The Gender Gap in Trade Preferences

The gender gap in mass attitudes toward protectionism was identified at least as early as Holsti (Reference Holsti1996). Because lowering trade barriers exposes domestic producers to international competition and therefore creates new wage pressures and displacement risk for domestic workers, subsequent efforts to explain this gap have focused on gender differences in observed labor market risk or participation (e.g., Mayda and Rodrik Reference Mayda and Rodrik2005). However, the gap persists even when accounting for these factors, suggesting that the divide may be driven by differences in perceptions of the consequences of international trade. For example, women might be less exposed to economic ideas about the benefits of free trade and thus perceive trade as more threatening (Mansfield and Mutz Reference Mansfield and Mutz2009).

Disentangling informational explanations, Guisinger (Reference Guisinger2016) shows that different risk perceptions, rather than differences in knowledge about trade, are at the core of the gender gap in trade preferences. Brutger and Guisinger (Reference Brutger and Guisinger2022) trace this gender gap to different responses to the labor market volatility induced by trade. Trade policy is often perceived as a main lever in the trade-off between prices and employment, and women tend to accept higher prices in exchange for reduced labor market volatility (Scheve Reference Scheve2004). Complementary work indicates that this increased risk sensitivity may be driven by gendered differences in competitiveness or willingness to relocate for employment (Mansfield, Mutz, and Silver Reference Mansfield, Mutz and Silver2015), and experiences of labor market discrimination (Guisinger and Kleinberg Reference Guisinger and Kleinberg2021).Footnote ² Indeed, both women workers and entrepreneursFootnote ³ may enjoy fewer benefits from globalization.

Although there is consensus that there is a gender gap in trade policy preferences, whether and how these differences affect policy outcomes is an open question. The only related research on the topic asks how women’s suffrage reshaped trade policy and yields competing results. Hall, Kao, and Nelson (Reference Hall, Kao and Nelson1998) argue that women gaining the franchise in the United States led to a reduction in trade barriers, whereas De Bromhead (Reference De Bromhead2018) examines 30 countries during the interwar period and concludes that women’s suffrage led to an increase in trade barriers. Neither examines the effects of women’s representation in parties or government.

Linking Women’s Representation and Protectionist Policy

Existing research suggests that women politicians should be more likely than men to support protectionism. First, women elites are more apt to share the policy preferences of women citizens (Clayton and Tang Reference Clayton and Tang2018; Wängnerud Reference Wängnerud2000). Indeed, our analysis of surveys of Members of the European Parliament and candidates running in national parliamentary elections in 11 countries suggest that, like women in the electorate, women politicians are more protectionist than men are (see Appendix SI.2). Second, even if women in the electorate and women elites have different policy preferences, mandate effects lead women in government to act on the specific interests of women in the polity (Franceschet and Piscopo Reference Franceschet and Piscopo2008). For example, the only research to date on women’s representation and trade policy shows that women in legislatures represent women’s interests by using their influence to ameliorate gender-discriminatory apparel tariffs (Betz, Fortunato, and O’Brien Reference Betz, Fortunato and O’Brien2021). Though women in the electorate are unlikely to support specific tariffs that actively discriminate against them, they do prefer protectionism overall. Women elites, in turn, should advance more protectionist platforms and policies. An analysis of roll-call votes on free trade agreements in the US House of Representatives lends individual-level support to this expectation (see Appendix SI.2).Footnote ⁴

We begin by examining the effects of women’s representation on parties’ policy platforms. Women’s influence here is a necessary condition for their influence on government trade policy—party platforms are antecedent to government action. This analysis also allows us to evaluate, across many countries and years, whether gender differences in trade preferences survive the political selection process. Given existing research on the influence of women’s presence in parties on their policy positions (e.g., Greene and O’BrienReference Greene and O’Brien2016; Kittilson Reference Kittilson2011; Weeks Reference Weeks2022), we expect that women’s representation is associated with more protectionist statements on party policy platforms.

