Food Policy 35 (2010) 154–162

Contents lists available at ScienceDirect

Food Policy

journal homepage: www.elsevier .com/ locate/ foodpol

What makes countries initiate WTO disputes on food-related issues?

Christian Götz a,*, Thomas Heckelei a,1, Bettina Rudloff b

a University of Bonn, Institute for Food and Resource Economics, Nußallee 21, D-53115 Bonn, Germanyb Stiftung Wissenschaft und Politik, Deutsches Institut für Internationale Politik und Sicherheit, Ludwigkirchplatz 3-4, 10719 Berlin, Germany

a r t i c l e i n f o a b s t r a c t

Article history:Received 23 April 2008Received in revised form 2 November 2009Accepted 21 December 2009

Keywords:WTO disputeAgro-food sectorBinary choice model

0306-9192/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.foodpol.2009.12.001

* Corresponding author. Tel.: +49 228 73 2323; faxE-mail addresses: christian.goetz@ilr.uni-bonn.de

ilr.uni-bonn.de (T. Heckelei), bettina.rudloff@swp-ber1 Tel.: +49 228 73 2332; fax: +49 228 73 4693.

This paper analyses relevant determinants for the probability to initiate a dispute on policy measuresunder the World Trade Organization (WTO) dispute settlement system. The empirical analysis differsfrom existing assessments by focusing on agro-food-related disputes and provides a more in-depth anal-ysis of specific country and sectoral characteristics not considered in previous studies. Contrary to recentanalyses of overall trade disputes, the results show that some determinants such as legal capacity andmonetary means are not statistically significant. Own protectionist behaviour, endured protectionism,and the duration of WTO membership, however, could be identified as relevant determinants with theexpected direction of impact.

� 2010 Elsevier Ltd. All rights reserved.

Introduction

The dispute settlement system of the World Trade Organization(WTO) was set into force as a part of the WTO agreement on Janu-ary 1, 1995. It is the device for the resolution of conflicts arising be-tween Members over the interpretation of their commitmentsunder the regime of the organization. Dispute settlement mustbe self-enforcing, i.e. from the consultation to the potential compli-ance phase all actions are driven by Members. Referred to as the‘‘central pillar of the multilateral trading system” (World TradeOrganization, 2007a) the design of the WTO dispute settlementsystem is central to the debate on institutional reforms of theWTO and has also been under negotiation at the current WTO Doharound. A major desire is to make the settlement system more effec-tive and to allow for the appropriate consideration of developingcountries’ demands. Reform proposals span a wide field from tight-ening time frames regarding panel proceedings over more effec-tively dealing with compliance and compensation procedures toassistance for developing countries ensuring their equality ofopportunity (Petersmann, 2003). However, the understanding ofthe factors that drive the system is required for targetedimprovement.

The question addressed in this paper is, therefore, which Mem-bers’ characteristics explain their activity as complainants in WTOfood-related trade disputes. Compared to previous empirical stud-ies, this investigation provides an in-depth analysis of food-related

ll rights reserved.

: +49 228 73 4693.(C. Götz), thomas.heckelei@

lin.org (B. Rudloff).

disputes and considers new potential determinants that may sup-plement the understanding of what drives the use of the disputesettlement system. The empirical investigation is based on a dis-pute distribution model developed and employed by Horn et al.(1999).

This paper is organised as follows: after a short description ofthe WTO dispute settlement’s features and the food-related case-load of the investigation period, a survey on existing empiricalstudies is provided. The model specification, including a discussionof considered determinants, is described in the next chapter. Sta-tistical implementation and estimation results are subsequentlypresented before concluding.

Facts and figures on WTO dispute settlement

All Members are provided with equal right to seek adjudicationthrough the WTO dispute settlement system. Acceptable reasonsfor filing a complaint are a trading partner’s measures that nullifythe benefits or impair the attainment of any objective of one ormore of the WTO agreements. The system’s rules and proceduresgenerally are administered by the General Council, i.e. the plenarymeeting of the WTO at the level of governmental officials, whichturns itself into the Dispute Settlement Body when adjudicatingtrade disputes. Dispute settlement procedures are stricter underthe WTO compared to those in force under its predecessor GATT1947. This is due to the elimination of blocking or delaying tacticsthrough the adoption of time limits for all stages of adjudication,the implementation of standard terms of reference for panels,and an improved mechanism for enforcement of compliance withpanel rulings (Hoekman and Kostecki, 2001). This is in particularowing to the implementation of the so called ‘negative consensus’,

Table 1Food-related dispute initiations by development classification until June 30, 2006.Source: Own compilation based on World Trade Organization (2007b) and UnitedNations (2008).

Group No. ofmembers

Share inmembership

No. ofdisputes

Share indisputes

Developed 31 0.23 97 0.66Developing 71 0.53 50 0.34Least developed 32 0.24 0 0

Total 134 147

Table 2Food-related dispute initiations by per capita income classes until June 30, 2006.Source: Own compilation based on World Trade Organization (2007b) and WorldBank (2008).

Per capita income class No. ofmembers

Share inmembership

No. ofdisputes

Share indisputes

High income 25 0.19 91 0.62Upper middle income 30 0.22 20 0.14Lower middle income 36 0.27 31 0.21Low income 43 0.32 5 0.03

Total 134 147

C. Götz et al. / Food Policy 35 (2010) 154–162 155

which means that a panel is established, a panel report adopted, orthe complainant authorized to trade related retaliation unless theDSB decides by consensus to reject.2

The consequences of this reform are reflected by economists’notion of the new multilateral trading system as a more ‘‘rules-ori-ented” compared to the more ‘‘power-oriented” system of theGATT 1947 (e.g. Jackson, 1997; Petersmann, 1997; Trebilcock andHowse, 1999). However, there exists empirical evidence on therelevance of power-based relationships between Members (e.g.Besson and Mehdi, 2004; Bown, 2004a), thereby challenging thatthe new system adequately considers the economic inequalitiesof its Members.

In the area of food-related disputes, 147 cases have been initi-ated over the period from January 1, 1995 to June 30, 2006.3

Regarding country participation the figures on current and previousfood-related WTO disputes reveal that the majority of cases are re-lated to the economically advanced countries. Table 1 shows the par-ticipation pattern by development classification of the UnitedNations (2008) for the period January 1, 1995 to June 30, 2006.4

The Members of the European Communities are not separately cap-tured in the statistic, thus reducing the number of WTO Members to134 instead of 149.5

Representing 23% of the WTO membership, the group of devel-oped countries are the most active users with 66% of all initiatedfood-related disputes. Thirty four percent of all disputes are initi-ated by the large group of developing countries. The group of leastdeveloped countries6 accounts for about 24% of WTO Members, butthey did not use the system at all for food-related issues. This isremarkable since food sector related exports generally represents asubstantial share in their export structure.

