I. Introduction

Over the last two decades, new trade restrictions have been introduced in many industrial countries principally through administrative channels. These include antidumping duties (AD), countervailing duties (CVD), voluntary export restraint agreements (VERs) ^1/, and so-called “safeguard” actions. ^2/ This development reflects the constraints placed on the use of quantitative restrictions and tariff barriers under GATT (now World Trade Organization (WTO)) rules. Under legislated protection, it is well-established that protectionist pressures tend to build under macroeconomic downturns. ^3/ Antidumping and countervail procedures are governed by specific rules requiring both injury (by reason of imports) and dumping/ subsidies, and, thus, might be expected to be different; to be less susceptible to pressures for protection from foreign competition arising from cyclical movements in a country’s economy. The purpose of this study is to investigate whether there is a connection between the state of domestic macroeconomic activity and pressures for protection under AD/CVD, in order to better understand the ebb and flow of these actions over time. Because of the ready availability of data on AD/CVD petitions and the body of empirical work done on similar questions, data from the United States is used. Of course, the kind of relationship tested for here need not be specific to the United States, and this analysis could be usefully applied to other countries that have been frequent users AD and/or CVD.

II. Use of Administered Trade Actions

In a frequently cited study, Takacs (1981) investigated the extent to which pressures for administered protection in the United States, as measured by the number of yearly petitions for escape-clause (also called safeguard) protection, responded to developments in the macroeconomic cycle. Her empirical results indicated that the pressure for administered protection was inversely related to the level of macroeconomic activity (as measured by the unemployment rate or the rate of industrial capacity utilization). Takacs’ focus on escape-clause protection was not apparently determined a priori, as she observed that “preliminary empirical tests using antidumping and countervailing duty cases in addition to escape clause cases failed to provide significant results” (p. 688). ^1/ Coughlin et al, (1989) corroborated the results of Takacs’ study using a number of alternative model specifications. Salvatore (1987), using a simultaneous equations model, found that macroeconomic activity, as measured by the level of real GNP, affected escape clause petitions, but he did not find a significant relationship when alternative indicators of economic activity (the unemployment rate and capacity utilization) were used. Magee and Young (1987) examined the proportion of yearly antidumping petitions that received an affirmative ruling during the years 1954-1981 and found evidence that antidumping protection was more likely to be granted following a petition during macroeconomic downturns.

Under the Trade Agreements Act of 1979 (which implemented the results of the Tokyo Round of multilateral trade negotiations), a number of significant changes were introduced affecting the way AD/CVD cases are handled in the United States. ^2/ Subsequently, AD/CVD petitions and actions increased significantly in the 1980s. ^3/ At the same time, safeguard petitions and actions waned, suggesting a high degree of substitutability between AD/CVD and safeguards. ^4/ These developments suggest there is no reason to expect the absence of a relationship expressed by Takacs based on earlier data would also describe the post-Tokyo Round period.

Unlike escape-clause protection, antidumping and countervail are intended to target so-called “unfair” trading practices: either “dumping” or subsidization. Thus, in addition to the injury criterion, ^1/ which is also necessary to obtain protection under the escape clause, antidumping and countervail also require a demonstration of dumping and subsidization, respectively. ^2/ As suggested by Finger and Murray (1993, p. 241 and p. 247), however, when AD/CVD petitions fail to result in a final duty in the United States, it is generally because the injury test is negative. ^3/ This suggests that given sufficient flexibility in the way in which dumping margins ^4/ and subsidies are calculated, AD/CVD cases can be considered roughly comparable to safeguard actions, in the sense that the outcome effectively hinges on the injury decision. In such circumstances, the relationship between macroeconomic fluctuations and pressures for escape-clause protection identified by Takacs could be expected to carry over to antidumping and countervail petitions. However, care has to be taken in extending this analysis to AD/CVD cases and in interpreting the results. It is possible that inasmuch as there is a degree of synchronization in the timing of downturns between the domestic and foreign economies, such a relationship could conceivably reflect a rise in the propensity of foreign producers to actually “dump” goods, or of foreign governments to grant new subsidies.

