The Political Economy of Nontariff Trade Barriers in Emerging Economies
Drope, Jeffrey M., Political Research Quarterly
This article examines the counterpoint to recent economic reform programs: policies that potentially defy the reform process. Policy "reversal" is particularly evident in the area of trade liberalization where nations reduced tariffs while simultaneously introducing nontariff barriers (NTB), particularly the widespread antidumping (AD) measure. This research seeks to identify the determinants of these new "postliberalization" trade policies by combining dominant interest group approaches with more state-centered explanations in an examination of two vigorous NTB users, South Africa and Mexico. The results of nested probit analyses suggest that private interests dominate the process in Mexico, while both interests and the state's overall strategic development objectives shape trade policy outcomes in South Africa.
Keywords: economic reform; Mexico; nontariff barrier; South Africa; trade policy
In the final decades of the twentieth century, many developing nations initiated and realized marketoriented economic reform. One cornerstone of the reform process was trade liberalization, and average tariff rates dropped considerably. Yet as tariff rates decreased, the provision of trade protection through nontariff barriers (NTBs) has continued to increase markedly throughout the developing world. While scholarship has focused on the paths to and the effects of reform, we know very little about the important and perhaps surprising counterpoint to these programs - nations' actions to slow or even reverse processes of economic restructuring. What are the determinants of the recent "reregulation" of trade policy? Utilizing a theoretical framework that emphasizes the roles of both powerful interest groups and the state, this research seeks to answer this question by examining trade policy implementation in South Africa and Mexico, which have recently become among the most vigorous users of NTBs in the world.
An important theoretical debate exists between scholars emphasizing how interests influence the implementation of major economic policy and others who suggest that the state may play a more active role in shaping policy outcomes. Trade researchers in particular have tended to privilege society-centered approaches that emphasize the role of interest groups in policy making. Moreover, there is a widespread perception in the era of economic reform that the policy implementation options open to governments are increasingly limited as the state's active role is strongly discouraged and perhaps even punished by international financial institutions, other powerful national governments, and/or private creditors. While the extant research reflects a bias toward interests' role in conditioning policy implementation across the globe, the role of the state may be underestimated, particularly in large emerging economies. Anecdotal evidence suggests strongly that some governments play more proactive "developmental" roles. By integrating these theoretical literatures, we can determine if these new nontariff policies are much more than just rents paid by the highest or most powerful bidders, particularly in the early years after initial economic restructuring.
The rise of NTUs as a possible counterpoint to trade liberalization is best illustrated by the recent widespread use of the antidumping (AD) measure. Until recently a rarely utilized NTB outside of a handful of developed nations, these measures now affect tens of billions of dollars in trade each year in both developed and developing nations, equal to roughly one-quarter of worldwide imports (Hindley and Messerlin 1996; Niels 2000). Recent scholarship has underscored both AD's new status as a significant barrier to trade and how little we know about these policies in the developing world (Mankiw and Swagel 2005; Prusa 2005). Furthermore, developing nations have recently become the most vigorous users of these measures, and this research addresses directly the puzzle of who or what is shaping AD policy in emerging economies in the postliberalization era.
Antidumping and Context
While the use of NTBs has recently become the subject of more scholarly investigations, the preponderance of trade research continues to focus on tariff levels in developed nations. Without question, tariff levels have plummeted in recent years. Twenty years ago, average tariff rates were higher than 40 percent in most developing nations, while today, most countries' average tariffs are below 10 percent. But NTBs demonstrate the opposite trend, and the widespread use of such measures globally points to a veritable sea change in trade policy preferences and is an important and relatively unexplored-and perhaps unanticipated-counterpoint to the economic reform process. In little more than a decade, developing nations have become aggressive users of AD measures, surpassing even the traditional major AD users in the developed world. Twenty years ago, AD was almost exclusively the policy domain of developed nations, with just three developing nations using the instrument, and only sparingly. Only fifteen years later, more than forty developing nations were employing the policy, and aggressively in many cases. Perhaps as many as an additional sixty countries have added AD laws in the past few years, a possible harbinger of even more use or possible abuse, and initiations and implementations have now reached into the hundreds annually. Yet, to date, we have very little insight into why there was such a dramatic shift in the use of this policy.
Originally, policy makers developed AD measures as a remedy against the unfair trading practice of "dumping," which is selling goods below the cost of manufacturing and/or the price in the domestic market of the producer. Some firms allegedly dump their goods to gain a foothold in a new foreign market, while at other times, more perniciously, they dump to eliminate the foreign country's domestic competition and establish market domination. International trade guidelines, enshrined in Article VI of the General Agreement on Tariffs and Trade (GATT), permit nations to levy substantial duties against these alleged "dumpers" in the form of AD measures. Yet most analysts agree that dumping is not nearly the common phenomenon that the number of these measures might suggest and have demonstrated empirically that the vast majority of AD cases are thinly veiled excuses to put in place barriers that are simply added protection for domestic industries that are actually not experiencing any dumping (e.g., Finger 1993; Gallaway, Blonigen, and Flynn 1999; Lindsey and Ikenson 2003; Prusa 2001, 2005). Moreover, though it is an instrument that in theory should be implemented in a straightforward manner using fairly strict economic criteria, analysts note that nations regularly interpret the rules to privilege strong domestic interests regardless of the economic merit of the case, which suggests that governments are susceptible to outside political and interest group pressure (Blonigen and Bown 2003; Drope and Hansen 2004; Hansen 1990; Hansen and Prusa 1997; Moore 1992). Demonstrating actual dumping, material injury and a causal link between the two-the bases of Article VI-has proven to be fraught with inconsistencies and vulnerabilities perhaps unforeseen by the designers of the statute. It appears that politics-not just economics-conditions many AD outcomes.