As there exists no established principle for the assignment ofthe developed or developing status in the United Nations system,a pattern which more precisely defines the economic status refersto the World Bank classification of income levels7. Table 2 depictsthe initiation pattern related to per capita income groups.

High income Members initiated 62% of all food-related disputes.Members of the Upper middle and Lower middle income groupsaccount for about 25% of all initiated disputes, whereas Membersof the Low income group are almost completely absent. The UnitedStates and the European Union dominate the High income group,followed by Canada, Australia and New Zealand. Brazil, Chile, Ecua-dor, Philippines and Thailand in the Middle income groups, and In-dia in the Low income group are examples for comparatively activeusers of the system.

As is evident from the brief description above, the WTO disputesettlement system’s rules themselves do not discriminate betweenMembers and offer equal opportunities for dispute initiation.Therefore, it is of some interest to investigate what are the majordeterminants for observing this pattern of dispute initiation across

2 The ‘Understanding on Rules and Procedures Governing the Settlement oDisputes’ is laid down in Annex 2 of the ‘Marrakech Agreement Establishing theWorld Trade Organization’ from April 15, 1994.

3 These 147 cases refer to nine different agreements related to issues on food, seeFootnote 8.

4 The developing status according to the WTO is based on members’ self-declaration and not on verified economic attributes.

5 The European Communities comprised 15 Members until April 30, 2004. Disputesof its 10 new Members from May 1, 2004 on are captured separately and not assignedto the EC in Table 1.

6 Thirty two of currently 150 WTO members are classified as least developedcountries. The identification of their status by the United Nations is dependent on thefollowing economic criteria: (1) a low per capita income criterion, (2) a humanresource weakness criterion and (3) an economic vulnerability criterion (UnitedNations, 2007).

7 Per capita income classification according to the World Bank (2008): Low income$875 or less, Lower middle income: $876-3465, Upper middle income: $3466–10,725High income: >$10,726.

f

:,

countries and whether other than the most prominent attribute‘‘income” may influence a Member’s decision to file a complaint.

Empirical analyses on general dispute initiation

A few empirical assessments on the WTO initiation of disputesexist considering various determinants, agreements referred to,and roles in a dispute (complainant, defendant, co-complainantand interested party). Table 3 depicts their investigation period,dispute coverage, main issue of analysis and the models used. Ta-ble 4 comprises the detected influences of determinants under pre-vious investigations.

Horn et al. (1999) mark the first empirical investigation byusing a binomial dispute distribution model. According to theiranalysis the dispute initiation pattern is to a large extent reflectedby the pattern of export diversity and value. GDP did not reveal asignificant influence, but a country’s legal capacity shows a slightpositive influence on its probability to complain. Besson and Mehdi(2004) find empirical evidence that legal capacity matters with re-spect to a country’s likelihood to win disputes. This supports theconclusion of Busch and Reinhardt (2003) that early settlementsof developing countries, i.e. in the consultation stage or in the Pa-nel stage before a ruling, are missing due to the lack of legalcapacity.

The self-enforcing nature of the dispute settlement system hasbeen the starting point for Bown (2004a, 2004b, 2005): A focus lieson costs of running a dispute and a country’s retaliation power tofinally enforce compliance by penalty tariffs on imports of the con-demned party. Bagwell and Staiger (2000) and Dam (1970) statethat the retaliation threat always has been a central componentof the GATT system. The success of this power is linked to thecountries’ relevance as trade partner and there exists also theoret-ical support that the retaliation threat is not uniformly distributedover Members and that imbalances relating to trade volume andmarket size shows influence on their force under trade disputes.Bown (2002) demonstrates that a country’s capacity to influenceits terms-of-trade determines the credibility of its retaliationthreat which is confirmed as well by Johnson (1953) and Kennanand Riezman (1988).

Table 3Survey on investigation period, dispute coverage, main issue and used model of previous empirical studies. Source: Own compilation.

Empirical study Investigation period and dispute coverage Main issue of analysis Used model

Horn et al. (1999) WTO disputes; 1995–1998; 155 complaints; all agreements Determinants for the initiation ofcomplaints

Binomial disputedistribution model

Holmes et al.(2003)

WTO disputes; 1995–2002; 279 complaints; all agreements Involvement in complaints (both sides) andsuccess in disputes

Descriptive statistics

Bown (2004a) GATT & WTO disputes; 1973–1998; 174 complaints; allagreements

Determinants for compliance after tradedisputes

Linear regression

Bown (2004b) GATT & WTO disputes; 1992–2003; complaints against US traderemedies

Initiation of complaints against US traderemedies

Probit model

Besson and Mehdi(2004)

WTO disputes; 1995–2002; 40 complaints of developing againstdeveloped countries

Success in disputes: developing againstdeveloped countries

Probit model

Bown (2005) WTO disputes; 1995–2000; 54 complaints; complaints againstimport protection on MFN-basis

Engagement as co-complainant orinterested third party

Ordered multinomial logitmodel

Table 4Survey on findings of previous empirical studies. Source: Own compilation.

Influence on the likelihood to

Determinant Empirical study Initiate disputes Partake in initiateddisputes

Win a dispute Reach complianceafter ruling

Export diversity Horn et al. (1999) +

Exporting interest Bown (2005) +

Export volume Holmes et al. (2003) +

Gross Domestic Product Bown (2005) +Horn et al. (1999) 0

Political economy relationship with respondent Bown (2004b) 0Bown (2005) +Besson and Mehdi (2004) �

Reliance on bilateral assistance Besson and Mehdi (2004) �Bown (2005) �

Legal capacity Horn et al. (1999) +Besson and Mehdi (2004) +

Military power Bown (2004b) +Besson and Mehdi (2004) +

Retaliatory capacity Bown (2004d) +Bown (2005) +Besson and Mehdi (2004) 0

+ Positive influence; � negative influence; 0 no influence.

