The Demand for Protectionism: Democracy, Import Elasticity, and Trade Barriers. Timothy M. Peterson University of South Carolina.

The Demand for Protectionism: Democracy, Import Elasticity, and Trade Barriers Timothy M. Peterson University of South Carolina and Cameron G. Thies University of Iowa Verso running head: The Demand for Protectionism Recto running head Timothy M. Peterson and Cameron G. Thies Numerous studies suggest that democracies employ lower trade barriers than nondemocracies. In this paper, we examine the conditioning role that the elasticity of import demand at the commodity level plays on the relationship between democracy and import barriers. Beginning with the assumption that democracies are more responsive than non-democracies to the preferences of mass publics, we demonstrate that the value of free trade as a public good depends on the elasticity of import demand. When import demand for a given commodity is inelastic, trade barriers are more harmful to consumers; as such, democracies will employ lower trade barriers than non-democracies. However, as import demand becomes more elastic, publics find it easier to adjust to higher prices; as a result, the difference in imposed trade barriers by regime type decreases. We find support for this argument in statistical analyses of cross-sectional data covering 4,656 commodities imported by 73 countries Furthermore, we find that democracies raise higher trade barriers than non-democracies on commodities for which import demand is very elastic. 1

Studies show that democracies tend to have lower trade barriers than non-democracies (Bliss and Russett 1998; Mansfield, Milner, and Rosendorf 2000; Milner and Kubota 2005), a phenomenon often attributed to the fact that democracies are more responsive to mass publics, which, in the aggregate, are harmed by protectionism. Yet, political competition within democracies sometimes leads to protectionist trade policy. Although often attributed to the influence of narrow interest groups (e.g., Stigler 1971, Grossman and Helpman 1994), higher trade barriers in democracies can also follow as a response to populist pressure (Rogowski 1987, 1989). Given the potential for majorities to allow perhaps even prefer trade barriers in some cases, the question arises: what factors influence citizens demand for protectionism? In this paper, we examine one such determinant of the public's trade policy preferences. We contend that, at the commodity level, the price elasticity of import demand influences the degree to which mass publics favor liberal trade policy, a preference to which democratic leaders are responsive. When demand is inelastic, protectionism harms populations more because they have little recourse other than to pay higher prices for desired commodities, the result of which is a loss of real income. However, as demand becomes more elastic, protectionism is less harmful to publics given the relative ease with which they can adjust to higher prices. As a result, leaders in democracies will employ less protectionism than non-democracies on commodities for which demand is inelastic in order to provide a public good to their citizens. However, as demand becomes more elastic, the value of free trade as a public good declines, and thus the difference in protectionism between democracies and non-democracies diminishes. This effect is reinforced by the fact that demand elasticity also has implications for the strength of domestic producers opposing free trade; all else equal, more elastic demand suggests the presence of more import 2

competitors who stand to gain from protectionism. Statistical tests confirm a hypothesis following from this argument. In fact, we find that democracy is associated with relatively higher levels of protectionism when import demand for a given commodity is very elastic. This study is designed to build upon extant research on the link between democracy and protectionism. Importantly, we examine this connection at the commodity level, contrary to the majority of studies that aggregate to the country or industry. Accordingly, we are able to distinguish whether protectionism (or a lack thereof) results from the distinguishing characteristics of potential imported goods. Additionally, in accordance with recent research (e.g., Kono 2006; O Rourke and Taylor 2005), we demonstrate that democracy is not necessarily synonymous with lower trade barriers. Rather, democracies are responsive to citizens, for whom the value of liberalized trade policy varies by commodity. We also demonstrate a somewhat counter-intuitive conclusion: although aggregate efficiency losses from protectionism are higher when demand is more elastic, a fact that leads all states to employ lower overall protectionism (Grossman and Helpman 1994), democracies raise relatively higher trade barriers than nondemocracies under this condition. We suggest that this result follows because voters are aware of price increases and loss of real income stemming from trade barriers, yet generally unaware of deadweight losses. Similarly, import competitors can reap high returns from protectionism on a commodity with elastic import demand; accordingly, their likelihood of lobbying for trade barriers is higher. This paper proceeds as follows. First, we review work examining democracy and protectionism, highlighting the relative lack of attention paid to commodity-level factors. Then, we describe the mechanisms by which import elasticity conditions the influence of democracy on the extent of import trade barriers at the commodity level. We outline our research design and 3

test a hypothesis following from our argument utilizing a cross-section of 4,656 commodities imported by 73 countries. Finally, we discuss the implications of our findings for the study of protectionism and the consequences for international security of trade policy. Democracy and Protectionism Kono (2006: 369) highlights three propositions linking democracy to liberalized trade. First, all types of government are lobbied by interest groups seeking protection, and all governments respond to some degree to such pressures (e.g., Lee and Swagel 1997). Second, mass publics generally prefer more liberal trade than would likely result from complete government accession to interest group demands (e.g., Milner and Kubota, 2005; Hermann et al. 2001). Third, democratic governments are more responsive to mass publics than autocracies. All three of these propositions are logically sound and have a substantial degree of empirical support; thus many have argued that democracies should unconditionally have lower trade barriers than autocracies (Bliss and Russett 1998; Mansfield, Milner and Rosendorff 2000; Milner and Kubota 2005). This conclusion that democracies employ less protectionist trade policy than nondemocracies tends to follow from the assumption that the benefits of protectionism are highly concentrated while its costs are diffuse Milner and Kubota (2005, 108). Trade barriers are more easily employed by autocratic regimes because the revenues they generate help to support the leader and his supporters as they seek to maintain power (e.g., Bueno de Mesquita et al. 2003). Taxes on trade for revenue are common in developing countries that have not experienced long periods of democratic governance; conversely, democracies tend to exchange direct income taxes for public goods in accordance with the fiscal contract theory of the state (e.g., Timmons, 2005). Similarly, democratic governments are typically considered to be less interested in maximizing 4

revenue extraction from their citizens, whereas autocracies are often little more than stationary bandits (Olson 1993; Przeworski, 1990). Given that democratization prompts political leaders to be sensitive to previously disenfranchised citizens, Milner and Kubota (2005) suggest that movement toward democracy in developing countries results in the diminished ability of leaders to use protectionist trade policies that impose higher costs on citizens. However, other studies have challenged the idea that democracy always leads to lower trade barriers on the grounds that the median voter does not always benefit from liberalized trade policy. For example, Rogowski s (1987) examination of political cleavages resulting from exposure to trade demonstrates that majorities could prefer protection from foreign competition. Rogowski demonstrates that in land-abundant, labor-scarce states, labor will seek protectionism while land seeks openness. If labor comprises a majority (as it tends to, especially in poorer countries), then democratic leaders should provide higher trade barriers when labor demands them. 1 Furthermore, Rogowski shows that if this high land-to-labor ratio occurs in less advanced economies, a labor-capital coalition favoring protectionism is likely to form, leaving owners of land isolated as the only advocates of openness. Indeed, Rogowski (p 1125) points out that this was the case in the United States leading up to the election of 1896, wherein rural interests favoring free trade lost to a protectionist, urban coalition of labor and capital. O Rourke and Taylor (2005) find cross-national evidence that democratization can actually lead to increased protectionism when labor is scarce while land is abundant, a conclusion that follows because labor tends to comprise a political majority even when it is a scarce factor endowment. However, the authors also find evidence that the US experience in 1896 is part of larger pattern in which 1 However, Ruggie (1982) points out that the losers to openness may increase their support for reduced trade barriers if the government compensates them. 5