South Africa and Mexico are both theoretically compelling cases not only because they are major users of AD policy-South Africa uses these measures the most on a per import dollar basis while Mexico implements the most measures as a proportion of initiated cases-but also because both countries have strong state apparatuses and private interests that clearly wield genuine political power. We might therefore expect a priori utility from both society- and state-centered explanations. Because the two countries have been and remain such vigorous users of AD, I have the luxury of examining all the relevant cases over a reasonably long period of time and testing for the many crucial subnational, national, and international factors that might condition outcomes-many of which vary both between the two countries and between the specific cases within them.
Like many emerging economies, South Africa and Mexico have notable histories of strong political institutions. Prior to the 1990s, both countries had highly centralized and nondemocratic governments. Furthermore, both countries have recent histories of experimentation with a variety of state-led economic policies including but not limited to import substitution industrialization (ISI). On one hand, as in many other large developing nations, governments from both countries have sought growth through continued "orthodox" economic restructuring including floating currencies and inflation-targeting policies. On the other hand, however, to engender economic growth, both countries' governments have employed heterodox policies, particularly to respond better to economically vulnerable political constituencies, including publicly funded job creation and public service provision programs in South Africa (e.g., Mattes and Thiel 1998; Lane and Ersson 1997) and the PROGRESA program in Mexico. Undoubtedly, these policy-making struggles and experiments are common across other important emerging economies suggesting strongly that this research has broad implications.
Like most large emerging economies, South Africa and Mexico are relatively highly industrialized and demonstrate powerful business sectors that have traditionally wielded influence over the government. Ironically, the birth and/or growth of these powerful private interests may have owed in considerable part to earlier state-led policies. After many years of often cozy relations, many of these major businesses have undoubtedly continued to seek to maintain or even augment the protection and/or other preferential policy treatment they once enjoyed. Moreover, many of these firms are part of the countries' relative economic stability and/or are very much the engines of economic growth (or perhaps in some cases, impediments to it). But national governments all over the developing world must strike a challenging balance between addressing the needs or desires of specific powerful private interests and fulfilling their own overarching economic goals, because the needs of one firm at any specific time may not necessarily be consistent with governments' overall plans for economic growth.
Literature and Theoretical Orientation
The existing trade literature employs both state- and society-centered theoretical approaches, though the preponderance of recent research, particularly on U.S. trade policy, has focused more on society. Typically, these approaches underscore interest groups' demands for trade protection, while portraying the state's role as either passive or reactive. In contrast, a smaller proportion of research presents the state more as an independent actor with its own sets of preferences and goals beyond that of a register of demands. Furthermore, some state-centered approaches focus specifically on states' relationships with other states, while other approaches emphasize the states' proactive policy decision making, particularly in terms of economic policy. Finally, a few researchers have sought to integrate these approaches (e.g., Hansen and Park 1995; Mansfield and Busch 1995).
The recent scholarship on nontariff trade protection and particularly AD policy generally focuses on the U.S. International Trade Commission (USITC) and takes a society-centered approach, emphasizing how different aspects and mechanisms of influence lead to specific policy outcomes-that is, more or less trade protection (e.g., Devault 1993; Drope and Hansen 2004; Hansen and Prusa 1996, 1997; Moore 1992). Studies examine the characteristics of interest groups involved in the policy process to determine which types of actors are most likely to achieve their preferred policy outcomes, particularly underscoring characteristics such as organization, resources, and geographical/political location.
The "supply" side of the trade protection dynamic presents passive politicians who desire réélection and seek to garner the support of powerful constituents by responding to their demands with desired protectionist policy outputs. Large and economically important industries are argued to be key constituents both politically and economically, and empirical research particularly in developed nations demonstrates that they are more likely to be the recipients of favorable policy outputs such as AD.
Similarly, general research on trade protection has emphasized the importance of macroeconomic context, particularly states' reactions to changes in the business cycle. When an economy is in decline it inevitably means that some or all industries are facing downward pressures on wage levels and/or the possible elimination of jobs, and one common way to obtain relief from these challenges is to seek protection; thus, overall "demand" for protection increases (Bergsten and Cline 1983; Ruggie 1983). On the supply side, the literature makes a reasonable case that the state does not necessarily react only to particularistic demands of constituents but also follows the business cycle and will make some decisions in crude reaction to aggregate economic performance. Notably, the empirical evidence to support this idea-at least in developed nations-is weak.1 The more volatile economic context of most emerging economies suggests, however, that it is a potentially important variable.
An alternative set of approaches to explaining trade protection posits that the relationship between the state and society is more complex than the vague system of a reactive quid pro quo that some interest group scholars hypothesize, and that the state plays a more proactive role in implementing policy that is not only targeted but also motivated by a different set of distinct goals (e.g., Ikenberry, Lake, and Mastanduno 1988; Krasner 1978; Lake 1988). It is reasonable to expect that some states generate policy outputs that are shaped more deeply by their overarching interests in economic development. Rather than simply reacting primarily to the exogenous demands of powerful domestic interests and/or a general perception of the economy, the state determines its greater interests and strategic goals from within itself, and then generates many of its policy outputs accordingly.
State-centered approaches focus on different elements of the state seeking to manipulate trade policy to achieve larger political or economic goals. Hansen and Park (1995) made the important distinction between policy goals that focus on economic welfare and those that focus more on political welfare. On the economic side, national governments attempt to manipulate trade policy to achieve larger economic goals, and particularly to maximize economic welfare (Conybeare 1987; Krugman 1986; Krugman and Smith 1994), presuming usually that freer trade is more "welfare maximizing" than more closed trade, though the actual empirical evidence is mixed (see Buffie 2001). It is reasonable to expect that some governments believe that the goals of improved economic welfare can be reached in a number of ways beyond simple open markets, including targeted sector-level assistance. Obviously, improving economic conditions also helps to reach political goals-most elected officials recognize that economic prosperity greatly conditions their future electoral success.