156 C. Götz et al. / Food Policy 35 (2010) 154–162

Bown (2005) concentrates on the question whether to join com-plaints as co-complainant or interested party and demonstrated apositive impact of the capacity to absorb legal costs on both deci-sions. Additionally, he identifies a positive effect of a Member’sretaliatory capacity in terms of its relevance as trading partnerand a negative impact of countries’ dependencies on bilateraldevelopment aid. Bown (2004c) shows that the threat of retaliationis significant for determining whether a government chooses toabide by its international obligations. Bown (2004b) demonstratesthat the successful economic resolution to disputes is influencedby the threat of retaliation by the complainant. With respect todeveloping countries’ success in disputes Besson and Mehdi(2004) discover empirical support for the influence of their traderetaliation power.

Market access and exporting interests are expected to be rele-vant for the decision on initiation or participation and there existsempirical substantiation for this. Bown (2005) provides support forthe positive impact of a country’s volume of exports at stake in itsdecision to attend disputes as co-complainant or interested thirdparty and Bown (2004d) demonstrates its positive influence onthe likelihood to complain against United States (US) imposedtrade remedies. In the broader sense there is evidence for the rel-evance of trade volume or share respectively. Holmes et al.(2003) reach the conclusion that a Member’s trade volume deter-mines its likelihood to file complaints on the basis of simple

descriptive statistics. This supports the findings of Horn et al.(1999) that trade volume and export diversity are closelycorrelated.

Bown (2004a) finds only limited confirmation that internationalobligations affect a country’s decision to fulfil its commitmentswhereas Bown (2005) finds empirical evidence on the positiveinfluence of a Member’s international economic relationships –measured by its engagement in preferential trade agreements –on its decision to formally engage in a dispute as co-complainantor interested third party. On the topic of success in disputes, the re-sults of Besson and Mehdi (2004) suggest that international eco-nomic relationships show influence on a Member’s likelihood towin and they conclude that the reliance on bilateral assistancehas a negative impact on the success. Further, they discuss the im-pact of military power and find confirmation for the negative influ-ence that military powerful defendants have on the performance ofdeveloping countries in dispute.

Assessing relevance of determinants

A binomial dispute initiation model

This analysis is based on the model first presented by Horn et al.(1999): the initiation decision is described through a binary choice

8 Agricultural and food related issues comprise initiated disputes which wereraised under the following agreements: agreement on agriculture, agreement onsanitary and phytosanitary regulations, agreement on technical barriers to tradeagreement on safeguards, general agreement on tariffs and trade, agreement onsubsidies and countervailing measures, agreement on trade related investmenmeasures, anti-dumping-agreement and the agreement on import licensingprocedures.

C. Götz et al. / Food Policy 35 (2010) 154–162 157

model in which the Member’s probability to complain against an-other Member is dependent on a set of the complainant’s traits orthe characteristics of its specific environment. The implicated con-ditional probability function for this binary choice situation is theBernoulli distribution

Fðyijjxi; bÞ ¼ piðxibÞyij ½1� piðxibÞ�1�yij ¼piðxibÞ for yij ¼ 1;1� piðxibÞ for yij ¼ 0:

(ð1Þ

where yij is the binary dependent variable which takes one for acomplaint and 0 for no complaint, b denotes the vector of K coeffi-cients, i and j indicate the complainant and the defendant respec-tively. The set of K influences is merged in vector xi. Functionpi(xib) calculates the individual probability to complain for a pro-spective complainant i which can be represented by any cumulativeprobability distribution function. Here, we use the widely employedconditional logistic distribution,

piðxibÞ ¼expðxibÞ

1þ expðxibÞ; ð2Þ

which would result in the well-known Logit model when applied tosingle trials.

The proceeding for the assessment of determinants is the repro-duction of the observed sample of dispute initiation over the peri-od from January 1, 1995 to June 30, 2006 based on a disputedistribution function which yields probabilities for positive inte-gers, i.e. the number of a Member’s initiated disputes. Assumingthat the probability for a litigation decision pi(xib) is constant fromone trial to the next and that successive trials are independent,Member i’s probability for ci complaints in ni trials against all otherWTO Members is then specified through the Binomial distribution

f cijxi;b;nið Þ ¼ni

ci

� �piðxibÞci ½1� piðxibÞ�ni�ci ; ð3Þ

where ci ¼P

jyij. The expected number of Member i’s complaintsagainst all other WTO Members is then given by the expected valueof the Binomial distribution,

EðciÞ ¼ nipiðxibÞ; ð4Þ

which is strictly proportional to the number of independent Ber-noulli trials ni. The applied method is maximum likelihood estima-tion. Assuming that the data drawn from this Binomial distributionis independent and identically distributed with unknown parame-ter b, the likelihood function, i.e. the joint probability density ofobserving the given sample of complaints (c1, c2, . . ., cm) is specifiedby

Lðbjci;xi;niÞ ¼Ymi¼1

ni

ci

� �piðxibÞci ½1� piðxibÞ�ni�ci : ð5Þ

Starting from the logarithmic likelihood function

lnLðb ci;xi;nij Þ ¼Xm

i¼1

lnni

ci

� �þ ci ln piðxibÞ½ �þðni�ciÞ ln 1�piðxibÞ½ �

� �;

ð6Þ

the first order conditions for a maximum are

@ ln Lðb �j Þ@b

¼Xm

i¼1

ci � ðni � ciÞ expðxibÞ½1þ expðxibÞ�

� �¼ 0: ð7Þ

Restricting the vector of determinants to a constant, the proba-bility to complain reduces to pi(xib) = p for all Members i and canbe determined analytically by solving the first derivative of Eq.

(7) with respect to p leading to p ¼P

iciP

ini

. Hence, for the restricted

model, the maximum likelihood estimator of the probability to ini-tiate a dispute is simply the number of observed complaints overthe total number of independent Bernoulli trials.

The definition of the number of independent Bernoulli trials re-quires information about the exact number of infringements thateach Member faces, as the aforementioned binary choice model re-fers to the litigation decision when WTO obligations are violated.For the reason that we have no a priori information about the exis-tence of inconsistent trade measures – their existence can merelybe assured after a positive Dispute Settlement Body or AppellateBody ruling – the analysis is based on an assumption about theirdistribution. For Horn et al. (1999) the number of independent Ber-noulli trials is dependent on a country’s export diversification, i.e.its number of different exported goods over all products and trad-ing partners under the regime of the WTO. Each counted bilateralexport flow is considered as one trial. They worked on the assump-tion that ‘‘disputable trade measures” (DTM) are uniformly distrib-uted over all bilateral export flows. The problem of this approach isthat the determinants for the occurrence of disputes cannot be sep-arately identified from the impacts on the existence of DTM, lead-ing to an ‘‘export diversity bias”, i.e. an increase in disputes withincreasing export diversity. This problem already was a centralcriticism of Holmes et al. (2003).

