2.1 The textual basis

The investigating authorities in antidumping and countervailing proceedings are required to examine the price effects of dumped or subsidized imports on domestic like products. The general obligation to examine such price effects is set out in ADA Article 3.1 and SCM Article 15.1, which contain identical provisions except for the reference to ‘dumped’ (in ADA) or ‘subsidized’ (in SCM) imports:

The requirement to examine ‘price effect’ is then elaborated in ADA Article 3.2 and SCM Article 15.2:

Thus, in accordance with ADA Article 3.2 and SCM Article 15.2, the investigating authorities need to consider three types of price effect of dumped or subsidized imports on domestic like products: (i) whether there has been ‘a significant price undercutting’ by such imports, (ii) whether the effect of such imports is otherwise to depress prices to a significant degree (i.e., significant price depression), or (iii) whether the effect of such imports is to ‘prevent price increases, which otherwise would have occurred’, to a significant degree (i.e., significant price suppression).

2.2 WTO case law

Articles 3.2 and 15.2 do not define the terms of ‘price undercutting’, ‘price depression’, or ‘price suppression’, or prescribe any particular methodology in determining these price effects. Nonetheless, WTO case law has established certain criteria for identifying the effect of price undercutting. According to the Appellate Body, price undercutting ‘involves situations where imports are being sold at prices lower than the domestic like products’, and the finding of price undercutting requires ‘a dynamic assessment of price trends’ over the period of investigation.Footnote ⁶

By comparison, the criteria for establishing price depression and price suppression are less clear. In China–GOES, the Appellate Body explained that the concept of price depression ‘refers to a situation in which prices are pushed down, or reduced, by something’.Footnote ⁷ The finding of price depression, therefore, involves an observation of ‘a price decline’, and an analysis of whether the price decline ‘stems from’ dumped or subsidized imports.Footnote ⁸ As for price suppression, the Appellate Body has established the following: (i) price suppression may occur with or without price undercutting or price depression; (ii) price suppression cannot be properly examined without a consideration of whether, in the absence of subject imports, prices ‘otherwise would have’ increased; (iii) the effect of price suppression stems from relevant aspects of dumped or subsidized imports, ‘including the price and/or the volume of such imports’.Footnote ⁹ Yet, the Appellate Body did not explain how the phenomenon of price suppression should be identified, or clarify how the causal relationship between dumped or subsidized imports and price suppression/depression should be established. The lack of a clear standard for finding the effect of price suppression and price depression under Articles 3.2 and 15.2 has led to confusion and uncertainties.

2.3 Main arguments advanced

In its Redetermination, MOFCOM found that subject imports (i) suppressed domestic prices in 2008 and (ii) suppressed and depressed domestic prices in Q1 2009. With respect to price suppression, MOFCOM found that the domestic industry was unable to raise prices despite rising costs, and cited the following as the causes: (a) increases in the volume of subject imports, (b) the loss of market share of the domestic industry predominantly to subject imports, and (c) price competition between subject imports and domestic like products.

The controversy over MOFCOM's finding of price suppression centred on the question of whether a comparison between prices of subject imports and domestic like products is required. MOFCOM did not make such a price comparison in the Redetermination. The United States argued that MOFCOM could not have concluded that subject imports had the effect of price suppression without comparing the actual prices of subject imports and the domestic like products.Footnote ¹⁰ According to the United States, MOFCOM failed to compare the prices of subject imports and domestic like products, despite evidence in the record indicating that there was a substantial divergence in the prices of subject imports and domestic like products.Footnote ¹¹ In response to the US argument, China contended that Articles 3.2 and 15.2 do not require investigating authorities to compare the prices of subject imports and domestic like products, citing the Appellate Body's statement that, in light of the text of Articles 3.2 and 15.2, price suppression or depression may stem from ‘the prices and/or volume of such imports’.Footnote ¹²

