Effects of U.S. trade remedy law enforcement under uncertainty: the case of steel.

1. Introduction

The history of trade protection for the U.S. steel industry is long and complex. The period of protection for steel is customarily broken down into three phases: 1969-74, protection principally by voluntary export restraint (VER); 1978-82, protection by a trigger price mechanism; and 1982 to the recent past, protection by both VER and trade remedy policies.(1)

Among the various trade remedy provisions, sections 701 (countervailing duty: CVD) and 731 (antidumping duty: AD) are the ones most frequently invoked in connection with steel imports.(2) These provisions are implemented through quasi-judicial processes conducted by the International Trade Commission (ITC) in collaboration with the International Trade Administration (ITA). Unlike trade policies in the past, which were implemented fairly systematically (through tariffs and VERs) in response to changes in the trade balance, the entire judicial process for CVD and AD cases begins with petitions filed randomly by firms, industries, and unions. American trade policy in the third phase therefore consists of both a systematic and an unsystematic component. This requires a researcher to incorporate uncertainty into any analytical model related to CVD and AD policies.

There have been numerous studies of U.S. trade remedy policies.(3) Most of them are concerned with the argument on "rules or discretion" for the ITC's decision-making criterion. References on the effectiveness of trade remedy policy are limited. An interesting study is that of Herander and Schwartz (1984), who suggest that foreign dumping margins were sensitive to U.S. threats by means of the AD policy during the period 1976-86.(4) However, it is questionable whether their model would be suitable to explain the cases in the third phase because they ignored the uncertainty dimension in their analyses.

The purpose of this paper is to explore the effectiveness of U.S. trade remedy law enforcement by sections 701 and 731 on the import penetration of foreign-manufactured steel into the U.S. market for the period 1983-92. I analyze this problem within the context of uncertainty. In addition, only three industries (food, chemicals, and steel) have been chiefly subject to sections 701 and/or 731 during the period of observation. Therefore, an industry-specific analysis is more appropriate. Given the short period of active implementation of the two provisions, it is necessary to specify the model as simply as possible. To meet both methodological needs, I separate the effect that is explained mainly by unsystematic policies from the total effect of both systematic (tariffs and VERs) and unsystematic (CVD and AD) policies on the steel industry. This generates the probability-augmented model to be used as the theoretical basis for the research. The model explains foreign steel import penetration into the U.S. steel market primarily in terms of unanticipated trade policy and expected import penetration.

Estimated results suggest that the CVD and AD policies were effective in protecting the U.S. steel industry, but the degree of effectiveness was small. The results provide valuable insights into the effectiveness of trade remedy policy. The low degree of effectiveness may imply a limit to administrative/legislative solutions.

2. The Theoretical Model

To derive a simplified model, I employ the rational expectations hypothesis and the innovation accounting and vector autoregressive methods.(5) I first lay out a model for bidirectional causation between import penetration and the terms of trade, each of which is influenced by trade policy instruments and various lagged variables. In essence, its underlying microeconomic rationale is the relationship between price and quantity.

The model is

[M.sub.t] + a[P.sub.t] = [summation over i] [b.sub.i][M.sub.t-i] + [summation over i] [c.sub.i][P.sub.t-1] + d[Z.sub.t] + [u.sub.t] (1)

[P.sub.t] + a[prime][M.sub.t] = [summation over i] [b[prime].sub.i][M.sub.t-i] + [summation over i] [c[prime].sub.i][P.sub.t-i] + d[prime][Z.sub.t] + [u[prime].sub.t] (2)

(i = 1, 2, ...)

where M = foreign import penetration into the domestic steel market, P = terms of trade, Z = trade policy, and u and u[prime] = random errors.

