INTRODUCTION

The past 20 years have been a period of intense transformation in the global governance of trade. Many aspects of trade policy have changed, but two major transformations stand out. First, the World Trade Organization has been experiencing a prolonged deadlock, and not much has happened in terms of multilateral trade liberalization since the conclusion of the Uruguay Round in 1994. Second, preferential trade agreements (PTAs) have dramatically proliferated since the mid-1990s and have become one of the trademarks of globalization. According to the Desta dataset (Dür et al. 2014), there were a little more than 100 PTAs in the 1990s, whereas there are more than 700 PTAs in force to date. Both developed and developing countries are heavily involved in preferential liberalization, and the number of North–South PTAs (i.e., PTAs between developed and developing countries) has boomed since the formation of the North America Free Trade Agreement (NAFTA). In summary, much of the trade liberalization that we have seen in the past 20 years is preferential rather than unilateral or multilateral.

While impressive, the growing number of PTAs is not the most defining transformation in the global governance of trade. Rather, the most important change is that modern PTAs not only reduce tariffs but also regulate investment, intellectual property rights, competition policy, government procurement, and many other matters. In other words, PTAs remove barriers not only at the border but also behind the border, producing what has been referred to as deep integration between countries (Lawrence 1996). An illustration of this change is the contrast between the PTA that the European Union signed with Egypt in 1972, which is 92 pages long, and the 2016 Comprehensive Economic and Trade Agreement between Canada and the European Union, which is 1,598 pages long. Since many of the provisions and regulations included in PTAs go beyond World Trade Organization commitments (Horn et al. 2010), it is fair to say that preferential liberalization shapes the global governance of trade in the twenty-first century.

Given the changing nature of preferential liberalization, scholars have embarked on important data collections on the design of PTAs in both economics and political science. The most extensive data set on the design of PTAs is Desta (Dür et al. 2014), which includes the entire universe of PTAs. The two main dimensions measured by Desta are depth (i.e., the strength of commitments) and flexibility (i.e., the degree to which countries can escape their commitments). Other data sets include a more limited sample of PTAs (Estevadeordal et al. 2009, Horn et al. 2010, Orefice & Rocha 2014). Some scholars have focused on the data in specific issue areas such as enforcement provisions (Allee & Elsig 2016) and nontrade issues (Lechner 2016, Morin et al. 2018). Armed with these novel data, scholars have produced a plethora of studies on PTAs in the past ten years.

This review surveys the academic literature that has attempted to explain the formation of PTAs, their design, and their consequences. Empirical and theoretical research on PTAs has evolved in symbiotic tandem, though the empirical studies appear to be predominant, especially in political science. Regarding the formation of PTAs, I document that studies have gradually moved from exploring the macro-foundation of preferential liberalization to focusing on its micro-foundation, a trend in line with other areas of international relations. Novel fine-grained data on the design of PTAs, on industries and firms, and on lobbying activities have helped this transition to micro-level analysis. The take-home message from this stream of research is that PTAs serve the interests of large, productive firms involved in offshoring activities (Chase 2005, Manger 2009, Blanchard & Matschke 2015, Baccini et al. 2017). Such firms are the key actors behind the proliferation of PTAs. In this regard, preferential liberalization moves hand in hand with the growing importance of foreign direct investment (FDI) and global value chains (GVCs).

With respect to the consequences of preferential liberalization, there is a shortage of rigorous studies regarding the empirical evidence on the welfare effect of PTAs, which is not limited to the volume of trade but also entails terms of trade and real wages. To date, the most sophisticated study is by Caliendo & Parro (2015), who find a small positive effect of NAFTA on Mexico's and the United States' welfare and a small negative effect on Canada's welfare. However, there is overwhelming evidence that PTAs substantively increase trade flows and FDI, especially deep PTAs (Büthe & Milner 2014, Dür et al. 2014, Baccini et al. 2017). Moreover, the previous literature shows that PTAs help lock in reforms in developing countries (Baccini & Urpelainen 2014a). Insofar as these reforms are welfare enhancing, these studies unveil further gains from preferential liberalization. The political benefits from preferential liberalization are more controversial. In particular, there is conflicting evidence that PTAs help countries to democratize, improve human rights compliance, and reduce conflicts among countries. Much of this literature provides analysis at the macro level with limited evidence of the mechanisms at play.