The role of the state in new or transitioning democracies also generates interesting theoretical questions. As noted above for these countries historically, and for many nondemocracies more generally, nondemocratic governments have frequently played significant roles in managing the economy through ISI and/or other policies. However, during nondemocratic times, governments also still rely on the support of powerful groups, often including key business people, who comprise what Bueno de Mesquita et al. (2003) termed the "selectorate." In fact, in many undemocratic contexts the goals of powerful private economic interests and the state may be difficult to distinguish. So when the selectorate expands to include new actors, particularly new voters broadly, there are likely new potentially conflicting pressures on the government from the old nondemocratic and the new democratic selectorates; it is therefore vital to consider whose preferences the state is likely to represent most.
Finally, moving from more domestic to international explanations, trading issues are almost without exception at the forefront of most nations' relations with other countries in today's global context. National governments must seek ways to navigate the specific issues stemming from confrontations over particular products or services with the broader trading relationships that often fundamentally shape a country's economy. To survive and even thrive in the international context, well-managed trading relationships with other countries are pivotal to any government's success.
Approach and Hypotheses
This research utilizes an integrated theoretical framework that combines state- and society-centered approaches since there is no compelling logic that suggests that they are mutually exclusive. Moreover, in the context of emerging economies generally, and postliberalization contexts more specifically, we should test for both explanations. In fact, if the "reform" process has been successful, neither the state's goals nor private interests' demands should have systematic effects on trade policy implementation; instead, we should only observe that case merit guides the decisions. Of course, we know that policy making rarely happens in a vacuum and that we can expect that politics shapes policy.
The first set of hypotheses examines the state's "reactive" responses to interest group pressures. The propositions seek to identify relative measures of the political power of particularistic (private) economic interests by utilizing measures of their political and economic characteristics. First, scholars have long argued that larger sectors are often more important politically because of the inherent advantages of size, especially because they employ more voting workers and wield significant financial resources that can be put toward supporting elected officials' campaigns (Finger, Hall, and Nelson 1982; Hansen 1990; Moore 1992), suggesting the following hypothesis:
Hypothesis 1: The greater the size of the petitioning firm's sector, the greater the likelihood that an AD case will end in protection.
Many analyses of protection and/or AD in other nations suggest that patterns of representation, and particularly breadth of distribution of an industry's operations across political constituencies, have an effect on case outcomes (Caves 1976; Hansen and Prusa 1997; Pincus 1975). Presumably, firms and industries can appeal to powerful subnational elected officials to lobby national-level politicians and/or bureaucrats on their behalf, and local and provincial governments in the interest of the economic well being of their region have considerable incentive to assist these petitioners. Thus, it is reasonable to expect that industries that are widely dispersed across political constituencies will have more politicians advocating their requests with the national government, suggesting the following:
Hypothesis 2: Petitioners from industries located in more provinces or states are more likely to receive a protectionist decision in an AD case.
Though analysts have long considered the effects of both import penetration and political organization on trade protection, the empirical results have been decidedly mixed for both variables in terms of direction and statistical significance. Grossman and Helpman (1994) addressed this confusion by arguing that the effects of import penetration will differ depending on the sector's level of political organization.2 In well-organized industries, protection should increase with a decline in import penetration because the specific-factor owners have more to gain from an increase in price (from protection) than the overall economy loses because of fewer imports, suggesting the following:
Hypothesis 3: Decreases in the import penetration ratio will condition a greater likelihood of a protectionist decision for AD petitions originating in organized sectors.
On the other hand, for less organized sectors, protection should decrease with less import penetration because the specific-factor owners' gains from a price increase would be less than the overall loss to the economy, suggesting the corollary hypothesis:
Hypothesis 3 Corollary: Decreases in sector-specific import penetration in cases involving firms from unorganized sectors are likely to lead to a denial of protection through AD.
The next hypothesis tests theoretical propositions consistent with approaches that emphasize the state's possible "proactive" pursuit of larger national and/or public interests including especially economic growth and stability. Scholars of AD policy in the United States and other countries have noted that government trade agencies often respond to dynamic changes-particularly growth and decline-in the performance of specific industries (Hansen 1990; Holden and Casale 2002; Sabry 2000). Not surprisingly, questions of employment matter enormously in national politics, particularly in emerging economies. In South Africa, where recent nonofficial estimates place unemployment at greater than 40 percent (Kingdon and Knight 2003), it is actually explicitly part of the main trade agency's mandate to consider this variable (Board on Tariffs and Trade [BTT] 1995). There are, however, two plausible reactions to employment struggles. In the United States, empirical studies consistently demonstrate that the USITC disproportionately rules in favor of industries that are losing employees and/or experiencing drops in production, whether or not the industry is showing any signs of recovery. In contrast, Holden and Casale (2002) found in their sample of protection requests that the South African government is more likely to assist growing industries. These differing reactions or actions suggest competing hypotheses:
Hypothesis 4: Negative changes in sector-level employment are more (less) likely to condition a protectionist response to an AD petition.
Last, I expect national governments to use trade policy to help shape relationships with other countries. First, governments often seek to address the negative public perception of trade deficits. Though some scholars suggest that trade deficits possibly lead to a weaker currency and/or economic instability, empirical evidence is uneven, and the consequences of such imbalances are far from understood. Nevertheless, politicians frequently present-and many citizens view-negative trade balances as problems that require remedy. At the same time, however, governments must successfully manage relationships with important trading partners, particularly the nations that buy their goods, to sustain export growth. Nations do not want to provoke retaliation against their trade policies or, worse yet, an actual "trade war" wherein protectionist actions against their key domestic industries increase dramatically. Accordingly, I propose the following two hypotheses:
Hypothesis 5: The greater the trade deficit with the home nation of the targeted foreign industry or firm, the greater the likelihood that the government will decide in favor of protection.
