trade related policy questions about whether the US government should increase limits on imports (LIMITS ON IMPORTS), increase subsidies to affected industries (SUBSIDIES), do more to stabilize the dollar (STABILIZE DOLLAR), increase unemployment benefits and job retraining programs (INCREASE BENEFITS) and ensure its trading partners protect basic rights of workers (WORKER S RIGHTS), the left-side of Figure 1 displays the average responses of the control group coded answers (-2 Disagree to 2 Agree) by gender. On each issue, the average response by women (denoted by a square marker) was more supportive than the average response by men (denoted by a diamond marker). Additionally, the same survey responses suggest that knowledge or perceptions about who ranks as the US s top trading partner also matters in the formation of preferences over trade policy. The right-side of Figure 1 breaks down the same responses by top trading partner response. Respondents selected among one of 5 choices: Canada, China, the European Union, Japan, and Mexico. The 20% who selected the correct answer of Canada on average were less supportive of trade ameliorating policies than the almost 80% who erroneously selected China, the European Union, Japan, or Mexico. The full 60% of the sample who identified China as the US s top partner were 16

17 equally or more supportive of the policies than those who selected the European Union, Japan, or Mexico 7. As presaged by previous studies of economic knowledge, the probability of a respondent correctly answering the top trading partner question varies according to gender category. As depicted in Figure 2, women were as a group less likely than men to correctly identify Canada as primary (14% of women to 29% of men) and more likely to incorrectly identify China as primary (65% of women to 54% of men). A simple chi-squared test shows these differences between men s and women s responses are significant (Pr = 0.000) 8. Since women are more likely to believe that China is the US s top trading partner and since those who believe that China is the US s top trading partner are significantly more protectionist, then the next step is to test whether women are more protectionist than men because they hold erroneous information about the US s major trading partner. To test this proposition that gender differences in economic knowledge lead to the differences in stated preferences on trade policy, I extend the basic observational analysis in two ways first entering in gender and other characteristics to check that the statistical significance of estimates on trade partner response are not due to omitted variable bias and second entering an interaction on gender and response to ensure that the finding on economic knowledge in the form of country response is not in fact hiding its own gender divide. If the results are simply driven by differences in economic 7 Regression analyses undertaken on data collected from respondents in the control group (those whose perceptions were not adjusted through a specific primer described below) supports the descriptive presentation in Figure 1. Recoded responses (-2 Disagree to 2 Agree) for each policy area (LIMITS ON IMPORTS; SUBSIDIES; STABILIZE DOLLAR; INCREASE BENEFITS; WORKER S RIGHTS) were regressed on respondents answer to the question of whom ranks as the US s top trading partner through use of a set of dummy variables denoting each possible answer (the excluded category was Canada). With one exception, those who had ranked China as the US s top trading partner were substantially and significantly more supportive of protectionist policies than those who answered Canada. The size of the increase in support ranged from.19 to.35, a 5 to 9% swing on the 4 point scale. 8 Analysis utilizing Mlogit and thus allowing for the inclusion of controls for age, partisanship, and skill-level, provides evidence that women are significantly less likely to answer Canada, the correct answer. 17

18 knowledge regardless of gender issues, entering gender should not remove the significance of the results. If gender plays a special role not only independently but also in conjunction with knowledge (in this case, top partner response), then an interaction should be significant. Second, I analyze the results including observations from the survey experiment in which half of those who responded China received the correction that Canada was actually the United States largest trading partner and half of those who responded Canada received the information about China and Mexico being the next two largest trading partners. Since priors on trading partner strongly effect preferences, there should be a strong treatment effect if the lack of shared information is the source of the disagreement. In particular women who receive the treatment should have effect sizes that counterbalance the observational findings on women in general. Quantitative analysis can more closely parse the relationship between the persistent gap and gender by incrementally incorporating the potential effects of personal characteristics and the treatment itself into the basic regression model. Table 2 presents the results of detailed analysis of this observational and experimental survey data on both the support for increasing limits on imports (Limits on Imports) and the support for ensuring trading partners protect basic rights of workers (Worker s Rights). The first is the most commonly conceived type of trade protection and the latter is often deemed a specific concern of women. Models 1 to 3 use only observations from the control set (with no attempt to manipulate beliefs about the top trading country); whereas models 3 through 4 include also data from the survey experiment in which half of the China respondents received the additional information that Canada was in fact the US s top 18