While covering some papers in economics, this survey engages mostly with the literature in international relations and comparative politics. PTAs are one of the many forms of international cooperation, and they face the same issues that other international institutions do. Building on seminal works in the legalization (Abbott et al. 2000) and rational design literatures (Koremenos et al. 2001), many international relations studies have explored the legal content of international institutions in areas other than trade (Allee & Peinhardt 2014, Tallberg et al. 2014). Moreover, much of the work reviewed here demonstrates that trade cooperation is highly intertwined with domestic politics, as is cooperation in other issue areas, such as human rights (Vreeland 2008, Simmons 2009), capital (Arias et al. 2018), and the environment (Tingley & Tomz 2014).

This review suggests directions toward which to push the research on PTAs. I offer some stylized facts supporting the argument that more research is required in some issue areas. Concretely, future studies would benefit from (a) paying greater attention to interdependent components of the design of PTAs, (b) exploring lobbying activities related to PTAs beyond the United States and merchandise, and (c) focusing on the distributional consequences of preferential trade liberalization. The last point is particularly relevant because it would allow scholars working on PTAs to engage with recent debates pointing to trade shocks as a key determinant of the backlash against globalization (Colantone & Stanig 2018a,b).

Methodologically, future research should continue delving into the micro-foundation of preferential trade liberalization, making use of new data at the firm level (now available for many countries), and exploiting variation at the subnational level to explore untested political mechanisms. Empirical studies should pay particular attention to the identification strategy in order to go beyond simple correlations, which characterize much of the previous literature. In the absence of a rigorous research design, scholars will struggle to understand the relationship between the economics and politics of PTAs and to provide valuable policy recommendations.

ECONOMIC AND POLITICAL DETERMINANTS OF PREFERENTIAL TRADE LIBERALIZATION

In order to understand the proliferation of PTAs, the first question to ask Is, “Why do countries engage in preferential liberalization?” To make sense of the vast body of studies that has attempted to answer this question, I divide it into two groups: studies that address the macro-foundation of PTA formation and studies that address the micro-foundation of PTA formation. Generally speaking, these studies build on the insights of theoretical models showing that PTAs are welfare improving either because they solve terms-of-trade externalities (Bagwell & Staiger 1999) or because they allow governments to commit vis-à-vis domestic demand for protection (Maggi & Rodriguez-Claire 2007).

The Macro-Foundation of Preferential Trade Agreement Formation

In the economic literature, the seminal paper explaining the formation of PTAs is by Baier & Bergstrand (2004). Their analysis focuses on the utility gains from a PTA between pairs of countries and identifies three factors as key economic determinants of trade agreements. First, countries are more likely to form PTAs if their transportation costs are low. This insight comes directly from the theory on natural trading partners developed by Krugman (1991). Second, larger economies are more likely to form PTAs. A PTA between two large countries increases the volume of trade in more ways than a PTA between two small partners. Moreover, PTAs between large economies lead to an expansion of demand and so a larger rise in real income than PTAs between smaller countries. Third, the more similar the economic sizes of the countries (controlling for GDP), the greater their utility gains if they form a PTA.

There is robust empirical evidence that economic determinants explain a large number of PTAs. Indeed, variables like distance, GDP, and similarity of economic size are not only significantly correlated with the formation of PTAs, but they also correctly predict more than 80% of the PTAs that are currently in force. If so, what is left to explain? The answer is simple: politics.

In a widely cited book and an article coauthored with Rosendorff (Mansfield et al. 2002, Mansfield & Milner 2012), Mansfield and Milner show that democratic countries are more likely to form PTAs than autocratic countries. The theory boils down to a signaling game. Democratic leaders are able to remain in power if and only if they improve voters' welfare. Changes in voters' welfare are a function of both trade policies implemented by leaders and exogenous systemic shocks over which leaders have no control. There are two types of leaders: those who implement policies to improve voters' welfare (the good type) and those who implement rent-seeking policies to increase their own wealth or the wealth of their cronies (the bad type). Voters are unable to observe a leader's true type, which generates a problem for good leaders. If a negative economic shock hits a country, voters' welfare plummets, and any leader (good or bad) will lose her office in the subsequent election, even if she is not responsible for the negative economic performance of her country.

PTAs may come to the rescue of this good but unlucky leader. How so? Signing a PTA signals to voters that the leader is truly committed to implementing policies that improve their welfare. Indeed, not only do trade agreements reduce tariffs (and thus prices of goods), but PTAs also generate an international commitment on which leaders cannot renege. In short, a PTA signals that a leader is rejecting rent-seeking behavior. After having observed this signal, voters are more likely to keep leaders in office even in the presence of negative economic shocks for which leaders are not responsible. Incentives to form PTAs to unveil the true type are present only in democratic regimes, in which leaders face regular and fair elections. Autocratic leaders do not have the same incentives, since, absent elections, their tenure does not depend on voters' welfare.