labor tends to win protectionism only in poorer countries in cases where the owners of capital also favor closure. The scholarly consensus regarding the proposition that democracies are more responsive than non-democracies to the demands of mass publics has led to inquiries into public sensitivity to trade policy. The transparency of trade policy plays an important role in this regard. Mansfield, Milner and Rosendorf (2002) demonstrate that preferential trade agreements (PTAs) serve as one means of policy transparency within democracies; PTAs signal to voters that elected leaders are committed to reducing protectionist trade policy since violations thereof lead to highly publicized international disputes that inform voters. Goldstein (1986) and Thies and Porche (2007) suggest that democracy facilitates free trade policy by providing the appearance of responsiveness to narrow demands for protectionism without actually translating those demands into substantive protectionist policy. Conversely, Magee, Brock and Young (1989) note the problem of optimal obfuscation whereby politicians may continue to supply protection through technical forms of non-tariff barriers (NTBs), with effects on consumers that might be more difficult for voters to detect. Kono (2006) finds evidence that democracies do engage in this form of obfuscation by enacting lower tariffs in order to provide the illusion of liberalization, while actually maintaining higher NTBs, which are less transparent. Examining consumer-based explanations of trade policy preferences, Baker (2003) demonstrates that prices, which are much more visible to consumer-voters, are instrumental in determining citizens' views on liberalization in cases where factor endowment and retrospective voting models fail to do so. When protectionism leads to higher prices for commodities citizens purchase, they are more likely to demand liberalization. Baker uses this model to explain the puzzle of why citizens supported liberalization in Latin America despite the perception that it fostered higher levels of 6

unemployment and income inequality; lower prices and greater availability of goods drove the public's view. Generally, the movement from studies positing an unconditional relationship between democracy and trade protectionism to those suggesting limits and conditions resulting from variation in public demand for trade barriers is a sign of a maturing field of study. While the assumption that democracies respond to public concerns appears nearly universally supported, there is more disagreement regarding the proposition that publics always prefer liberal trade policy, or even that publics are aware of the potential for increased welfare due to liberalization. We seek to further our knowledge about the boundary conditions that affect the relationship between regime type and protectionism by exploring the commodity-level conditioning effect of import demand. A Commodity-level Determinant of Preferences for Protectionism: the Price Elasticity of Import Demand Given that democracies are shown to be more responsive than non-democracies to the preferences of mass publics, it is somewhat surprising that few scholars have looked at whether there is variation in the degree to which free trade benefits or harms citizens. Instead studies have focused on its impact on aggregate economic efficiency (e.g., Grossman and Helpman 1994), despite the fact that economic gains and losses could be highly concentrated in some cases. Although the availability of cheaper imports following from liberalization should always result in an increase in consumers' real income, its bottom-line impact will vary depending on how much consumers use imported commodities (Baker 2003). If domestic substitutes of an imported commodity are easily available, or if consumers can easily reduce usage of the commodity when prices rise, then protectionism leading to higher prices on imports of that commodity will harm 7

consumers relatively less. For example, sugar tariffs in the United States are not very harmful to citizens because sugar is not a necessity, and because cheap substitutes such as corn syrup are widely available. Conversely, if domestic substitutes are unavailable or very expensive, and if the commodity is vital to consumers' well-being, then protectionism leading to higher prices for such commodities will harm these consumers relatively more. For example, fuels tend to be important for consumers directly, as gasoline for personal transportation, as well as indirectly, given their use in industry. In the United States, an increase in trade barriers on fuels, even if designed with the aim of encouraging investment in alternate energy sources or exploitation of domestic reserves, would probably be broadly criticized by consumers seeing rising prices for many indispensable goods. 2 In other words, the degree to which consumers benefit from liberalization depends on the price elasticity of demand of the commodity in question, typically defined as the percentage change in quantity demanded given a 1% change in price. In the examples above, demand for sugar is elastic while that for (some) fuels is inelastic. Similarly, import demand elasticity refers to the percentage change in imports given a change in import price. Yet there is relatively little attention paid to the influence of import demand elasticity on the resulting trade barriers for a given commodity. The best known model of this relationship comes from Grossman and Helpman's (1994) protection for sale model, in which inelastic import demand produces higher tariffs, ceteris paribus, given that higher prices for goods facing inelastic demand will cause less trade distortion and lower deadweight costs relative to cases in which demand is elastic, while 2 This isn't to say that democratic governments would never levy taxes on fuels. However, these taxes usually are aimed at offsetting negative externalities associated with the use of fuels (e.g., infrastructure wear and pollution) rather than to protect domestic industry from competition. These taxes tend not to be imposed specifically on imports. 8

higher tariffs also provide the importing government with revenue. This argument has garnered some empirical support in studies of the United States (e.g., Goldberg and Maggi 1999; Gawande and Bandyopadhyay 2000). However, this expectation is grounded in the assumption that governments are concerned primarily with (1) aggregate market efficiency and (2) tariff revenue, weighing these benefits against campaign contributions. 3 Yet, when demand is inelastic, consumers have little recourse other than to pay the higher prices that would result from protectionism, the result of which is a decline in real income. Although market distortions from protectionism may be relatively lower when demand is inelastic, consumers of the protected commodity are nonetheless hurt more. 4 Furthermore, democracies are less reliant on trade taxes for their revenues and less likely to use tariffs to extract rents from their populations. As such, the expectation that democracies will enact higher trade barriers than non-democracies on commodities for which demand is inelastic does not account fully for what we know about regime type: that democracies are more responsive to mass publics. Import Demand Elasticity and the Effect of Democracy on Protectionism Beginning with the premise that democracies are more responsive than non-democracies to mass publics, we contend that a commodity s import elasticity influences the degree to which publics will suffer from higher trade barriers on that commodity. As a result, the value of liberalized trade as a public good varies with the structure of import demand; and democratic leaders provide liberalized trade policy accordingly. Furthermore, the elasticity of import demand also holds implications for the prevalence and preferences of domestic producers, which 3 Grossman and Helpman s model assumes a democratic form of government in which leaders balance interest group campaign financing against voter welfare. Non-democracies and mixed regimes are ignored. 4 For majorities to prefer protectionism in this case, governments would have to redistribute tariff revenue to compensate consumers paying higher prices. 9