As third parties in the dispute, Japan and the European Union expressed their respective opinions. Agreeing with the United States, Japan argued that an increase in the volume of subject imports or parallel pricing between subject imports and the domestic like products only establishes the potential for price suppression or depression, and not actual price suppression or depression by subject imports.Footnote ¹³ The EU, on the other hand, was of the view that an increase in the volume of subject imports and a decline in the domestic industry's market share, taken together with the existence of price competition between subject imports and the domestic like products, will generally indicate that subject imports suppress or depress domestic like products. However, the EU also added that, while in principle it may be possible to establish price competition between subject imports and the domestic like products without comparing the prices of subject imports and the domestic like products, it is unclear whether in practice it is possible to do so.Footnote ¹⁴

2.4 The Panel's ruling

Unlike the parties, the Panel did not consider the issue in this case to be whether price comparisons are mandated for the finding of price suppression under Articles 3.2 and 15.2. According to the Panel, since Articles 3.2 and 15.2 do not prescribe any particular methodology for the consideration of the effect of price suppression or depression, ‘investigating authorities retain some degree of discretion in adopting a methodology they deem fit’; but this discretion ‘is qualified by the overarching obligation set out in Articles 3.1 and 15.1 that investing authorities shall base their determinations on an objective examination of positive evidence of, inter alia, the effect of subject imports on the market prices of domestic like products’.Footnote ¹⁵

Consequently, the Panel did not consider it necessary to determine whether price comparisons are required in finding price suppression under Articles 3.2 and 15.2 ‘as a general matter of law’. Instead, it would consider ‘whether MOFCOM's failure to make price comparisons in this particular case, in light of the evidence and arguments before it, meant that it failed to make reasoned determinations regarding price effects on the basis of an objective examination of positive evidence’.Footnote ¹⁶ The Panel answered the question in the affirmative, citing two reasons. First, the Panel faulted MOFCOM for failure to consider the impact of non-subject imports (i.e., GOES imports not targeted in this particular AD/CVD investigation) on prices of domestic like products.Footnote ¹⁷ Second, recalling that there was information before MOFCOM on the prices of subject imports, non-subject imports, and domestic like products, the Panel found that ‘MOFCOM could not have reached a reasoned and adequate conclusion that the domestic industry's prices were suppressed by the volume of subject imports, having failed to consider the relative prices of subject imports and the domestic like product at all’.Footnote ¹⁸ This second reason, however, appears to be circular: MOFCOM's failure to make price comparisons in this case is wrong because MOFCOM failed to make such price comparisons.

In sum, the Panel declined to decide whether, as a matter of law, price comparisons between subject imports and the domestic like products are required in identifying the effect of price suppression under Articles 3.2 and 15.2, but nonetheless concluded that MOFCOM's failure to make such price comparisons in this particular case is inconsistent with the Articles because MOFCOM's finding was not ‘based on an objective examination of positive evidence’ as required by Articles 3.1 and 15.1. In other words, as a matter of applying the law, the Panel held that price comparisons between subject imports and the domestic like products are required for finding the effect of price suppression in this particular case.

Unfortunately, the Panel's reasoning does not help clarify the criteria for finding the effect of price suppression. By design, Articles 3.1 and 15.1 set out the overarching obligation on injury determination in antidumping and countervailing investigations, and Articles 3.2 and 15.2 elaborate what such obligation entails with respect to the price effects of dumped or subsidized imports on domestic like products. Instead of clarifying what is required in identifying the effect of price suppression, the Panel relied on the general principle set out in Articles 3.1 and 15.2 to find non-compliance with Articles 3.2 and 15.2. The Panel's approach begs the question: what constitutes ‘an objective examination of positive evidence’ for the purpose of determining the effect of price suppression?

2.5 Why a counterfactual analysis should be required

We submit that for determining the effect of price suppression under Articles 3.2 and 15.2, a counterfactual analysis should be required as a matter of law. Our reasons are as follows.

First, the text of Articles 3.2 and 15.2 implies a requirement to consider a counterfactual situation. As already noted, Articles 3.2 and 15.2 provide that, with respect to price suppression, the investigating authorities ‘shall consider’ whether the effect of dumped or subsidized imports is to ‘prevent price increases, which otherwise would have occurred, to a significance degree’. The subjunctive mood of the provision suggests that the investigating authorities should find out where the prices of domestic like products would have been without the presence of dumped or subsidized imports. Obviously, this level of domestic price cannot be observed in reality, but is a counterfactual situation.