The empirical evidence supporting this argument is convincing. The correlation between democratic pairs and the probability of forming PTAs is positive and significant across different data sets and model specifications. A key feature of the original game theoretic model is that voters are more open to trade than their government is (Mansfield et al. 2002). This appears at odds with what we have been recently observing in developed countries, in which growing demands for protectionist policies are coming from voters. At the very least, the micro-foundations of voters' preferences remain largely unexplored in Mansfield and Milner's work, calling for more research at the individual level, a point to which I return below.

The point here is that PTAs do not necessarily increase voters' welfare in general, but rather they benefit that part of the population whose support is necessary for leaders to remain in power. Moreover, not all democratic leaders are the same and not all face the same political challenges, something that has been overlooked by previous studies. Similarly, there is great variation in the design of PTAs, which is not accounted for by the aforementioned literature. For instance, PTAs formed by the European Union and the United States contain a range of provisions that prescribe deep economic reforms across the board, e.g., constraining public subsidies to domestic companies, expanding protection of investment and intellectual property rights, and requiring opening the financial sector to foreign banks. The same provisions are not present in many PTAs in the global South.

Baccini & Urpelainen (2012, 2014a,b,c) advance the argument that leaders in developing countries sign into PTAs with the European Union and the United States to credibly commit to liberal policies and to create domestic political support for reform. However, not every leader has incentive to do so. A couple of conditions must apply. First, an implicit assumption is that implementing reforms is a way to please the leader's constituency and to improve her political fortunes. Second, Baccini & Urpelainen (2014a,b) contend that new leaders in democratizing countries are significantly more likely to form PTAs with the European Union and the United States because they face both lack of credibility and domestic political opposition.

In what ways does a PTA with the European Union or the United States help a new leader in a democratizing country? First, PTAs contain legally binding provisions that tie leaders' hands in relation to economic reform. These commitments are particularly credible in the case of PTAs with the European Union and the United States, since reneging on them leads to severe retaliation and reputation costs. Second, PTAs can help the leader create domestic political support for economic reform when interest groups that stand to lose from economic reform are more vocal than those that expect gains. In these cases, PTAs can tilt the balance in favor of those that support reform. Thus, new leaders can strategically use the international institution to create benefits for influential constituencies and interest groups as well as side payments for the losers.

Baccini & Urpelainen (2014a,b) test this argument by using a combination of quantitative methods that rely on the Desta data set (Dür et al. 2014). They complete the analysis with case studies on Colombia, Chile, Croatia, and South Africa. Both quantitative and qualitative analyses associate newly elected leaders in democratizing countries with a higher probability of signing PTAs with the European Union and the United States compared to established leaders in countries that are not going through a democratization process. Importantly, they also find that leaders entering into PTAs are more likely to implement economic reform than are leaders not joining PTAs, a point to which I return below.

Building on the empirical fact that the universe of PTAs is very heterogeneous, several studies have attempted to explain why countries commit to deep integration. Scholars have mostly focused on the presence of GVCs to explain deep commitments in PTAs (Chase 2008, Manger 2009). The argument is that the provisions included in trade agreements grease GVCs' wheels. For instance, by including provisions that remove technical barriers to trade, PTAs allow parts and components to move more freely from one country to another, so that economies reap the full benefits of locational advantages. Recent studies have also explored the determinants of deep integration in nontrade issues such as environmental clauses and labor provisions, confirming the importance of interest groups in shaping the design of PTAs (Lechner 2016, Raess et al. 2018).

Other studies focus on other dimensions of the design of PTAs. For instance, Kucik (2012) explores the role of interest groups in relation to flexibility provisions included in PTAs. He finds that import-competing industries demand escape clauses, which allow them to keep their share of the markets. Conversely, exporters push for limiting flexibility, which increases the uncertainty of trade relations and thus imposes a cost on exporting industries and firms. Building on the seminal work by Mansfield & Milner (2012), who find that PTAs are less likely to be signed with a large number of veto players, Allee & Elsig (2017) show that more veto players are associated not only with fewer commitments, but also with greater flexibility and weaker dispute settlement provisions.

The studies that have been reviewed so far assume that the formation of PTAs is a function of economic and political factors at the domestic level. However, trade agreements, like virtually every other policy, are interdependent. For instance, Elkins et al. (2006) shows that bilateral investment treaties diffuse because of competition over capital and investment among developing countries. PTAs are no exception; countries are more likely to form PTAs if other competing countries have previously done so. In short, the diffusion of international economic agreements in general, and PTAs in particular, is a force not only of domestic factors but also of systemic forces.