in turn could affect the level of lobbying in favor of trade barriers. A higher trade barrier on some commodity with inelastic demand will impose higher costs on consumers, who cannot easily reduce consumption of the commodity. All else equal, this cost should lead citizens to demand reduced protectionism from elected leaders (Baker 2003). Although several studies contend that import competitors can nonetheless win protectionist policy because consumers face a collective action problem that precludes lobbying (e.g., Alt and Gilligan 1994), elected leaders will nonetheless be aware that consumers will realize a loss in real income stemming from higher trade barriers. Research shows that, in democracies, leaders' electoral fortunes depend on economic conditions (e.g., Powell and Whitten 1993). Accordingly, leaders should be responsive to incentives to provide the public good of liberalized trade policy on commodities in high demand, or risk facing an electorate perceiving declining real income if trade barriers are raised. While a government could, in theory, raise a tariff when demand is inelastic and then redistribute the revenue to offset consumers' losses, the analysis by Baker (2003) suggests that consumers would nonetheless prefer lower prices and the availability of a wider selection of imported commodities. Accordingly, we expect, all else equal, that democracies will provide lower trade barriers as a public good when demand is inelastic. This relationship can be illustrated by protectionism on fuels in the United States, which tends to be low; in fact, the median rate is 0. We contend that this lack of trade barriers follows at least in part from the fact that US import demand is very inelastic for fuels. 5 On the other hand, autocrats are unconcerned with providing public goods; instead, they 5 This pattern is most striking for propane, import demand for which is nearly perfectly inelastic, while trade barriers are equal to 0. Outside of fuels, rare earth metals are a salient US example in which inelastic demand corresponds to very low trade barriers. High trade barriers on rare earth metals would likely lead to higher prices for high-technology goods, on which consumers and US businesses depend. Understanding this fact, the US government is unlikely to pursue such protectionism. Indeed, China s announcement of export controls on rare earth metals (which similarly impose costs on US consumers) provoked considerable discontent from the US on behalf of its consumers and businesses. 10

are concerned with maximizing their own rents, which are used to maintain power through the provision of private goods to a narrower selectorate (e.g., Bueno de Mesquita et al. 2003). Trade barriers on commodities with inelastic import demand could therefore appeal to leaders in autocratic states as a source revenue with low aggregate welfare loss (Grossman and Helpman 1994). 6 Accordingly, we expect that non-democratic states will raise relatively higher trade barriers than democracies when import demand is inelastic. 7 The benefit to consumers of liberalized trade policy declines as import elasticity increases. A more elastic import demand for some commodity suggests that the cost to consumers associated with trade barriers will be lower because either (1) domestically produced substitutes exist or (2) consumers can adjust with less difficulty to lower consumption when prices rise. Although aggregate welfare loss due to protectionism increases as import demand becomes more elastic (Grossman and Helpman 1994), consumers might be less aware of this inefficiency because they are not obliged to incur a loss in real income, their sanguineness reflecting the availability of domestic substitutes, or a lack of reliance on the imported commodity. US trade barriers on sugar, as mentioned above, appear to match this expectation. In the US, sugar (which includes a range of sucrose products) is in the 95 th percentile of US commodities with the most elastic demand (Kee et al. 2008); simultaneously, sugar is one of the most protected US commodities, with ad valorem rates more than 9 times the US median, approaching the 99 th 6 Importantly, Grossman and Helpman's model examines tariffs, which provide direct revenue to governments. However, subsequent work has shown that governments can extract some (although typically not all) of the rents from non-tariff barriers (Facchini, Van Bieseboeck, and Willmann 2006). 7 This argument is not necessarily opposed to Grossman and Helpman's (1994) influential protection for sale model, in which inelastic demand leads to higher tariffs. Specifically, it could be that all states employ more protectionism on commodities with inelastic demand, but that the magnitude of this relationships is smaller for democracies. Importantly, we do not argue that interest groups will be ignored, or that democratic governments will adopt no protectionism when demand is inelastic, just that democracies should enact lower trade barriers than nondemocracies for specific commodities having inelastic demand because they wish to provide public goods to their citizens. 11

percentile for US protectionism. The Role of Domestic Producers There is a similar price effect for domestic producers using imported commodities as inputs. 8 Trade barriers should harm these firms more when demand is inelastic because this condition implies that these firms have few alternatives other than to pay higher prices. Conversely, when demand is elastic, firms can more easily adjust to protectionism. For example, in the United States, soft drink firms adapted to high sugar tariffs by switching to cheaper, domestically-produced corn syrup. Importantly, collective action problems regarding lobbying for liberalization, while varying, should be lower for domestic producers than for consumers. Accordingly, proposed trade barriers on commodities with inelastic demand will likely meet resistance from importers of that commodity. Furthermore, import elasticity could have implications for the existence of import competitors, which typically support protectionism. Inelastic import demand suggests a lack of domestic firms producing substitute goods, all else equal, and therefore implies that there are fewer interests to lobby in favor of protectionism. 9 Elastic import demand, conversely, could result specifically because import competitors do exist. In fact, it is possible for overall demand for a commodity to be inelastic while import demand is elastic specifically because domestic producers exist as an alternative to foreign firms. These import competitors may even be more willing to lobby for trade barriers in the presence of elastic demand because higher prices for 8 It is important to point out that we cannot distinguish empirically between the effects of consumer demand and the demand of importing firms. Ultimately, we expect that the same causal mechanisms to function for both actors. In the research design, we discuss our attempt to control for the existence of producers; however, we cannot obtain cross-national data on political organization at the commodity level employed in this study. 9 However, this same condition could increase the willingness of an industry to lobby, given that the gains would be distributed among a smaller number of firms. As such, there is more ambiguity in expectations regarding producers, whereas consumer expectations are more straightforward. 12

imports would lead consumers to switch to domestic substitutes. 10 Democracies typically have a greater number of access points through which these domestic groups can attempt to influence policy (Ehrlich 2007), which in turn suggests a greater likelihood that such lobbying would be successful. 11 Additionally, given that autocratic leaders tend to be responsive only to a narrow selectorate, an import competitor's ability to secure protectionism in an autocratic state could depend on its role in the regime. Given these complementary expectations for consumers and producers, we contend that leaders in democratic states will provide lower trade barriers than their non-democratic counterparts when import demand is inelastic, but that this effect will diminish as import demand becomes more elastic; in fact, democracies could raise relatively higher trade barriers on commodities with very elastic demand. 12 This argument leads to our main hypothesis. Hypothesis 1: Democracy is associated with lower levels of protectionism when import demand is inelastic. As demand becomes more elastic, this effect diminishes. Research Design To assess the relationship between democracy, import elasticity, and trade barriers, we utilize panel data of states and commodities, specifically, 4,656 traded commodities imported by 73 states. Many states import the same commodities; accordingly, these levels are non-nested. Our unit of analysis is the country-commodity, with commodities disaggregated to the Harmonized System 6-digit (HS6) level. All variables are coded for the year 2004 except where 10 Again, however, the potential for collective action problems complicates this relationship. 11 Ultimately, as Alt et al. (1996, 711) note, an interest group approach, whether it focuses on specific factors or on broad classes, does not account for why interests against trade are invariably dominant over interests in favor of expanding trade. All else equal, more mass preference for low trade barriers should suggest lower trade barriers, holding protectionist interests constant. Even Grossman and Helpman (1994) suggest that political leaders balance the need to obtain campaign financing with the need to represent citizens preferences. 12 Again, this argument does not necessarily contradict Grossman and Helpman (1994). It could be that there is a negative association between import elasticity and trade barriers for states, but that this relationship is weaker for democracies than for non-democracies. 13