19 trading country. Model 1 repeats the results of the base analysis presented in Table 1 and discussed previously. Model 2 includes individual characteristics that may determine trade protection preferences: a seven point party-identification measure (PID7), age (AGE), gender (FEMALE), and skill-level (SKILLED) with skilled defined as some education after high school. 9 Of these individual-characteristics, skill-level has received the most theoretical and empirical attention. Heckscher-Ohlin based assumptions of international trade generate the expectation that countries will export goods which use intensively abundant resources and import those which use scare resources. Thus in a developed country such as the US, high-skilled workers will benefit from the ability to export, while low-skilled workers will find themselves facing increased competition from imports. Model 3 further incorporates an interaction term combining trade partner response and gender which allows for independent varying effects of top trading partner priors by gender. For both LIMITS ON IMPORTS and WORKER S RIGHTS, a similar pattern emerges: adding individual characteristics and thus allowing for an independent estimate of each gender s propensity to support protectionist measures initially diminishes the size and wipes out the significance of RESPOND CHINA suggesting that it is gender, not trade partner knowledge that is the source of differences in trade preferences. However, model 3 presents a fascinating gender difference: for LIMITS ON IMPORTS, the coefficients on RESPOND CHINA, FEMALE and RESPONDED CHINA & FEMALE were all significantly different from the base category of male participants answering Canada. In particular, the interaction between RESPOND CHINA and FEMALE is large, significant, and negative which in conjunction witha positive estimate on RESPOND CHINA in model 3 suggests that women responding China are in fact no more (or 9 The smaller number of observations prevented the inclusion of employment status in models

20 less) protectionist than women in general. However, men who respond China are in fact more protectionist than other men. Thus it is men, not women, who are reacting to their belief about the identity of the US s primary trading partner. The analysis on WORKERS RIGHTS shows the same pattern although the significance on the coefficient for RESPOND CHINA is not robust (although the estimated coefficient is itself fairly stable). Thus, observational data disputes the hypothesis that it is women s lack of knowledge about trade that is driving the difference in men and women s preferred levels of protectionism. Clearly a subset of those responding China are far more protectionist, but that subset is men. Men appear far more affected by country-type than women. Including data from the survey experiment in the analysis (Models 4 to 6) further confirms the findings from the observational data. Providing shared information on the nature of the US top trading partner does little to change the gender gap between men and women: the treatment effect by itself (Model 4), interacted by gender (Model 5), and interacted by skill-level do not substantially or significantly diminish the gap between men and women on their trade policy preferences and the coefficient on FEMALE remains stable once priors on trade country are interacted with gender (Models 3-6). Even if the interaction of treatment with gender were significant, at the current estimate, knowledge would account for less than 17% of the gender gap. Priors about trading partners matter for a subset of individuals but not in a way that explains higher levels of support for protectionist policies by women. To illustrate the persistence of the gender gap, Figure 3 displays for the control group ( no treatment ) and treatment groups ( China/Mexico treatment and Canada treatment, the 20

21 average response by gender group and by trade partner response to the question The US government should increase limits on imported goods and services. The gap between women s and men s responses across treatment conditions is striking. Treatments did shift average responses. Those who responded Canada but were informed about China and Mexico s rank supported limits on imports more than those who did not receive the treatment. Those who responded China and received the correction that Canada was in fact the US s top trading partner supported limits on imports less than those who did not receive the treatment. Yet, the information provision only marginally diminished the gap between men s responses and women s responses, suggesting a durability of gender differences which is unaffected by knowledge about trading partners. Analysis of Different Processing of Political Knowledge The fact that men s trade preferences varied so greatly depending on their perceptions of trading partner, suggests that men and women may process political knowledge differently. In the second experiment, prior to questions concerning the effect of trade on their own employment, employment of their friends and family, employment in their region, and employment in the United States, all participants were informed of continuous government efforts to expand international trade. In addition, a randomly selected half received a positive framework which highlighted the importance of trade to the United States economy, the US world dominance in the service sector, the recent growth in US exports in services, and the importance of the export trade not only in service jobs but also manufacturing jobs. 21