Studies showing the importance of PTA interdependence abound. Qualitative evidence shows that the creation of NAFTA prompted Japan (Manger 2005) and the European Union (Dür 2007) to sign PTAs with Mexico. Similarly, Gruber (2000) argues that NAFTA may be explained by Mexico's reaction to the PTA signed by Canada and the United States in 1988. Quantitative evidence also demonstrates the importance of PTA interdependence. For instance, using a spatial econometric model to operationalize trade diversion, Baccini & Dür (2012) find strong evidence that competition among exporters has been a key driver of the proliferation of trade agreements over the past two decades. Studies in economics find similar results relying on analogous econometric techniques (Baldwin & Jaimovich 2012, Baier et al. 2014b). Moreover, several studies find evidence that specific provisions—clauses to protect investment or to liberalize services, for instance—diffuse from one PTA to another, explaining the proliferation of deep integration (Baccini et al. 2014, Baccini & Dür 2015, Kim & Manger 2016).

While the aforementioned studies highlight some macro-correlations between the probability of forming PTAs and economic, political, and systemic variables, this body of research suffers from two main problems: the identification strategy is often unsuitable to assess causality, and the research design does not allow the researcher to capture the micro-foundation of PTA formation. There is limited empirical evidence of industry, firm, and individual preferences as well as of lobbying activities related to preferential trade liberalization in these studies, though their theoretical framework often hinges on interest groups' and voters' utility functions. In short, the mechanisms at play remain often untested in this body of research. Exploring the micro-foundation of PTA formation is the task taken up by the literature surveyed in the next section.

The Micro-Foundation of Preferential Trade Agreement Formation

There is a vast literature on endogenous tariff formation in both economics and political science. This literature builds on the insights of the Grossman & Helpman model (1994). In a nutshell, the model predicts that the level of protectionism granted to industries is a function of the deadweight loss for consumers generated by tariffs and the benefits that industries reap from tariffs, as well as a function of industries' political contributions to politicians. Convincing and robust evidence supports the main insights of the Grossman & Helpman model (Gawande & Bandyopadhyay 2000).

Because the primary function of trade agreements is to reduce tariffs in a preferential way, it is no surprise that several studies have applied the endogenous tariff formation framework to PTAs. One of the pioneering studies in political science is by Chase (2003, 2005). Building on the influential work by Milner (1997), his argument is that PTAs in general, and North–South PTAs in particular, help large exporters to reap the benefits of economies of scale and also help firms that are involved in offshoring activities. He tests his argument by looking at data on tariff transition (i.e., the number of years that it takes an industry to have zero tariffs) for NAFTA. He shows that preferential tariffs go to zero faster in those industries that have large economies of scale and high levels of intrafirm trade. Relatedly, Chase (2003, 2005) shows that the same industries are more likely to support the formation of NAFTA.

Building on seminal papers claiming that intraindustry trade is associated with lower adjustment costs and so greater support for trade liberalization (Milner 1999), Manger (2015) finds that tariff reductions are larger in industries with high levels of intraindustry trade. His analysis relies only on trade agreements signed by the United States. Manger's findings are at odds with recent evidence from a newly collected data set on preferential tariffs of all PTAs signed by Australia, the European Union, Canada, China, Japan, South Korea, and the United States. Specifically, Baccini et al. (2018a) find little evidence that intraindustry trade facilitates preferential trade liberalization, a result in line with Gilligan (1997). Perhaps more important, their analysis unveils that tariffs on intermediates, which are a proxy for trade in parts and components, are liberalized more rapidly than tariffs on finished goods, highlighting the importance of GVCs to explain preferential trade liberalization.

All these empirical studies explore variation at the level of industries. However, the trade literature has moved away from industries to focus on firms over the past two decades. Building on the insights of the New New Trade Theory (Melitz 2003), we observe that only large, productive firms are involved in exporting activities and in FDI, since they can afford the extra fixed and variable costs of competing in foreign markets. Moreover, there is solid empirical evidence that trade liberalization benefits large, productive firms at the expense of small, unproductive firms (Bernard et al. 2012). In short, there is a great deal of firm-level heterogeneity within the same industry, which is masked by an analysis at the sectoral level.