Contextually, the concept of price suppression also exists in SCM Article 6.3(c), which has been interpreted by the Appellate Body to require a counterfactual analysis. SCM Article 6.3 details the meaning of ‘serious prejudice’, one of the adverse effects that may be caused by subsidies as identified by Article 5(c). Article 6.3(c) states that serious prejudice may arise where:

In United States–Upland Cotton (21.5–Brazil), the Appellate Body has interpreted the concepts of ‘price suppression’ and ‘price depression’ contained in Article 6.3(c). The following statement of the Appellate Body is worth quoting in full:Footnote ¹⁹

The Appellate Body then stated that one way to undertake the counterfactual analysis is ‘to use economic modelling or other quantitative techniques’, because these techniques ‘provide a framework to analyse the relationship between subsidies, other factors, and price movement’.Footnote ²⁰ According to the Appellate Body, ‘the relative complexity of a model and its parameters is not a reason for a panel to remain agnostic about them’.Footnote ²¹

In EC and Certain Member States–Large Civil Aircraft, the Appellate Body confirmed again that the use of a counterfactual analysis provides an adjudicator with a useful analytical framework to isolate and properly identify the effects of the challenged subsidies. In that context, the Appellate Body further explained:Footnote ²²

Thus, the Appellate Body has by now established a clear standard under SCM Article 6.3(c): In order to find if a subsidy has the effect of ‘significant price suppression’, the adjudicator must undertake a counterfactual analysis that involves a modelling exercise.

It should be noted that, in the original proceedings of China–GOES, the Appellate Body did refer to the concepts of price suppression and price depression in SCM Article 6.3(c), including a citation to its interpretation of the concepts in US–Upland Cotton (21.5–Brazil) quoted above.Footnote ²³ The Appellate Body made the references in the context of discussing the relationship between dumped or subsidized imports and domestic prices. By referring to its interpretation of the same concepts in SCM Article 6.3(c), the Appellate Body reinforced its view that Articles 3.2 and 15.2 require the investigating authorities to establish a causal link between subject imports and price suppression or depression. By focusing on the issue of relationship between subject imports and domestic prices, however, the Appellate Body did not discuss whether a counterfactual analysis should be required in finding the effect of price suppression or depression under Articles 3.2 and 15.2.

In this regard, one might question whether it is appropriate to set the same standard for the determination of price suppression and price depression under Articles 3.2 and 15.2 as under SCM Article 6.3. Other commentators have suggested that the Appellate Body might hold the Panel determination under Article 6.3 to a higher standard, since the Panel operates as the initial trier of facts in the ‘serious prejudice’ determinations, whereas in the context of antidumping and countervailing investigations, the Appellate Body seemed satisfied to have panels engage in procedural rather than substantive review of national agency determinations (Sapir and Trachtman, Reference Sapir and Trachtman2008; Davey and Sapir, Reference Davey and Sapir2010).

In our opinion, requiring the national authority to undertake a counterfactual analysis under Articles 3.2 and 15.2 does not necessarily mean the review standard for the national agency determinations must be the same as that for the Panel determinations under SCM Article 6.3. What at issue here is not the review standard, but how the concept of price suppression should be interpreted. Does the concept of price suppression have an inherent economic logic, i.e., does it presuppose a comparison between a factual situation and a counterfactual situation? Does a counterfactual analysis necessarily entail a modelling exercise? If the answer to these questions is affirmative, then there should be no reason not to require the national authorities to undertake such counterfactual analysis under Articles 3.2 and 15.2. Put differently, whether the effect of price suppression requires a counterfactual analysis is a matter of interpreting the term of Article 3.2 and 15.2, not a matter of setting a review standard. So long as the national authority conducts the required analysis, the Panel and Appellate Body can review such analysis according to whatever standard of review that is deemed proper under those provisions.

