Chasing Phantoms: The Political Economy of USTR

Forthcoming, International Organization. I would like to thank Chi Zhang, Li-Gang Liu, and Sophie Louvel for research assistance. I would also like to thank Bill Cline, Kim Elliott, Gary Hufbauer, Masahiro Kawai, Peter Orszag, Jeff Schott, Chris Udry, John Odell, and three anonymous referees for helpful comments on an earlier draft of this paper, without implicating them in the final outcome.

Abstract

During the last decade the US government has taken numerous bilateral actions intended to reduce other countries' border or internal regulations that were said to restrict US exports, investments, and property rights. Judging from the data for 1994-1993, USTR paid attention to countries with large bilateral trade surpluses with the US, and to those with larger and faster-growing economies, and not particularly to those with higher barriers. Washington actually took greater action, on the other hand, towards those whose identifiable barriers were high. These patterns of behavior do not appear to vary by presidential administration. Japan received more bilateral pressure than its economic characteristics would appear to warrant, but there is little evidence that this pressure had its desired effects. The US achieved greater change in countries whose economies were more dependent on the US market and on issues covered by international rules.

Introduction

Trade barriers, though usually considered welfare-reducing, are globally ubiquitous. In light of the enormous clash between what is and what ought to be, researchers have begun addressing the question of why countries so consistently pursue such apparently counterproductive policies.

Early formalizations (Krueger 1974, Bhagwati 1982) emphasized the divergence between private and social interests in the formulation of trade policies. Empirical work (Lavergne 1983; Baldwin 1985; Destler and Odell, 1987) applied a rational cost-benefit approach to political organizing and identified characteristics such as the degree of industry concentration, trade dependence etc. which were important in determining success in securing (or opposing) protection. More recently, researchers such as Magee, Brock, and Young (1989), Rodrik (1994), and Grossman and Helpman (1994) have begun to formally model the institutional features of democracies such as political parties, lobbyists, and elections in determining the equilibrium structure of protection.

At the same time the character of trade policies has been changing. Successive GATT rounds have constrained the ability of countries to impose the border impediments to trade such as the tariffs or quotas typically modelled by theorists, and actual protection has shifted toward more bureaucratically controlled forms of contingent protection such as antidumping actions. In addition, in the United States at least, the emphasis of trade policy appears to have shifted from import protection to export promotion through the removal of barriers to trade in foreign markets.

Cognizant of these changes, researchers have begun to develop a body of literature analyzing the determinants of these new policies. The results of Trefler (1993) for US non-tariff barriers to trade largely reconfirmed earlier results that the success or failure to secure protection was strongly correlated with buyer (and seller) concentration, barriers to entry, and other industry characteristics. Bayard and Elliott (1994) and Elliott and Richardson (1996) classified US unilateral actions to open foreign markets pursued under the auspices of Section 301 of US trade law as successes or failures, then regressed a number of variables against these determinations. They found that the dependence of a target country on US trade and the size of the other country's bilateral surplus with the US were positively correlated with success, as was whether the partner's practice in question was a transparent border measure.¹

This paper proceeds in a three step fashion to analyze the determinants of US bilateral export, and investment policies (recognizing that these alone do not constitute the totality of US trade policy). The paper first asks given the scarcity of bureaucratic resources and political capital, what determines the priorities of US bilateral export and investment policy making? What issues get the attention of the Office of the United States Trade Representative (USTR) and how does it go about prioritizing its activity with respect to trade partners? Second, once an issue is on the agenda, what determines whether the USTR proceeds with formal action? Lastly, what are the determinants of success in resolving these issues? To this end, bureaucratic attention, action, and success rates are analyzed statistically.

The results of this analysis indicate that bilateral trade imbalances get the attention of policy makers, but that actual formal trade actions are more closely correlated with the existence of impediments to trade. There is some evidence that Japan is singled out, beyond what would be expected on the basis of its economic characteristics. The preponderance of evidence suggests that this pattern of behavior appears to be invariant to the Administration in office, and ineffectual, at least at the level of overall bilateral trade and investment flows. Successful elimination of other countries' trade barriers appears to be related to the degree to which foreign practices violate internationally accepted norms, and to the extent that partner countries are economically dependent on the US market.