III. Are Antidumping and Countervail Economically Significant?

The USITC (1995) recently published a staff study examining the net social cost of antidumping and countervail actions in the United States. Using a static computable general equilibrium model, based on 163 AD and 76 CVD orders, it was estimated that the net welfare loss (equivalent variation basis) in 1991 U.S. dollars was $1.59 billion (0.03 percent of 1991 GDP). This number is small and may cast some doubt on the economic significance of the measured sensitivity of AD/CVD petitions to the U.S. business cycle. However, the USITC’s estimate did not include “170 orders that were revoked, 9 suspended and 37 terminated investigations, and 41 orders in which subject imports stopped completely after their imposition.” It did not include the effects of voluntary restraint agreements negotiated under actual or threatened AD/CVD petitions, and it did not attempt to measure other threat effects that might accompany these instruments. ^1/ The remainder of this section briefly reviews a number of models which suggest that the trade restraint induced by AD/CVD is likely to go beyond that suggested by the observable AD/CVD duties imposed. Thus the economic significance of the measured sensitivity of AD/CVD to the macroeconomic cycle must be assessed with these less transparent effects in mind.

The theoretical literature on the firm-level incentive effects of an AD/CVD regime identifies a number of channels through which export restraint might occur in the absence of AD/CVD duties. A complete review of this literature is beyond the scope of this paper, but some examples will help to flesh out these trade-inhibiting effects. According to this literature, the prospect of protection under the rules governing antidumping and countervail may alter the conditions of competition in a way that can induce restraint of trade.

Staiger and Wolak (1994), for example, using a duopoly model under stochastic demand conditions show that the threat of AD actions, made credible by occasional AD petitions (but not necessarily actual AD duties), could help support successful collusion and greater market share for domestic firms by providing a means of punishing incipient defections from a tacitly collusive agreement. Indeed, the Staiger and Wolak model shows specifically that a domestic import-competing firm will never petition for antidumping in a “high demand” state. A domestic firm has an incentive to file an AD petition in “low demand” states in order to credibly inform a foreign competitor of his intention to punish a defection. This is sufficient to prompt the foreign competitor to compete less aggressively than otherwise, and this effect occurs “in spite of the fact that AD duties are never actually imposed” (p.252). Based on related empirical work at the industry level, Staiger and Wolak conclude that “a petitioning firm can expect to receive roughly half the import relief during the period of investigation that it would have received if AD duties had been imposed from the date of filing. (p.257)”

Prusa (1994) and Leidy (1994a) investigated the change in the incentive structure induced by an antidumping/countervail regime in a duopoly setting. Using a model of Bertrand competition ^1/, Prusa showed that the introduction of an AD regime, and the attendant sanctions against dumping margins, had the effect of imposing a binding price floor above the pre-regime equilibrium price. Leidy examined the effect of introducing an antidumping and/or countervail regime using a model of Cournot competition in which the criteria for establishing material injury affect the output decisions of both import-competing and foreign exporting firms. ^2/ Leidy showed that the incentives facing foreign and domestic firms to manage the prospect/ threat of antidumping and/or countervail would tend to produce restraint of trade and production that moved in the direction of the industry-wide cartel solution. Moreover, the cartel-like equilibrium was shown to be stable, in the sense that it was free of the usual incentives to defect.

Prusa (1996) examined empirically the effects of AD actions (petitions and duties) on prices and exports from both named and unnamed countries. Using industry level data on AD petitions filed between 1980 and 1988, he found that whether cases were ultimately rejected or ended in final duties, prices increased on goods originating in both named and unnamed exporting countries.

These and related papers suggest that the social costs of antidumping and countervail extend well beyond what is implied by the imposition of final duties. ^3/ Thus in order to gain a sense of the economic significance of the macro-cyclical effects analyzed in this paper, petitioning activity should be viewed in terms of its possible connection to the strategic behavior of domestic import-competing firms and their foreign competitors.