As Prusa and Vermulst (Reference Prusa and Vermulst2014: 198) pointed out insightfully in their comment on the original proceedings of China–GOES, the Panel in the case did not address the critical question of what constitutes a proper price effects analysis under Articles 3.2 and 15.2. Consequently, it remains unclear what China, or for that matter any other Member, needs to do in future cases in order to produce a satisfactory price analysis. According to their observation, the problem with China's price effect analysis in the case was to a large extent due to its poor documentation of the basis for determining that subject imports had caused price depression and/or price suppression. In practice, ‘neither the US nor the EU rigorously assess the impact that alternative factors might have played in the evolving pricing dynamics’, but their facility with public reporting largely obscures the fact. Having identified the lack of a real standard as the key issue in the price effects analysis, Prusa and Vermulst (Reference Prusa and Vermulst2014: 259–261) also recommended the use of econometric techniques for the future.

3.1 MOFCOM's causality analysis

In a nutshell, the reasoning put forward by MOFCOM for causality of price effects goes as follows. Imports from subject countries (USA and RussiaFootnote ²⁸ ) rose by 60% in 2008. This resulted in an increase of 5.56% in the market share of subject countries, which then suppressed the prices of China's domestic producers. While the factual evidence presented by MOFCOM (see Table 1) indicates an actual increase in prices for the domestic Chinese GOES of 14.53% in 2008, this increased level in prices was allegedly much below what should have been the level of Chinese GOES prices in the absence of subject imports, hence Chinese domestic GOES suffered price suppression.Footnote ²⁹ MOFCOM reasoned that in 2008, unit production costs of domestic producers rose by 63%, which should have resulted in a much higher increase in the prices of domestic GOES if domestic producers could have passed through the increase in costs to market prices as in times of normal competition. MOFCOM determined that the inability for domestic producers to raise the prices of GOES was due to the strong rise in the volume of subject imports.

Table 1. GOES facts

Source: MOFCOM Redetermination (2013)

The Panel did not find this sufficient to conclude that subject imports resulted in price suppression of domestic GOES. One reason given by the Panel that questioned the linkage between subject imports and price suppression was that non-subject imports were almost double the size of subject imports (Table 1, col. 4) and that the prices of non-subject imports did not differ much from subject imports (Table 1, col. 8). Hence the Panel questioned how MOFCOM could come to the conclusion that the inability to raise domestic prices was entirely driven by subject imports. Another concern the Panel raised was that in 2008 a new Chinese GOES producer, Baosteel, had entered the market. This new entry may have resulted in overcapacity and could subsequently also have lowered Chinese GOES prices, thus questioning whether subject imports were the main cause of domestic price suppression. Moreover, as noted in Section 2, the Panel faulted MOFCOM for not making a price comparison between subject imports and domestic like products, which the Panel deemed necessary in this particular case to provide conclusive evidence on price supression.

3.2 Alternative methods for establishing causality

The question we aim to address here is not whether MOFCOM was right or wrong in claiming price suppression – we simply do not have conclusive evidence to answer that question. Instead, we ask ourselves what MOFCOM could have done differently to make a more convincing case for domestic price suppression. What would have been needed to convince the Panel that subject imports have ‘explanatory force’ for the suppression of domestic prices?

As already pointed out, establishing price suppression requires the construction of a counterfactual. A counterfactual analysis demonstrates what Chinese domestic GOES prices would have been in the absence of subject imports. But how to engage in such a counterfactual analysis? As noted above, since WTO law does not prescribe any particular methodology in determining price supression, WTO members retain a certain degree of discretion in choosing their own methodology. Although such discretion ‘is not without limits’,Footnote ³⁰ the Panel did not specify these limits. Consequently, it is up to WTO adjudicators to decide what these limits should be in each particular case.

Below we present a number of techniques that have been used by other WTO members to make a convincing case of the causal link between subject imports and domestic prices. For each method, we assess its usefulness in determining causality.