The Machinery of Trade Policy

The Trade Expansion Act of 1962 established a formal interagency process for the formulation of discretionary US trade policies. Over time this has evolved into three organizational tiers. At the bottom of this hierarchy is the Trade Policy Staff Committee (TPSC) and its associated subcommittees, made up of senior civil servants. Next is the Trade Policy Review Group (TPRG) at the Deputy USTR/Under Secretary level. USTR chairs both of these groups (and the associated subcommittees), and although the bodies make decisions on a consensus basis, in practice USTR dominates the bureaucratic process due to informational asymmetries arising from its primary responsibility for policy implementation. The few issues that cannot be resolved at the TPSC or TPRG level are raised to the final, cabinet-level decision making group. In the Bush administration this was the Economic Policy Group chaired by the President, with the Secretary of the Treasury serving as chairman pro tempore. In the Clinton Administration, formally the final tier of the trade policy decision making process is the National Economic Council (NEC), chaired by the President. In practice, NEC Deputies Committees have been convened to consider issues before being sent to the cabinet-level body. USTR, while not formally chairing these higher level bodies, remains the most influential single voice due to its bureaucratic stake in trade policy formation and implementation.

These interagency groups cover the entire panoply of US trade policies: multilateral, regional, and bilateral negotiations; implementation of US trade laws such as section 301 of the 1974 Trade Act regarding unfair foreign trade practices, section 182 (the so-called "special 301") on intellectual property rights, Sections 1377-1881 of the 1988 Trade Act regarding foreign trade practices in the telecommunications sector, Title VII on foreign government procurement practices, section 337 of the Tariff Act of 1930 on unfair practices by foreign exporters, as well as programs such as the Multifibre Arrangement and Generalized System of Preferences. Actions under these laws typically can come through two channels: an industry petitions the US government through USTR to take some action to achieve redress, or, in the absence of such a petition, the US government decides to self-initiate action in the national interest. The one major trade policy that is really outside the interagency process (and beyond the control of USTR) is the implementation of antidumping and countervailing duty laws which is handled by the International Trade Commission and the International Trade Administration of the Department of Commerce.

Table 1 lists all formal unilateral actions completed by the US government to achieve enhanced foreign market access between 1984-1994, as listed in the USTR's annual report.²

(These actions were typically formal "identifications" of a particular partner country practice which, under US trade law, would initiate legal procedures which could potentially involve the imposition of sanctions against the offending country's imports into the US. In most cases, the disputes were resolved prior to the imposition of sanctions.) The list is striking in its relative brevity. Contrary to its popular image, the US government pursues few formal unilateral actions, averaging around four per year.

This gives a misleading impression of bilateral disputes over exports and investment, however. For every case in which the US government takes formal action, there are tens if not hundreds of cases that never reach this stage, either because the US government is able to reach some resolution with foreign governments that avoids the necessity of formal designation and action, or, alternatively, because policy makers dismiss industry petitions or they dissuade industry from filing petitions. (The latter might occur if the policy makers convince industry that redress could be more easily achieved through alternative channels such as multilateral negotiation or that the industry's case is weak and unlikely to succeed.)³ Yet another possibility is that industry may convince policy makers to self-initiate in an attempt to distance themselves from the US government action in the eyes of the foreign government. Like the proverbial tip of the iceberg, the observable actions of Table 1 only hint at the volume of unobservable activity beneath the surface.

It might be useful then to distinguish between attention and action. While the latter can be observed directly, the former can be usefully proxied through the USTR's National Trade Estimate Report on Foreign Trade Barriers. This report mandated by the 1974 Trade Act provides "an inventory of the most important foreign barriers affecting U.S. exports of goods and services, foreign direct investment by U.S. persons, and protection of intellectual property rights. Such an inventory may facilitate negotiations aimed at reducing or eliminating these barriers...Information is also included on actions being taken to eliminate any act policy or practice identified in the report" (USTR, 1994, p.1). It has been published annually (except 1988) since 1984.⁴ It is drafted by USTR and prior to publication circulated among TPSC and TPRG agencies for comment and can thus be regarded as an accurate indication of the revealed attitudes and interests of the interagency trade policy groups under the leadership of USTR.⁵ For purposes of analysis, the number of pages in this report devoted to individual partner countries will be used as a proxy for the government's unobservable level of attention to bilateral trade problems.

Hypotheses

The literature on trade policy formation suggests a number of hypotheses regarding trade conflict.

1) Trade barriers. The most obvious hypothesis is that the degree of trade conflict is positively related to barriers to trade. Operationally barriers to trade will be measured in two ways. The first is through the trade-weighted tariff level (TARIFF).⁶ A related hypothesis is that conflict may be related to the existence of particularly high tariffs in some sectors or the variability of tariffs across sectors. To investigate this possibility the standard deviation of tariffs (STNDTAR) was calculated.⁷ These measures could be interpreted in two ways. Most obviously they are measures of actual tariff barriers. However, if governments act in a consistent manner, they could also be interpreted as proxies for a variety of other non-tariff barriers which governments can impose. These have become relatively more important as successive GATT rounds have constrained countries' use of tariffs.