IV. Empirical Tests

The measure of protectionist pressures under AD/CVD used in the analysis is the yearly number of petitions for protection as reported by the Office of the United States Trade Representative (USTR). ^4/ Because of the high degree of substitutability between AD and CVD, the sum of all AD and CVD petitions is treated as the basic dependent variable (although AD and CVD petitions were also examined separately as dependent variables and these results are presented in the appendix tables ^1/). Since the Tokyo Round of multilateral trade negotiations (the period 1980-95), there were 752 antidumping and 407 countervailing duty petitions filed in the United States. ^2/

The analysis focuses on “petitions” rather than final “actions” because focusing on final actions could understate underlying protectionist effects for several reasons. First, during the post-Tokyo Round period, AD/CVD cases were frequently withdrawn because a settlement was reached, whereby foreign producers would “voluntarily” restrain exports. ^3/ Second, the literature on the trade-inhibiting effects of AD/CVD threats discussed above strongly suggests that petitions for AD/CVD protection can be used strategically to influence the decisions of foreign exporters, and as such no duty need be imposed and no formal VER need be negotiated to induce trade-inhibiting effects. Third, the lag between filing a petition and its ultimate disposition is generally long (typically more than a year) and variable.

Two indicators of macroeconomic activity are considered in the analysis: the civilian unemployment rate and the rate of industrial capacity utilization. Many of the regression results reported below also control for exchange-rate developments by including measures of the real effective exchange rate (REER) of the dollar. ^4/ Testing for the effect of REER movements on AD/CVD petitions was deemed particularly relevant in view of an opinion expressed by the USTR that the surge in AD cases from 1981-85 was due in large part to “the dollar’s high value.” ^1/ To control for effects that might be induced by aggregate trade developments, the role of import penetration is also investigated using the ratio of imports to GDP; alternatively, the merchandise trade balance relative to GDP is included. Also, as mentioned previously, there may be a strong correlation between macroeconomic downturns in the United States and downturns among trading partners. Because such external factors could induce petitions for reasons unrelated to domestic macroeconomic conditions, results are also reported controlling for the macroeconomic conditions in major trading partner countries (as measured by their unemployment rates). ^2/ Chart 1 displays the time series for unemployment and AD/CVD petitions over the sample period. ^3/

The basic equation (estimated in row one of Table 1) posits an inverse relationship between domestic macroeconomic activity and pressures for protection under antidumping and countervail, with causality running from the general state of the economy to protectionist pressures. ^4/ Lagged petitions are also included as an explanatory variable in order to capture what might be described as a depletion effect. Specifically, a relatively high number of petitions in one year would, other things equal, partially deplete the stock of potential petitions in the following year. This is because the stock of petitioners is finite and because many of the previous year’s petitions will remain under investigation during the following year. ^1/ Augmented Dickey-Fuller tests indicated that the presence of a unit root could not be rejected for any of the variables considered, but these tests also suggested all series were stationary in first differences (Table A1). Accordingly, all equations were estimated in first differences.

Table 1.

United States: Dependent Variable: Total AD and CVD Petitions Per Year ^1/

(Annual observations, 1980-95)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

Table 1.

United States: Dependent Variable: Total AD and CVD Petitions Per Year ^1/

(Annual observations, 1980-95)

U.S.				Merchandise		Trading	Industrial
Unemployment	Capacity	REER	Import	Trade Balance	Petitions	Partner	Country 2/
Rate	Utilization	(CPI)	Penetration	Percent of GDP	Lagged	Unemployment Rate		adj-R2	BG 3/
54.1					-0.76			0.72	0.15
(4.84)**					(-4.93)**
51.7		2.2			-0.81			0.79	0.08
(5.22)**		(2.04)			(-5.84)**
	-15.8				-0.69			0.51	0.02
	(-2.90)*				(-3.38)**
	-15.2	2.5			-0.74			0.58	0.65
	(-3.00)*	(1.68)			(-3.88)**
47.9			-16.5		-0.77			0.70	0.02
(2.59)*			(-0.54)		(-4.79)**
56.4				-20.9	-0.80			0.72	0.01
(4.81)**				(-0.94)	(-4.98)**
38.3					-0.80	33.3		0.74	—
(2.28)*					(-5.19)**	(1.23)
49.1		2.04			-0.81	5.9		0.76	0.07
(2.78)*		(1.44)			(-5.54)**	(0.19)
49.1					-0.79		23.7	0.72	0.01
(4.00)**					(-5.03)**		(0.99)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

View Table

Table 1.