Some recent contributions engage directly with this renewed emphasis on firms to understand the micro-foundation of preferential liberalization. Using firm-level data on US subsidiaries operating in host markets, Blanchard & Matschke (2015) show that preferential tariff reductions are more likely in industries with large amounts of vertical FDI, whereas horizontal FDI is not associated with preferential concessions. By relying on instrumental variables, the authors are able to rule out reverse causality. Similarly, Kim et al. (2019) find that multinational corporations (MNCs) are not only concerned about tariff cuts but also care about strong dispute settlement procedures to protect their assets abroad.

Focusing on MNCs, Manger (2009) provides a compelling argument on the formation of PTAs and, more specifically, on the diffusion of specific trade-related provisions included in PTA treaties. According to Manger (2009), PTAs are not about traditional trade flows but rather about FDI. Thus, he claims that looking at tariff reductions does not capture the essence of the current wave of PTAs. His argument is simple but powerful: PTAs are instruments used by MNCs to discriminate against competitors. Specifically, MNCs are the main advocates of the inclusion of provisions protecting investment and intellectual property rights and liberalizing services to gain an edge over MNCs from other countries, which are excluded from the protection granted by PTAs. In turn, those MNCs from countries without PTAs lobby their governments to form PTAs with the same design as those enjoyed by their competitors to level the playing field. This mechanism explains the proliferation of deep integration over the past two decades. In summary, firm-level analyses confirm the strong relationship between PTAs and FDI.

What is striking from the survey of the literature conducted so far is the lack of empirical evidence on lobbying activities related to PTA formation, Chase's (2003, 2005) work on NAFTA being a notable exception. The issue is, of course, data availability, since lobbying data are famously difficult to collect in virtually every country. In this regard, the United States stands out as a clear exception. In particular, the Lobbying Disclosure Act database is a source of detailed information on lobbying activities since the late 1990s. Scholars in international political economy have recently taken advantage of this wealth of micro-data to explore the micro-foundation of lobbying on trade policy (Kim 2017).

Using the Lobbying Disclosure Act database, Osgood (2017, 2018) explores lobbying on preferential trade liberalization. His goal is to explain which firms and associations support PTAs and the degree of industrial disagreement over these policies. His main finding is that vertical FDI and input-sourcing are largely responsible for explaining support over trade liberalization, a result that echoes Chase's (2003, 2005) finding. Moreover, Osgood shows that global sourcing is also a key determinant of industrial disagreement. His findings point out that firm heterogeneity in global engagement is a source of industrial and therefore political conflict.

While studies in trade focus generally on industries and firms, some recent contributions have explored legislators' and individuals' preferences in trade agreements. Owen (2017) has made one of the few efforts to link international trade to the labor market literature. Her argument is that PTAs improve the opportunity to outsource jobs abroad due to lower preferential tariffs. This potential job loss triggers opposition to PTAs from workers, to whom legislators have to respond in order to keep their seats. Indeed, Owen finds that US legislators are significantly less likely to vote in favor of PTAs in those counties that are heavily affected by offshorable jobs. These findings are important because they provide the basis to link conflicts over preferential liberalization to voting behavior.

Finally, while there is a large literature on individual preferences over trade liberalization (Scheve Slaughter 2001, Mansfield & Mutz 2009, Rho & Tomz 2017), there is little work on individual preferences over PTAs. For instance, we know surprisingly little about the determinants of voters' support for and opposition to PTAs in democracies. Naoi & Urata (2013) explore the support for the Trans-Pacific Partnership Agreement (TPP) in Japan and find that partisanship rather than economic self-interest is the relevant determinant of individual attitudes toward the TPP. Similarly, Spilker et al. (2018b) provide evidence of individual attitudes toward PTAs from three middle-income countries: Costa Rica, Nicaragua, and Vietnam. Their result shows that sympathy or antipathy toward particular countries seems to matter even more than purely economic considerations.

The negotiation of the Transatlantic Trade and Investment Partnership (TTIP) has spurred a few analyses at the individual level. In particular, Dür (2018) shows that the argument that the TTIP would allow foreign companies to sue domestic governments had a large negative effect on public opinion, whereas the promise of jobs created by the trade agreement hardly mattered for attitudes toward the TTIP. Similarly, Jungherr et al. (2018) find that deep cooperation increases opposition against the TTIP and that geopolitical considerations matter in explaining support for the TTIP. It is debatable whether, in explaining PTA formation, individuals really matter and are the relevant actors. Nonetheless, results at the individual level confirm that political variables are of paramount importance to understanding preferential trade liberalization.

ECONOMIC AND POLITICAL EFFECTS OF PREFERENTIAL TRADE LIBERALIZATION

The literature related to the consequences of preferential trade liberalization can be grouped in three broad streams of research: (a) the effect of PTAs on trade and investment, (b) the effect of PTAs on multilateral liberalization, and (c) the effect of PTAs on economic and political reform implemented domestically. Below, I review each strand of the literature.