Non-tariff barriers pose a variety of problems for researchers attempting to estimate their impact.⁸ One approach is to attempt to measure them inferentially. The usual procedure is to estimate econometrically a model of international trade, and then to ascribe to trade policy the differences between actual and predicted trade flows.⁹ Conventional "gravity model" regressions were estimated for US bilateral exports and direct foreign investment for the period 1984-1993 and are reported in Table 2. The residuals from these regressions XRESID and INVRESID are measures of the implicit trade and investment barriers facing US firms.¹⁰ One would expect a positive partial correlation between the explicit trade barriers and trade conflict, holding market size constant. The relation between the implicit barrier measures and trade conflict should be a simple positive relationship since the implicit barriers are naturally scaled.

2) Market size. Trade conflict may be positively associated with market size for two reasons. First, aggregate market size (measured as gross domestic product, GDP) is presumably correlated with market power and countries with market power could be expected to bargain strategically generating conflict. A second reason is that the payoff to removing a barrier is greater the larger the market is and should attract the attention of trade policy makers. Note that the first argument provides a justification for a positive correlation between conflict and country size, and the second provides a justification for a positive partial correlation between conflict and size, holding trade barriers constant.

3) Market growth. If policy makers are maximizing the present discounted value of benefits, as a corollary to the arguments regarding market size, there will be a positive correlation between market growth and conflict (policy makers heading off emerging problems), and a positive partial correlation between growth and conflict, holding the level of trade barriers constant. (The latter effect would be particularly pronounced if the barriers took the form of quantitative restrictions which would become more binding as the economies grew.) The growth of national income over the previous decade (GDPGROWTH) is taken as a proxy for policy makers' expectations of future growth.

4) Trade balance. Casual observation of US political discourse suggests that there is an inverse relationship between bilateral trade balances (TBAL) and trade conflict, although in strictly economic terms there is only a very limited case to be made linking national welfare to bilateral balances per se.

5) Exports and Investment. Although it might seem logical that if exports and investment is high, there should not be much trade conflict, some argue that greater exports or investment may enlarge the potential number of domestic firms which may demand further opening. As a consequence, the expected sign is ambiguous.

6) Intraindustry trade. Analysts at least as far back as Balassa (1967) have argued that intraindustry trade (IIT, measured by the Grubel-Lloyd index adjusted for trade imbalances) will ameliorate trade conflict by facilitating adjustment to increased trade volumes. Consequently, for a given level of trade, there should be fewer trade conflicts the greater the share of intraindustry trade. This implies a negative partial correlation between IIT and trade conflict, holding the volume of trade constant.

7) Regional effects. Some have argued that US trade policy may be subject to regional effects in which the US develops unusually serene or acrimonious relations with particular parts of the world. The argument is sometimes made that the US singles out Japan for special scrutiny. This will be taken up in the succeeding section through the use of dummy variables. These hypotheses are not exhaustive. For example, it has been argued that the Section 301 cases undertaken against Brazil and India in 1989 were motivated by a desire on the one hand to shield Japan from being singled out in the Super 301 process while on the other to signal displeasure with the two countries negotiating stance in the Uruguay Round. Nevertheless, these seven hypotheses would appear to be a good starting place for the quantitative analysis.

Statistical Analyses

Correlation coefficients between the number of pages devoted to a country in the National Trade Estimate Report on Foreign Trade Barriers (ATTENTION) and a binary variable based on Table 1 (ACTION) and the variables discussed above for 1984-1993 are reported in Table 3. ¹¹

Four variables-GDP, exports, FDI, and the trade balance-are all strongly correlated with both ATTENTION and ACTION. All could be considered scale variables: the larger the partner country or the larger US exports or investment to that country, the more attention or action is directed at that partner country. The trade balance variable correlation implies that the larger the US deficit is with a country the greater the attention and action.

The results for the trade barrier variables are less compelling. There is a weak correlation between other countries' tariff levels and attention. Partner tariffs are uncorrelated with actions. These results are not particularly surprising since these are simple correlations, not partial correlations that hold partner country size constant. The standard deviation of tariffs does not appear to be correlated with either attention or action.

With regard to the implicit barriers derived from the gravity model (which are naturally scaled by country size) the export equation residual is correlated with attention, but it has the "wrong" sign-higher than expected exports are positively associated with attention. Otherwise, the residuals from the export and investment equations do not appear correlated with either attention or action.