22 The descriptive data provided by the control group (those not provided a positive framework before being asked to state their beliefs on the effect of international trade on their and other s employment) suggests only small differences between individual types (low skilled, high skilled, male, female) but dramatic differences between individuals beliefs about the effects of trade on their own employment and on US employment. Figure 4 displays a breakdown of responses both by question and by individual type. On average only 25% of respondents believe that trade hurts their own employment, but 62% of all respondents believe that trade hurts employment in the United States. Disaggregating the data by groups results in only minimal variations: compared to high-skilled workers, low-skilled workers are slightly more likely to think that trade hurts employment, more likely to think it makes a difference, and less likely to think it helps, regardless of which type of employment is under discussion; compared to women, men are more likely to think trade helps their own and US employment; compared to men, women are slightly more likely to think that trade makes no difference to their own or US employment prospects. These differences are minor however to the overwhelming difference between individuals expectations for themselves and for others. On its face, the descriptive data provides support for not only socio-tropic explanations of trade preferences but also gendered differences. Since the majority of individuals do not think trade affects them, it is unsurprising that individual trade preferences appear more linked to sociotropic concerns than individualistic ones. That women are slightly less likely to think that trade hurts their own employment and more likely to think that trade hurts others provides evidence for the hypothesis that women s altruism underpins the gender divide in trade preferences. 22

23 Analysis of the survey experiment results complicates such a characterization of the causal link between gender and differences in trade preferences. To see whether women s concerns about general unemployment are the source of the gender differences in trade once other factors are considered, I analyze data collected from both the control group and the treatment group, including the dummy variable TREATMENT and the interaction of the treatment dummy variable with FEMALE to allow for the estimation of the difference created by a positive treatment and to allow for the potential of a heterogeneous treatment effect. Not surprisingly, a positive framing of trade dramatically increased respondent perceptions about the benefit of trade on employment and particularly employment in the region and in the United States as a whole. Figure 5 presents a comparison between the control group and treatment group for all four questions on employment ( Your employment, Employments of friends and family, Employment in you region, and Employment in the United States. Again the square marker denotes the average response by women and a diamond marker denotes the average response by men with the line between the two demarking the difference for both the control group and the treatment group. Surprising, the positive message dramatically increased the average differences between men and women: increasing a.02 to.12 difference between genders to a.25 to.34 difference between genders. Across the board, men responded far more positively to the treatment than women. Such findings run counter to the expectation of experimental economists and psychologists that women s preferences are more situationally specific and their social preferences are more malleable than those of men (Croson and Gnezzy 2009). 23

24 Table 3 presents regression analysis which allows for additional variable of interest to also be included: as before a seven point party-identification measure (PID7), age (AGE), gender (FEMALE), and skill-level (SKILLED). With the exception of party-identification, all are significant. As theoretically expected, skilled workers are less concerned about the effect of trade on their own employment and this difference carries across the board to beliefs about the effect of trade on others. Women, with a few exceptions, appear less certain of the benefits of trade on their own employment and this difference from men increases as they consider employment more generally. The average treatment effect has the expected positive effect on individuals perception of trade s effects on employment. Comparing across Model 1 s, the treatment effect on individual benefits appears somewhat smaller than those for societal benefits; however, this is only when treatment effects are assumed to be homogenous. Allowing for a heterogeneous treatment effect through the inclusion of the interaction between TREATMENT and FEMALE (Model 1) finds significant gender differences in the reaction to the treatment: for own employment and employment of friends and family women, the treatment effect is substantially smaller on men, about 2/3 smaller (comparable in size to the difference between high and low skilled workers.). The gendered treatment effect is smaller and not significant for the more general societal concerns. This evidence suggests that it may be women s concerns about their individual risk rather than socio-tropic concerns that lead to the difference in trade preferences - a reverse expectation to that presented by the literature to date. 24