While there is a limited literature on the welfare effects of trade agreements (Trefler 2004, Caliendo & Parro 2015), there is a vast literature investigating whether international economic agreements increase trade and investment between member countries. The answer is generally positive. For instance, Baier & Bergstrand (2007) find that PTAs doubled two members' bilateral trade after ten years, and Büthe & Milner (2008) show a strong association between forming PTAs and attracting FDI. In addition to increasing trade between member countries, Mansfield & Reinhardt (2008) show that PTAs reduce trade volatility. Furthermore, recent papers show that PTAs have a positive effect not only on the intensive margins of trade, i.e., the quantity of a given good exported after the PTA, but also on the extensive margins of trade, i.e., the number of products in which member countries trade (Baier et al. 2014a, Spilker et al. 2018a).

Moreover, recent literature has explored the effect of the design of PTAs on trade flows and FDI. It turns out that what countries write down in PTA treaties matters a great deal and that tariff cuts are not solely responsible for the increase in trade and investment. These studies respond to new challenges generated by the current wave of PTAs, which go far beyond traditional trade restrictions at the border and remove many behind-the-border barriers.

Using the universe of PTAs signed after World War II, Dür et al. (2014) show that deep PTAs increase trade flows between member countries significantly more than shallow PTAs, a result in line with what Hicks & Kim (2012) find for a smaller subsample of PTAs. Similarly, Büthe & Milner (2014) find that PTA design matters for attracting FDI as well. In particular, only PTAs with clauses protecting investment and with a strong dispute settlement mechanism increase FDI between member countries, whereas Büthe & Milner find no effect for PTAs without investment clauses and dispute settlement mechanisms. Deep integration is not the only driver of trade flows. Bearce et al. (2016) have recently shown that PTAs including a smaller degree of flexibility increase trade more than PTAs with a large number of escape clauses, though they document that the effect is not linear; too much rigidity in PTAs reduces their ability to promote trade.

A second body of research explores whether preferential trade liberalization facilitates or undermines multilateral trade liberalization. Two camps hold very different views. Some scholars, like Bhagwati (2008) and Krishna (2013), argue that PTAs may undermine multilateral trade liberalization and that they have a “stumbling block” effect. Other scholars, like Summers (1991) and Baldwin (1997), claim that PTAs do not impede and, in fact, facilitate multilateral negotiations, exerting a “building block” effect. Aghion et al. (2007) rely on a model of international trade to illustrate equilibrium outcomes in which both effects are possible.

With these contradictory theoretical contributions, what do the empirics say? The empirical evidence has been limited so far by data availability problems, mostly on the tariff side. One of the most convincing tests supporting the stumbling-block effect is by Limão (2006). His argument is that major powers have incentives to maintain high external Most-Favored-Nation tariffs, to extract policy concessions from South countries. Limão's empirical analysis shows that US PTAs were indeed a stumbling block to multilateral liberalization during the Uruguay Round negotiations. Other studies cast doubt on the generalizability of Limão's findings. For instance, Estevadeordal et al. (2008) and Calvo-Pardo et al. (2011) find that preferential tariff cuts lead to a reduction of Most-Favored-Nation tariffs in, respectively, Latin America and Southeast Asia. This is clear evidence in favor of a building-block effect.

A third stream of research explores the relationship between PTAs and reform. The idea that PTAs can affect policies and regulations implemented at the domestic level is not novel. The seminal work by Ethier (1998) advances the argument that preferential trade liberalization is not about reducing tariffs, which are already quite low, but facilitating reform in South countries. Fernandez & Portes (1998) push this argument further by claiming that there are sizable nontraditional gains from PTAs. In particular, developing countries use PTAs to signal their commitment to liberal economic policies, to enhance their reputation, and to lock in reform. While compelling, these claims have remained untested for a long time.

Baccini & Urpelainen (2014a,b,c) have implemented a thorough empirical investigation of the relationship between PTAs and reform. Using a mixed-method approach, they show that a South country forming a PTA with the European Union or United States significantly increases the probability of passing pieces of legislation to protect investment, of implementing capital liberalization, and of privatizing state-owned enterprises. Moreover, Baccini & Urpelainen (2014c) show that South countries are strategic in timing reform. Specifically, developing countries tend to change regulations—related to environmental protection, for instance—between the signing and the ratification of PTAs to persuade North countries that South countries are reliable trade partners.