Hypothesis 6: The proportion of total exports to the target country will vary negatively with the government's decision to grant protection.
As a part of economic reform programs in the late 1980s and 1990s, South Africa and Mexico lowered tariffs, but as discussed above, they also began to utilize NTBs regularly.3 To address-or perhaps facilitate-these new policy trends, both governments created national agencies to implement the policies. South Africa's Board on Tariffs and Trade (BTT) and Mexico's Unit for International Trade Practices (in Spanish, the Unidad de Prácticas Comerciales Internacionales; herein, the UPCI) bear significant responsibility for implementing AD and related policies. Notably, the agencies are structured differently. The executive of the BTT is a semiautonomous board with four or more members appointed by the president of the Republic. The members are required to have experience in commerce, industry, and/or economics; serve a five-year term; and can be reappointed by the president. The president, with "sound reason," is also permitted to remove members. The Board is authorized to appoint a committee and/or an "investigating office" to consider each petition, and any firm or industry association can petition. The investigating office reports to the committee and/or the Board, and has the rights of a regular South African court to compel witnesses to testify at open public hearings and/or produce relevant documents. The committee presents its findings and preliminary verdict to the Board, which can uphold the decision or turn it over. A "turned-over" decision usually translates into further investigation. The committee and/or the Board are, by statute, expected to consider any or all of the following: political influences, state of the economy, labor matters, boycotts, product quality and range, delivery periods; technology utilized, capacity, policies of industry concerning production, marketing, and finance (BTT 1995). Finally, the BTT employs a permanent bureaucracy that helps to facilitate the resolution of petitions, and the senior bureaucrats serve three-year terms.
The structure of the UPCI is more fluid-and less transparent-and falls more directly under the control of the Secretariat of Commerce and Industrial Development (SECOFI), which is ultimately principally under executive control. Though plans to restructure and make the agency more autonomous have been proposed by various high-ranking officials, final decisions still rest with SECOFTs chief minister. Ostensibly, the lack of autonomy is a result of SECOFTs self-proclaimed and seemingly fiercely protected role as guardian of trade liberalization-wherein it can ensure that AD is used only as an "escape valve" in the broader liberalization process and not as a political tool. The flipside of this contention is that there is no guarantee that SECOFI is itself immune to political pressure or that liberalization is its ultimate goal. In fact, many including those in industry suggest that it is vulnerable to political pressures.
Antidumping measures affect between 15 and 25 percent of the two countries' imports as measured by the four-digit Harmonized Tariff Schedule (HS) codes, which is consistent with the estimates for worldwide trade distortion caused by AD (Hindley and Messerlin 1996; Niels 200O).4 Such levels of protection, while perhaps not at the level of distortion of tariffs from twenty or thirty years ago, are still high, particularly for any nation that claims to be liberalizing trade. The size and scope of AD suggest that in most circumstances it is likely more than just a small escape valve redistributive program, or a policy designed to "pay off' possible opponents to economic reform (Etchemendy 2001).
This study focuses specifically on AD cases in the era after major tariff rationalization. Both the quantity of initiations and the number of new measures implemented by the two countries have increased dramatically through this time. For example, in 1993, there were only two cases initiated by the South African government, while at the height of petitioning in 1998, there were forty-five initiations. The number of petitions has remained high since. The percentage of affirmative decisions also increased dramatically from 20 percent in 1994 and 1995 to a pinnacle of 100 percent in 1999, and has remained above the two-thirds mark in subsequent years. In Mexico, initiations hovered around ten annually until 1992 and 1993 when they spiked to twenty-five and fifty, respectively. Initiations returned to a lower but steady level subsequently.
In terms of the geographical distribution of target countries by AD initiations, certain regions are disproportionately represented in the cases. In particular, East Asia is targeted at a level much greater than the volume of imports from these countries, which is perhaps a consequence of direct-and-vigorous-competition in many sectors. Interestingly, South Africa does not target many of its African trading partners, which is likely a result of the low volume of imports from these nations, though recent increased economic integration may also be a factor. While Mexico does target its largest trading partner, the United States, it is at a much lower proportion (less than one-third of cases) than the volume of imports (roughly three-quarters of total imports) might suggest.
Though the sectors targeted by these two countries vary widely, the steel and steel product industries participate prominently and dominate the process particularly in Mexico. Their numerous-almost half in Mexico-petitions in part likely reflect the competitive nature of the market for these global commodities, and perhaps also the use of AD by the industry around the world. More generally, the characteristics of the individual firms involved in the cases also vary significantly, though more so in South Africa, where the firms include global conglomerates down to small, family-owned firms. In contrast, in Mexico, one firm, Altos Hornos, one of the largest steel manufacturers, is the most active private participant, having filed more than a third of cases.
Model and Measures
Since most governments including Mexico's and South Africa's generally initiate formal AD cases only after a firm or industry petitions for protection, many scholars argue that the two-step nature of the process necessitates the use of a selection model (e.g., Hansen 1990). In short, researchers suggest that certain types of firms or industries may be more likely to apply for protection-a dynamic that may affect the sample of firms or industries involved in the actual cases. Perhaps because of the complexity and data-gathering burden, few analysts employ a two-step approach despite the fact that results from these analyses frequently demonstrate robust findings in both stages. I utilize the required data to perform two-step tests for both countries.5 Furthermore, because endogeneity could potentially present problems-that is, AD duties may affect some of the independent variables-I use measures for these variables only from the year before (i.e., lagged) the outcomes.6
In the decision stage, this study's principal focus, the dependent variable is a binary indication of whether the BTT or UPCI provided/recommended protection for a petitioning industry through an AD measure for all petitions between 1993 and 2002 (inclusive). I use individual BTT and UPCI case reports to determine which industries petitioned and the specific case outcomes, and cross-reference these data with WTO semiannual AD reports for accuracy.