United States: Dependent Variable: Total AD and CVD Petitions Per Year ^1/

(Annual observations, 1980-95)

U.S.				Merchandise		Trading	Industrial
Unemployment	Capacity	REER	Import	Trade Balance	Petitions	Partner	Country 2/
Rate	Utilization	(CPI)	Penetration	Percent of GDP	Lagged	Unemployment Rate		adj-R2	BG 3/
54.1					-0.76			0.72	0.15
(4.84)**					(-4.93)**
51.7		2.2			-0.81			0.79	0.08
(5.22)**		(2.04)			(-5.84)**
	-15.8				-0.69			0.51	0.02
	(-2.90)*				(-3.38)**
	-15.2	2.5			-0.74			0.58	0.65
	(-3.00)*	(1.68)			(-3.88)**
47.9			-16.5		-0.77			0.70	0.02
(2.59)*			(-0.54)		(-4.79)**
56.4				-20.9	-0.80			0.72	0.01
(4.81)**				(-0.94)	(-4.98)**
38.3					-0.80	33.3		0.74	—
(2.28)*					(-5.19)**	(1.23)
49.1		2.04			-0.81	5.9		0.76	0.07
(2.78)*		(1.44)			(-5.54)**	(0.19)
49.1					-0.79		23.7	0.72	0.01
(4.00)**					(-5.03)**		(0.99)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

Over the period 1980-95, the estimated regressions (Table 1) indicate that the number of AD/CVD petitions responded, in a statistically significant way, to the state of domestic macroeconomic activity. Other things equal, fewer antidumping and countervailing duty cases were initiated when the unemployment rate was low and the rate of capacity utilization was high. ^2/ For example, the first equation in Table 1 indicates that a 1 percentage point increase in the unemployment rate would result in an additional 54 AD/CVD petitions in the first year (in an expected value sense), with a steady-state increase of 31 additional petitions. This is clearly not an economically trivial amount, especially given that it also probably reflects a heightened threat effect as discussed in section III.

The inclusion of the real effective exchange rate does not alter this result; the estimated coefficient is always positive as expected, but this effect is not statistically significant at the 5 percent level in any of the regressions examined in Table 1. While a real effective appreciation (depreciation) of the dollar appears to contribute to a rise (fall) in petitions, the effect of changes in macroeconomic activity persists when this effect is taken into account. It is noteworthy that, in the sample, changes in the REER appear to be associated more strongly with changes in the number of countervail petitions than with antidumping petitions (Tables A2 and A3). The statistical significance of the REER is revisited. in Table 2, and discussed below, using a sample that is purged of the June 1992 AD/CVD petitions linked to the end of the wide-ranging VERs in steel.

Table 2.

United States: Dependent Variable: Total AD and CVD Petitions Per Year, Excluding June 1992 Steel Petitions ^1/

(Annual observations, 1980-95)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

Table 2.

United States: Dependent Variable: Total AD and CVD Petitions Per Year, Excluding June 1992 Steel Petitions ^1/

(Annual observations, 1980-95)

U.S.				Merchandise		Trading	Industrial
Unemployment	Capacity	REER	Import	Trade Balance	Petitions	Partner	Country 2/
Rate	Utilization	(CPI)	Penetration	Percent of GDP	Lagged	Unemployment Rate		adj-R2	BG 3/
49.3					-0.84			0.80	2.52
(5.66)**					(-6.26)**
46.6		2.5			-0.89			0.92	0.10
(8.56)**		(4.31)**			(-10.57)**
	-15.6				-0.75			0.66	0.14
	(-3.84)**				(-4.45)**
	-14.9	2.8			-0.81			0.80	2.47
	(-4.78)**	(2.97)*			(-6.22)**
37.3			-32.1		-0.86			0.80	2.50
(2.73)*			(-1.14)		(-6.43)**
53.8				-21.3	-0.89			0.81	0.54
(5.86)**				(-1.27)	(-6.54)**
32.6					-0.87	34.8		0.83	0.14
(2.68)*					(-7.57)**	(1.81)
45.9		2.48			-0.89	5.9		0.91	0.09
(4.75)**		(3.22)*			(-10.02)**	(0.19)
44.6					-0.87		22.2	0.81	1.17
(4.80)**					(-6.53)**		(1.24)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

View Table

Table 2.