I assume that trade policy Z consists of both systematic and unsystematic components and that the policy authority uses the following policy rule:

[Z.sub.t] = g[G.sub.t-i] + [[Zeta].sub.t] (3)

where [G.sub.t-i] = the systematic trade policy vector at time t - i such as predetermined tariffs and import quotas (or VER in case of steel), and [[Zeta].sub.t] = the random component of trade policy at time t, such as CVD and AD policies implemented through quasi-judicial processes.

The reduced form of Equations 1 and 2 yields the following equation:

[M.sub.t] = [[Sigma].sub.i][[Alpha].sub.i][M.sub.t-i] + [[Sigma].sub.i][[Beta].sub.i][P.sub.t-i] + [Gamma][Z.sub.t] + [[Epsilon].sub.t] (4)

where [[Alpha].sub.i] = ([b.sub.i] - a[b[prime].sub.i])/(1 - aa[prime]), [[Beta].sub.i] = ([c.sub.i] - a[c[prime].sub.i])/(1 - aa[prime]), [Gamma] = (d - ad[prime])/(1 - aa[prime]), and [[Epsilon].sub.t] = ([u.sub.t] - a[u[prime].sub.t]/(1 - aa[prime]).(6)

I take expectations throughout Equations 3 and 4 to have

E([M.sub.t]) = [[Sigma].sub.i][[Alpha].sub.i][M.sub.t-i] + [[Sigma].sub.i][[Beta].sub.i][P.sub.t-i] + [Gamma]E([Z.sub.t]) (5)

and

E([Z.sub.t]) = g[G.sub.t-1] (6)

where E = expectations operator; E([M.sub.t]), E([u.sub.t]), and E([u[prime].sub.t]) are, respectively, conditional on information available at time t - 1 in accordance with the rational expectations theory; that is, E([M.sub.t]) = E([M.sub.t][where][I.sub.t-1]); E([u.sub.t][where][I.sub.t-1]) = E([u[prime].sub.t][where][I.sub.t-1]) = 0 and thus E([[Epsilon].sub.t]) = 0; and thus E([[Zeta].sub.t][where][I.sub.t-1]) = 0 with some explicit adjustment for standardization so that the standardized random variable has zero mean and unit variance; that is, [Zeta] [similar to] N(0, 1). I now subtract each side of the expected version from the corresponding side of the original version and combine two resulting equations to obtain

[M.sub.t] - E([M.sub.t]) = [Gamma][[Z.sub.t] - E([Z.sub.t])] + [[Epsilon].sub.t] = [Gamma][[Zeta].sub.t] + [[Epsilon].sub.t]. (7)

In the rational expectations context, a forecast error that corresponds to the random component of trade policy in Equation 3 is

[[Zeta].sub.t] = [[Pi].sub.t] - E ([[Pi].sub.t][where][I.sub.t-1]), (8)

where [[Pi].sub.t] = the ITC's affirmative determination, which is a standardized random variable; that is, [Pi] [similar to] N(0, 1) and thus E([[Pi].sub.t][where][I.sub.t-1] = 0. I view a chance for the ITC to make an affirmative decision as an ex ante probability for the outcome. Given the rational expectations, however, the ex ante probability distribution of future outcome conditional on the available information has a connection with the ex post probability distribution of actual outcome conditional on that information.(7) Equation 7 is then rewritten in terms of ex post probability:

[M.sub.t] - E([M.sub.t]) = [Gamma][[Pi].sub.t] + [[Epsilon].sub.t]. (9)

For a simple explanatory model, the terms included in information set [I.sub.t-1] for E([M.sub.t]) are limited to [M.sub.t-1] and [P.sub.t-1], two fundamental variables in international trade that comprehensively represent any directly and indirectly relevant factors, such as trade balance, foreign exchange rates, profit, unemployment, capacity utilization, and so on, at t - 1. The final equation is

[M.sub.t] = [Gamma][[Pi].sub.t] + f([X.sub.t-1]) + [[Epsilon].sub.t] (10)

where [X.sub.t-1] = ([M.sub.t-1], [P.sub.t-1]).