3.2.1 Trends analysis

In several antidumping cases initiated by the EU and the US, the investigating authorities used ‘trends analysis’ as a way to provide evidence of linkage between subject imports and evolution in domestic prices, as required by ADA Articles 3.2 and 3.5. For example, Prusa and Sharp (Reference Prusa and Sharp2001) report that for the United States, the traditional injury analysis conducted by the International Trade Commission (USITC) has long consisted of ‘trends analysis’. While this type of analysis varies from case to case, its overall structure typically consists of a review of time series data pertaining to (1) the imports under investigation; (2) the prices for sales of the relevant imports and their domestic counterparts; and (3) lost sales or revenue on domestic transactions. More specifically, in a ‘typical’ antidumping case, the investigating authority considers in parallel, the evolution of imports and the evolution of domestic variables. Often, these parallel trends are illustrated graphically during the course of the investigation period, which typically involves time series data up to three years prior to the initiation of an AD case during which suspicions of dumping of subject imports arise.

Consider the following hypothetical example illustrated in Figure 1(a) where it is clear that a rising trend in subject imports during the investigation period (from t-3 onwards) coincides with a falling trend in domestic market share of a particular product. Moreover, note from the example that during the investigation period, a clear reversal of trends occurred. Before t-3, subject imports were stable and domestic market shares were on a postive trend, whereas during the investigation period after t-3, subject imports rose strongly, while the domestic sales’ trend reversed and started falling.

Source: Panel Report, China–GOES 21.5, para. 7.150.

Figure 1. (a) Quantity evolution. (b) Price evolution

Additionally, suppose that in the same example the evolution of subject import prices were given as in Figure 1(b), where a fall in subject import prices would be followed by a parallel trend in domestic prices. Again, visual inspection of Figure 1(b) clearly suggests that from t-3 onwards, foreign subject import prices started falling strongly which coincided with a reversed trend for domestic prices that also started falling.

To claim causality between subject imports and trends in domestic prices, the investigating authority in this hypothetical example would then also need to show that non-subject import trends evolve differently than subject imports. Again consider Figure 1(a), where non-subject imports are also shown, but where it is clear that their time trend did not change drastically during the investigation period. And when considering Figure 1(b), we also note that the trend in prices of non-subject imports evolved differently than for subject imports, i.e. remained relatively stable even during the investigation period.

Based on the hypothetical example presented in Figures 1(a) and 1(b), an investigating authority could make a convincing case that the downward trends in domestic sales and domestic prices is caused by subject imports which becomes clear through visual inspection of parallel trends. Note that it is immaterial whether domestic prices lie above or below the foreign prices, but what is important is the parallel movement in subject import prices and domestic prices. In other words, when trend reversals occur for both import prices and domestic prices around the same time during the dumping period, there is a strong suspicion of injury.

Now how does trend analysis provide for a counterfactual? A counterfactual can be constructed from the pre-investigation period trends of foreign and domestic variables. In other words, counterfactual values are given if the trends would have continued as they did before the dumping of subject imports. Such trends analysis provides a simple way to construct counterfactual prices or even counterfactual imports.

To illustrate what we mean by that, consider the dashed lines in Figure 1(a) and 1(b). The dashed lines show a continuation of the trend for domestic sales along the lines of pre-investigation period trends. And similarly for domestic prices, the dashed lines show how domestic prices would have evolved in the absence of a rise of subject imports and what domestic prices would have been in any particular year after the rise in subject imports. Hence, based on trends analysis one can construct counterfactual market shares and prices based on time series trends in the data observed prior to the investigation period. If, in any particular year, the actually observed domestic prices lie substantially below what the domestic prices would have been according to pre-investigation period trends, as shown on the dashed line as is the case in Figure 1(b), the investigating authority can convincingly claim price suppression/depression. Domestic price suppression/depression then refers to the lack of increase or decrease in domestic prices from t-3 onwards.

MOFCOM did not present a ‘typical’ trends analysis as the one illustrated in Figures 1(a) and 1(b). This may have been the reason that the evidence presented by MOFCOM was not considered convincing by the US and by the Panel. The question then is what the evidence would have looked like in the GOES case if MOFCOM had turned to a trends analysis. The investigation period, i.e. the period during which the US was accused of dumping, is very short and only involved 2008 and the first quarter of 2009, which makes it difficult to observe particular trends. Based on the data presented by MOFCOM and shown in Table 2 below, we graph the pattern of subject and non-subject imports in Figure 2(a) and 2(b).