25 While the experiment does not directly explore why men reacted more positively to the positive trade information, insight is provided by responses to follow-up questions. Individuals were prompted to explain why they thought that trade benefitted or hurt their own employment and US employment and were provided both set answers and the opportunity to provide an open answer. In comparing men s and women s responses both from the control and treatment group, a few patterns emerge. In the control group, men were far more likely than women to explain their individual employment benefits in terms of their own skills (45% versus 13%); women were far more likely to explain the benefit of trade in terms of the success of their company internationally and their ability to change jobs. In the treatment group, differences between men and women were in fact minimized as many more men expressed benefits in terms of international competitiveness. In comparison, men and women s explanations for why trade hurt their employment were very similar in the control group and diverged in the treatment group: in particular, the treatment group had almost half the percentage of men who thought that trade hurt employment (26% control, 12% treatment) but those who continued to think trade hurt overwhelmingly expressed concerns about outsourcing. Women s concerns about outsourcing varied only minimally. Also, in both the control group and the treatment group, men were less likely to express concerns about moving for work than women were. Conclusion Why do men and women differ in their preferences for trade protection? Although women and men differ in tests of economic knowledge, particularly in questions pertaining to trade, these 25

26 informational disparities do not appear to drive preference differences. Instead men and women appear to respond differently to positive messages about trade. Contrary to those who suggest that women s altruism drives their greater desire for protectionist policies, positive messages appear to mainly affect individual s beliefs about employment, suggesting that individual concerns may be the primary source of preference differences between the sexes. 26

27 References Andersen, Kristi Gender and Public Opinion. in Understanding Public Opinion, ed. Barbara Norrander and Clyde Wilcox. CQ Press Washington, DC: Anker, Richard Gender and Jobs: Sex Segregation of Occupations in the World. Geneva: International Labor Organization. Baker, Andy. (2005). Who Wants to Globalize Consumer Tastes and Labor Markets in a Theory of Trade Policy Beliefs? American Journal of Political Science. 49(4): Beaulieu, Eugene, Ravindra A. Yatawara, and Wei Guo Wang Who supports free trade in Latin America? World Economy. 28 (7): Blekesaune, Morten and Jill Quadagno Public Attitudes toward Welfare State Policies: A Comparative Analysis of 24 Nations. European Sociological Review. 19 (5): Carroll, Susan J Women s Autonomy and the Gender Gap: 1980 and In The Politics of the Gender Gap, ed. Carol Mueller. Newbury Park, CA: Sage. Ceccoli, Steven J., Ronald D. Gelleny, and Karl C. Kaltenthaler Explaining citizen support for trade liberalization. International Studies Quarterly. 48: Croson, Rachel and Gneezy, Uri Gender differences in preferences. Journal of economic literature. 47.2: Darian, Jean C Convenience of work and the job constraint of children. Demography (May): Diebold, Francis X., David Neumark, and Daniel Polsky Job Stability in the United States. Journal of Labor Economics. Vol. 15: Eagly, Alice H. and Amanda B. Diekman Examining Gender Gaps in Sociopolitical Attitudes: It's Not Mars and Venus. Feminism Psychology. 16(1): Edlund, Jonas Trust in government and welfare regimes: Attitudes to redistribution and financial cheating in the USA and Norway. European Journal of Political Research. 35; Gidengil, Elisabeth and André Blais, Richard Nadeau, Neil Nevitte (m.s.). Women to the Left? Gender Differences in Political Beliefs and Policy Preferences. Chapter prepared for Manon Tremblay and Linda Trimble (eds.) Gender and Elections in Canada Gidengil, Elisabeth Economic Man Social woman? The Care fr the Gender Gap in Support for the Canada0nited States Free Trade 27