United States: Dependent Variable: Total AD and CVD Petitions Per Year, Excluding June 1992 Steel Petitions ^1/

(Annual observations, 1980-95)

U.S.				Merchandise		Trading	Industrial
Unemployment	Capacity	REER	Import	Trade Balance	Petitions	Partner	Country 2/
Rate	Utilization	(CPI)	Penetration	Percent of GDP	Lagged	Unemployment Rate		adj-R2	BG 3/
49.3					-0.84			0.80	2.52
(5.66)**					(-6.26)**
46.6		2.5			-0.89			0.92	0.10
(8.56)**		(4.31)**			(-10.57)**
	-15.6				-0.75			0.66	0.14
	(-3.84)**				(-4.45)**
	-14.9	2.8			-0.81			0.80	2.47
	(-4.78)**	(2.97)*			(-6.22)**
37.3			-32.1		-0.86			0.80	2.50
(2.73)*			(-1.14)		(-6.43)**
53.8				-21.3	-0.89			0.81	0.54
(5.86)**				(-1.27)	(-6.54)**
32.6					-0.87	34.8		0.83	0.14
(2.68)*					(-7.57)**	(1.81)
45.9		2.48			-0.89	5.9		0.91	0.09
(4.75)**		(3.22)*			(-10.02)**	(0.19)
44.6					-0.87		22.2	0.81	1.17
(4.80)**					(-6.53)**		(1.24)

^1/OLS regressions performed on first-differences; a constant term was included. In no case was the constant term significantly different from zero. The t-statistics are in parenthesis below the estimated coefficients. Coefficients significantly different from zero at the 1 percent level are indicated by (^**), and at the 5 percent level by (^*), using a two-tailed test.

^2/Excluding the United States.

^3/The Breusch-Godfrey statistic tests for the presence of autocorrelation when there is a lagged dependent variable in the regression equation. The tests performed here used one lagged residual. The BG test statistic with one lagged residual is asymptotically distributed as chi-squared with 1 degree of freedom. In none of the cases examined here is the null hypothesis (H₀: errors are serially uncorrelated) rejected.

The alternative indicator of overall macroeconomic activity (capacity utilization) yields essentially similar results. In the basic equation (Table 1, regression 3), a 1 percentage point increase in the rate of utilization is associated with about 16 fewer AD/CVD petitions in the first year, other things equal, and a steady-state decline of about nine petitions.

When the ratio of imports to GDP (import penetration in percent change, exclusive of military goods) is included in the standard model, the data continue to suggest a significant role for the unemployment rate in determining the number of AD/CVD petitions, despite a high degree of collinearity between import penetration and the unemployment rate. ^1/ High (low) import-to-GDP ratios are strongly associated with high (low) levels of domestic aggregate economic activity. The estimated coefficient for import penetration is negative, indicating that higher levels of import penetration are associated with fewer AD/CVD petitions, but the estimated effect is not statistically significant. The negative sign is the opposite of what would be expected if domestic macroeconomic downturns tended to be associated with an across-the-board surge in dumped and subsidized imports. ^2/

Because macroeconomic downturns (upturns) at home may be associated with downturns (upturns) in trading partner countries and because downturns could be associated with attempts to increase foreign market share through “dumping” and/or new subsidies, several reported regressions attempt to control for the effects of macroeconomic conditions in trading partner economies. ^3/ When either a trade-weighted average foreign unemployment rate or the average unemployment rate in industrial countries excluding the United States were included as regressors along with the U.S. unemployment rate, the foreign indicator was not statistically significant. The coefficient on the U.S. unemployment rate, however, remained significant at the 5 percent level.

Turning to the “depletion effect”, it is clear that the previous year’s petitions significantly affect the expected value of petitions in the current year. This result probably reflects not only the depletion of a finite stock of petitioners—as described above, but also the “safety valve” nature of AD/CVD petitions. As general pressures for protection build, those pressures are vented by AD/CVD petitions which, in turn, implies reduced pressures in the subsequent period. Conversely, if little protectionist pressure was vented in the previous period it is more likely than otherwise to build and to be released in the current period. For forecasting purposes, the depletion effect would appear to be quite useful.