If reform is famously difficult, what do developing countries really gain from changing policies and regulations in conjunction with PTA negotiations? Baccini & Urpelainen (2012) demonstrate that, to make reform more palatable, developing countries enjoy a short-term surge of aid from the European Union or United States after forming a PTA, especially in those areas that are most affected by preferential liberalization. Gray (2009) shows that spreads on sovereign debts substantively decreased when Eastern European countries received the seal of approval from the European Union. Specifically, the reduction of perceptions of default risk coincided with closed negotiation chapters on domestic economic policy reforms implemented by Eastern European countries.

PTAs also may affect political reform through issue linkage and membership conditionality. Pevehouse (2002) shows that regional trade agreements lead to democratization, though Poast & Urpelainen (2013) have recently qualified his findings. In addition, Limão (2005) finds that PTAs foster cooperation on environmental issues. Furthermore, Hafner-Burton (2013) provides some evidence that PTAs help improve human rights in developing countries, though correcting for selection bias seems to weaken the correlation between PTAs and human rights compliance (Spilker & Böhmelt 2013). In summary, this literature indicates that preferential liberalization goes far beyond the removal of traditional barriers to trade and that it is a crucial instrument to enhance global governance and to provide a regulatory framework to globalization.

A FUTURE RESEARCH AGENDA FOR TRADE AGREEMENTS

I see three critical issue areas in which future research could improve our understanding of the effect of PTAs. My suggestions are informed by an overarching conviction related to this literature: Scholars interested in PTAs will always benefit from engaging with broader debates animating economics, political science, and international relations. Given the renewed interest in trade policy and its salience in recent elections, especially in developed countries, there is a great opportunity to address core issues in social science through the lens of preferential liberalization. Below, I detail my suggestions, offering some stylized facts supporting the argument that more research is required in each of these areas.

Interdependent Components of Preferential Trade Agreement Design

Despite impressive progress in coding PTAs and measuring economic integration over the past ten years, our understanding of the design of PTAs is still limited. In particular, the interdependence of different dimensions of PTAs—depth and flexibility, for instance—has been largely overlooked by the previous literature with few exceptions (Baccini et al. 2015). In explaining the design of PTAs, scholars have typically focused on one dimension of integration, e.g., depth of commitment or degree of flexibility, without exploring how these different dimensions coevolve across PTAs and over time. However, both theory and empirics demonstrate the interdependence between different dimensions of international institutions.

Figure 1 shows the correlation between depth and flexibility for more than 20,000 dyads that signed at least one PTA between 1948 and 2014. A couple of observations stand out. First, the correlation is very high, ρ > 0.8; in fact, it is significantly higher than any other economic, political, or systemic variable commonly included as a predictor of PTAs. We observe similar relationships between other components of the design of PTAs. For instance, the correlation between depth and enforcement is also very high, ρ > 0.9, suggesting that countries bother to negotiate deep PTAs if and only if they can enforce them. Second, the correlation between depth and flexibility is not linear, suggesting decreasing flexibility for very deep PTAs, a result at odds with previous theoretical contributions (Rosendorff & Milner 2001, Johns 2014).

All in all, there is little doubt that different dimensions of PTA design are interdependent. Moreover, the text of PTA treaties diffuses from one to another (Baccini et al. 2014, Allee & Lugg 2016, Allee et al. 2018). Using text-as-data methods is a powerful, promising tool to understand interdependence among PTAs and textual diffusion, a point also forcefully made by international economic lawyers (Alschner et al. 2017). Ideally, future papers should implement full-fledged models that incorporate both mechanisms, i.e., interdependence and diffusion, to explain how the design of PTAs has evolved over time.

Lobbying Activities Beyond the United States and Manufacturing

A second critical issue for future studies is which pattern of lobbying on trade agreements we observe in countries other than the United States and in industries other than manufacturing. While we have a good understanding of how lobbying works in the United States, we know surprisingly little about lobbying activities on trade agreements in European countries and other developed economies, not to mention in developing countries. Since the United States has a very specific institutional environment, there are reasons to believe that lobbying may operate differently in other democracies. Again, the issue is the lack of lobbying data. In this regard, the Transparency Register database, which is freely available to scholars (http://ec.europa.eu/transparencyregister/public/homePage.do), could provide a powerful tool to explain lobbying activities in trade in general, and PTAs in particular, in the European Union.