Following the approach of Horn et al. (1999) we try to mitigatethe problem of missing information about the distribution ofinfringements by incorporating two new indicators: Endured Pro-tectionism by Trade Partner and Own Imposed Protectionism. Inaddition to this information on the likelihood of DTM in exportflows, the attempt of Horn et al. (1999) to select the relevant ex-port flows is slightly modified by taking empirical instead ofparameterized values for average induced litigation costs intoaccount.

Determinants considered

Deviating from existing studies, this paper focuses specificallyon agricultural and food-related disputes in order to develop anin-depth analysis of determinants relevant in this sector and toadditionally introduce new potential determinants. The set ofdeterminants or countries’ traits already used in prior studies is re-flected by agricultural trade flows characterising the export diver-sity, a country’s wealth and its legal capacity. Due to limited dataavailability for some determinants under investigation the Mem-bers sample is limited to 53 while maintaining the distributionover income classes.

Disputes dataDispute initiations were collected that affected products of the

food sector.8 The investigation covers the period from January 1,1995, to June 30, 2006. Each initiation is counted once to avoid dou-ble counting, thus omitting re-uptakes of disputes that occur whenthe consultation period of 12 months is exceeded. For jointly filedinitiations, each participant is assigned one dispute. When one Mem-ber simultaneously requests for consultations on the same subjectbut with different defendants each one is counted on its own. Sincethe European Communities (EC) is a single customs union with aharmonised trade policy and common tariffs all disputes initiatedby its Members are assigned to the EC. On the other hand, when dis-

,

t

158 C. Götz et al. / Food Policy 35 (2010) 154–162

putes are initiated against several EC Members there is only one dis-pute assigned, including all defendants. The number of disputes foreach Member is related to the whole investigation period,

y�i ¼ yoi � T; ð8Þ

where y�i is the time-corrected number of disputes of Member i, yoi

assigns Member i’s observed disputes over its WTO membershiptime ti and T stands for the investigation period. This proceedingis self-evident, since the number of filed disputes ought to be linkedto a Member’s membership time in the WTO. By this means thetime-bias is taken care of.

Export diversityHere we adopt the approach first presented by Horn et al.

(1999). Members’ export diversity is calculated as the averagenumber of their bilateral agro-food-related export flows per year.The average is taken over the investigation period 1995–2006and different thresholds on bilateral export value apply. Strictlyspeaking, export diversification is not an explanatory variable,but an intrinsic component of the underlying binomial dispute dis-tribution model as the total number of trials depends on the num-ber of export flows. Horn et al. (1999) found empirical support forthe dependency of a Member’s activity as complainant from its ex-port diversity, i.e. its number of different exported goods over alltrading partners. The underlying principle lies in the expectationof an increased probability to encounter infringements if a Mem-ber’s export diversity increases. This is self-evident if we assumeinfringements to be uniformly distributed over all markets, prod-ucts and trading partners. Hence, we expect the number of dis-putes to be positively related to Members’ amount of differentbilateral export flows. The export diversification factor’s explana-tory contribution content is just confirmable by excluding all othervariables as the expected number of complaints is proportional tothe number of a country’s export flows.

With this approach, export diversity might be underestimatedfor countries experiencing banned trade on SPS-grounds. This is alimitation of the analysis, however, it is mitigated to some extentby taking the average of Members’ number of export flows peryear. Export flows come from EuroCARE (2006) available at HS9-4-level.

Induced costs of litigationHorn et al. (1999) were the first analysing the litigation costs in-

volved and demonstrated their relevance. Their approach is fol-lowed through the implementation of a threshold for counting aMember’s bilateral export flows, thus excluding flows under a cer-tain value not being worth to fight for: as said above, a Member’sexport diversity, i.e. its number of independent Bernoulli trials ni,is obtained by averaging the number of all agro-food-related bilat-eral export flows over all trade partners across years. The thresh-olds for the incorporation of export flows are imposed bilaterally,which means that Members’ bilateral export flows are counted iftheir value exceeds the applied threshold. According to calcula-tions of Nordström (2005), average costs for dispute settlementproceedings range from $128 K to $706 K, dependent on the degreeof their complexity and the per hour rate of engaged lawyers.Hence, the analysis is conducted for four different litigation costlevels, i.e. excluding all flows below the respective threshold: $0when no threshold is applied, $300 K for low costs, $500 K for med-ium costs and $700 K for high litigation costs. The impact of theadopted cost-thresholds is shown for the restricted model, i.e. tothe exclusion of all explanatory variables, thus comparing different

9 Harmonized commodity description and coding system of the World CustomsOrganization (WCO).

cost-thresholds with respect to the corresponding model’s predic-tion quality.

Endured protectionism by trade partnerThis is to our knowledge the first empirical effort to incorpo-

rate information about the distribution of WTO-inconsistent tradebarriers to reduce the lack of information about the existence ofactual infringements which is the precondition to each dispute.It is assumed that the more protective the trade policy of a coun-try’s trading partners is, the higher the probability that it facesdisputable trade barriers. Hence, we expect the number of initi-ated disputes to be positively related to a country’s faced traderestrictiveness. For this purpose the Market Access Overall TradeRestrictiveness Index (MA-OTRI) provided by Kee et al. (2006) isused. It comprises a tariff equivalent of all barriers in the agro-food sector that exporters of the respective country face onaverage across the rest of the world. The indicator refers to datastemming from 1995 to 1998 concerning the non-tariff compo-nent and from 2000 to 2004 for the tariff component of theaggregated MA-OTRI.

Own protectionist behaviourAnother hypothesis is that the number of its filed disputes is

negatively related to a country’s tendency towards protectionism.The rationale behind this is the assumption that a more protectiveMember faces also a greater likelihood to become ‘‘victim” of anaccusation. We presume a more protective country to pursue adefensive and peaceful strategy to not provoke to be challenged it-self. On the other hand we hypothesize that more protective coun-tries have a lower propensity to fight for market liberalisation. Forthis purpose the Overall Trade Restrictiveness Index (OTRI) by Keeet al. (2006) is used as a measure for a country’s inclination torestrictive policies. It is a tariff equivalent for all trade barriers inthe agro-food sector which the respective country imposes in aver-age upon the rest of the world. Consequently, it provides the mirrorimage of the aforementioned MA-OTRI indicator, measuring thetrade restrictiveness from the potential complainant’s perspectiveand refers on the same period.