To disentangle possible cross-correlations ATTENTION was regressed against the explanatory variables in combination. Six specifications are reported. In the first (4.1), ATTENTION was regressed against the explicit trade barrier variables (TARIFF and STNDTAR), market size (GDP) and growth (GDPGROWTH), the trade balance (TBAL), intraindustry trade (IIT), and the levels of exports (EXPORT) and US direct investment in other countries (FDI).

In the second regression (4.2) the explicit trade barrier variables were replaced with the implicit trade barrier indicators (XRESID and INVRESID). In the third (4.3) and fourth (4.4), dummy variables for Europe, Asia, Central America and the Caribbean, and South America were added to specifications 4.1 and 4.2. In the final two specifications (4.5 and 4.6), a dummy variable for Japan is added.¹²

The results reported in Table 4 are striking: ceterus paribus for every one billion dollars of US trade deficit, the USTR devotes roughly a third of a page in its annual compendium to trade problems in that country. This coefficient estimate is highly robust, and is obtained all six specifications reported in Table 4. Beyond the trade balance result, the results in Table 4 indicate that ATTENTION is strongly positively correlated with partner GDP and GDP growth.

ATTENTION appears to be completely uncorrelated with either implicit or explicit trade barriers, nor to any significant regional effects emerge from these results. The results also indicate that ATTENTION is positively correlated with FDI, though not correlated with exports. ATTENTION also appears uncorrelated with intra-industry trade, although for a shorter, more recent time period (1989-1993), IIT was strongly negatively correlated with ATTENTION (Noland, 1996). Moreover, the coefficient of determination fell from around 0.6 in the regressions over the more recent subsample reported in Noland (1996) to approximately 0.2 for those in Table 4, suggesting that the determinants of ATTENTION may have become more systematic in the latter part of the sample. These observations suggest that the determinants of ATTENTION may have changed (and become more systematic) over time, a topic that will be taken up in a subsequent section.

A similar set of regressions on the ACTION variable are reported in Table 5. The striking results here are that ACTION is strongly correlated with tariffs and the standard deviation of tariffs. The Japan dummy is also positive and significant in both specifications, indicating that more actions are taken against Japan than its economic characteristics would appear to warrant. Intraindustry trade (IIT) is also positively and robustly correlated with ACTION. This is a bit surprising since one might expect that cross-penetration would be associated with low adjustment costs and little trade friction. It could also be a proxy for rivalry, however. The other variables, including the trade balance, are uniformly insignificant.¹³

To summarize, the simple correlation analysis suggests that USTR targets large countries with bilateral trade surpluses for their attention and action. However, the multiple regression analysis which identifies the impact of variables, holding others constant, suggests a somewhat different picture: attention is driven by country size, growth, and the existence of bilateral trade imbalances, but formal actions are more closely tied to the existence of observable impediments to trade.

Does Badgering Your Trade Partner Make a Difference?

The analysis thus far has focused on the determinants of bilateral disputes over trade and investment policies. The question naturally arises as to whether the attention and action actually make any difference in terms of observed outcomes. The XRESID and INVRESID variables provide a convenient way of testing whether bilateral trade conflict has any effect on target country behavior or observed trade flows. To test this proposition, lagged values of ATTENTION and ACTION were regressed against XRESID and INVRESID and the year-to-year change in XRESID and INVRESID. In other words, do ATTENTION and ACTION influence either the level of implicit trade barriers in partner countries or subsequent change in the level of these barriers?

The results are reported in Table 6. The evidence is mixed. ATTENTION and ACTION never are significant within a year, nor do ATTENTION and ACTION have a significant effect on the year to year change in the implicit barriers. However, when the cumulative impact of ATTENTION and ACTION over a four year period is taken together, both ATTENTION and ACTION have a statistically significant impact. Unfortunately, the pattern of this impact appears inconsistent: ATTENTION and ACTION over the previous four years increase FDI higher than otherwise would have been expected, but reduce exports lower than what would have been predicted.

These results are not particularly satisfying, and it could be argued that attention or actions typically focus on a particular industry or practice and may not have an appreciable impact on the overall level of exports or investments.¹⁴ To address this concern the descriptions of partner country trade practices in the National Trade Estimates were broken down into nine issue areas: import policies (including tariffs and quotas); standards, testing, and certification requirements; government procurement; export subsidies; intellectual property rights; countertrade; services; investment; and other trade policies. A panel data set was then constructed indicating whether each of these was raised as an issue with a particular country in a particular year. Altogether this yielded a sample of 1,508 observations.