We next examine whether gender differences in representation leave a mark on real trade policy outcomes, focusing on tariffs. Here, existing work offers less conclusive expectations. On the one hand, a large body of research finds correlations between women’s descriptive and substantive representation, and documents women legislators’ (Barnes Reference Barnes2016; Barnes, Beall, and Holman Reference Barnes, Beall and Holman2021; Betz, Fortunato, and O’Brien Reference Betz, Fortunato and O’Brien2021; Clayton and Zetterberg Reference Clayton and Zetterberg2018; Franceschet and Piscopo Reference Franceschet and Piscopo2008) and cabinet members’ (Atchison Reference Atchison2015) efforts on behalf of women. On the other hand, trade policy can be subject to intense political conflict. Even if gender differences in the mass public result in differences in party platforms, women’s descriptive representation may have a more limited effect on government policy.

We thus expect a nuanced relationship between women’s representation and tariff rates. Specifically, we posit that women’s presence has different effects across products, depending on their visibility to citizens. Visibility is highest for consumption products, such as automobiles, which citizens encounter in their daily lives. It is lowest for intermediate inputs used by firms in the production process, like drive-axles or airbag inflators, which citizens rarely observe directly. Visibility in turn shapes the politics surrounding trade policy on consumption products and intermediate inputs for two related reasons.

First, because of their visibility, consumption products are the most politically expedient place to implement a protectionist agenda. It is easier for politicians to portray protectionism as benefiting domestic producers of tangible goods familiar to consumers than the producers of inputs that most consumers never encounter directly. Of course, higher tariffs on consumption products might lead to higher household expenditures and therefore a preference for lower tariffs. Yet, the consumption effects of trade frequently play only a limited role in citizens’ evaluations of trade policy (Bearce and Moya Reference Bearce and Moya2020; Hiscox Reference Hiscox2006), likely because few are aware of the link between tariffs and prices (Davenport, Dorn, and Levell Reference Davenport, Dorn and Levell2021; Rho and Tomz Reference Rho and Tomz2017) and many (including women in particular) are supportive of higher prices in exchange for economic stability (Scheve Reference Scheve2004). These findings suggest that policy makers can implement a protectionist agenda without fear of immediate reprisals from price-conscious citizens—and instead garner increased support from protectionist voters.

Second, firm lobbying tends to be more intensive and effective on intermediate inputs, which places more constraints on a protectionist agenda. Recent work shows that multinationals and importing firms dominate lobbying on trade policy, including in the context of trade agreements, pushing for lower tariffs for intermediate inputs in their production process (Osgood Reference Osgood2018).Footnote ⁵ This is also reflected in trade policy choices. Across countries, both in the context of unilateral trade policy and in reciprocal trade agreements, we observe lower tariffs and faster tariff cuts on intermediate inputs than on consumption products (Anderer, Dür, and Lechner Reference Anderer, Dür and Lechner2020; Baccini, Dür, and Elsig Reference Baccini, Dür and Elsig2018; Betz and Pond Reference Betz and Pond2019; Shapiro Reference Shapiro2021).

Several explanations account for this pattern. In contrast to consumer goods, production inputs are frequently sourced by a small number of large firms (Bernard et al. Reference Bernard, Jensen, Redding and Schott2018). The concentration of benefits on a small set of firms solves collective action problems in support of trade liberalization, and, relative to consumption products, tilts the political environment against protectionism. Reinforcing these effects, and related to our argument, firm lobbying tends to be more effective for less visible policy issues, where policy makers face fewer constraints from the public (Hillman and Hitt Reference Hillman and Hitt1999). Of course, individual wholesalers and retailers have lobbied for lower tariffs on behalf of, and in concert with, consumers. An example is the efforts by US retailer, JoAnn Fabrics, to rally support against tariffs during the US–China trade war.Footnote ⁶ Yet, these cases are rare. For many consumer goods, firms are able to pass on the costs from higher tariffs (Amiti, Redding, and Weinstein Reference Amiti, Redding and Weinstein2019)—reducing the incentives for firms to lobby for lower tariffs on these goods. For policy makers who are pursuing a protectionist agenda, this implies greater constraints on higher tariffs on (less visible) intermediate inputs and more political opportunities for tariffs on (visible) consumption products. Therefore, we expect that women’s representation is associated with higher tariffs on consumption products and that this effect decreases for intermediate inputs.