Relevance of the agricultural sectorIndependent from a country’s contact to a trading partner we

expect the overall importance of the agricultural sector having apositive influence on initiating a case: the higher the overall eco-nomic relevance, the more sensitive a country may be regardingviolations. To quantify the sector’s importance the agriculturalshare of a Member’s GDP is employed. This rather crude indica-tor is used due to missing data on the value of the countries’food industry. An improved measure should comprise informa-tion on the relevance of a Member’s whole agro-food sector.The data is drawn from the United Nations Conference on Tradeand Development (2002, 2003).

Importance of the agro-food export sectorSince disputes are dependent on agro-food-related trade, the

relevance of Members’ agricultural sector is complemented by anindicator on the importance of their agro-food exports. Thehypothesis is that the more important Members’ agro-food exportsector, the higher their general incentive to fight for their agro-food-related export interests. Hence, their number of observeddisputes is expected to be positively related to this importance.Members’ importance of the agro-food export sector is measuredas the share of agro-food-related export value in their total exportvalue. The data comes from World Bank (2007). The indicator rep-resents an average over the investigation period, i.e. the averageshare of Members’ export value in total exports.

C. Götz et al. / Food Policy 35 (2010) 154–162 159

It has been suggested by an anonymous referee that the incen-tive to complain could also be positively related to country size orthe share of a Member’s agro-food exports in world trade. We gen-erally agree with this possibility, but would argue that the relativeimportance of the agricultural food export sector is more closelyrelated to the incentive than the size. Furthermore, in the next sub-section, we use the GDP to measure the capacity to absorb legalcost. This is clearly a variable measuring country size. Conse-quently the GDP potentially also picks up an effect related to aMember’s incentive to complain that is not captured by the relativeimportance of the agro-food export sector.

Capacity to absorb legal costs/wealthThe capacity to absorb legal costs is supposed to be essential for

the accomplishment of disputes as explicit compensation for litiga-tion costs is not intended by the system. Even though the expectedgains from removing the trade barrier exceed the induced litigationcosts, this potential payoff lies ahead and is uncertain. For this rea-son each potential complainant must anticipate substantial coststhat are involved by prosecution and, if applicable, also by enforce-ment of compliance. It is assumed that the number of complaints ispositively related to a Member’s capacity to absorb legal costs. Asproxy for such financial means we use a country’s Gross DomesticProduct in US-Dollars, provided by the World Bank (2007). Theindicator is an average of Members’ yearly reported GDP over theinvestigation period.

Legal capacityOne argument often raised to explain the limited access of the

system to developing and low income countries is their lack of hu-man and legal capacity (see e.g. Whalley, 1996). Horn et al. (1999)found empirical evidence on the matter of a country’s legal capac-ity in respect of initiating disputes. We hypothesise that the largera country’s endowment with skilled legal personnel, the higher itscapability to challenge arguable trade measures of its trading part-ners and we expect the number of bilateral complaints to be pos-itively linked. The respective determinant should comprise thewhole extent of a country’s trade administration, i.e. its budget,its staff’s size and quality. Since there is no differentiated informa-tion on Members’ legal capability we use, like Horn et al. (1999),their delegation size at Geneva as proxy. The respective informa-tion comes from the United Nations (2004). The indicator providesinformation on the status of Members’ permanent delegation in2004, however it is alleged that this endowment represents theiradministrative power over the investigation period.

Some other potential indicators on Members’ ability to followthrough with a case have been considered as well but were dis-

Table 5Overview on explanatory variables, data and expected sign. Source: Own compilation.

Explanatory variables Data

Export diversity* Census of different export flows on HS-Capacity to absorb legal costs/wealth* Gross Domestic ProductLegal capacity* Size of permanent delegation at GenevaInfluence of private actors Measure of legal dimensions of undue p

influence by the private sectorGovernmental efficiency Measure of effectiveness and integrity o

and judicial systemRelevance of the agricultural sector Percentage share of GDP produced in agImportance of agro-food export sector Share of agro-food-related export value

export valueEndured protectionism by trade partner Average endured tariff equivalent

Own imposed protectionism Average imposed tariff equivalent

WTO membership time Index based on a Member’s percentage mshare over investigation period

* Influencing factors already integrated in previous empirical investigations.

carded due to data availability issues. Since disputes on SPS-mat-ters are incorporated in the sample, Members’ scientific capacitycould play a role to make a good case in such disputes. A possibleobjection to this rationale is that skilled personnel can reside inone Member state but be under contract for another Member inWTO matters. To capture the aspect of scientific capacity, indica-tors on Members’ expenditure on research and development, ontheir total number of researchers and on their number of publishedscientific and journal articles have been compiled from World Bank(2007). Unfortunately, the three indicators are not available for allMembers in the sample and could therefore not be incorporated.However, it is likely that almost all information will be capturedby Members’ GDP as this is highly correlated with the three indica-tors for the overlapping observations (pair wise correlation all>0.98). Consequently, a potentially significant influence of theGDP will include the influence of available legal capacity to someextent.

Influence of private actors and governmental efficiencyThe influence of private pressure groups on the government is

relevant as only the government may finally enter a dispute butcan be persuaded by private actors in doing so. This power may dif-fer among countries depending on the national framework fororganizing private lobby activities and on their respective rele-vance. It is increasingly seen as especially relevant for developingcountries in determining the use of the settlement system (Shaffer,2003a; Bown and Hoekman, 2005). Shaffer (2003a,b) demonstratethe relevance of private–public partnerships for the initiation andprosecution of trade disputes at the WTO and Besson and Mehdi(2004) argue that domestic variables should be incorporated tohandle the potential sources of distortion of the dispute settlementprocedure.

