International Institutions and Issue Linkage February 2004 TABLE 3.Ordinal Probit Regression Models Japan EU Variable Base Model Fixed Effects Base Model Fixed Effects Cross-sector linkage 0.635* 0.914m 0.490° 0.609* (0.183) (0.211) (0.203) (0.215) Threat 0.388* 0.508 0.079 0.148 (0.112) (0.165) (0.123) (0.136) Trade balance -0.010 -0.026 0.053* 0.056* (0.011) (0.015) (0.015) (0.015) Product value to the U.S. -0.059 -0.077 0.001 0.004 (0.033) (0.046) (0.022) (0.025) Budget growth -0.023 -0.006 -0.025* -0.032 (0.017) (0.022) (0.007) (0.008) Economic growth 0.017 -0.015 0.571* 0.632* (0.062) (0.069) (0.171) (0.172) LDP strength 0.117 0.139 (0.027) (0.040) French Council presidency -0.906* -1.068* (0.303) (0.329) Export subsidy -1.114* -1.693* (0.267) (0.413) Multicountry demand -0.360 -0.271 -0.791 -0.648 (0.307) (0.454) (0.527) (0.551) Count -0.170 -0.004 0.140 -0.132 (0.116) (0.123) (0.101) (0.116) First threshold 6.699 7.386 -0.885 -0.234 (1.502) (2.195) (0.669) (0.798) Second threshold 7.690 8.770 0.391 1.169 (1.522) (2.243) (0.655) (0.783) log-likelihood -137.283 -102.399 -99.608 -90.952 T statistic 0.665 0.860 0.754 0.790 (0.075) (0.042) (0.076) (0.072) N 153 153 114 114 Note:The constant is set to zero.Robust standard errors are in parentheses.The fixed effects models include 17 commodity categories for the Japan data and 13 commodity categories for the EU data,although these coefficients are not reported here.Significant at the 5%level.Significant at the 1%level. indicator variables.This model attempts to capture through selecting the agricultural sector for analysis product-specific effects related to the unobserved po- and including a commodity fixed effects model to allow litical or economic characteristics of the commodity be- for variation among agricultural products.17 Examples ing negotiated.For example,some products with more such as the decision to negotiate Japan's rice import ban political clout in terms of the number of farmers or in the Uruguay Round also indicate that cases with high campaign contributions,etc.,could be less likely to be li- domestic opposition can arise in linkage negotiations. beralized.Following the OECD Agricultural Accounts This reduces concern that only cases that will be easy to commodity categories,I classified the 153 Japanese liberalize have a strong linkage.Nevertheless,more sys- cases into 17 commodity groups and the 114 EU cases tematic study is necessary to understand the selection into 13 commodity groups.16 The fixed effects models mechanism for the institutional setting of negotiations. give results similar to the base models used for the anal- Table 4 shows how a shift in the level of one variable ysis below.Given the possibility for some correlation will affect the predicted probability of major liberaliza- among negotiation cases,the estimates in Table 3 are tion while all other variables are held constant.18 Use based on Huber/White robust standard errors. It is possible that there is endogeneity in the agenda- setting process so that the level of domestic opposition 17 An alternative approach would be to use a two-stage selection model where the choice of negotiation forum is modeled in the first to liberalization influences whether there is a strong stage and the effect of linkage is estimated in a second stage.The linkage.This article has addressed this issue in part use of such selection models requires a valid instrument,a variable that affects the negotiation outcome only through the linkage choice. The recent literature has found that without such an instrument,the 16 See the Appendix for a list of commodity categories In some cases, model could yield misleading results (e.g.,Heckman et al.1998). I made a deviation from the OECD categories,such as separating rice In this case,it seems that any variable that is related to choice of into a separate category for Japan while leaving it in the cereals group negotiation agenda is also related to the other variables that influence for the EU,where there is less reason to expect special treatment. liberalization outcomes. Processed products are included in the closest raw material category. 18 Holding other variables constant involves setting variables to their e.g.,pasta is in the cereals group.Industrial crops is used as the base mean levels and multiplying these means by their coefficient esti- category. mate.In Table 4,I only summarize the change in the top category of 162 Reproduced with permission of the copyright owner.Further reproduction prohibited without permission.Reproduced with permission of the copyright owner. Further reproduction prohibited without permission