Women’s Representation and Protectionism in Party Platforms

We begin by assessing the relationship between women in political parties and parties’ trade preferences in three different samples.Footnote ⁷ Our first analysis examines party platforms in 81 national elections across 20 democracies from 1996 to 2017 (474 total party platforms).Footnote ⁸ We combine data on women’s descriptive representation in parties’ parliamentary delegations taken from Adams et al. (Reference Adams, Bracken, Gidron, Horne, O’Brien and Senk2022)Footnote ⁹ with estimates of parties’ trade preferences based on content codings of their electoral manifestos provided by Volkens et al. (Reference Volkens, Burst, Krause, Lehmann, Matthieß, Merz, Regel, Weßels and Zehnter2020). Following Lowe et al. (Reference Lowe, Benoit, Mikhaylov and Laver2011), we aggregate all positive and negative statements regarding trade, where higher values reflect a preference for greater trade restrictions (including domestic protections, quotas, tariffs, etc.).

We estimate ordinary least squares models using party platforms as our unit of analysis and constraining estimates to within-election (e.g., Israel 1996) variation using fixed effects. This holds constant the effect of potential confounders that vary over time and space (e.g., trade integration, unemployment) on both predictor and outcome, identifying the estimate only with variation in women’s representation and trade preferences across parties within a given election. We also distinguish between mainstream and “niche” parties, as niche parties are substantially more likely to make hostility toward trade (and immigration) a central focus of their platforms.Footnote ¹⁰

The results in Table 1 support our first expectation. We find a significant, positive association between women’s representation and the level of protectionism in the party’s platform. The effect of increasing the proportion of women in the party by 20 percentage points (i.e., from 30% of the legislative delegation to 50%) is 0.12, which is larger than the difference between the 50th and 75th percentile on the dependent variable.

Table 1.

Women’s Representation within Parties and Protectionism in Party Platforms

Note: Larger values of the dependent variable indicates stronger support for trade barriers. Coefficient estimates, standard errors in parentheses; *p < 0.10, ^**p < 0.05, ^***p < 0.01, two-tailed tests.

Comparing parties in national elections across space and time suggests that women are influencing the platforms of their parties, making them more protectionist on average. At the same time, these results are subject to typical concerns about design and inference. We thus supplement the cross-national results with two additional analyses, drawing on the 2014 European Parliamentary (EP) election and national elections in postwar Sweden. Studying the EP election allows us to observe multiple parties within several countries competing in a single contest for the same policy authority.Footnote ¹¹ The Swedish case allows us to estimate a causal effect, via difference-in-differences, of the imposition of a gender parity rule by three major parties.

We once again combine information on parties’ stated trade preferences with information on women’s descriptive representation. For the EP, data on women’s representation come from Aldrich (Reference Aldrich2020), who measures women’s share of the party list. Euromanifesto data come from Schmitt et al. (Reference Schmitt, Braun, Popa, Mikhaylov and Dwinger2016) and protectionist trade positions are measured as above.Footnote ¹² Swedish quota data are taken from IDEA (2009). Quotas mandating gender parity (“50/50” split) in party lists were adopted by the Social Democratic Party and the Left Party in 1993 and by the Green Party in 1997; the remaining parties (Centre, Christian Democrats, Moderates, New Democrats, People’s Party, and the Sweden Democrats) are the control group.Footnote ¹³ Here, we again use manifesto codings from Volkens et al. (Reference Volkens, Burst, Krause, Lehmann, Matthieß, Merz, Regel, Weßels and Zehnter2020) to measure preferences for protectionism and we analyze parties’ positions in 23 elections between 1944 and 2018.

As before, the results are found in Table 1. These analyses further support the expectation that women’s representation increases parties’ preferred level of protectionism. The 2014 EP analysis, which constrains estimates to within-country variation, shows a significant and positive association between women’s representation on party lists and protectionism. In this case, the effect of increasing the proportion of women in the party by 20 percentage points is 0.31, which is larger than the difference between the 50th and 80th percentile on the dependent variable. For the Swedish study, we apply the typical two-way (party and year) fixed effects specification. The estimate shows evidence for a positive, causal effect of the gender parity rule on protectionism—once women are granted equal representation within parties, their parties’ policy preferences grow significantly more protectionist (p = 0.06, two-way test). The effect of the quota, about 0.31, is larger than the difference between the 50th and 90th percentile on the dependent variable. Taken together, these three analyses provide strong evidence that women’s presence in parties’ legislative delegations drives those parties to adopt more protectionist trade policy.Footnote ¹⁴