28 Gilens, Martin Political Ignorance and Collective Policy Preferences. American Political Science Review. 95: Glass, Jennifer and Valerie Camarigg Gender, parenthood, and job-family compatibility. The American Journal of Sociology (July): Hainmueller, Jens and Michael J. Hiscox Learning to Love Globalization: Education and Individual Attitudes Toward International Trade. International Organization. 60: Hainmueller, Jens and Michael J. Hiscox Attitudes toward Highly Skilled and Lowskilled Immigration: Evidence from a Survey Experiment. American Political Science Review. 104: Hall, Robert E. (1972), "Turnover in the Labor Force," Brookings Papers on Economic Activity. 3: Herrmann, Richard K and, Philip E. Tetlock, Matthew N. Diascro How Americans Think About Trade: Reconciling Conflicts Among Money, Power, and Principles International Studies Quarterly (June): Hiscox, Michael and Brian Burgoon The Mysterious Case of Female Protectionism: Gender Bias in Attitudes Toward International Trade Working Paper: Harvard University. Hiscox, Michael Through a Glass and Darkly: Attitudes toward International Trade and the Curious Effects of Issue Framing. International Organization. 60(Summer): Kono, Daniel y Optimal obfuscation: Democracy and trade policy transparency. American Political Science Review : Lee, Hyunji. m.s. The source of female protectionism: The consequence of the knowledge gap in economics. Presented at the Canadian Political Science Association Annual Conference, Mansfield, Edward D. and Diana c. Mutz Support for Free Trade: Self-Interest, Sociotropic Politics, and Out-Group Anxiety. International Organization. 63: Mayda, Anna Maria and Dani Rodrik Why are some people (and countries) more protectionist than others? European Economic Review. 49: Mondak. Jeffery J. and Mary R. Anderson The Knowledge Gap: A Reexamination of Gender- Based Differences in Political Knowledge. The Journal of Politics (May): O Rourke, Kevin and Richard Sinnott The Determinants of Individual Trade Policy Preferences: International Survey Evidence. Brookings Trade Forum: Padavic, Irene and Barbara Reskin Women and Men at Work, second edition. Thousand Oaks, California: Pine Forge Press. 28

29 Presser, Harriet B Employment schedules among dual-earner spouses and the division of household labor by gender. American Sociological Review (June): Sapiro, Virginia It s the Context, Situation, and Question, Stupid; The Gender Basis of Public Opinion. in Understanding Public Opinion, ed. Barbara Norrander and Clyde Wilcox. CQ Press Washington, DC: Scheve, Kenneth and Matthew Slaughter What determines individual trade policy preferences? Journal of International Economics. 54: Schlesinger, Mark and Caroline Heldmen Gender Gap or Gender Gaps? New Perspectives on Support for Government Action and Policies. The Journal of Politics (February): Shapiro, Robert Y. and Harpreet Mahajan Gender Differences in Policy Preferences: A Summary of Trends From the 1960s to the 1980s. The Public Opinion Quarterly (Spring): Sheeran, Timothy J Title VII and Layoffs under the Last Hired, First Fired Seniority Rule: The Preservation of Equal Employment. Case Western Reserve Law Review. Vol. 26: Spence, Michael and Sandile Hlatshwayo The Evolving Structure of Global Growth The New Wealth of Nations. Feb 14, Project Syndicate. Verba, Sidney, Nancy Burns, and Kay Lehman Schlozman Knowing and Caring About Politics: Gender and Political Engagement. Journal of Politics. 59 (November): Welch, Susan and John Hibbing Financial Conditions, Gender, and Voting in American National Elections. The Journal of Politics, 54.1: Welch, Susan, and John Hibbing Financial Conditions, Gender, and Voting in American National Elections. Journal of Politics 54 (February): Wolbrecht, Christina Gender differences in Public Opinion. Polling America. Ed. Samuel Best and Ben Radcliffe. Westport CT: Greenwood Press: Wolfe, Robert and Matthew Mendelsohn Values and Interests in Attitudes toward Trade and Globalization: The Continuing Compromise of Embedded Liberalism. Canadian Journal of Political Science. 38:1 (March):

30 Figure 1: Support for trade related policies by gender and by perception of top trading partner 30