Note that the two lagged terms play two different roles: lag processes (a lag process with greatest weight in the most recent past in the case of [M.sub.t-1]) and expectations (accumulation of information). It is reasonable to regard them as integral components in the rational expectations framework.

3. Methodology

In this section, I state the definitions of variables and the sources of data.

Definitions

(i) Import penetration (M) = (consumption of imported steel)/(consumption of domestically produced steel + consumption of imported steel), where consumption of domestically produced steel = domestically produced steel - exports of domestic steel. This definition is similar to that used by the U.S. Bureau of Labor Statistics.

(ii) Probability ([Pi]) = probability of the ITC's final injury determination in connection with Title VII, Subtitle A (section 701: Prob701) or Subtitle B (section 731:Prob731), which is the measure of trade remedy enforcement. A petition filed by any interested party with the ITA or the ITC against a foreign subsidy or dumping, simple or complicated, is subject to the ITC's final determination, affirmative or negative, to be made through the four-stage statutory process, which includes three stages for the ITA's preliminary determination. The statutory time limits to complete the entire judicial process for sections 701 and 731 are long. For section 701, the process requires 205 days to complete a simple case and 300 days for a complex case. The time limit for section 731 ranges from 280 to 420 days. On the average, a typical case encompasses two years. I have therefore taken two-year moving averages of the total number of petitions, withdrawals, and ITC affirmative determinations for both steel and metal manufactures. I then computed the probability in each case. The formula is (the average number of ITC final injury determinations)/(the average number of petitions - the average number of withdrawals).(8) Since the judicial processes for the two provisions are basically the same except for slight differences in statutory time limits, I also calculate the overall probability for sections 701 and 731 combined to measure their joint effect. This approach also resolves the econometric problem of multicollinearity.

(iii) Expected import penetration (E([M.sub.t])) is E([M.sub.t][where][I.sub.t-1]), where the information set includes [TABULAR DATA FOR TABLE 1 OMITTED] [M.sub.t-1] and [P.sub.t-1], two basic variables in international trade that collectively represent other relevant variables mentioned in section 2.

(iv) Terms of trade (P) = price of domestic steel (taken from producer price index series)/price of imported steel.

Data

Table 1 lists the variables and sources of data for empirical study.

4. Empirical Results

Before estimating Equation 10, it is necessary to take two preliminary steps. First, I assume that random errors [[Epsilon].sub.t] satisfies the conventional assumptions. Second, I have taken first-order differencing of all data to avoid the spurious results frequently generated from time series. The decision on differencing was based on the results of the Dickey-Fuller (DF) along with the augmented Dickey-Fuller test (ADF) - tests on stationarity of the series - and the cointegration test - test on possible loss of valuable long-run relationships among the series due to differencing.(9, 10)

I estimate the first-order differenced form of Equation 10 by both the ordinary least squares method (OLS) and the maximum likelihood method (ML). Table 2 shows three different ML results: sets 1, 2 and 3. I discuss only the ML results here because their overall statistics for goodness of fit are more satisfactory than the OLS results. The coefficients of determination ([R.sup.2]'s) are high. The values of the DW statistic (in the absence of [Delta][M.sub.t-1]) fall in the inconclusive [TABULAR DATA FOR TABLE 2 OMITTED] range at the 5% significance points of [d.sub.L] and [d.sub.U], or else they are close to [d.sub.U], minimizing the likelihood of serial correlation. The values of Durbin's h-statistic (in the presence of [Delta][M.sub.t-1] as a regressor) are mostly less than the critical value (1.645); we thus accept the hypothesis of zero autocorrelation at the 5 % level. There are three different null hypotheses ([H.sub.0]) to test against the alternative hypothesis ([H.sub.A]):

