[First published September 12, 2005] This is Part II of my response to Gartzke’s defense against my critique, “The CATO Institute Gets It All Wrong,” (here) of his chapter 2 published in the CATO Annual Ranking of Economic Freedom. In reading my response, keep in mind how this Gartzke chapter was trumped in the very first lines of CATO’s news release:
I will present the rest of Gartzke’s reply that I didn’t cover yesterday, and will intersperse it with my comments in green.
Dr. Rummel argues that I am doing democratization injustice by using the term “impose.” He suggests no alternative term, but references another blog post titled “Unchaining Human Rights, Not Imposing Democracy.” Certainly, “unchaining” sounds more affirmative, just as “freedom fighter” sounds more affirmative than “terrorist.” By “imposed,” I meant situations like Iraq, where democracy has not evolved endogenously. [RJR: What difference should this make in understanding that Iraqis were freed from a bloody tyranny?] In Iraq, for example, unless democratic peace exists and is general (monadic), there can be no robust effect of democratization because other states in the region (besides Israel and Turkey) are not democracies. Research by Hegre (2004) shows that increasing democracy when few states are democratic tends to increase, not decrease, conflict. Even many advocates of democratic peace doubt that democratization in the Middle East will lead to peace in anything but the very long run. [RJR: This misses what is perhaps most important about a democratic Iraq: people are free; their government will not murder, rape, and arbitrarily imprison them; it will not support terrorism, and aggrandize against its neighbors] This, of course, also requires that we assume that US efforts to democratize Iraq will succeed, a debatable claim in its own right.
Dr. Rummel takes my study to task because I point out that the democratic peace observation has recently been limited to prosperous states [RJR: All my research and most I know of have been done on all democracies, regardless of development and prosperity]. Here again, I am simply reporting the evolving consensus of democratic peace researchers themselves. Mousseau (2000) and Hegre (2000) report that an interaction term between variables for democracy and economic development leads the democracy term to become no longer statistically significant. In a newer study, John Oneal himself collaborates with Mousseau and Hegre in further substantiating this conclusion. As the result makes clear, democratic peace, if it exists, is conditioned by economic development. My view is that it is development itself, along with economic liberalization, that explains the peace. [RJR: Leave all these studies aside and just look at he world today. Among the 117 democracies now existing, which range across all levels of development and prosperity, there is no war between any two of them, no expectation of war, and no arming against another. Moreover, there is no violence between any two. Also, I’ve tested this in a variety of ways, and found that even holding development constant, the relationship between democracy and war is robust. See again the chart on this I showed yesterday (here), where “wealth” in equivalent to economic development. See also my Appendix to Saving Lives (here)]
Dr. Rummel claims that my assertions are falsified in my own data. As evidence, he argues that there are no “wars” between democracies. The specific claims that I make, and the data that I use involve militarized interstate disputes (MIDs), a broader category of conflict behavior. Wars are very rare. There are just 44 state participations in wars beginning in 1970, the earliest date for which the Index of Economic Freedom supplies data. Less than 1% of state years (think “man hours”) involve a war. For this reason, democratic peace researchers and others studying conflict among nations have overwhelmingly preferred in recent years to examine MIDs [RJR: Yet, Gartzke has much to say about democracy or economic freedom and war based on this analysis of a data set in which there are no wars for democracies]
Still, it is not difficult to have a look. I examined the Correlates of War project listing of wars (conflicts involving at least 1000 battle deaths per year per participant). I find no statistical relationship between either the index of economic freedom, or the democracy variable, either separately or together, using these data. [RJR: not clear — is this a monadic or dyadic analysis? In any case, there were no wars between democracies in these data either.] The effect of capitalism is either more subtle, reducing conflicts only over a lower intensity, or the sample of wars is too small, or both. In any case, democracy does not have the effects Dr. Rummel claims in these data, even when it is left by itself in the regression. [RJR: I don’t understand Gartzke’s reasoning, when in the data set he originally used and in this one, there is NO WAR between democracies] As a further check on these findings, I also examined data from the Stockholm International Peace Research Institute (SIPRI). These data report conflicts involving at least 25 fatalities. Thus, they are clearly conflicts involving “violence.” Using