noted below. 13 Although the use of data for only one year of data is limiting, it also reduces the possibility of omitted variable bias, which could follow from the repeated measurement of each unit. To test our hypothesis, we use regression models with fixed effects by commodity. 14 We use robust standard errors to account for possible heteroskedasticity; in alternate models, we also include region dummy variables to control for omitted variable bias; 15 Measuring Protectionism Unlike prior studies, which tend to examine determinants of individual trade barriers, our goal is to understand how democracy influences total protectionism at the commodity level. Towards this end, we use measures of protectionism developed by Kee, Nicita, and Olarreaga (2008, 2009), and available from the World Bank. Kee, Nicita, and Olarreaga derive a measure of non-tariff barriers (NTBs) by country and commodity in terms of their ad valorem equivalent (AVE) 16 such that this indicator can be combined with tariff rates in order to measure aggregate trade restriction. The authors use trade barrier data that are not time-varying, taken primarily from The United Nations Conference on Trade and Development (UNCTAD)'s Trade Analysis and Information System (TRAINS) database for the most recent year available between 2000 and 2004 (Kee, Nicita, and Olarrega 2009, 181). Our primary dependent variable is total protectionism, defined as the sum of ad valorem tariffs and the AVE of NTBs, from Kee et al. (2009). We also disaggregate this measure to look 13 Data limitations preclude the use of a time-series cross-section model. Specifically, the data on ad valorem equivalent of non-tariff barriers is available only for one year (Kee, Nicita, and Olarreaga 2009). 14 This is equivalent to the inclusion of commodity dummy variables; however, we use the xtreg command in Stata to avoid creating the more than 4,000 dummy variables required. Elasticities vary by country as well as by commodity, so we retain variation despite using fixed effects at the commodity level. 15 Our democracy variable is measured at the country-year level. Accordingly, because we have only one year of data, we cannot include country dummy variables. 16 Specifically, the authors estimate separate measures of the quantity impact of core NTBs and agricultural subsidies, and then transform these into price equivalents (Kee, Nicita, and Olarreaga 2009, 178). 14

directly at ad valorem tariff rates and AVE NTBs separately. All of these variables are tradeweighted averages at the Harmonized System 6-digit (HS6) level. Although trade weighting is employed, the HS6 level is considerably disaggregated, typically breaking down into few (at times only one) lower-level tariff lines. Our tariff measures are import-weighted average tariff rates for all tariff lines within each HS 6-digit level commodity. This variable captures the percentage of import value paid as a tariff. In theory, these rates may vary by trade partners (exporters) as well by commodity. However, we restrict the analysis to most favored nation (MFN) tariffs. 17 By granting a trade partner MFN status, a state agrees that it will offer it tariffs as low as those granted to any other trade partner. 18 Importantly, however, MFN rates are set by states themselves (although some rates are bound; i.e., the World Trade Organization sets a maximum tariff). As a result, this variable captures the extent to which a state allows others outside of its preferential trade partners, with which it may have political reasons for liberalizing, access to its market for a given commodity. 19 Our variable for the ad valorem equivalent of non-tariff barriers estimates the tariff on a given commodity that would produce the same level of trade restriction caused by an in-place NTB. This variable also comes from Kee, Nicita, and Olarreaga (2009), who in turn take raw data from TRAINS. NTBs include quantity restrictions (quotas) as well as quality standards. As with tariff rates, this variable is import-weighted. NTBs are used by scholars with increasing 17 For each importer, we look only at the World as a trade partner, rather than at dyadic relationships. 18 However, some states grant lower rates on some commodities to preferred partners. Variation arises from Preferential Trading Arrangements, in which states agree to reduced (or zero) trade barriers, and from the Generalized System of Preferences (GSP), which grants lower tariff rates to developing countries. Our results are unchanged when we create a trade partner-weighted applied tariff using bilateral trade and tariff data, also from Kee et al. (2009). The trade partner weighted tariff is, as expected, lower on average than the MFN tariff. Specifically, the trade partner weighted tariff has a mean of 5.4% and a standard deviation of 20.2%, whereas the MFN tariff has a mean of 8.6% and a standard deviation of 21.3%. 19 We do not examine dyadic protectionism because AVE NTBs are available only by state and commodity. 15

frequency in studies of protectionism, 20 given concerns that results obtained using average tariff rates are biased. Specifically, the use of tariff rates alone could be problematic because patterns of trade liberalization have resulted in increased usage of non-tariff barriers, particularly by democratic states (e.g., Kono 2006). Both tariff rates and AVE NTBs alone are limited given the potential for these barriers to be used as substitute trade policies (Kono 2006). 21 As such, our total protectionism variable sums the tariff rate and ad valorem equivalent of non-tariff barriers for each HS-6 level commodity, by state. Regardless of whether tariffs and NTBs are used as substitutes or complements (a phenomenon that could vary by commodity), this measure incorporates the true degree of protectionism states enact on a given commodity. A Note on Trade Weights Although our research design utilizes import-weighted average protectionism measures, it is in fact useful to mitigate the problems typically associated with the use of trade-weights. Typically, these weighted measures are problematic given that higher import barriers tend to result in lower imports, ceteris paribus. As such, higher trade barriers will be given smaller weighs than lower ones if they succeed in limiting imports more (e.g., Anderson and Neary 1995). At the extreme, a trade barrier that eliminates all imports would be weighted to 0. Tariffs, AVE NTBs, and elasticities are all components of the Trade Restrictiveness Index (TRI) (e.g., Anderson and Neary 2005; Kee, Nicita, and Olareaga 2008, 2009; Irwin 2007), which is 20 Importantly, however, most studies utilizing NTBs as a measure of protectionism specifically examine NTB coverage ratios, which are also imperfect measures of protectionism, given that they assume all tariff lines are equally important (Ehrlich 2007, 590). 21 But see Ehrlich (2007, 589) and Kee et al. (2009), who argue that tariffs and NTBs tend to be complements rather than substitutes. Ehrlich cites Ray (1981; see also Ray and Marvel 1984; Trefler 1993; and Lee and Swagel 1997), who demonstrate that tariffs and NTBs are correlated even within industries. Kee et al. (2009, 187) regress (logged) tariff rates on (logged) NTBs at the commodity level, finding strong evidence of a positive association. In our data, tariffs and NTBs correlate at 0.06, a very small positive association, yet significant at the 0.0001 level. 16