Future research would also benefit from exploring lobbying activities related to trade agreements beyond the manufacturing sector. Figure 2 shows differences in lobbying behavior between the merchandise and service sectors in relation to the implementation bills of all PTAs signed by the United States after 1995 (Baccini et al. 2018b). Two facts are salient. First, the US service sector is highly involved in lobbying on PTAs. Strikingly, services invest in lobbying for PTAs as much as merchandise does, though many services are not tradable. Still, few studies explore incentives for services to push for preferential trade liberalization (Chase 2008, Manger 2009).

Second, the US service sector tends to lobby cohesively in trade associations, whereas manufacturing firms operating in the same industries are more likely to lobby alone, which has been interpreted as a sign of conflicting preferences over preferential liberalization (Bombardini & Trebbi 2012). In other words, we do not observe in services the same degree of industrial disagreement that we observe in merchandise when it comes to implementing PTAs (Baccini et al. 2018b). These differences between services and merchandise beg further investigation, given that services are by far the most important sector in every developed economy in terms of both share of GDP and number of workers employed. More generally, the trade literature should devote more attention to how the politics of trade policy play in services and how they differ from the politics of trade policy in manufacturing (Weymouth 2017).

The Distributional Consequences of Preferential Trade Agreements

A third critical issue for future studies relates to the distributional consequences of PTAs. There is now strong evidence that PTAs increase trade and investment between member countries. In short, PTAs seem welfare enhancing at the aggregate level. However, we know significantly less about winners and losers from preferential trade liberalization at the level of the firm. There are reasons to be concerned about uneven gains from PTAs. For instance, Baccini et al. (2017) show that only the most productive US MNCs increase their sales after PTAs, whereas many smaller, unproductive subsidiaries gain nothing from preferential liberalization and, in fact, some face losses. These results are in line with the predictions of the New New Trade Theory (Melitz 2003) and echo recent claims that PTAs are tools serving the interests of large corporations (Rodrik 2018).

Figure 3 suggests that the results from US MNCs are generalizable to other firms operating in other countries. In particular, the figure shows the marginal effect of preferential tariff cuts on revenue for different levels of firm productivity. The sample includes more than 700,000 firms operating in the 28 EU countries from 1995 to 2014. The figure shows that only the most productive EU firms increase their revenue after preferential liberalization, whereas revenue shrinks for the unproductive firms.

While these results do not necessarily imply negative economic effects generated by PTAs, they unveil clear losers from preferential trade liberalization, which should be a concern especially for political scientists. Indeed, the current backlash against globalization that developed countries are experiencing is also a result of uneven benefits from international economic cooperation, with concentrated gains among large corporations and diffuse losses among blue-collar workers employed in the manufacturing sector. This growing trend of negative attitudes toward trade liberalization puts at risk regional integration, as we see with unprecedented conflicts among EU partners and the very tense renegotiation of NAFTA. Thus, implementing policies to compensate losers from globalization seems of paramount importance to keep the process of economic integration going and to avoid the collapse of preferential liberalization.

Building on these insights, I see a few concrete avenues that future studies could pursue. First, using firm-level data sets with variation across countries, scholars in international political economy may engage with the literature of the varieties of capitalism and explore if domestic political institutions—for instance, labor market and welfare institutions—mitigate these uneven distributional consequences of preferential liberalization.

Second, political scientists may be interested in examining whether losers from preferential liberalization develop negative attitudes toward globalization in general and trade liberalization in particular. Ultimately, since losses from preferential liberalization are concentrated among a specific set of workers (Hakobyan & McLaren 2016), future research would benefit from investigating whether this affects voting behavior in democratic countries. Electoral victories of protectionist parties and candidates in both Europe and the United States, as well as the unexpected outcome of the Brexit referendum, have recently unveiled that globalization in general, and trade liberalization in particular, has the potential to fuel nationalism and populism across developed democracies (Colantone & Stanig 2018a,b).

Third, making use of fine-grained data on the design of PTAs, scholars may explore whether removing behind-the-border barriers generates the same uneven distributional consequences as cutting tariffs. Since removing behind-the-border barriers is more likely to affect the fixed costs of exporting than the variable costs of exporting, it may impact the extensive margins of trade more than the intensive margins of trade. Exploring the effect of behind-the-border barriers' removal would also allow bringing other economic and political groups into the picture in addition to firms and business groups. Indeed, civil society organizations have become increasingly engaged in trade policy lobbying in developed countries. For instance, there is evidence that civil society groups oppose the TTIP due to concerns that provisions included in the agreement would lower safety and environmental standards (Young 2016).

Exploring the distributional consequences of PTAs would help us understand for whom PTAs really work and whose interests they serve. That would make preferential liberalization more central in the debate about challenges and opportunities of international cooperation and global governance in the twenty-first century.

disclosure statement

The author is not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.