(i) [H.sub.0]: coefficient of [Delta][Prob701.sub.t] = 0

(ii) [H.sub.0]: coefficient of [Delta][Prob731.sub.t] = 0

(iii) [H.sub.0]: coefficient of [Delta][Prob.sub.t] = 0

against [H.sub.A]: coefficients [not equal to] 0 in all three cases. The individual two-tailed test rejects [H.sub.0] for (i) and (ii) at the 1% and 5% levels, whereas the same test accepts [H.sub.0] for (iii) even at the 10% level. The results suggest that the effect of either [Prob701.sub.t] or [Prob731.sub.t] measured alone is significant, whereas the effect of [Prob.sub.t] is less significant. Obviously, computation of Prob for its effect has decreased the value of the estimated coefficient associated with [Delta][Prob.sub.t] and increased the standard error of the estimate. The lagged terms of trade, [Delta][P.sub.t-1], is generally significant at the 1% level, whereas the lagged market penetration, [Delta][M.sub.t-1], is insignificant at all levels. The statistical significance of [Delta][P.sub.t-1] confirms empirically the lag structure and the rational expectations hypothesis, as specified.

Table 3. Elasticity of [M.sub.t] with Respect to [Prob701.sub.t] and
[Prob731.sub.t] (at means)

[Prob701.sub.t]                 -0.082
[Prob731.sub.t]                 -0.039

The statistical significance of the coefficients associated with [Delta][Prob701.sub.t] and [Delta][Prob731.sub.t] is clear evidence of effective trade remedy law enforcement. The magnitudes of the elasticity of import penetration [M.sub.t] (not [Delta][M.sub.t]) with respect to [Prob701.sub.t] (not [Delta][Prob701.sub.t]) and [Prob731.sub.t] (not [Delta][Prob731.sub.t]) with or without term [Delta][M.sub.t-1] in the regression equation are very close to each other. I present the values of the elasticity for the latter case in Table 3. It is notable that they are not large. As the ITC, in collaboration with the ITA, intensifies trade remedy enforcement of section 701 and section 731 by 1%, import penetration decreases by only 0.082% and 0.039%, respectively. Given the value of elasticity with respect to enforcement of section 701 (0.082%), there was a decrease in the average degree of import penetration by 0.012 (from 14.406 to 14.394). This amount corresponds to a decrease in the average quantity of imports by 0.019 net million tons (from 23.630 to 23.611), which is equivalent to a decrease in primary metal imports for consumption by 0.012 billion dollars in customs value basis (from 14.544 to 14.532) and a decrease in primary metal and fabricated metal combined together by 0.019 billions of dollars (from 23.114 to 23.095). The degree of effectiveness in case of enforcement by section 731 was even smaller. Given the value of elasticity with respect to enforcement of section 731 (0.039), the average degree of import penetration decreased only by 0.0056 (from 14.406 to 14.400). The average quantity of imports decreased by 0.0092 net million tons (from 23.630 to 23.621), and the average values of imports of primary metal by 0.0057 billion dollars (from 14.544 to 14.538) and both primary metal and fabricated metal by 0.009 billion dollars (from 23.114 to 23.105).

5. Concluding Remarks

This paper investigates the effects of U.S. trade remedy enforcement on steel imports by means of section 701 and section 731 under conditions of uncertainty. Given the relatively short period of active implementation and the quasi-judicial processes for the two provisions, I use a simple, probability-augmented model that can explain import penetration primarily in terms of unanticipated trade policy and rational expectations variables.

Statistical inferences on estimated coefficients provide clear insight into the effectiveness of CVD and AD policies. The measured degree or amount of effectiveness was small. The low degree of effectiveness may mean that administrative/legislative solutions are of limited benefit.

I thank Chung K. Min and Willem Thorbecke for helpful comments. I am also grateful to anonymous referees for valuable comments. Michael Morrison provided valuable research assistance.

1 See Crandall (1987) for a study on the effects of U.S. trade protection for autos and steel. Other earlier studies related to the steel industry are Tarr (1979) and Walter (1983). Feenstra (1988, 1995) suggests an additional deadweight loss due to the quality change under quotas (VER) in many industries.

