4 - Statistical tests

Summary

Chapter 3 provided a plausibility probe of existing statistical analysis and well-known case studies. It found the existing literature on economic sanctions inadequate to the task, while the conflict expectations model received qualified support. A second cut is warranted. This chapter uses events data from Hufbauer, Schott, and Elliott (HSE) to explicitly test the hypotheses developed in chapter 2. It starts by describing the criteria used to include or exclude observations. It then details how the dependent and independent variables are operationalized, using data from HSE and other sources. Finally, the hypotheses predicting the causes and outcomes of coercion attempts are evaluated.

Selecting the appropriate sample

HSE's events data will be used for testing the model's hypotheses. The HSE database has the advantage of a relatively large sample size and an impressive inventory of recorded independent variables. The drawback is that this data set has been criticized for controversial codings and questionable methodology. In addition, the selection of cases could be open to bias. HSE admit that their sample, “probably omits many uses of sanctions imposed between powers of the second and third rank … also, we have overlooked instances in which sanctions were imposed by major powers in comparative secrecy to achieve relatively modest goals. To the extent of these omissions, our generalizations do not adequately reflect the sanctions experience of the twentieth century.”