To our knowledge this is the first empirical attempt to capturesome aspects of the aforementioned interaction between the pub-lic and the private sector regarding dispute initiation. For this pur-pose two domestic variables are included which are provided byKaufmann (2004): (i) the Corporate Legal Corruption Component(CLCC), measuring legal dimensions of undue political influenceby the private sector and (ii) the Judicial/Legal Effectiveness Integ-rity Index (JLEI), assessing the effectiveness and integrity of the le-gal and judicial system. The greater the influence of lobbyists, e.g.by legal political finance or by the voice of interests of powerfulfirms, the more successful the private sector is supposed to be inachieving its export interests. Accordingly, the number of chal-lenged disputes should be positively correlated to the amount ofundue influence, aggregated in the CLCC variable. It is hypothe-sized, that the higher the efficiency and integrity of the legal and

Source Expected sign

4 level EuroCARE (2006) (+)World Bank (2007) +United Nations (2004) +

olitical Kaufmann (2004): Corporate Legal CorruptionComponent (CLCC)

+

f the legal Kaufmann (2004): Legal and JudicialEffectiveness and Integrity Index (LJEI)

+

riculture UNCTAD – statistical yearbook (2002, 2003) +in total World Bank (2007) +

Kee et al. (2006): Overall Trade RestrictivenessIndex (OTRI)

+

Kee et al. (2006): Market Access Overall TradeRestrictiveness Index (MA-OTRI)

�

embership World Trade Organization (2007c) +

Table 6Results for the restricted model subject to different thresholds for export flows. Source: Own compilation.

Threshold on exportflow values

Number of export flows Mean of absolutedeviations

Fraction of exactpredictions

Min Max Avg

$0 Beta 0 �5.28Prob 0.0006 126 54,965 4254 2.00 0.13

$300 K Beta 0 �5.28Prob 0.0035 23 15,514 712 2.02 0.43

$500 K Beta 0 �5.28Prob 0.0043 19 12,821 572 2.00 0.45

$700 K Beta 0 �5.28Prob 0.0051 18 11,146 489 2.04 0.45

Number of observations: 53.

0 Estimation, selection of variables, the likelihood ratio test and the bootstrap re-mpling and testing procedure are implemented in GAMS (General Algebraicodeling System), see Brooke et al. (1998): GAMS – A User’s Guide, GAMSevelopment Corporation, Washington, DC. The standard errors of the coefficientsre calculated for 2000 re-sampling iterations.

160 C. Götz et al. / Food Policy 35 (2010) 154–162

judicial system of a country, the higher its ability to identify illegaltrade measures and to pursue a legal action. Hence, the probabilityfor litigation is presumed to be positively dependent on the JLEIvariable.

Membership timeThe time of membership may be negatively related to the costs

of filing a dispute as learning occurs. Hence, we suspect a Mem-ber’s experience through its membership in the WTO to be posi-tively related to its number of filed disputes. An index is createdover the time since the inception of the organization until June30, 2006, relating each Member’s membership time to the wholeobservation period. The associated data is from World TradeOrganization (2007c).

Table 5 provides an overview on all explanatory variables withtheir respective data source and expected impact on the initiationof disputes.

Statistical implementation and results

For the restricted model, the probability to complain is identicalfor all Members and its estimate only dependent on the number ofall observed disputes and of the sum of bilateral export flows be-tween all trading partners. Hence, improved model prediction ismerely owing to changes in the distribution of export flows overMembers by weighing the relevant exports flows, i.e. introducingthresholds for accounting only export flows beyond a certain value.The average number of export flows declines from 4254 in case ofno threshold to 489 when the highest threshold of $700 K is used.The fit of the model is measured by two different indicators: thefraction of exact predictions and the mean of absolute deviation(MAD) between observed and predicted disputes

MAD ¼ 1m

Xjci � bci j; ð9Þ

where ci denotes the number of observed and bci the number of pre-dicted disputes of Member i and m assigns the sample size of 53Members.

The fraction of exact predictions (FEP) rises from 0.13 to 0.43and thereby shows an increased fit of the model by imposing athreshold of $300 K but the MAD remains almost unchanged. Themodel’s fit is not improved by imposing higher thresholds than$300 K as the MAD remains almost unchanged around two andthe FEP is only raised from 0.43 under threshold $300 K to 0.45 un-der threshold $700 K. Contrary to Horn et al. (1999) this resultshows that the selection of trade flows is only relevant for the dif-ference between no threshold and the lowest threshold of $300 K,but seems not to be relevant for the selection at higher values oftrade in the agro-food sector. Hence, their findings that the pattern

of dispute initiation is to a large extent reflected by differences inMembers’ diversity and value of trade is only partially supportedby our result for the agro-food sector. Table 6 comprises the resultsfor the restricted model.

For the unrestricted model, the Akaike information criterion isused to select the relevant variables. Based on this, the incorpora-tion of additional variables is traded off against the increased fit ofthe model. By incorporating additional explanatory variables thegoodness of fit is improved regardless of the number of free param-eters in the data generating process. The indicator penalizesincreasing complexity thus mitigating the danger of over-fitting.It is then sought after the model specification showing the lowestinformation criterion value. All different model specifications areevaluated, i.e. all specifications are estimated and their corre-sponding information criterion value calculated. For each thresh-old the best specification, i.e. that one yielding the lowestinformation criterion value, is then selected and subject to a teston joint significant influence. For the best model under eachthreshold, standard errors of the coefficients are derived usingbootstrap methods. The quality of the unrestricted model is furtheron validated by a likelihood ratio test. In this process the logarith-mic likelihood function value of the unconstrained ML estimator bb,is compared with the likelihood function value of the constrainedML estimator eb, which is obtained by maximizing the logarithmiclikelihood function subject to the linear restrictions fbk ¼ 08k–0.The LR test statistic is computed as

LR ¼ 2 ln L bb ci;xi; nij� �

� ln L eb ci; nij� �h i

; ð10Þ

which has a Chi-squared distribution with degrees of freedom equalto the number of imposed restrictions.10

According to this proceeding six of the considered determinantsare retained in the final model: (1) endured protectionism, (2) ownimposed protectionism, (3) legal capacity, (4) influence of privateactors, (5) WTO membership time and the (6) importance of theagro-food sector result in a sufficient increase in the goodness offit for no threshold. For the application of all higher thresholdsthe variable legal capacity is discarded in the selection processwhereas the other indicators are included in the specification.Table 7 comprises the results for the selected specifications ofthe unrestricted model subject to different thresholds for exportflows. The standard errors are given in brackets behind the respec-tive coefficients. All included variables show the hypothesizedsign. The influence of endured protectionism, own imposed

1

saMDa

Table 7Results for unrestricted specification selections subject to different thresholds forexport flows. Source: Own compilation.