designed to overcome problems associated with using trade-weighted trade barriers as a measure of welfare loss. 22 However, by looking at the impact of elasticities on protection level, we also avoid this problem because we are testing whether cases in which trade barriers would lead to more or less welfare loss (based on the shape of import demand) are associated with higher or lower levels of protection. Importantly, however, the definition of welfare loss used here refers to reduced trade volumes and deadweight loss: i.e., market efficiency. We contend that the consequences of protectionism for the real income of mass publics and hence their toleration of protectionism should be examined along with market efficiency considerations. Primary Explanatory Variables Our primary independent variables capture the interactive effect of democracy and the price elasticity of import demand. First, we code democracy using the 21-point, combined democracy-autocracy score from the Polity IV project (Marshall and Jaggers 2010), 23 adjusted by adding 10 such that the most autocratic states are coded as 0. We interact each state's polity score with import demand elasticity. The elasticity variable captures the extent to which import demand for a given commodity and a given state adjusts down as price rises. As with our measures of protectionism, we take this variable from Kee, Nicita, and Olarreaga (2008, 2009s), who estimate demand elasticities for 142 countries. 24 The raw version of this variable ranges 22 Economists use elasticities to determine how much welfare would be gained if current protectionism levels were reduced to zero. With inelastic demand for imports, welfare gains from reduced trade barriers are small, as liberalization is unlikely to translate into higher volumes of imports. Conversely, when demand is elastic, lower tariff rates suggest considerably increased import flows, and higher welfare gains as a result. 23 Results were equivalent using Freedom House measures of democracy. However, we contend that Polity 21-point variable, coded to capture competitiveness of elections and constraints on the executive, is a more useful measure of the extent to which leaders will be responsive to public pressures relative to measures that primarily incorporate political freedoms. 24 Specifically, the authors construct over 377,000 HS6-level estimates using an instrumental variable technique to account for endogeneity and measurement error (Kee, Nicita, and Olarreaga 2008, 675). These estimates are assumed to vary by state and commodity, but not over time. 17

from -372 (the most elastic) to -0.000031 (the most inelastic). We adjust this scale, multiplying by -1 such that higher numbers represent more elastic demand. Furthermore, we take the natural log of this value (plus one) because the variable is skewed (with a mean equal to 3.12 and standard deviation equal to 14.05). 25 With this interactive specification, the coefficient for polity score represents its effect when import elasticity equals 0: i.e., when import demand for a given commodity is perfectly inelastic. Similarly, the coefficient for import demand elasticity represents its impact when the polity score equals 0: i.e., for the most autocratic states (see Braumoeller 2004 and Brambor et al. 2006 for discussions of interactive specifications). Other Explanatory Variables We control for economic and political determinants of the demand for protectionism. First, we capture intra-industry trade using Grubel and Lloyd's (1975) index. 26 Intra-industry trade is associated with exchange of similar, yet distinctive commodities (e.g., Hondas for Volkswagens), whereas inter-industry trade is associated with exchange of different commodities (e.g., oil for wheat). Conventional wisdom suggests that intra-industry trade is associated with lower levels of protectionism because lessened distributional consequences associated with reducing trade barriers render liberalization politically more feasible (e.g., Balassa 1966; but see Kono 2009, who shows that intra-industry trade maybe associated with higher levels of protectionism when electoral institutions reward narrow interests). We use 2004 trade data by 25 The supplemental appendix presents descriptive statistics for all variables, as well as histograms for the raw and logged version of import elasticity, demonstrating the utility of the logged measure. 26 The intra industry trade index for a given country and commodity is equal to 1 minus the quantity: the absolute difference in imports vs. exports, divided by the sum of imports and exports. The value ranges between 0 (where a country engages in no intra-industry trade of this commodity) to 1 (where all trade of this commodity is intraindustry trade). 18

country and commodity from the United Nations (UN) Comtrade database to create this measure. We also control for import penetration and export dependence, coded as the value of imports and exports, respectively, of a given commodity for each importer divided by the importer's GDP. Ultimately, interpretation of these variables is difficult given the potential for endogeneity, particularly with regard to import penetration (e.g., Trefler 1993). For example, it may be that imports are higher because demand for protectionism is lower, or that higher trade barriers reduce the volume of imports. However, controlling for import demand elasticity may increase the explanatory power of these measures of trade interaction. To code this measure, we use 2004 trade data from the UN Comtrade and 2004 GDP data from the Penn World Table (Heston, Summer, and Aten 2012). To capture political determinants of protectionism (in addition to democracy), we include dummy variables for the count of allies exporting a given commodity, and the count of rivals exporting the commodity. Research shows that states tend to trade more with allies and less with rivals, specifically shaping trade patterns in order to obtain positive security externalities (Gowa 1994; Gowa and Mansfield 1993). Similarly, a large literature suggests that conflict expectations result in lower levels of trade (e.g., Long 2008); they could also invoke higher trade barriers. Accordingly, we take alliance data from ATOP (Leeds et al. 2002) and strategic rivalry data from Thompson (2001), using the latest available year from each dataset (2003 for alliances and 2001 for rivalries). We also control for factors associated with the importer s economy that may influence trade barriers. There is potential for the influence we attribute to democracy to result instead from the fact that many democracies are also developing states, which have been forced to liberalize by outside forces. For example, a large number of developing states have undergone 19

trade liberalization in accordance with development plans promoted by international institutions. Indeed, some are subject to liberalization as part of the structural adjustment conditions imposed by IMF loans (e.g., Kahler 1995). To isolate the impact of democracy from this developing country effect, we code developing country as a dummy variable equal to 1 if the International Monetary Fund (IMF) lists the country as emerging or developing in their World Economic Outlook survey. 27 Finally, we include a dummy variable indicating whether the state is a WTO member. We take this variable from the Correlates of War International Organizations data (Pevehouse, Nordstrom, and Warnke 2004), using the value for 2004. Analysis We find strong support for our hypothesis that the price elasticity of import demand conditions the influence of democracy on protectionism. Specifically, we find that democracy is associated with lower trade barriers on commodities for which demand is inelastic, but that this effect diminishes as demand becomes more elastic. In fact, when demand is very elastic, we find that the influence of democracy switches, becoming associated with relatively higher levels of protectionism. However, this effect occurs only at high levels of demand elasticity. Furthermore, an interpretation of the marginal effect of demand elasticity demonstrates that elastic demand is associated with lower overall protectionism regardless of regime type; the positive conditional impact of democracy suggests that this negative effect of elastic demand is smaller for more democratic states. Table 1 presents coefficients and standard errors for six models assessing the relationship between democracy and protectionism levels, conditional on the shape of import demand. We 27 Primary results are substantively equivalent in models including a dummy variable specifically for recipients of IMF conditionality. Additionally, primary results are unchanged when we substitute the developing state variable with (logged) GDP per capita, as well as when we include both a developing state dummy variable and a continuous indicator of (logged) GDP per capita. 20