Events in world steel trade raise a qualifying note for interpreting these empirical results. The two peak episodes in AD/CVD petitioning activity occurred in 1982 and 1992. Both years had a particularly high number of petitions from producers of various steel products. ^1/ These two years also corresponded to local peaks in the U.S. average yearly unemployment rate. The rise in steel AD/CVD petitions in 1992, in particular, is certainly attributable to some extent to the expiration at end-March 1992 of the far-reaching multilateral VERs in steel; although the macroeconomic downturn almost certainly added additional impetus to these petitions. Thus, it is unclear to what extent events specific to developments in steel trade might be distorting the estimated effects over all AD/CVD petitions.

In order to clarify the extent to which the end of the system of steel VERs might be unduly affecting the regression results, the regression equations reported in Table 1 were re-estimated using a data set that excluded every steel-related AD/CVD petition that occurred in the month of June of 1992—the month during which steel AD/CVD petitions surged dramatically following the expiration of the VER agreements. ^2/ These results are presented in Table 2.

Several aspects of the results are noteworthy. First, qualitatively the influence of the macroeconomic cycle on total AD/CVD petitions remains unchanged. Second, the influence of the real effective exchange rate is now more firmly supported by the data. Third, each of the regressions reported in Table 2 have an adjusted R² statistic that is higher than those reported in Table 1. Thus, the essential results are clearly robust to the elimination of the special circumstances surrounding the end of the steel VERs in 1992. The first equation in Table 2 suggests that a 1 percentage point increase in the unemployment rate would result in an additional 49 AD/CVD petitions in the first year (in an expected value sense), with a steady-state increase of 27 additional petitions. After controlling for changes in the real effective exchange rate, a 1 percentage point increase in the unemployment rate would result in 47 AD/CVD petitions in the first year, with a steady-state increase of 25 additional petitions.

V. Concluding Comments

The evidence presented here suggests that pressures for protection in the United States under AD/CVD since the Tokyo Round of multilateral trade talks have heightened during macroeconomic downturns and receded with higher levels of resource utilization. This relationship is robust to a variety of specifications of the basic model, including those controlling for real exchange rate movements, indicators of trading-partner economic activity, the state of the U.S. trade balance, and the degree of aggregate import penetration. When the data are purged of the happenstance timing of the end of the steel VERs and the 1992 macroeconomic downturn the data continue to offer evidence of cyclically-induced protectionist pressures. Moreover, the estimates presented here of the sensitivity of AD/CVD petitions to changes in the unemployment rate and industrial capacity utilization, especially given the link between AD/CVD and induced restraint of trade independent of actual duties being imposed, would appear to be highly significant from an economic perspective.

The relatively low level of AD/CVD petitions in the United States during 1995 (see Table A4) is consistent with the robust macroeconomic performance during that year, with the average monthly unemployment rate falling from 6.1 percent in 1994 to 5.6 percent in 1995. Is there reason to suspect that the relatively low level of petitions in 1995 might signal the occurrence of a structural change? It is, of course, impossible to say at this stage whether the implementation of Uruguay Round reforms in January 1995 might alter the cyclical relationship identified here. However, the year 1995 was by no means an outlier; the fitted value using the basic equation from Table 1 shows a decline of 21 AD/CVD petitions in 1995 compared to an actual decline of 29, and the residual for 1995 is among the smallest in the sample. Moreover, while a review of Uruguay Round reforms is beyond the scope of this paper, others have concluded that the reforms would imply little substantive change in the use of these measures. ^1/

It is important to note that nothing in this analysis suggests that the U.S. application of AD/CVD measures is, or has been, inconsistent with GATT/WTO rules. Rather, the implication is that despite conformity with multilateral rules, protectionist pressures under the rules for administered protection still ebb and flow with the state of the macroeconomy. It is quite possible that similar results would be found if this type of analysis were to be extended to other countries that, like the United States, have undertaken extensive liberalization commitments in the GATT/WTO and thus have induced a reorientation of protection-seeking away from legislative and toward administrative channels.