Threshold on export flow values

Explanatory variables $0 $300 K $500 K $700 K

BetaO �13.46 �11.45 �11.29 �11.17Endured protectionism by trade

partner

*2.54(1.57)

***3.60(1.35)

***3.73(1.39)

***3.83(1.41)

Own imposed protectionism *�2.34(1.21)

**�2.44(1.21)

**�2.45(1.21)

**�2.43(1.21)

Legal capacity 0.92(1.04)

Notincluded

Notincluded

Notincluded

Influence of private actors 0.96(0.82)

0.45(0.67)

0.37(0.67)

0.31(0.67)

WTO membership time 4.55(3.83)

*5.07(3.60)

*5.13(3.62)

*5.18(3.64)

Importance of agro-food exportsector

1.34(1.50)

0.67(1.22)

0.65(1.23)

0.62(1.23)

Mean of absolute deviations 1.13 1.09 1.11 1.06Fraction of exact predictions 0.49 0.51 0.51 0.53Level of significance for

likelihood ratio test onspecification

1% 1% 1% 1%

Standard errors are given in parentheses.Number of observations: 53.* Significant at the 10% level.** Significant at the 5% level.*** Significant at the 1% level.

11 The ACWL is a WTO institution supporting developing countries with generalegal advice on WTO matters and was established in 2001 at Geneva. Its ‘‘Roster oExternal Legal Counsel” is available at: http://www.acwl.ch/e/news/news-00003.html, 20.01.2010.

C. Götz et al. / Food Policy 35 (2010) 154–162 161

protectionism and WTO membership time is proven to be statisti-cally significant for the thresholds $300 K–$700 K; for the lowestthreshold only endured protectionism and own imposed protec-tionism show a significant influence. The variables’ joint significantinfluence is verified by an asymptotic significance test based on thebootstrapped sampling distribution of the estimator (see Efron andTibshirani, 1993). Compared to the restricted model, the FEP is sub-stantially higher for the lowest and slightly higher for the remain-ing thresholds. The MAD between observed and predictedcomplaints is substantially lower under all thresholds comparedto the restricted model. This is mainly due to improved modelbehaviour for Members with a large number of observed disputes,predominantly for the EC and the US. Both measures show that themodel amendment is much higher for the $300 K threshold and thespecification without threshold.

The probability to complain per export flow covers a widerange: for the specification without threshold the highest probabil-ity is 489 times, for the highest threshold it is 589 times the lowestprobability. However, Member’s activity in dispute initiation can-not be inferred from its probability to complain without consider-ing the number of its export flows: Being one of the two mostactive users of the system, the probability to complain for the ECfalls into the lower third whereas the probability of Australia, Can-ada and the US belong to the highest for all thresholds. For the$500 K threshold the probability of the US constitutes 3.77 timesthe EC’s probability. Corresponding to their reciprocal ratio withrespect to their bilateral export flows (1:28) this results in 31 pre-dicted disputes for the US (but 34 actually observed) and 30 for theEC (only 24 actually observed).

The likelihood ratio test proves a significant amendment of themodel based on the incorporation of the addressed determinants.For all thresholds the concerned variables’ contribution could besubstantiated at a 1% level.

The findings of Horn et al. (1999) on a significant influence oflegal capacity could not be supported in our analysis of food-re-lated disputes. The variable is selected for the lowest thresholdbut does not survive the significance test. This may be explainedby the fact that legal capacity increasingly becomes an internation-ally tradable good such that each Member can purchase legal

expertise, provided that it has sufficient financial resources. TheAdvisory Centre on WTO Law (ACWL)11 lists eleven law firms andfour individuals on its ‘Roster of External Legal Counsel’. On the otherhand, the findings of Bown (2005) with respect to the influence ofmonetary means is not confirmed by our results either. Therefore,it seems more likely that legal capacity and monetary means aremore relevant determinants for the overall number of dispute initi-ations but simply less important for the variation of probabilitiesacross countries for the smaller food sector.

The indicators Influence of private actors and importance of theagro-food export sector are selected under all thresholds based onthe Akaike information criterion but their statistically significantinfluence could not be substantiated.

The indicators on governmental efficiency and relevance of theagricultural sector did also not survive the variable selection pro-cess The latter might simply be an insufficient proxy for the rele-vance of a Member’s agro-food-industry.

Conclusions

This paper presented an analysis of the determinants for initiat-ing WTO disputes related to the agro-food sector. Apart from thisnew sectoral focus, the analysis extended the literature with amore in-depth analysis of potentially relevant determinants. Theempirical model representing the number of initiated disputes bycountry as a sequence of Bernoulli trials – with probabilities mod-elled by a logistic distribution – was applied to 53 WTO Membercountries.

The results show that some of the determinants relevant in pre-vious dispute studies such as legal capacity and monetary meanscould not be confirmed as statistically relevant in the context ofthe agro-food sector. It could be shown that increasing own protec-tionist attitude lowers the probability to complain and the level ofprotection faced by a country leads to an increase as both variablesprove to be statistically significant determinants of dispute initia-tion in the agro-food sector. At the same time, the duration ofWTO membership clearly contributes to a larger likelihood to ini-tiate a WTO dispute. Though selected under all models with theirexpected sign, the Influence of private actors and the Importanceof the agro-food export sector do not turn out to show a significantinfluence. This is also the case for the indicator legal capacity, thathas only been selected under the model incorporating all exportflows.

The outcome on endured and imposed protectionism showsthat trade restrictive policy measures and the willingness to fightfor market liberalisation are important drivers in WTO disputeswhich is intended by the Dispute Settlement Understanding. Thevariables legal capacity and influence of private actors are selectedwith their expected sign but the hypothesis of no statistical influ-ence could not be rejected. From the validation of legal capacitywould follow that there are resource constraints in WTO disputesand that reform approaches should consider to strengthen devel-oping country Members in the process of preparing and pursuingdisputes, e.g. by means of the ACWL institution. A validation ofthe influence of private actors would suggest to promote proposalsof encouraging Members’ private sector in providing dispute set-tlement related legal assistance, as analysed by Bown and Hoek-man (2005).

Further research should focus on the improvement of data qual-ity to validate or disprove the findings on insignificant influencesof some variables, for example the importance of the agro-food

lf

162 C. Götz et al. / Food Policy 35 (2010) 154–162

sector for the country considered. A generalisation of the modelallowing to simultaneously incorporate characteristics of thedefendant country would also be very useful. Currently, the im-plied assumption that probabilities to be a defendant is equalacross all countries could only be partially mitigated by includingthe determinant endured protectionism of the complainingcountry.