Successful resolution of an issue was defined as its removal from the agenda the following year. Table 7 reports significance tests by country and by policy of the likelihood of success against the null hypothesis of failure. (That is the larger the figure in Table 7 the more successful USTR was in resolving issues.)¹⁵ As can be seen in Table 7, in no case could the null hypothesis of failure be rejected in favor of success at the conventional 10 percent level of statistical significance. When the test was relaxed to the weaker 25 percent level of significance, in two cases, the "other policies" issues area with respect to Turkey, and countertrade for all countries, the null hypothesis could be rejected.

This formulation could be too stringent. Success is defined as resolution of an issue within the following year. As seen in Table 6, it might be possible that the impact of attention or action in regard to a foreign practice might have an eventual impact that would not be apparent within a year. In response the test was reformulated, with success defined as the resolution of an issue at any point of time after its appearance on the agenda. A variant of this approach would be to define the sample as the set of issues identified in the first report (1984) and define success as the resolution of these issues at any point in the succeeding years. These results are reported by country in Table 8.

Under these reformulations, the null hypothesis of failure can be rejected for Switzerland in both samples and for Finland in the 1984 subsample, at the 10 percent level for the EU and Venezuela for the 1984 subsample, and for a number of countries when the significance level is lowered to 25 percent.¹⁶ Indeed, under this more generous formulation, the null hypothesis can be rejected for the sample as a whole at the 25 percent level. Interestingly, in the case of Japan, which the results of Table 5 suggest may have been singled out for bilateral trade action, the null hypothesis of failure cannot be rejected for either sample.

The tests can also be calculated by issue area as shown in Table 9. The null hypothesis of failure can be rejected for both samples for countertrade, at the 10 percent level for export subsidies under both formulations, and can be rejected at the 10 percent level for standards, testing, and certification requirements, and other barriers, when the sample is limited to issues raised in 1984.¹⁷

Given that USTR has had some success measured against these more relaxed standards, the obvious question is what determines the variability of the probability of success across issues and countries. In the case of the different issue areas, it appears that the existence of international norms embodied in the GATT/WTO system is associated with success. Issues such as export subsidies and countertrade where the international rules have been fairly clear appear to lend themselves to successful resolution, whereas issues such as government procurement and intellectual property, which until the Uruguay Round international obligations were less well-defined, were less amenable to successful resolution.

Regarding partner countries, the literature suggests a variety of possible determinants of outcomes. From a bargaining perspective, McMillan (1990) and Elliott and Richardson (1996) have argued that success is likely to be correlated with the costs and benefits of compliance from the partner countries standpoint. The argument is that if the partner does not comply the US may retaliate by restricting access to the US market. Partner export dependence on the US market (defined as the share of exports to the US in partner country income) could be interpreted as a proxy for the implicit threat behind US demands. Elliott and Richardson argue that the bilateral balance could be interpreted as reciprocity: the US is more likely to get compliance in countries in which it runs a bilateral deficit and the likelihood of carrying out the threat is greater. Bayard and Elliot suggest that there is some evidence that countries that have been "bullied"-singled out for more than their share of Section 301-cases may be less compliant.

To test these propositions, the success rates underlying Table 8 were regressed against a set of variables plausibly relating to the costs and benefits of compliance. (Again, this approach is not exhaustive: there are obviously other non-economic considerations that may strongly influence outcomes.) The explanatory variables included LEXPRAT, the logarithm of the share of partner country exports to the US in partner country income, a measure of partner country dependence on the US market; LIMPRAT, the logarithm of the share of imports from the US in partner country GDP, arguably a measure of US dependence on the foreign market, or of reciprocity of access; and LINVRAT, the logarithm of the share of US foreign direct investment in partner country income. Also included were GDP and GDP per capita in logs. Presumably the US would have less success with larger targets and possibly less with richer ones. The trade balance, Elliott and Richardson's measure of reciprocity, and the usual variables relating to the existence of trade barriers.¹⁸ The results are reported in Table 10.

These results conform to expectations in certain respects and are surprising in others. Partner country dependence on the US market (LEXPRAT) is the single most robust explanator in variations in success rates across countries. The share of imports from the US in partner country income (LIMPRAT) is highly significant in regressions (10.1) and (10.2) (the regressions which use success rates for the entire sample as the dependent variable) though less so in regressions (10.4) and not at all in (10.3), the regressions on the 1984 subsample success rates. Oddly, the trade balance (TBAL) is positively associated with success: the larger the US bilateral trade balance the more likely the US was to be successful in removing a barrier. This is the opposite of what one would expect if the trade balance was measuring reciprocity (or the lack thereof) and raises questions about possible simultaneity biases in these regressions.¹⁹

Country size as measured by GDP was not significantly correlated with success in any of the specifications, though per capita income was positively correlated with success in all four. That is, the richer the country the greater the likelihood of success in removing a trade barrier. One possible explanation for this is that richer developed countries had more significant obligations under the GATT during the sample period, and GDP per capita may be acting as a proxy for the commitment to liberal trading practices and the existence of GATT remedies for disputes.