Source: Data in Table 2.

Figure 2(a). Trends of subject and non-subject GOES imports (tons) (b) Trends in prices of subject and non-subject imports in GOES (RMB/ton)

Table 2. Volumes and prices of subject and non-subject imports

Source: Panel Report, China – GOES 21.5, para 7.150.

A visual inspection of Figure 2(a) does not suggest a different trend in subject imports (in black) versus non-subject imports (in gray) of GOES, which is what would be required for a linkage between subject imports and the evolution of domestic prices. Similarly Figure 2(b) does not show a markedly different trend in import pricing between subject (black) and non-subject imports (gray), except for the first quarter of 2009 in which subject import prices lie below that of non-subject imports, but the 2009 prices are still higher than the 2006 prices, so one cannot really speak of a trend reversal.

More importantly, what is lacking in Figures 2(a) and 2(b) is an evolution of the domestic sales in tons and levels in domestic prices (RMB/ton) of GOES, respectively. Hence, the trends in subject and non-subject variables cannot be compared with the corresponding Chinese values as this information was lacking and not revealed by MOFCOM. The only numbers made available by MOFCOM were a relative evolution (in %) of domestic market shares and of changes in domestic prices, as shown in Table 1 in columns (5) and (9), respectively. MOFCOM argued that it did not reveal prices because the Chinese petitioners requested confidentiality. To make a convincing case, however, MOFCOM could have shown the evolution of domestic sales in tons and the domestic prices to the Panel on a confidential basis.Footnote ³¹ In short, the evidence presented by MOFCOM does not allow for a simple trends analysis to be made in this case.

Trends analysis, however, has its limits. Because trends analysis heavily relies on visual inspection of trends, it is considered as ‘more art than science’ (Prusa and Sharp, Reference Prusa and Sharp2001). While these trends are clearcut in our hypothetical example, such is not always the case in reality (Ahn and Zanardi, Reference Ahn and Zanardi2016).

More importantly, trends analysis is not sufficent to demonstrate causality between subject imports and price effects. That is because trends analysis does not go beyond documenting correlations in the data, and a correlation does not necessarily prove causality. Put differently, documenting that the two variables – subject import prices and domestic prices – move in the same direction is not conclusive evidence of the subject import prices causing a drop of domestic prices. The reason is that correlations do not exclude the possibility that third variables, other than subject import prices, also have an effect on domestic prices, e.g., overproduction by the domestic industry, or non-subject import prices, or macro-economic variables (financial crisis), which may also be at play to a small or large extent. Trends analysis is not equipped to disentangle the importance of each of these factors, since it does not attribute weights to the different factors that may explain domestic prices. Thus, trends analysis is not suitable to serve as the standard methodology for identifying the effect of price suppression/depression under ADA Article 3.2 and SCM Article 15.2.

MOFCOM, therefore, would best have chosen alternative methods to claim the effect of price suppression/depression to begin with. In fact, in the late 1980s, the USITC in several of its cases already started questioning the validity of trends analysis due to its subjective interpretation. Below we describe two alternative methods that economists consider to correspond to empirical economic modeling and constitute an objective way of engaging in injury investigations in general.

3.2.2 Regression analysis

The subdiscipline in economics that offers a toolbox of statistical techniques and regression analysis to establish causality between variables is called econometrics. Econometrics allows economists to determine whether there is an economically significant impact of one variable on another. This is typically done by first collecting time series data on the variables and then applying regression analysis on these data.

