6 This figure is based on a number of Library of Congress contract documents approved by the 111th, 112th, 113th or 114th U.S. Congress. See www.congress.gov. 67 The treaties are in force: a list of treaties and other international agreements concluded by the United States (1929-2017) [tif]. The dataset is limited to agreements concluded since 1982, as its construction requires cross-references with the Kavass`Guide of Treaty in Force, first published in 1982. For the first time since 1957, the U.S. State Department has not published a separate edition of the TIF for 2013 and 2014, making it difficult to give the exact year of the expiry of the agreements during this two-year period. The analysis ends in 2012. Figure 3 shows the survival estimate and cumulative risk curves for the preferred model (5), which corresponds to a contract and an executive agreement. Digital covariates have been centered around their average. For category variables, the most common value is used.
Footnote 100 Keep in mind that the Grambsch and Therneau tests were developed in a medical context where sample sizes are generally less than one hundred, making the test insensitive to slight disproportion. For sample sizes as large as those in this study, small confidence intervals give significant p values, even if the data show negligible anomalies. In addition to formal tests, it is therefore recommended to conduct a visual examination of the residues of Schonfeld. Footnote 127 Such a visual investigation does not result in significant violations of proportional risk acceptance for one of the material covariates and a violation only for a handful of countries with which the United States generally has few agreements, such as Burma, Ecuador or New Caledonia. The corresponding charts are included in the online appendix. This section presents the results of the analysis. First of all, I consider the differences between contracts and executive agreements in general, without distinguishing between different types of executive agreements. I will check whether the results change if I continue to review the various executive agreements, the ex ante agreements between congresses and the ex post-Congress executive agreements. 103 Formally, it is necessary that the cariats of interest, here the indicator of the contract, are not correlated with the potential result, here the sustainability of an agreement, after the integration of all control variables. This is also called the acceptance of “observable selection.” By definition, if the choice between congressional contracts and executive agreements were random, the community of interest would not be correlated with the potential outcome.
However, without randomization, the choice of acceptance of observables remains unverifiable and is the subject of theoretical debate. Future research could remedy this limitation by sorting agreements to a more detailed level than is possible for the excitement of TIF data. A particularly promising approach could result from the detailed content of individual agreements.