Implicit trade barriers were never correlated with success in removing barriers, and explicit barriers were marginally negatively correlated in two cases.

Do Elections Matter?

One strand of the theoretical literature on endogenous policy formation (e.g. Magee, Brock, and Young) models trade policy formation in a democracy as a process in which competing parties announce their positions, industry lobbies make donations on the basis of their internal calculation of gain, and the parties use the donations to sway imperfectly informed voters. In another strand of the literature (e.g. Grossman and Helpman) the role of direct competition and elections is discounted and instead the incumbent political party selects from a menu of detailed policy choices to maximize political support among lobbies. Grossman and Helpman are able to derive an equilibrium structure of protection as a function of the state of industry organization, trade dependency, and the elasticity of import demand or export supply. These models imply significantly different things about trade policy. The Magee-Brock-Young view of the world implies that trade policy may change dramatically with changes in government. The Grossman-Helpman model suggests that political capture gives rise to an unchanging (or at least slowly changing) equilibrium trade policy. Moreover, as noted earlier, there is some evidence that the determinants of trade conflict may have changed over time. The obvious question is whether changes in Administration mark regime changes, or are the estimated coefficients stable across presidencies?²⁰

Table 11 summarizes statistical tests of coefficient stability for the regressions reported in Tables 4 and 5. (The relatively small sample size and way the success is defined precludes similar regime change tests for the Table 10 results.) Again, the results are mixed. In the case of the Table 5 regressions on ACTION, the Bush Administration appears to be indistinguishable from both the Reagan Administration which preceded it, and the Clinton Administration which followed. However with respect to the ATTENTION regressions reported in Table 4, the hypothesis of coefficient stability between the Bush and Reagan years is decisively rejected. In the test of Bush against Clinton, the hypothesis of no regime change could not be rejected. There are obviously multiple interpretations of these results. One is that while the pattern of formal trade actions has remained stable for more than a decade, the more informal process proxied by ATTENTION changed significantly in the late 1980s. Another possibility is that it has changed again under the Clinton Administration but the relative lack of observations for the Clinton subsample means the power of the second set of tests may be low. When the requisite data is available it would obviously be desirable to extend the panel for subsequent years of the Clinton Administration as a check on the robustness of these results.

Conclusions

Trade policy is an important public policy affecting everyone in a society. In the case of the US, trade policy is the outcome of complex private multiparty negotiations, much of which occurs beyond public scrutiny. The result is that it is difficult to divine the priorities and motivations of policy makers in this important arena. This paper has gone beyond the previously used measures of observable outcomes to model the determinants of US bilateral export and investment policy attention, action, and success.

The evidence suggests that bilateral trade imbalances get the attention of policy makers, but that formal actions are more closely correlated with the existence of formal barriers to trade. There is some evidence that Japan is the target of a greater than expected number of trade actions. (At the same time the null hypothesis that with respect to Japan these actions were inefficacious could not be rejected.) The question of why bilateral imbalances would be so strongly correlated with attention is unclear. Certainly such surpluses can act as a political lightning rod, and the trade policy makers may go after surplus countries under political pressure. Yet in the absence of a compelling case why such balances are economically meaningful, this response should be viewed as a failure on the part of economic policy makers. So too, the emphasis on trade barriers as a cause of bilateral imbalances: if policymakers want to reduce trade imbalances, macroeconomic policy would appear more efficacious. Indeed, the emphasis on trade barriers as the cause of bilateral imbalances and the subsequent failure to achieve the mistakenly expected results could drive policy makers to adopt increasingly undesirable policies.

This paper has also analyzed the determinants of success in eliminating partner country barriers. With respect to different kinds of barriers, there is some evidence that success is positively associated with the extent of international norms against the disputed practice. With respect to countries, success is positively related to the degree of partner country dependence on the US market as well as a variety of other factors.

However, it does not appear that these bilateral trade policies have a significant distinct impact on the level of exports or investment, or the growth of export and investments. This is not to say that US bilateralism is totally ineffective-it could be that US pressure changes behavior in particular sectors, but these effects are simply too small or too random for the econometric model to capture. The model also implicitly takes target country behavior as given; it could be that in the absence of such behavior by the US partner country behavior might be vastly different leading to significantly different outcomes. A more sophisticated model which would treat partner country behavior as endogenous would obviously be a significant step forward.