2 The principal trade remedy provisions are section 301 (enforcement of U.S. rights and response to foreign trade practices,) Super 301 identification of trade liberalization priorities), Special 301 (identification of intellectual property rights priority), section 701 (countervailing duty), section 731 (antidumping duty), and section 201 (escape clause). See U.S. House of Representatives (1995). Cases under sections 301 and 201 during the sample period are relatively few in number. Super 301 is concerned with foreign trade barriers to U.S. exports rather than U.S. import restrictions. It was in effect for 1989-90 and was reinstated in 1994. During this period, the Office of the United States Trade Representative (USTR) identified only six cases involving three priority countries: Japan, Brazil, and India. None of the six cases was concerned with steel. See the Office of the USTR (various years) for details. Special 301 is irrelevant to steel imports.

3 They include Takacs (1981); Finger, Hall, and Nelson (1982); Eichengreen and van der Ven (1984); Herander and Schwartz (1984); Baldwin (1985); Feigenbaum and Willett (1985); Grossman (1986); Crandall (1987); Salvatore (1987); Coughlin, Terza, and Khalifa (1989); Hansen (1990); Boltuck and Litan (1991); Prusa (1991); Moore (1992); Anderson (1993); Nivola (1993); Baldwin and Steagall (1994); Krupp (1994); and Krueger (1995).

4 Herander and Schwartz (1984) use the predicted values of two main explanatory variables - the likelihood of filing "less than fair value" (LFV) complaints by U.S. firms and the probability of ITC's affirmative decision - as instrumental variables. They use a dummy variable for the latter to be measured by the logit technique.

5 Since the rational expectations hypothesis, innovation accounting technique, and vector autoregression are widely known, references for these are not provided here except the one in note 7.

6 In terms of the business cycle theory, [[Epsilon].sub.t] corresponds to the reduced-form innovations.

7 See Sheffrin (1985, pp. 1-26).

8 It is desirable to incorporate knowledge about specific cases and their merits in the probability measure employed here as an indicator of the stringency of U.S. trade sanctions. However, knowledge that actual agents might have is not accessible to researchers.

9 The period of observation utilized in this paper is insufficient to have a meaningful test for the unit roots. The short period is dictated by the brief period of active implementation of sections 701 and 731 and availability of data. Evans and Savin (1981, 1984) suggest that insufficient observation may cause a lower power for the tests. In light of the reported critical values corresponding to the number of observations adjacent to the number under consideration, however, I found the overwhelming evidence to confirm stationarity of the data. Each series was I(1) (integrated of order one), meaning that its first difference did not contain a unit root. I also found that the first differencing did not lose long-run relationships among the variables. The test results are not presented in this paper because of space limitations but are available from the author on request.

10 The DF test and the cointegration test are well-known in time series analysis and therefore not discussed here in detail. For a reference, see Greene (1993).

References

American Iron and Steel Institute. 1992. Annual statistical report. Washington, DC: American Iron and Steel Institute.

Anderson, Keith B. 1993. Agency discretion or statutory direction: Decision making at the U.S. International Trade Commission. Journal of Law and Economics 36:915-35.

Baldwin, Robert E. 1985. The political economy of U.S. import policy. Cambridge, MA: MIT Press.

Baldwin, Robert E., and Jeffrey W. Steagall. 1994. An analysis of ITC decisions in antidumping, countervailing duty and safeguard cases. Weltwirtschaftliches Archiv Heft 2:290-308.

Boltuck, Richard, and Robert E. Litan, eds. 1991. Down in the dumps: Administration of the unfair trade laws. Washington, DC: Brookings Institution.

Coughlin, Cletus C., Joseph V. Terza, and Noor A. Khalifa. 1989. Determinants of escape clause petitions. Review of Economics and Statistics 71:341-7.

Crandall, Robert W. 1987. The effects of U.S. trade protection for auto and steel. In Brookings papers on economic activity 1, pp. 271-88.