For the relevant independent variables, I first develop a number of suitable measures to explore interest group/"reactive" approaches. For sector size, I use the total number of workers in the petitioning firm's industry as a percentage of total national employment. The data (South Africa's Standard Industrial Classification-five-digit [SIC-5] and the International SIC at the three-digit level [ISIC-3] for Mexico) are drawn from South Africa's Trade and Industry Policy Strategies (TIPS; 2004-2006) economic database and the Mexican National Institute of Statistics, Geography and Information's (Instituto Nacional de Estadística Geografia e Informática; INEGI's) main online database. To measure geopolitical dispersion of the industry, I use the number of provinces/states in which the petitioner's sector has operations using data from the "Regional Output by Group" in the South African Census of Manufacturing (1996) and INEGI's Indicadores Regionales de Coyuntura (n.d. [various years]).
To examine the endogenous (i.e., simultaneous) effects between nontariff protection and import elasticities, import penetration, and political organization, most recent studies of developed nations have the luxury of using both lobbying expenditures (usually political contribution measures) and crossindustry trade protection levels. Unfortunately, meaningful political spending data do not exist for these countries. Moreover, I am not estimating models with industry levels of protection (e.g., average tariffs or NTB coverage ratios), but rather discrete and important-but not necessarily systematic-protection cases. I can however partially examine these same concepts by drawing on measures and methods developed by Goldberg and Maggi (1999) and particularly by Holden and Casale (2002). Accordingly, I first use a measure of the change in the ratio of output to imports (effectively, the "inverse" import penetration ratio) in the two years before the protection decision (SIC-5 for South Africa from TIPS, and ISIC-3 for Mexico from INEGI).7 To incorporate political organization more meaningfully, I add a measure that is the interaction between a dichotomous measure for "organized sectors" and the change in output-import ratio. Though not a perfect technique, to identify organized and unorganized sectors, Gawande (1997) and others have utilized industrial concentration as a proxy for actual political spending. Accordingly, I use the median of Fedderke and Szalontai's (2004) Gini-type measure of concentration and the median of INEGI's broad industry concentration measures to determine which sectors are organized and unorganized.
To explore state-centered approaches, I utilize a number of measures to represent sector-level dynamic change and foreign relationships. First, I use changes in sector-level employment in the two years before the decision (South Africa-SIC-5 and Mexico-ISIC-3), drawn again from the TIPS and INEGI databases. For the foreign relations variables, I use two straightforward measures. To measure trade deficit, I use the difference between imports and exports with the target country in the year before the decision using data from the International Monetary Fund's (IMF) Direction of Trade Statistics (n.d. [various years]). I use the same data to calculate the percentage of total goods and services exported to the target country in the year before the case.
To control for actual dumping in a case, I use the measure that most AD analysts and agencies use to measure case merit: changes in product-level imports in the year before the case. Sudden and significant changes in import flows are widely considered to be the most reliable indicator that dumping is taking place. While change in imports is not a perfect measure, because occasionally import flows can change for unrelated reasons such as a sudden shift in demand, across a larger sample of cases, it is generally accepted by trade scholars and policy makers to be prima facie evidence of dumping. I use a measure of the change in imports at the HS-4 level as the BTT reports describe the involved products in the year before the decision (from TIPS).8 I use the UPCI's six-digit HS product descriptions and Dataintal's import database for Mexico (Interamerican Development Bank 2004).
Finally, it is important to control for macroeconomic context. I use change in annual GDP to represent it broadly (World Bank). While it is difficult to isolate precisely how macroeconomic context is affecting policy outputs like AD, the measure at very least acts as an important check of the robustness of the other variables. In addition to broad growth, some analysts suggest that changes in currency values will affect the provision of protection (e.g., Bergsten and Williamson 1983; Dornbusch and Frankel 1987). Like many emerging economies, South Africa's and Mexico's currencies have experienced considerable volatility in recent years. The theory suggests that positive changes in the real effective exchange rate will make imports more affordable and more competitive, and accordingly, domestic import-competing producers will put pressure on the government generally and the trade agencies specifically to make decisions in favor of protection. Thus, protection will naturally increase. Conversely, when the currency depreciates, imports become more expensive, thereby increasing the competitiveness of national producers and decreasing the demand for protection. Protection, therefore, will decline. I use change in real effective exchange rate from the previous year to capture this dynamic (Deutsche Bank 2006).
Finally, in specifying the petition stage of the process, I follow the general AD literature. The dependent variable in this stage is a simple binary indication of whether any firm or association in each major industry filed an AD petition in a given year. First, scholars have found that struggling industries are more likely to perceive that they have a plausible case and are more likely to petition (see Hansen 1990). I use the change in sector-level employment data described above. Second, there is a general perception that governments will favor industries that have made greater capital investments, causing these industries to apply more often (Blonigen and Prusa 2003). I use a measure of each industry's share of total fixed capital stock across all sectors to represent capital investment. Third, industry size may condition the likelihood of petitioning for two reasons: larger industries believe that governments will favor them because of their inherent economic importance, and practitioners and bureaucrats have noted that larger firms and industries are better able to afford the substantial financial commitment of pursuing AD petitions (Brink 2003). I use the percentage of total annual national output of each sector as a measure of size. Finally, agencies frequently use import flows as a tool to assess the validity of cases because large increases in imports are perceived to be a possible indication of dumping. Therefore, firms in industries experiencing increased imports may be more likely to apply believing that the government will look more favorably on their case. Accordingly, I use change in sector-level imports. All measures for South Africa are at the SIC-5 level (from TIPS) and ISIC-3 (from INEGI) for Mexico.
