The Behavioral Origins of War

Notes

249 NOTES CHAPTER 1 1. Some recent theoretical and simulation-based research lends credence to Riker’s pessimism. See Lewis and Schultz 2002. 2. For a detailed discussion of this debate and an alternative perspective, see Walt 1999. 3. See, for instance, Brooks 2000. CHAPTER 2 1. Ray and Wang (1998) also advocate the use of dyads as a way of integrating levels of analysis in international politics. For a critical perspective on the use of dyad-years, see Raknerud and Hegre 1997. 2. Of course, many actual models are nonlinear in form; we use the typical regression equation to illustrate our argument. 3. There is a variety of statistical tests to help judge the severity of the problem , including the Hausman test, which we discuss in chapter 3. Unfortunately, the treatments are sometimes worse than the disease. 4. Of course, exceptions to this generalization exist. For example, Oneal and Russett (2001) go to some length to lay out the complementary effects of international organizations, democratic institutions, and international trade. 5. It would also be appropriate to use the median value as the baseline, as some of the variables have highly skewed distributions. For our analytic cut points, we used plus or minus one standard deviation. If we used plus or minus two standard deviations, the explanatory power of some of the variables would appear somewhat stronger. CHAPTER 3 1. This approach does pose an interesting dilemma from the perspective of Kuhn’s (1986) argument about paradigmatic shifts in the evolution of knowledge. 2. The list of books and articles that we associated with each theoretical perspective is in table 3.1, along with our annual count of citations. 3. Oneal and Russett (1999b) drop the ongoing years of the world wars, leading to a data set with some, but not all, ongoing dispute/war years included. 4. We also considered Jackman’s methods (1998, 2000). He examines several computational methods of dealing directly with errors and correlations over time in limited dependent variable data. Unfortunately, the solutions he presents are not practical, given data sets of the size we employ here. 5. See . See also Neuhaus , Kalbbeisch, and Hauck 1991, which considers the interpretations of the regression parameters in these two general approaches. Neuhaus, Kalbbeisch, and Hauck show that the parameters of cluster-speciac (the axed-effects estimator ) and population-averaged models for correlated binary data describe different types of effects of the covariates on the response probabilities. 6. It might also be that the heterogeneity lies solely in the disturbance term. We discuss this possibility and corrections later. 7. Many statistics texts discuss multinomial logit in more detail, including Hanushek and Jackson 1977; Maddala 1983; and Greene 1993. 8. We use Stata 6.0 and 7.0 for our estimation. For the splines, our quartile cut points for the spline fell at six, fourteen, and twenty-six years. 9. More speciacally, the steps involved in this procedure are as follows. (1) Start with the initial data and estimated coefacients. (2) Change the value on an independent variable of interest to a given value for all cases in the data. (3) Use the coefacients from the model to create predicted outcome probabilities for each case. Each case now has probabilities computed as if the case had the modiaed value of the independent variable of interest but with all other variables at their actual values. (4) Compute the average probabilities across the data set, giving an overall estimate of outcome probabilities at the speciaed variable value. (5) Repeat with different independent variable values and compare. CHAPTER 4 1. For the arst data sets on alliances, see Small and Singer 1969; Singer and Small 1966. On the size of alliances, see Olson and Zeckhauser 1966 for the seminal argument therein. For a complete survey of many of these issues, see Snyder 1990. On the origins of alliances, see Walt 1987. On the links between alliances and sovereignty, see Morrow 1993; Gartzke 1998; Siverson and Emmons 1991. For more on the pernicious effects of secret alliances, see Ritter 2001. 2. See Adler and Barnett 1998 and Deutsch 1957. 3. On the nature of signaling games, see Fearon 1994b, 1997; Morrow 1992; Schultz 1998; Kydd 1997. 4. See, for example, Lebow 1981 on the role of psychological processes in conbict escalation. For a more detailed explication of psychological constraints on decision making in foreign policy, see McDermott 1998. 5. We tried alternative measures, including a continuous measure and a three-year moving average...