Eichengreen, Barry, and Hans van der Yen. 1984. U.S. antidumping policies: The case of steel. In The structure and evolution of recent U.S. trade policy, edited by Robert E. Baldwin and Anne O. Krueger. Chicago: University of Chicago Press, pp. 67-104.

Evans, G. B. A., and N. E. Savin. 1981. Testing for unit roots: 1. Econometrica 49:753-79.

Evans, G. B. A., and N. E. Savin. 1984. Testing for unit roots: 2. Econometrica 51:1241-69.

Feenstra, Robert C. 1988. Quality change under trade restraints in U.S. autos. Quarterly Journal of Economics 103:131-46.

Feenstra, Robert C. 1995. Estimating the effects of trade policy. In Handbook of international economics 3, edited by Gene M. Grossman and Kenneth Rogoff. Amsterdam: Elsevier, pp. 1553-95.

Feigenbaum, Susan, and Thomas D. Willett. 1985. Domestic versus international influences on protectionist pressure in the United States. In Exchange rates, trade and U.S. economy, edited by Seven W. Arndt, Richard J. Sweeney, and Thomas D. Willett. Cambridge, MA: American Enterprise Institute and Ballinger Publishing, pp. 181-90.

Finger, J. Michael, H. K. Hall, and D. R. Nelson. 1982. The political economy of administered protection. American Economic Review 72:452-66.

Greene, William H. 1993. Econometric analysis. Englewood Cliffs, NJ: Prentice-Hall.

Grossman, Gene M. 1986. Imports as a cause of injury: The case of the U.S. steel industry. Journal of International Economics 20:201-23.

Hansen, Wendy L. 1990. The international trade commission and the politics of protectionism. American Political Science Review 84:21-46.

Herander, Mark G., and J. Brad Schwartz. 1984. An empirical test of the impact of the threat U.S. trade policy: The case of antidumping duties. Southern Economic Journal 51:59-79.

International Trade Commission. Various years. Annual report. Washington, DC: U.S. Government Printing Office.

International Trade Commission. Various years. Antidumping duty determinations. Washington, DC: U.S. Government Printing Office.

International Trade Commission. Various years. Countervailing duty determinations. Washington, DC: U.S. Government Printing Office.

Krueger, Anne O. 1995. American trade policy. Washington, DC: AEI Press.

Krupp, Cornnie. 1994. Antidumping cases in the US chemical industry: A panel data approach. Journal of Industrial Economics 42:299-311.

Moore, Michael. 1992. Rules or politics? An empirical analysis of antidumping decisions. Economic Inquiry 30:449-66.

Nivola, Pietro. 1993. Regulating unfair trade. Washington, DC: The Brookings Institution.

Office of the United States Trade Representative. Various years. Trade policy agenda and annual report. Washington, DC: U.S. Government Printing Office.

Prusa, Thomas J. 1991. The selection of antidumping cases for ITC determination. In Empirical studies of commercial policy, edited by Robert E. Baldwin. Chicago: University of Chicago Press, pp. 47-71.

Salvatore, Dominick. 1987. Import penetration, exchange rates, and protectionism in the United States. Journal of Policy Modeling 9:125-41.

Sheffrin, Steven M. 1985. Rational expectations. Cambridge, UK: Cambridge University Press.

Takacs, Wendy E. 1981. Pressures for protectionism: An empirical analysis. Economic Inquiry 29:687-93.

Tarr, David G. 1979. Cyclical dumping: The case of steel products. Journal of International Economics 9:57-63.

U.S. House of Representatives. 1993-1995. Overview and compilation of U.S. trade statutes. Washington, DC: U.S. Government Printing Office.

Walter, Ingo. 1983. Structural adjustment and trade policy in the international steel industry. In Trade policy in the 1980s, edited by William R. Cline. Washington, DC: Institute for International Economics, pp. 483-525.