The analysis also shows some interesting results on the effect of the control variables. First, based on the expectation that ministerial pattern of designing independent agencies may matter, dummy variables have been included for the five ministries to which more than ten independent agencies are linked. Looking at the regression results, such patterns do not appear to play an important role in the explanation of variation in formal political accountability. The only ministry which stands out is the Ministry of Transport, Public Works and Water Management. Agencies linked to this ministry are subject to significantly lower degrees of formal political accountability than agencies linked to other ministries. The question is whether this is because the differences in accountability across ministries are not significant or because the variation in accountability across ministries can be ascribed to other variables which are included in the analysis. Analysis of variance (ANOVA) may offer us more insight into this issue. Looking only at the five ministries to which more than ten agencies are linked, the F-statistic in the analysis is significant (F (4, 72) 5 4.99, p , 0.01), which indicates that there are significant differences between the mean accountability scores across ministries.11 However, a Bonferroni test, which analyzes which groups differ significantly from each other, indicates that it is only the Ministry of Transport, Public Works and Water Management which differs significantly from other ministries: the independent agencies linked to this ministry have a mean political accountability score which is significantly lower than the mean score of organizations linked to the Ministry of Education, Culture, and Science and the Ministry of Health,Welfare and Sport (see, for details on the analysis, Table C in the Appendix).