Table 1 presents the results for selection models for the two countries. The results for the BTT/UPCI decision stage demonstrate markedly different results for the two countries.9 In short, for South Africa, the results bear preliminary support for an explanation that combines both state- and society-centered approaches, though there is no evidence for the "state as an international actor," while for Mexico, there is strong evidence of interest group influence and no evidence for a strong state role. Briefly, the results for the petition stage in the two models are relatively similar for the two countries; I will return to the results of this stage at the end of this section.
First, the results for the sector-level performance measures are robust and notably divergent. In regard to the competing employment hypotheses, I find country-specific evidence for both. The coefficient for the South African data is positive and statistically significant, supporting the notion that the government is more likely to provide protection for industries that are adding jobs, which is consistent with Holden and Casale's (2002) related finding. For Mexico, the coefficient is negative and significant, suggesting that the UPCI is more likely to grant protection when an industry is in decline, which is much more consistent with studies of the USITC.
The coefficients of the variables that represent interest group approaches vary in terms of theoretical expectation. For both countries, the coefficients for the sector size measures are statistically significant (5 percent level for South Africa and 10 percent level for Mexico), but again are in opposite directions. The negative coefficient of the sector size variable for South Africa suggests that firms in smaller sectors are more likely to receive relief, challenging the longheld theoretical notions that governments have more incentive to aid firms in larger industries and/or larger industries always wield more political clout. For Mexico, however, the positive direction suggests that larger sectors are more likely to receive protection, which is more consistent with the extant research.
The directions of the coefficients of the change in output-import ratio variables are all consistent with theoretical expectation, though only the coefficient for the broader measure of output-import ratio in the South African model is statistically significant, suggesting that for cases involving the unorganized sectors in South Africa, increases in the output-import ratio (or decreases in import penetration) are more likely to lead to a rejection of the protection request. These results are fairly consistent with Goldberg and Maggi (1999).10 Importantly, the results may be affected by potential difficulties stemming from the use of market structure measures as a proxy for political organization." For example, in the United States, Hansen, Mitchell, and Drope (2005) have found that many industries that are not highly concentrated are politically active, and vice versa, suggesting that the variable may be unreliable. In contrast to analyses on the United States, however, there are no actual lobbying expenditure data; with better data, scholars should reexamine this dynamic considering its potential utility.
Oddly perhaps, the coefficients of the variables measuring the number of provinces are both negative-contrary to theoretical expectation-though neither is significant statistically. This finding adds to the existing inconclusive evidence for effects of political geography on protection.
Importantly, the steel dummy for the Mexican case is positive and significant. A steel case in Mexico is almost 20 percent more likely to receive a favorable decision from the UPCI. Considering also that it is a foundering industry, it is especially important to note the contrast with the industries receiving assistance in South Africa. For South Africa, the steel dummy was not significant, and a log-likelihood ratio test suggests that the steel variable was not adding to the overall explanatory capacity of the model. Accordingly, I elect to drop the variable from the South African analysis.12
The weak results for both countries for the more "international" state-centered variables suggest that foreign policy considerations do not strongly condition AD outputs. Except for the Mexican trade deficit variable, the coefficients of the two relevant variables are in directions opposite of theoretical expectation-negative for trade deficit and positive for export share-while none of the four are statistically significant.13
The results for the control for the actual merit of the case are notable for their lack of significance: though the coefficients are positive as anticipated for both countries, neither is significant. This finding-or lack thereof-is inconsistent with almost all of the literature on AD in developed nations and suggests that the actual economic facts of many cases may not matter as much as other factors in the decision making. Adding some intrigue to these results, it is interesting to note that in at least three South African cases (Suspension PVC from China, Hydrogen Peroxide from Korea, and Acetimenophol from Singapore), there are no records of any imports at the officially reported HS code for the named country in the three years prior to the case.14 In other words, duties have been levied against countries that are not exporting the particular goods. In light of the possible strategic policy behavior discussed above, these are notable case decisions.
In terms of controlling for macroeconomic context, the results for the business cycle control variable are consistent with theoretical expectation-the coefficients are negative and significant at the 1 percent level in both models. In contrast to most of the empirical studies of postindustrial countries, the results suggest that these trade agencies are more likely to decide in favor of protection in AD cases during times of economic decline. Similarly, the coefficient for the change in real effective exchange rate variable is positive as anticipated and significant for South Africa, suggesting that appreciating currencies heighten the demand for and/or the incentive to supply protection. I did not include a similar measure for Mexico because it correlates highly with the macroeconomic growth measure (>.9).15
I also test a number of alternative models that explore both new hypotheses and other hypotheses common in the wider literature. First, recent literature has suggested that reciprocal trade agreements (RTAs) help to facilitate trade between RTA partners and/or can distort trade with nonpartners. Relevant to this study, we might expect that cases involving RTA partners will be less likely to end in protection because the partners resolve their differences within the RTA's framework. When I include an RTA dummy, I do not find a significant relationship. That being said, there is very little trade between South Africa and most of its RTA partners, which might help to explain the lack of a systematic relationship. For Mexico, though the RTA relationship between it and the United States does not seem to affect AD policy decisions, it might help to explain the much lower proportion of cases compared to the enormous volumes of imports. Future research on NTBs might instead examine the effects of RTAs on decisions to initiate disputes.