Women’s Representation and Protectionist Tariffs

We next assess whether women’s representation in government is associated with more protectionist policy choices. The analysis combines data on women’s political representation with tariffs at the level of Broad Economic Categories (BEC) for 141 countries from 1991 to 2019.Footnote ¹⁵ We focus on representation in the legislature (using log-transformed seat shares) and in the executive (using the log-transformed share of women in the cabinet), taken from Hughes et al. (Reference Hughes, Paxton, Clayton and Zetterberg2017) and Nyrup and Bramwell (Reference Nyrup and Bramwell2020).Footnote ¹⁶ Because we posit differential effects for women’s representation for consumption products and intermediate inputs, we distinguish between the two product categories using the BEC classification with a dummy variable, derived from end-use classifications from the System of National Accounts (United Nations Department of Statistics and Social Affairs 2002).Footnote ¹⁷ Accordingly, our dataset is in the country-year-product category format. The dependent variable is the (log-transformed) applied tariff rate from the World Bank’s World Integrated Trade Solution.

We estimate linear regression models, with standard errors clustered by country. All models include logged GDP, GDP per capita, unemployment rates, and polity scores. All models also include year and country fixed effects. Our research design thus takes advantage of country-specific changes in women’s political representation over time and assesses their differential effects across product types while accounting for global trends in women’s representation and tariff rates. We can therefore rule out that our results are driven by country-specific preferences for—or global trends toward—protectionism and women’s representation.

Table 2 reports results for the association between women’s representation and tariff rates. Odd columns report results for legislatures, even columns for executives. The findings suggest that a higher share of women legislators and cabinet members translates into more protectionist trade policy on higher-visibility consumption products. As posited, this effect does not hold for lower-visibility intermediate inputs, which are also subject to greater lobbying for trade liberalization. Columns 1 and 2 present the results from the base model, with the control variables mentioned above included. An increase in women’s legislative seat shares of 10% is associated with an increase in tariff rates of about 1.40%. For cabinet members, the effect is about half as large, with an increase of about 0.72%. At the same time, the interaction terms indicate that women’s representation has small and statistically insignificant effects on the tariff rates of intermediate inputs (p = 0.71 for legislatures, p = 0.21 for cabinet members).

Table 2.

Women’s Representation and Protectionism in Trade Policy

Note: Coefficient estimates, standard errors clustered by country in parentheses; *p < 0.10, ^**p < 0.05, ^***p < 0.01, two-tailed tests.

The remaining models indicate similar effects. Columns 3 and 4 show the results when incorporating political variables: the electoral rule, system of government, and partisanship of the largest party in governmentFootnote ¹⁸ (from Cruz, Keefer, and Scartascini Reference Cruz, Keefer and Scartascini2021). In columns 5 and 6, we include secondary school enrollment, to capture skill levels of the work force (from the World Bank); the Women, Business, and Law Index, an index of eight indicators of how legal rules shape women’s economic opportunities (from the World Bank); and women’s labor market participation (from the International Labour Organization), to account for the correlation between women’s participation in economic and political markets. The substantive results hold: women’s representation has positive effects on the tariff rates of consumption products and small and statistically insignificant effects on the tariff rates of intermediate inputs, and the difference between the two effects remains statistically significant. Consistent with our emphasis on the role of politically influential pro-trade firms, and confirming earlier work, the results also show that intermediate inputs indeed have lower tariffs than consumption goods.

The appendix includes a series of additional results, three of which we highlight here. First, in SI.4 we present results from additional fixed effects specifications, such as including fixed effects for higher-order BEC groups (e.g., “food and beverages”),Footnote ¹⁹ fixed effects for each product category, and country-year fixed effects. The results remain robust to these demanding fixed-effects models.

Second, in SI.5, we account for demand-driven explanations of protectionism and women’s representation, showing that our results are robust to import shocks and to limiting the sample to years after 2001, when the “China shock” led to a protectionist turn across countries (Colantone and Stanig Reference Colantone and Stanig2018). Drawing on survey data from Pew, we also show that our results are not explained by a growing preference for protectionist policies in the population.

Third, in SI.6, we show that consumption products and intermediate products do not differ systematically in their reliance on women as employees, suggesting that our results reflect support for protectionist trade policy that extends beyond its immediate labor market effects on women.