Hence, to an economist, the causality test is a matter of regression analysis and econometrics which uses time series data (indicated by t). A relatively simple approach would be to estimate a regression equation as in (1) in which the evolution of Chinese domestic prices of GOES would be the dependent variable (y) on the left side of the equation which varies in every year t and for every firm i. All the explanatory variables that potentially can have an impact on domestic prices (y _it ) would be on the right-hand side of the regression equation and would include variables such as US import prices (x _1t ), Russian import prices (x _2t ) but would now also include non-subject import prices (x _3t ), Chinese production (x _4t ), and Chinese demand for steel (x _5t ). One cannot exclude that other variables may also affect domestic prices, other than x _1t to x _5t . To address that, equation (1) also includes year effects ( $\vartheta _t $ ) and firm-level fixed effects (θ _i ) and a rest category (ε):

(1) $$y_{it} = \alpha + \beta. x_{1t} + \gamma. x_{2t} + \delta. x_{3t} + \mu. x_{4t} + \rho. x_{5t} + \theta _i + \vartheta _t + \varepsilon _{it} $$

Any statistical software package designed for regression analysis allows one to estimate a regression equation as in equation (1) above. It will generate estimates for the relevant regression coefficients (β, γ, δ, μ, ρ) whenever sufficient data observations are available.Footnote ³² These regression coefficients then provide a useful indication on how important each of the right-hand side variables (x ₁, x ₂, x ₃, x ₄, x ₅)Footnote ³³ are for explaining Chinese domestic prices of GOES (y _it ), the left-hand side variable of regression equation (1).

The regression coefficients of explanatory variables can then be tested for their statistically significant impact on the left-hand side variable. Put differently, regression analysis allows the econometrician to determine whether a variable such as ‘subject import prices’ (x _1t ) has a significant effect on ‘Chinese domestic prices’ (y _it ), and compare the significance relative to that of other right-hand side variables in the regression. This way it is relatively straightforward to determine whether ‘non-subject import prices’ (δ) has more of an effect on the domestic market share of steel in China, as the US claimed, than the ‘subject import prices’ (β). A negative and significant value for the β-coefficient resulting from the regression analysis can then be interpreted as evidence in line with the allegation that the ‘subject import prices’ had a negative impact on domestic prices and domestic prices would have been higher in the absence of dumped or subsidized imports.

A significant difficulty with such a regression analysis is that most right-hand side explanatory variables are ‘endogenous’, which would result in biased regression coefficients that may not be trusted. For example, there can be a problem of ‘reverse causality’ between Chinese domestic prices and Chinese production: Chinese production is likely to affect domestic prices in China, but simultaneously, the prices at which GOES can be sold in China will determine the number of producers offering it to the market, and hence will determine total domestic production, which is the basic law of supply and demand in economics. In the case of this potential two-way causality, one cannot correctly estimate the regression coefficient on the importance of ‘Chinese production’ as a determinant on Chinese domestic steel prices. The theory of econometrics however provides several possible ways to deal with endogenous variables in order to get unbiased estimates for regression coefficients. One common solution is to replace each endogenous variable with a good instrument (alternative variable) that does not suffer from the two-way causality but is highly correlated with the endogenous variable.Footnote ³⁴ For example, in the case of ‘Chinese production’ (x _4t ), one can consider taking as an alternative variable Chinese GOES exports abroad. It is clear that Chinese exports are likely to be highly correlated with Chinese production, thus making it a good potential candidate as an instrumental variable. In addition, Chinese exports will not suffer from two-way causality as the demand for Chinese GOES exports is unlikely to be affected by the domestic prices of Chinese GOES, potentially making it a good instrument to deal with the endogeneity of the variable ‘Chinese production’.

To estimate all the regression coefficients in equation (1) correctly, therefore, one would have to find good ‘instruments’ for all the endogenous variables. While such an econometric exercise is by no means simple or straightforward, it is nevertheless feasible thanks to the techniques offered by the theory of econometrics. A clear indication of the merits of econometrics and the benefits it can render is that other fields of the law have long been using it. For instance, in the field of competition policy and merger analysis, economic and econometric modeling plays an important role in decision-making (Budzinski and Ruhmer, Reference Budzinski and Ruhmer2010). Thus, in our opinion, the application of econometric modelling can also make AD/CVD practice more objective, transparent, and less idiosyncratic.