Second, specifically for South Africa where firmlevel data are available, I build upon industry-focused studies by examining the role of firms more systematically. In particular, there are pervasive societal perceptions of large businesses' strong influence on government and that a "revolving door" exists between business executives from high-profile firms and senior government positions. Thus, we might expect that larger firms and/or firms with direct personnel ties to the government-particularly the number of directors who recently held high-level government posts-are potentially in a better position to receive preferential treatment. Using financial data from McGregor's Who Owns Whom (n.d. [various years]), I include a measure of the size of the firm (by revenue). Second, I cross-reference board membership data from the McGregor financial reports with biographical data from various official and more informal sources including the Who's Who of Southern Africa (Hayes n.d. [various years]) and the South African government's official Web sites and quantify the direct relationships between the firms and the government by tracking the careers of all of the participating firms' directors. I include in the model the number of directors who have recently (previous ten years) held a high position in the government or have participated actively in national advisory committees. In this alternative model, the coefficients of both new measures are positive as expected, but are not statistically significant, suggesting that firm size and the director-government relationship do not strongly condition AD outputs overall. The results for the other coefficients in the new model are similar to the main sector-level model. Even though anecdotally many South Africans maintain that the country is affected by the revolving door between government and business, at least in terms of AD, there is no clear empirical evidence supporting a systematic connection.
Third, many studies suggest that nonmarket economies are the disproportionate losers in AD cases because the GATT's Article VI permits countries to "construct" values for indicators such as costs and prices to approximate the economic conditions in a so-called market economy. The ambiguity of these constructions is thought to lead to systematic prejudice against these nonmarket economies (Drope and Hansen 2004; Lindsey and Ikenson 2003). I test for this possibility by including a dummy variable for cases that involve nonmarket economies and find that although the coefficients are positive, as theory would predict, they are never statistically significant. This is a particularly interesting finding for Mexico because of the large group of cases filed against China in the 1990s. While there may have been more cases involving China, they did not have a greater likelihood of ending in protection (systematic bias may however still be present in the petition stage).
Finally, NTB researchers have recently examined "retaliation" (e.g., Blonigen and Bown 2003). When I add a dummy variable for nations that were initiators of recent AD cases (less than three years), the coefficients are positive for both country models, but not significant. Similarly, it is reasonable to expect that formal and informal World Trade Organization disputes might condition strategic AD action. In this time period, South Africa was involved in only one case at the WTO (involving Indian pharmaceuticals), but the case never moved beyond the preliminary consultation stage. Thus, negotiations at the WTO do not appear to be a factor in the BTTs AD decision calculus. Though Mexico's involvement in the WTO is much greater (twenty cases, mostly involving the United States, the EU, and Guatemala), when I include a dummy variable for previous involvement at the WTO with the target country, the coefficient is not significant. Perhaps more importantly however, at least four cases in this study's empirical analysis led directly to disputes at the WTO. Thus, the causal arrow may be going the other way: decisions at the UPCI affect Mexico's interaction with the WTO.
Briefly, the results for the petition stage in the two models are notable. In terms of industry size, the coefficients are positive as expected but only significant for South Africa (at the 1 percent level). In South Africa at least, the results suggest that firms from larger industries are more likely to petition. The coefficients of the industry health variable-change in employment-are negative in both models, as anticipated, and significant at the 10 percent level for South Africa and 5 percent for Mexico. Clearly, firms in sectors experiencing decline in employment are more likely to petition. Third, the coefficients of the change in imports measures are negative-contrary to expectation-but are not significant in either model. Finally, the coefficient of the capital stock variable is negative in the South Africa model, contrary to expectation, but not significant, and positive as expected in the Mexico model, and significant. This result suggests that in Mexico the firms in industries that invest more are also more likely to petition.
The possibly "contradictory" findings for particularly South Africa between the two stages of the nested models are interesting because they suggest that firms and industries may be somewhat out of touch with government decision making. The results of the petition stage suggest that firms in both larger industries and/or sectors shedding jobs are more likely to apply, while the results for the decision stage suggest the opposite in terms of who receives protection. Perhaps petitioning behavior has not quite caught up to new trends in policy implementation.
Discussion and Conclusion
Importantly, the overall findings of the study demonstrate differing explanations for the two countries, suggesting that the politics of trade in emerging economies is complex. In Mexico, large and/or struggling industries and the dominant steel sector appear to have disproportionate influence over the AD decision-making process-results that are remarkably consistent with the interest-driven results of previous research, particularly in the United States. In contrast, in South Africa, a combination of sectorlevel economic performance and interest group pressures appear to shape the BTTs decisions to provide protection through AD, preliminarily suggesting a different government orientation toward using the policy. The results suggest that while petitions from unorganized sectors are less likely to end in protection when import penetration is decreasing, petitions from firms in growing (and/or smaller) sectors more generally are more likely to end in protectionist decisions. South Africa's and a number of other nations' AD statutes clearly state that the overall economic health of the concerned industry is to be pivotal in their decision, but it has been unusual in previous empirical work that the results bear it out as a major consideration; in this case, the outcomes appear surprisingly systematic. More importantly in broad theoretical and substantive terms, and in light of substantial speculation in the political economy literature that the state is becoming increasingly irrelevant in the era of economic reform-particularly in the developing world-this is preliminary and important evidence to the contrary.
The pharmaceutical and glass industries in South Africa and the steel industry in Mexico provide appropriate qualitative illustrations of these disparate findings. The two South African industries in comparison to others in the country are small (mining and basic industry such as steel and chemicals continue to dominate the economy) but are growing in general, and perhaps more importantly are also consistently adding jobs. These two sectors also happen to be particularly successful at securing assistance through the AD statute despite the fact that there is no unequivocal evidence of dumping. Considering the enormous scope of their infectious disease treatment programs, the pharmaceutical sector is surely positioned for substantial future growth. The glass industry, as a major supplier to a nation currently investing heavily in infrastructure, and to the continent more generally, is similarly situated for sustained growth. The South African government appears to be "placing bets" on industries that it thinks might be economic winners in the long run, a type of strategy more consistent with an East Asian model of industrial policy implementation. In contrast to-or perhaps particularly to-many postindustrial nations, the large and/or struggling industries are not systematically more likely to receive favorable AD policy decisions. Additionally, the alternative firm-level results suggest that there is not a clear pattern of the ubiquitous "revolving door" between government and business greatly affecting policy outputs. In light of the considerable literature on Africa and most developing nations that underscores the preponderance of neopatrimonial systems of governance (or in the case of much of Africa, predatory), we might reasonably have expected that even in more developed South Africa, powerful economic interests might affect policy more systematically.