A final issue is whether elections matter to trade policy. While it might appear obvious that they do, some have argued that incumbents and lobbyists will behave the same way regardless of which party is in power yielding a unique equilibrium set of policies. This proposition was tested formally, by examining US trade policy over the period 1984-1993, and the preponderance of evidence suggests that behavior is invariant across Administrations. Statistically speaking, the behavior of the first Clinton Administration was indistinguishable from that of its predecessor.

Appendix

Data on tariffs were provided by the USTR. Data for some countries was missing and the USTR data was supplemented by data from the WTO. Data on exports, investment, the trade balance, and attention are from USTR, National Trade Estimate Report on Foreign Trade Barriers, various issues. Data on GDP, GDP per capita, and GDP growth are from the World Bank, World Development Report, various issues. Data on trade actions were compiled from USTR, Trade Policy Agenda and Annual Report, various issues. The trade data used to compute the intraindustry trade index are from the U.S. Department of Commerce, International Trade Administration, U.S. Foreign Trade Highlights, various issues. Distance is defined as the great circle distance between Chicago and partner country economic centers as listed in Frankel (1993).

The gravity regressions in Table 2 were estimated using generalized instrumental variables estimator: the endogenous variables were regressed on the instruments, and the regressions estimated using the constructed values of the endogenous variables and corrected for heteroscedasticity. The two equations were then jointly re-estimated. The instruments were log population, log distance, an English language dummy, a landlocked country dummy, and a common border dummy.

The sample for these regressions (and the subsequent analysis) consists of the United States' 50 largest trade partners, less oil-exporting countries, leaving a sample of 37 countries, accounting for approximately 90 percent of US trade and investment: Argentina, Australia, Austria, Brazil, Canada, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Egypt, European Union, Finland, Guatemala, Hong Kong, India, Indonesia, Israel, Jamaica, Japan, Korea, Malaysia, Mexico, New Zealand, Norway, Philippines, Poland, Russia, Singapore, South Africa, Sweden, Switzerland, Taiwan, Thailand, Turkey, and Venezuela. In some parts of the analysis Russia is dropped due to missing data. The time period for these regressions is 1984-1993. It would be possible to add 1994 to the sample, but changes in the composition of the EU together with the relative sizes of the cross-section and time-series dimensions of the panel means that more information would be lost (through losing independent observations for the new entrants) than gained (by adding one additional sample year).

The regressions in Table 4 were estimated using a conventional fixed-effects panel estimator. The regressions in Table 5 are probits. As indicated in footnote 18 the regressions in Table 10 are OLS; results from MLE regressions were not robust to minor changes in search algorithm parameters.

References

Balassa, Bela. 1967. Trade Liberalization Among Industrial Countries. New York: McGraw-Hill.

Baldwin, Robert E. 1985. The Political Economy of U.S. Import Policy. Cambridge: MIT Press.

Bayard, Thomas O. and Kimberly Ann Elliott. 1994. Reciprocity and Retaliation in U.S. Trade Policy, Washington: Institute for International Economics.

Bhagwati, Jagdish. 1982. "Directly Unproductive, Profit-Seeking DUP Activities," Journal of Political Economy, 90, 988-1002.

Destler, I.M. 1995. American Trade Politics, third edition. Washington: Institute for International Economics.

Destler, I.M. and John S. Odell. 1987. "Anti Protection: Changing Forces in United States Trade Politics," Policy Analyses in International Economics Number 21, Washington: Institute for International Economics.

Elliott, Kimberly Ann and J. David Richardson. 1996. "Determinants and Effectiveness of "Aggressively Unilateral" Trade Actions," in Robert C. Feenstra, The Effect of U.S. Trade Protection and Promotion Policies, Chicago: University of Chicago Press, forthcoming.

Frankel, Jeffrey A. 1993. "Is Japan Creating A Yen Bloc in East Asia and the Pacific," in Jeffrey A. Frankel and Miles Kahler editors, Regionalism and Rivalry, Chicago: University of Chicago Press.

Grossman, Gene M. and Elhanan Helpman. 1994. "Protection for Sale," American Economic Review, 84, 833-850.

Krueger, Anne O. 1974. "The Political Economy of a Rent-Seeking Society," American Economic Review, 64, 291-303.

Lavergne, Real P. 1983. The Political Economy of U.S. Tariffs: An Empirical Analysis. Toronto: Academic Press.

Magee, Stephen P. William A. Brock, and Leslie Young. 1989. Black Hole Tariffs and Endogenous Policy Theory. Cambridge: Cambridge University Press.

McMillan, John. 1990. "Strategic Bargaining and Section 301," Politics and Economics, 2 45-58.