present both models excluding and models including region dummy variables intended to mitigate state-level heterogeneity that might bias our results. In Models 1 and 2, the dependent variable is total protectionism the sum of tariff rate and AVE NTB. We disaggregate this variable in Models 3 through 6. Models 3 and 4 assess the conditional impact of democracy on ad valorem equivalent (AVE) of non-tariff barriers (NTBs), while Models 5 and 6 examine tariff rates. In all six models, the coefficient for democracy is negative and significant at the 0.001 level, suggesting that democracies have lower levels of protectionism than autocracies under the condition that demand is at its most inelastic (i.e., demand elasticity is equal to 0). The coefficient for import demand elasticity is negative and significant at the 0.001 level in Models 1 and 2, examining total protectionism level. As such, we find that elastic demand is associated with lower trade barriers in the most autocratic states (when democracy is equal to 0). This result partially confirms the expectations of Grossman and Helpman (1994). 28 In the disaggregated models, we find that the coefficient for import demand elasticity is negative significant in models examining AVE NTBS (p<0.001 in Models 3 and 4), but is not significant when looking only at tariffs. The interaction term is positive and significant in all six models, suggesting that as demand becomes more elastic, the negative association between democracy and protectionism diminishes towards zero and, potentially, that democracy becomes associated with higher levels of protectionism. 29 [Table 1 about here] Coefficients convey only a small part of an interaction effect. Figure 1 displays the marginal effect of democracy over the entire range of demand elasticity for each of our three 28 However, Grossman and Helpman do not consider the possibility that the impact of import demand elasticity is conditional on democracy. 29 The positive interaction term also indicates that the negative association between import demand elasticity and protectionism decreases as democracy increases. We discuss this conditional effect below. 21

dependent variables. 30 The left-hand column presents the conditional marginal effect of a state's polity score on total protectionism i.e., the sum of tariffs and AVE NTBs (from Model 2). The center column presents the conditional marginal effect of democracy on AVE NTBs only (from Model 4); and the right-hand column shows this effect with respect to tariff rates (from Model 6). This figure shows that, consistent with the interpretation of coefficients, the effect of democracy is negative when elasticity equals zero, 31 but that this effect diminishes toward zero as demand becomes more elastic. In fact, for all three dependent variables, democracy becomes associated with higher levels of protectionism when demand is relatively elastic (approximately equal to its mean plus one standard deviation); however the positive marginal effect for democracy is not significant with respect to tariffs. This striking result may signify that the democratic process actually facilitates relatively higher protectionism for commodities for which demand is very elastic, possibly because domestic substitutes are easily available, leading consumers not to notice higher prices resulting from protectionism that benefits import competitors. However, it is important to note that this effect occurs only at extreme levels of elastic demand. Approximately 11 percent of imported commodities in our sample fall into the range in which the marginal effect of democracy is positive and significant. Furthermore, this results does not necessarily imply that elastic demand is associated with higher levels of protectionism in democracies, a relationship that would be counter to the expectations of Grossman and Helpman's (1994) model. Instead, the positive conditional coefficient for democracy could indicate only that the negative impact of elastic demand is smaller in magnitude for more democratic states. 30 We use the margins command in Stata 12 to obtain conditional coefficients. 31 The value of the graph is equal to the coefficient for polity score when import elasticity is equal to 0 (from Models 2, 4, and 6). 22

[Figure 1 about here] The substantive impact of democracy changes considerably over the range of import demand elasticity. To illustrate, we examine predicted levels of total protectionism in states with polity scores in the 25 th and 75 th percentiles, over a mean +/- two standard deviation range of import elasticity, with all other explanatory variables held at their means (from Model 2). When import demand for a given commodity is perfectly inelastic (i.e., equal to 0), overall protectionism levels will be equivalent to a 22% ad valorem for a state with a polity score at the 25 th percentile (i.e., polity equal to 13, which is equal to the value for Nepal), but only 20% ad valorem for a state with polity score at the 75 th percentile (i.e., polity equal to 20, the maximum value). All states' protectionism levels fall as import demand becomes more elastic (as expected by Grossman and Helpman's model); however, this effect is more dramatic for non-democracies. When import demand is equal to its mean plus two standard deviations, a state with a polity score at the 25 th percentile will have protectionism equal to 13% ad valorem. Conversely, a state with polity score at the 75 th percentile will have protectionism levels relatively higher than the less democratic state, equal to 17% ad valorem. 32 Our analysis focuses primarily on the changing influence of democracy on protectionism. However, interactive specifications model two mutually conditioning variables. Given that Grossman and Helpman's (1994) model suggests that inelastic demand is associated with higher protectionism, and in accordance with the suggestion in recent studies to interpret fully interactive specifications (Berry, Golder, and Milton's 2010), we also examine the changing marginal effect of demand elasticity over the range of polity scores. As examined above, the coefficient for import demand elasticity is significant in all models except Models 5 and 6. As 32 The supplemental appendix presents a figure illustrating the difference in these point predictions over a mean +/- two standard deviations of import elasticity. 23

such, we find some evidence that elastic demand is associated with lower protectionism in the most autocratic states (i.e., when polity score equals 0). However, the positive interaction term suggests that, as a state becomes more democratic, the association between elastic demand and lower protectionism diminishes. 33 For total protectionism, and NTBs specifically, the marginal effect of elasticity remains negative regardless of regime type; however the magnitude of the marginal effect is smaller for more democratic states. This is not to imply that non-democracies necessarily care more about aggregate welfare or economic efficiency than democracies. Rather, we suspect that import competitors are relatively more able to lobby for protectionism in democracies because (1) democracies typically have more interest group access points (Ehrlich 2007), and (2) import competitors in non-democracies may be less able to achieve protectionism if they do not comprise the leaders' selectorate (Bueno de Mesquita et al. 2003). Nonetheless, this striking pattern is worth future study. Surprisingly, we find that the conditional marginal effect for import elasticity is always positive with respect to tariffs regardless of regime type (in Models 5 and 6), although it is not significant for less democratic states. We suspect that this somewhat counter-intuitive result stems from the fact that many states have transitioned from using tariffs towards using NTBs in recent years (e.g., Kono 2006). In fact, this odd result for tariffs specifically serves as an example of why it is critical to examine total levels of protectionism rather than individual trade barriers, given their possible substitutability. Furthermore, this counter-intuitive result is not necessarily contradictory with the results of prior empirical analyses of Grossman and Helpman's (1994) model, because these prior studies tend to examine NTB coverage ratios rather than tariffs (e.g., Goldberg and Maggi 1999; Gawande and Bandyopadhyay 2000). 33 The supplemental appendix presents a figure illustrating the marginal effect of elasticity conditional on polity score. 24