In our view, MOFCOM could have turned to regression analysis without violating the confidentiality restriction requested by the Chinese petitioner. This could have been done in one of two ways. First, MOFCOM could have defined domestic prices, the dependent variable y in the regression equation (1), as an index variable, where the actual values of Chinese domestic prices could have been normalizedFootnote ³⁵ such that confidentiality would not have been breached. Second, it could have estimated regression equation (1) entirely in ‘first differences’ or in ‘growth rates’.Footnote ³⁶ Econometric theory can show that estimating equation (1) in first differences should yield the same estimated regression coefficients (β, γ, δ, μ, ρ) but with the additional advantage that it accounts for all omitted variables that are not time varying,Footnote ³⁷ such as the constant of the regression equation, given by α and all other variables that may affect domestic prices but that do not vary over time.Footnote ³⁸

However, the difficulty that MOFCOM would have faced in this particular case is that there were only two Chinese petitioners, which largely limited the number of data points available. The only data variation in regression equation (1) would come from the variation in domestic prices over time, which may not have been sufficient to get reliable regression coefficients.Footnote ³⁹ Therefore, while regression analysis in general can be regarded as much better equipped to address the causality issue, it requires a sufficient number of observations,Footnote ⁴⁰ which in this particular GOES case may not have been the case.

4.1 Tightening WTO disciplines on antidumping and countervailing measures

The methodology we proposed would have the effect of tightening WTO disciplines on injury determination in AD/CVD investigations, since it would narrow the scope of discretion currently enjoyed by the national authority in choosing methodologies to determine the price effects of subject imports under ADA Article 3.2 and SCM Article 15.2. But is such a discipline-tightening effect intended by the ADA and SCM Agreement?

In the original proceedings of China–GOES, China had submitted that the object and purpose of Articles 3.2 and 15.2 is to leave the authorities with discretion on the issue of price effects. The Appellate Body disagreed. According to the Appellate Body,

If the Appellate Body is correct, then the discipline on conducting the price effect analysis under Articles 3.2 and 15 must be tightened. That is so because the scope of discretion currently enjoyed by the investing authorities in choosing methodologies for determining price effects cannot ensure that the price effect analysis and the conclusion drawn therefrom is ‘robust’. The approach we proposed, in contrast, is capable of meeting the standard articulated by the Appellate Body.

4.2 Feasibility of econometric modeling

Econometrics is the application of mathematics and statistical methods in economics, which is of general utility in testing economic hypotheses and in helping distinguish between alternative explanations. It is true that econometricians can develop alternative solutions to test the same hypothesis. But if, no matter which econometric method is used, ‘subject imports’ always show up with a significant and negative coefficient in a regression analysis or structural modeling that aims to explain the evolution of domestic prices, one can truly speak of a ‘robust’ result which is unlikely to be questioned.Footnote ⁴³

We note that, in trade law and in AD/CVD law in particular, the use of econometrics by investigating authorities and by WTO panels remains relatively rare. This is in stark contrast with other fields of law. For example, in the field of competition law and policy, which is closely related to the field of industrial economics, the use of econometric modeling is more prevalent and econometrics has been widely accepted as a useful tool to support decision-making. If such is the case in related fields of law, it is hard to see why the same cannot, or should not, be done in AD/CVD law, given what econometrics can potentially offer to decision-making in trade law and policy.

4.3 Defense against criticism of ‘judicial activism’

When there is no clear standard on how the effect of price suppression/depression should be determined, it is ultimately within the discretion of WTO adjudicators to decide whether a particular determination by the national authority is consistent with Articles 3.2 and 15.2. In exercising this discretion, however, the adjudicators are susceptible to criticisms that their decision is biased or that their decision amounts to ‘adding to or diminishing the rights and obligations’ provided in the ADA and SCM Agreement, in violation of DSU Article 3.2. By adopting the methodology that is grounded in econometric modeling, the adjudicators will forgo most of their (unwanted) discretion in deciding what is inherently a matter of economic evaluation rather than legal reasoning. By turning to expert opinions, therefore, WTO judges will be able to insulate themselves from potential accusation of ‘judicial activism’ in the interpretation and application of Articles 3.2 and 15.2. Needless to say, this can only help enhance the legitimacy of the WTO dispute settlement system.