References

Bagwell, K., Staiger, R.W., 2000. GATT-Think. Working Paper 8005, Cambridge.Besson, F., Mehdi, R., 2004. Is WTO dispute settlement biased against developing

countries? an empirical analysis. In: Paper Presented at the SecondInternational Conference on ‘‘European and International Political andEconomic Affairs”, Athens, Greece, May 27–29.

Bown, C.P., 2002. The economics of trade disputes, the GATT’s article XXIII, and theWTO’s dispute settlement understanding. Economics and Politics 14 (3).

Bown, C.P., 2004a. Developing countries as complainants and defendants in GATT/WTO trade disputes. The World Economy 27 (1), 59–80.

Bown, C.P., 2004b. On the economic success of GATT/WTO dispute settlement. TheReview of Economics and Statistics 86 (3).

Bown, C.P., 2004c. Trade disputes and the implementation of protection under theGATT: an empirical assessment. Journal of International Economics 62 (2), 263–294.

Bown, C.P., 2004d. Trade Remedies and WTO Dispute Settlement: Why are so FewChallenged? Working Paper, Brandeis University.

Bown, C.P., 2005. Participation in WTO dispute settlement: complainants, interestedparties, and free riders. The World Bank Economic Review 19 (2), 287–310.

Bown, C.P., Hoekman, B.M., 2005. WTO dispute settlement and the missingdeveloping country cases: engaging the private sector. Journal ofInternational Economic Law 8 (4), 861–890.

Busch, M.L., Reinhardt, E., 2003. Developing countries and general agreement ontariffs and trade/world trade organization disputes settlement. Journal of WorldTrade 37 (4), 719–735.

Dam, K.W., 1970. The GATT: Law and International Economic Organization.University of Chicago Press, Chicago.

Efron, B., Tibshirani, R.J., 1993. An Introduction to the Bootstrap, first ed. Chapmanand Hall/CRC Press LLC, Boca Raton. 436p.

EuroCare, 2006. Consolidating Bilateral Trade Quantities and Values for AgriculturalTrade Based on an HPD Estimator. Internal Working Paper, Institute for Foodand Resource Economics, University of Bonn.

Hoekman, B.M., Kostecki, M.M., 2001. The Political Economy of the World TradingSystem, second ed. Oxford University Press, Oxford, 547p.

Holmes, P., Rollo, J., Young, A.R., 2003. Emerging Trends in WTO Dispute Settlement.Policy Research Working Paper No. 3133. The World Bank DevelopmentResearch Group Trade.

Horn, H., Mavroidis, P.C., Nordström, H., 1999. Is the Use of the WTO DisputeSettlement System Biased? CEPR Discussion Paper No 2340.

Johnson, H.G., 1953. Optimum tariffs and retaliation. Review of Economic Studies1953 (21), 142–153.

Jackson, J.H., 1997. The World Trading System: Law and Policy of InternationalEconomic Relations. second ed, MIT Press, Cambridge, MA. 453p.

Kaufmann, D., 2004. Corruption, governance and security: challenges for the richcountries and the world. In: World Bank Group, (Ed.), Global CompetitivenessReport 2004/2005.

Kee, H.L., Nicita, A., Olarreaga, M., 2006. Estimating Trade Restrictiveness Indices.World Bank Policy Research Working Paper No 3840, 36p.

Kennan, J., Riezman, R., 1988. Do big countries win tariff wars? InternationalEconomic Review 29 (1), 81–85.

Nordström, H., 2005. The Cost of WTO Litigation, Legal Aid and Small ClaimProcedures. Working Paper. Swedish National Board of Trade. Stockholm.

Petersmann, E.U., 1997. The GATT/WTO Dispute Settlement System: InternationalLaw, International Organizations and Dispute Settlement. Kluwer LawInternational, London. 368p.

Petersmann, E.U., 2003. WTO negotiators meet academics: the negotiations onimprovements of the WTO dispute settlement system. Journal of InternationalEconomic Law 6 (1), 237–250.

Shaffer, G., 2003a. How to make the WTO dispute settlement system work fordeveloping countries. In: Towards a Development-Supportive DisputeSettlement System in the WTO. ICTSD, Geneva.

Shaffer, G., 2003b. Defending Interests – Public–Private Partnerships in WTOLitigation. Brookings Institution Press, Washington, DC. 227p.

Trebilcock, M., Howse, R., 1999. The Regulation of International Trade, second ed.Routledge, London. 612p.

United Nations Conference on Trade and Development, 2002. UNCTAD Handbook ofStatistics. New York and Geneva, 439p.

United Nations Conference on Trade and Development, 2003. UNCTAD Handbook ofStatistics. New York and Geneva, 449p.

United Nations, 2004. Missions Permanentes Auprès Des Nations Unies À Gèneve.Directory Service and Telephone Directory of the United Nations, No. 94.

United Nations, 2007. Criteria for Identifications of LDCs. <http://www.unohrlls.org/en/ldc/related/59/>, (20.01.2010).

United Nations, 2008. Standard country or Area Codes for Statistical Use. Series M,No. 49, Rev. 4 (United Nations Publication, Sales No. M.98.XVII.9), <http://unstats.un.org/unsd/methods/m49/m49regin.htm>, (20.01.2010).

Whalley, J., 1996. Developing countries and system strengthening in the Uruguayround. In: Martin, W., Winters, A.L. (Eds.), The Uruguay Round and theDeveloping Countries. Cambridge University Press, Cambridge. 478p.

World Bank (2007). World Bank Development Indicator Database 2007, <http://web.worldbank.org/WBSITE/EXTERNAL/DATASTATISTICS/0,,contentMDK:20398986~menuPK:64133163~pagePK:64133150~piPK:64133175~theSitePK:239419,00.html>, (20.01.2010).

World Bank (2008). ‘‘Country Classification”, Data and Statistics, <http://web.worldbank.org/WBSITE/EXTERNAL/DATASTATISTICS/0,,contentMDK:20420458~menuPK:64133156~pagePK:64133150~piPK:64133175~theSitePK:239419,00.html>, (20.01.2010).

World Trade Organization, 2007a. WTO News: 1995–1999 Speeches. <http://www.wto.org/english/news_e/sprr_e/seoul_e.htm>, (20.01.2010).

World Trade Organization, 2007b. Dispute Settlement Gateway. <http://www.wto.org/english/tratop_e/dispu_e/dispu_e.htm>, (20.01.2010).

World Trade Organization, 2007c. Understanding the WTO: The Organization –Members and Observers. <http://www.wto.org/English/thewto_e/whatis_e/tif_e/org6_e.htm>, (20.01.2010).