Control variables generally behave as expected. The coefficient for imports/gdp is negative and significant only in Model 1, providing little evidence that the level of imports is associated with protectionism levels. However, as mentioned above, the result for imports is difficult to interpret given the likelihood of endogeneity between the tariff rate and import penetration. The coefficient for exports/gdp is not significant in any model. We find some evidence that a higher proportion of intra-industry trade for a given commodity is associated with less total protectionism and lower tariffs specifically, but higher NTBs. This result suggests some evidence in support of the conventional wisdom that intra-industry trade facilitates liberal trade policy; 34 however, the coefficient is significant only in 3 of 6 models. We find that developing states tend to have higher tariffs and lower NTBs but lower protectionism overall than developed states. As expected, WTO members tend to have lower overall protectionism levels than non-members. Interestingly, this relationship also extends to NTBs, to which some states could turn when the WTO binds tariffs. Finally, the effects of strategic determinants of protectionism follow expectations. In every model, rivals exporting a given commodity are associated with higher protectionism, while more allies exporting the commodity are associated with lower protectionism. 35 Discussion and Conclusion In this paper, we examine how the shape of import demand for a given commodity conditions the influence of democracy on trade barriers. We find that democracies are more likely to have lower trade barriers than autocracies when demand for a given commodity is inelastic. However, as import demand becomes more elastic, this effect diminishes. When import 34 However, the substantive effects of this variable appear very small. 35 The substantive effects of allies and rivals exporting a given commodity can be quite large as the number of such states increases. 25

demand for a given commodity is very elastic, democracies actually tend to employ higher levels of protectionism than autocracies, particularly with respect to non-tariff barriers. Our results support those of previous studies suggesting that democracy is associated with higher responsiveness to the preferences of mass publics. However, because consumervoters benefit most from liberalized trade when import demand is inelastic, the liberal trade policy premium in democracies is most dramatic in this circumstance. Importantly, we uncover conditions in which more democratic states employ relatively higher levels of protectionism than less democratic states. Our empirical results suggest that mass publics could be largely indifferent to the protectionist behavior of import competitors, leading policy-makers not to fear a public backlash against higher trade barriers on commodities for which import demand is very elastic. Future research can benefit from looking at import demand elasticity and protectionism with regard to the differences within democracies. Democratic institutions take many forms, and by extending the models we introduce here to include a number of political institutions, we may better understand the impact of political institutions on trade policy. For example, inelastic demand may be associated with higher levels of protectionism in the specific case that electoral institutions reward narrow interests because consumers could not easily reduce consumption, leading import competitors to gain considerably from higher prices. Additionally, variation within democracies (as well as between democracies and non-democracies) could be better understood if future scholars collect cross-national data on the political organization of producer groups. Extending our analysis to examine dyadic protectionism levels may also improve our understanding of the relationship between democracy and trade barriers. It could be that 26

commodities with elastic demand tend to be exported by non-democracies. If democracies do not lower trade barriers universally, instead giving preference to other democracies, then the effect we find in this paper could be an artifact of the correlation between regime type and commodity elasticity. 36 Additionally, this paper has important implications for studies of coercion and conflict. Previous research finds that inelastic import demand suggests potential for states to become vulnerable to trade partners (Crescenzi 2003; see also Hirschman 1945). As such, our finding that democracies employ lower trade barriers when demand is inelastic could suggest that democracies are more willing to tolerate dependence on trade when potential exit costs for that trade which could be leveraged by the trade partner are high. Trade between democracies and autocracies could generate vulnerability that sparks political tension and, perhaps, militarized conflict. Simultaneously, however, the tendency of democracies to employ less protectionist trade policy when potential security considerations are high may facilitate mutual dependence between jointly democratic trade partners, which could foster enduring peace between them (e.g., Peterson Forthcoming). 36 A first glance at trade data does not uncover a significant difference in average elasticity of exports by regime type. However, more rigorous analysis is needed. 27

References Alt, James E., Jeffrey Frieden, Michael J. Gilligan, Dani Rodrik, and Ronald Rogowski. (1996) The Political Economy of International Trade: Enduring Puzzles and an Agenda for Inquiry. Comparative Political Studies 29:689 717. Alt, James, and Michael Gilligan. (1994) The Political Economy of Trading States: Factor Specificity, Collective Action Problems and Domestic Political Institutions. Journal of Political Philosophy 2: 165-192. Anderson, James E., and J. Peter Neary. (2005) Measuring the Restrictiveness of International Trade Policy. Cambridge: The MIT Press. Baker, Andy. (2003) Why is Trade Reform so Popular in Latin America? A Consumption- Based Theory of Trade Policy Preferences. World Politics 55: 423-455. Balassa, Bela. (1966) Tariff Reductions and Trade in Manufactures among the Industrial Countries. American Economic Review 56 (3): 466 473. Berry, WilliamD., Matt Golder, and Daniel Milton. (2010) The Importance of Fully Testing Conditional Theories Positing Interaction. Unpublished manuscript. Bliss, Harry, and Bruce Russett. (1998) Democratic Trading Partners: The Liberal Connection, 1962-1989. Journal of Politics 60(4): 1126-47. Brambor, Thomas, William Roberts Clark, and Matt Golder. (2006) Understanding Interaction Models: Improving Empirical Analysis. Political Analysis 14: 63-82. Braumoeller, Bear. (2004). Hypothesis Testing and Multiplicative Interaction Terms. International Organization 58: 807-820. Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. (2003) The logic of political survival. Cambridge, MA: MIT Press. Crescenzi, Mark J. C. (2003) Economic Exit, Interdependence, and Conflict. Journal of Politics 65(3): 809-832. Ehrlich, Sean D. (2007) Access to Protection: Domestic Institutions and Trade Policy in Democracies. International Organization 61: 571-605. Facchini, Giovanni, Johannes Van Biesebroeck, and Gerald Willmann. (2006) Protection for sale with imperfect rent capturing. Canadian Journal of Economics 39 (3): 845-873. 28

Gawande, Kishore, and Usree Bandyopadhyay (2000) "Is protection for sale? Evidence on the Grossman-Helpman theory of endogenous protection." Review of Economics and Statistics 82, 139 52. Goldberg, Pinelopi Koujianou, and Giovanni Maggi (1999) "Protection for sale: an em- pirical investigation." American Economic Review 89, 1135 55. Goldstein, Judith. (1986) The Political Economy of Trade: Institutions of Protection. American Political Science Review 80(1): 161-184. Gowa, Joanne. (1994) Allies, Adversaries, and International Trade. Princeton: Princeton University Press. Gowa, Joanne, and Edward D. Mansfield. (1993) Power Politics and International Trade. American Political Science Review 87: 408-420. Grossman, Gene M., and Elhanan Helpman. (1994) Protection for Sale. The American Economic Review 84: 833-850. Grubel, H.G., andp. J. Lloyd. (I975) Intra-Industry Trade. NewYork: JohnWiley. Hermann, Richard K., Philip E. Tetlock, and Matthew N. Diascro. (2001) How Americans Think about Trade: Reconciling Conflicts Among Money Power, and Principles. International Studies Quarterly 45: 191-218. Hirschman, Albert O. (1945) National Power and the Structure of Foreign Trade. Berkeley: University of California Press. Heston, Alan, Robert Summers, and Battina Aten. (2012) Penn World Table version 7.1. Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. http://pwt.econ.upenn.edu/ Irwin, Douglas A. (2007) Trade Restrictiveness and Deadweight from US Tariffs, 1859-1961. NBER Working Paper 13450. Kahler, Miles (1995) International Institutions and the Political Economy of Integration. Washington: Brookings Institution Kee, Hiau Looi, Alessandro Nicita, and Marcelo Olarreaga. (2008) Import Demand Elasticities and Trade Distortions. The Review of Economics and Statistics 90(4): 666 682. Kee, Hiau Looi, Alessandro Nicita, and Marcelo Olarreaga. (2009) Estimating Trade Restrictiveness Indices. The Economic Journal 119(1): 172 199. Kono, Daniel Y. (2006) Optimal Obfuscation: Democracy and Trade Policy Transparency. The 29