Noland, Marcus. 1996. "Trade, Investment, and Economic Conflict Between the US and Asia," Journal of Asian Economics, 7:3 435-458.

Noland, Marcus. 1997. "Public Policy, Private Preferences, and the Japanese Trade Pattern," Review of Economics and Statistics, forthcoming.

Rodrik, Dani. 1994. "The Rush to Free Trade in the Developing World: Why So Late? Why Now? Will it Last?" in Stephan Haggard and Steven B. Webb editors, Voting for Reform, Oxford: Oxford University Press.

Trefler, David. 1993. "Trade Liberalization and the Theory of Endogenous Protection: An Econometric Study of U.S. Import Policy," Journal of Political Economy, 101, 138-160. USTR, 1994. National Trade Estimate Report on Foreign Trade Barriers, Washington: Office of the United States Trade Representative.

Notes

1. This approach, while intuitively appealing, has what appears to be an insoluble sample selection problem, namely that only cases in which observable actions were undertaken are included in the sample. There is no way to include the cases that did not result in observable actions, indeed there is no way to define the relevant universe of possible cases. The approach adopted in this paper avoids the sample selection problem inherent in the Bayard and Elliott and Elliott and Richardson approach at the cost of loss of industry specificity.

2. Note that the US government took no formal actions in 1984, hence no listing for that year in Table 1.

3. An example of the former would be when the Bush Administration successfully dissuaded the US Rice Millers Association from filing a Section 301 complaint against Japanese rice barriers tending the conclusion of the Uruguay Round of multilateral trade negotiations. USTR officials frequently discourage industries with weak cases from pressing their cases formally, and occasionally decline to accept formal petitions.

4. It is unfortunate that the report was only published beginning in 1984, effectively precluding an examination of the possible trade policy regime change in 1985 as argued by Destler (1995) and Odell and Destler (1987).

5. It should be noted that this compendium of barriers is not the product of any formal economic analysis. Rather, it is the product of subjective appraisals by bureaucrats and industry lobbyist submissions.

6. See the appendix for a description of data sources and estimation techniques.

7. All of the subsequent analysis was also carried out using simple, instead of trade-weighted, tariffs. The use of this alternative made no qualitative difference (and scarcely any quantitative difference).

8. It would be desirable to have formal measures of NTBs. UNCTAD began compiling data on industrial country NTBs to support the developing country negotiators in the Uruguay Round, but this data is available only for a limited number of countries and not useful in this application.

9. See Noland (1997) for an example and additional references.

10. These residuals take both positive and negative values. The statistical analysis was also done setting residuals with positive values equal to zero. Truncating the series in this way reduced the correlations to insignificance in all applications.

11. The binary variable ACTION takes the value 1 if a formal action was taken against a particular country in a particular year and 0 otherwise.

12. In theory, the coefficients GDP, GDPGROWTH, EXPORTS, and FDI in the regression with the implicit trade barrier variables represent only the direct partial though not total impact of these variables on ATTENTION, since the trade and investment residual variables XRESID and INVRESID implicitly contain elements of partner GDP, GDPGROWTH, EXPORTS, and FDI. Algebraically one can derive the total impact of these variables by back-substituting the Table 2 specification. In practice, since the coefficients on XRESID and INVRESID are statistically not different from zero these indirect effects can safely be ignored.

13. Again, this contrasts with Noland (1996) where the trade balance is a significant explanator of ACTION. The possibility of regime shift is taken up in a subsequent section.

14. It might also be argued that the gravity model that underlies the proxies for implicit trade and investment barriers is too crude to measure these accurately.

15. In the cases where USTR had no successes, the test statistic is reported as 0.0, though strictly speaking it is undefined. Cases in which particular policies were not at issue are labelled not applicable.

16. In the cases of Switzerland and the 1984 Finland subsample there were only successes so the conditional variance is undefined. In this case the value of test statistic has been listed as infinite.

17. As in the case of Switzerland in the previous table, all countertrade cases were successfully resolved.

18. The regressions were also estimated with regional dummy variables (including a dummy for Japan, which according to Table 5, has been "singled out" for trade actions) included among the regressors but these variables were never significant and these regressions are not reported for the sake of brevity.

19. These are OLS regressions. In theory the estimated parameters may be biased due to the truncation of the dependent variable at 0 and 1. Maximum likelihood regressions were estimated in an attempt to address this issue, but yielded parameter estimates that were not robust to changes in estimation algorithms.

20. Regressions were also estimated with a dummy variable for election years to test whether trade conflict was linked to elections. The estimated coefficients results were either insignificant or negative (less attention in election years).