In Mexico, not unlike in the United States, the enormous steel industry appears to be attempting to stall its ultimate demise partly through AD, and the government appears willing to provide such assistance. Unable to compete in the new global market place with countries such as China and India, the industry continues to turn to the government for policies favorable to its survival. Unfortunately, "survival" does not necessarily equate to high performance, and despite repeated aid through AD, the industry has yet to reinvent itself successfully. These findings are particularly notable in light of the fact that Mexico's UPCI has less institutional autonomy-ostensibly in the name of promoting liberalization-than its South African counterpart. Future research should examine systematically the impact of these agencies' institutional designs.
The overall findings that both state- and societyfocused approaches offer utility in understanding the politics of trade in South Africa are particularly notable because these results are not consistent with many previous studies of AD in other nations, particularly developed ones, in which interests most clearly dominate. While there does not appear to be a monolithic explanation of the politics of trade in emerging economies in the era of postliberalization, the mixed explanation for South Africa suggests that there is a meaningful place for both interests and the state in understanding these important processes in any nation. Many critics have disparaged the literature that "brings the state back in" because it fails to explain the complex relationships between state and society, but this analysis integrates theoretical approaches and finds that states can and do play key roles in making policy. Rather than acknowledge and attempt to incorporate what is undoubtedly a considerable amount of overlap between the two nebulous entities, researchers have instead tended to oversimplify the issue by making facile distinctions between them. When seeking to define variables that represent elements of either, it is important to be both firm in the theory guiding the definition process, yet remain realistic and flexible in allowing that the two are not always mutually exclusive. This research is an important systematic attempt to tease out the relative impacts of these important actors.
Finally, future research on emerging economies should reach beyond trade policy and examine other policy areas to determine where states might be behaving "developmentally" and where strong private interests dominate the process. Similarly, we need to carry this research to other regions and countries, because these issues are germane to many other key emerging economies including China, Brazil, and India, among others, where governments appear to play more active economic-planning roles.
Author's Note: I thank the National Science Foundation (SES 0214097) and the Latin American and Iberian Institute at the University of New Mexico for helping to fund this project. I am indebted to Johannes Fedderke for introducing me to many of the key economic actors in South Africa and the Trade and Industry Policy Strategies for supplying many data. Finally, I appreciate excellent suggestions on various pieces of this project from Alok Bohara, Wendy Hansen, Daniel "Yuch" Kono, Neil Mitchell, Ken Roberts, Tom Roehl, and Dale Smith.
1. Drope (2006) found that macroeconomic context does help shape nontariff barrier (NTB) policy outcomes in Argentina.
2. For a more in-depth discussion and derivations of these models, see Goldberg and Maggi (1999), Grossman and Helpman (1994), and Helpman (1995).
3. Beyond antidumping (AD), both countries have also employed other NTBs including countervailing duties, safeguards and phyto-sanitary regulations, but AD has by far played the largest role in the "post-tariff" era.
4. The method may still underestimate the effects because it does not account for cases that are abandoned; Prusa (1992, 2001) suggested that these cases have a significant distortive effect.
5. 1 use STATA's heckprob command to execute the models in Table 1.
6. In the case of "change" variables, I use the difference between / - 2 and t - 1, which are the closest approximations of what the trade agencies had at the time of the decisions.
7. Unfortunately, finding output and import measures for Mexico reported by the same industry classifications is very challenging. To develop a meaningful output-import ratio, I use Maskus's (1989) SITC-ISIC concordance system for imports.
8. Frequently, scholars use a sector-level measure of imports (e.g., Holden and Casale 2002). I use the specific product codes because it is reasonable to expect that the change vector will be stronger with the disaggregated data (and these are the data reported by the agencies). Notably, sector-level and productspecific measures of import change generate similar results.
9. Likelihood ratio chi-squared tests suggest that the explanatory capacities of the two models are statistically significant at the 1 percent level.
10. Scholarly discussion about interaction terms in models often argues for the inclusion of all lower-order terms, but Franzese and Kam (2005) made a compelling case that the real issue is the possible exclusion of theoretically relevant variables. In this case, the relevant literature argues specifically that the reason for the previous mixed results concerning political organization is the lack of attention paid to theory. In other words, it is the interactive term that is theoretically relevant, not necessarily both underlying variables on their own or, more specifically, the dummy for organized sectors. Accordingly, I choose to exclude the dummy in the models in Table 1 . Nevertheless, I run alternative specifications with the dummy, and when I include it, the other results are not affected appreciably. The coefficient is modestly significant (10 percent level) and negative in the South Africa model, and is positive and not significant in the Mexico model.
11. Using concentration measures from the South African Census of Manufacturing (1996) instead of the Fedderke and Szalontai (2004) measure to identify the organized and unorganized sectors generates similar results.
12. I also try including sector dummies other than steel and year dummies to determine if other sectors receive special treatment and/or if the results are time sensitive. Only one coefficient is ever significant, the year dummy for South Africa- 1997 (positive), which is not surprising considering that every decision was affirmative that year. Including this variable did not change the other results in any meaningful way.
13. It is possible that sets of cases that involve the same industry but different nations will lack