American Political Science Review 100(3): 369-384. Kono, Daniel Y. (2009) Market Structure, Electoral Institutions, and Trade Policy. International Studies Quarterly 53: 885-906. Lee, Jong-Wha, and Phillip Swagel. (1997) Trade Barriers and Trade Flows Across Countries and Industries. Review of Economics and Statistics 79 (3): 372 382. Leeds, Brett Ashley; Jeffrey M. Ritter, Sara McLaughlin Mitchell, and Andrew G. Long. (2002) "Alliance Treaty Obligations and Pro- visions, 1815 1944." International Interactions 28 (3): 261 284. Magee, Stephen P., William A. Brock, and Leslie Young. (1989) Black Hole Tariffs and Endogenous Policy Theory: Political Economy in General Equilibrium. New York: Cambridge University Press. Mansfield, Edward D., Helen V. Milner, and Peter Rosendorff. (2000) Free to Trade: Democracies, Autocracies, and International Trade. American Political Science Review 94 (2): 305-321. Mansfield, Edward D., Helen V. Milner, and Peter Rosendorff. (2002) Why Democracies Cooperate More: Electoral Control and International Trade Agreements. International Organization 56: 477-514. Marshall, Monty G. and Keith Jaggers. (2010) Polity IV Project: Political Regime Characteristics and Transitions, 1800-2009. Milner, Helen V., and Keiko Kubota. (2005) Why the Move to Free Trade? Democracy and Trade Policy in the Developing Countries. International Organization 59: 107 44. Olson, Mancur, Jr. (1993) Dictatorship, Democracy, and Development. American Political Science Review 87(3): 567-576. O Rourke, Keven H., and Alan M. Taylor. (2005) Democracy and protectionism in the nineteenth century. Unpublished manuscript, Trinity College Dublin and University of California, Davis (September). Peterson, Timothy M. (forthcoming). Dyadic Trade, Exist Costs, and Conflict. Journal of Conflict Resolution. Pevehouse, Jon C., Timothy Nordstrom, and Kevin Warnke. 2004. "The COW-2 International Organizations Dataset Version 2.0," Conflict Management and Peace Science 21(2):101-119. Powell, G. Bingham, and Guy D. Whitten. (1993) A Cross-national Analysis of Economic 30

Voting: Taking Account of the Political Context. American Journal of Political Science 37(2): 391-414. Przeworski, Adam. (1990) The State and the Economy under Capitalism. Chur, Switzerland: Harwood Academic Publishers. Ray, Edward J. (1981) The Determinants of Tariffs and Nontariff Trade Restrictions in the U.S. Journal of Political Economy 89 (February): 105 21.Ricardo, D. [1817] 1981. The Principles of Political Economy and Taxation. Cambridge: Cambridge University Press. Ray, Edward J., and Howard P. Marvel. (1984) The Pattern of Protectionism in the Industrialized World. Review of Economics and Statistics 66 (3): 452-458. Rogowski, Ronald. (1987) Political Cleavages and Changing Exposures to Trade. American Political Science Review 81: 1121 1137. Rogowski, Ronald. (1989) Commerce and Coalitions: How Trade Affects Domestic Political Alignments. Princeton: Princeton University Press. Ruggie, John Gerard. (1982) International Regimes, Transactions, and Change: Embedded Liberlism in the Postwar Economic Order. International Organization 36 (2): 379-415 Stigler, George. (1971) The Theory of Economic Regulation. The Bell Journal of Economics and Management Science 2: 3-21. Thies, Cameron G., and Schuyler Porche (2007) Crawfish Tails: A Curious Tale of Foreign Trade Policy Making. Foreign Policy Analysis 3: 171 187. Thompson, William R. (2001) Identifying Rivals and Rivalries in World Politics. International Studies Quarterly 45: 557-586. Trefler, Daniel. (1993) Trade Liberalization and the Theory of Endogenous Protection: An Econometric Study of U.S. Import Policy. Journal of Political Economy 101: 138-160. 31

Table 1: Fixed effects (by commodity) regression coefficients and robust standard errors for regime type, import demand elasticity, and import barriers. Total protectionism AVE NTBs Tariffs 1 2 3 4 5 6 Polity score -0.006*** -0.003*** -0.005*** -0.002*** -0.001*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) ln Import elasticity -0.092*** -0.084*** -0.088*** -0.084*** -0.004 0.000 (0.006) (0.006) (0.004) (0.004) (0.004) (0.004) Policy X Elasticity 0.004*** 0.004*** 0.003*** 0.003*** 0.001*** 0.001** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Import penetration -2.958* -2.436-1.726-1.549-1.233-0.887 (1.459) (1.307) (0.955) (0.901) (0.673) (0.588) Export dependence 0.016-0.003-0.104-0.119 0.120 0.116 (0.266) (0.262) (0.247) (0.246) (0.167) (0.168) Intra-industry trade -0.010** 0.004-0.000 0.005* -0.010*** -0.001 (0.003) (0.003) (0.003) (0.003) (0.002) (0.002) Rivals exporting 0.059*** 0.054*** 0.041*** 0.034*** 0.018*** 0.020*** (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) Allies exporting -0.003*** -0.004*** -0.003*** -0.003*** -0.001*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Less developed state -0.029*** -0.081*** -0.054*** -0.086*** 0.025*** 0.005*** (0.003) (0.004) (0.003) (0.004) (0.001) (0.001) WTO member -0.087*** -0.148*** -0.080*** -0.124*** -0.007*** -0.024*** (0.004) (0.004) (0.004) (0.004) (0.001) (0.001) Europe -0.023*** -0.024*** 0.000 (0.003) (0.002) (0.002) Middle East 0.058*** 0.055*** 0.003 (0.004) (0.003) (0.002) Africa 0.038*** 0.022*** 0.017*** (0.004) (0.003) (0.002) Asia -0.054*** -0.010** -0.044*** (0.004) (0.004) (0.002) South America 0.025*** 0.008* 0.017*** (0.004) (0.003) (0.002) Oceania -0.194*** -0.133*** -0.061*** (0.007) (0.006) (0.003) Constant 0.477*** 0.542*** 0.387*** 0.417*** 0.089*** 0.125*** (0.007) (0.008) (0.006) (0.007) (0.003) (0.004) Observations 202,077 202,077 202,077 202,077 202,077 202,077 R 2 0.039 0.050 0.030 0.037 0.022 0.030 HS6 groups 4,656 4,656 4,656 4,656 4,656 4,656 *** p 0.001, ** p 0.01, * p 0.05; two-tailed tests 32

Figure 1: The marginal effect of polity score by import elasticity, with 95% confidence bounds. 33