SIPRI conflicts as the dependent variable [RJR: I also analyzed a larger data set involving these data for the years 1973-2003 and found NO VIOLENCE causing deaths between democracies (see table here).], I am again unable to find a statistically significant relationship linking democracy and peace [RJR: This is an incredible statement, considering the all the data sets he consults show NO WAR between democracies]. I can, on the other hand, find weak support for the suppression of major violence by the economic freedom variable. [RJR: The methodological error is in turning a point prediction into a linear correlation one. The best way of testing a point prediction is by a contingency count, and this is what in effect I did with the table I referenced above] This variable is just short of the 5% significance threshold in a quick statistical comparison of democracy and capitalism as determinants of peace. [RJR: Gartzke is again applying statistical significance inappropriately to a population, and not to a sample that is assumed in statistical inference– there is, however, a way of applying the test, and that is if one is assessing the probability of getting a specific combination of data, but this is not what Gartzke is doing]
So, to summarize, Dr. Rummel’s critique that I should look at wars seems unfounded, though it did not hurt to check. The claim that democracy generally causes peace is again unsupported. [RJR: Again, in all the data he uses, there is either no wars, or no violence, although in a much longer set of data over almost two centuries, as I indicated in Part I, there are three minor cases of violence between democracies. In other words, the democratic peace holds, regardless of what Gartzke says]
Dr. Rummel claims I am using the wrong data and that my study “confounds nonviolence with violence.” I am not sure what this means. Every Correlates of War Project MID involves threats or acts of a militarized nature, almost all of which involve violence (the threshold for inclusion in the dataset is high, resulting in relatively few threats and more “uses of force”) [RJR: The data he is using involves both violence and nonviolence as a test of a point prediction that democracies do not commit violence against each other, or have less violence than other regimes. This is the dependent variable. It includes nonviolence, and thus as a test, confuses violence with nonviolence in whatever relationship is found may involve to some unknown degree nonviolence. This means that it is impossible to say then what is found about the relationship between democracy and violence]. Again, I rely on the same data as democratic peace researchers, the most widely used and referenced data, in fact, in the quantitative study of international relations. For Dr. Rummel to claim that the MIDs data are not an appropriate framework for testing the democratic peace is to reject most studies of democratic peace out of hand, something I, and most other researchers, are unwilling to accept [RJR: If they do regression analysis on data that combines violence and nonviolence, then I do reject it as inappropriate to the democratic peace hypothesis. I’m unimpressed with how many do this. ] Still, it would be nice to establish that my findings do not depend on a particular kind of data source. MIDs, COW wars, and the SIPRI data code conflict behavior of a given intensity level or higher. The Interstate Crisis Behavior dataset, on the other hand, examines crises. This can be useful because some conflicts, even relatively violent ones, do not involve direct leadership decisions. Suppose some sergeant decides to lob mortar shells at the enemy, perhaps because he is tired, irritated, or afraid. This would be a MID, and possibly a SIPRI conflict, depending on casualties, but it would not be an ICB crisis if the actions of the sergeant were not initiated by national leaders. The ICB data have also been used in studies that support the democratic peace (Hewitt and Wilkenfeld 1996), and potentially better reflect some of the arguments made about why democracies should be more peaceful. If democracies are more peaceful in any context, it should be in situations where decision making is explicit, conscious, deliberate, and not the result of accidents on the front lines. Results using the ICB dataset, however, are largely the same as those I report for MIDs in my chapter in the 2005 edition of Economic Freedom of the World. [RJR: Hardly the same. This is what Jonathan Wilkenfeld, Michael Brecher, and Sheila Moser, who created the data set, say in their book, Crises in the Twentieth Century:
The more authoritarian a regime the greater was the probability of violent crisis triggers [provoking crisis by the use of violence ] (democratic–37%, civil authoritarian–49%, military–56%)…. As expected, there was a higher frequency of violence responses by military regimes (50%) than by democratic (30%) and civil authoritarian regimes (33%). Non-violent military responses were most often employed by democratic regimes (32%), compared to 20% for the other two types . . . . in short, the effect of type of regime on an actor’s responsive behavior was evident for violent response–the more authoritarian a regime the more likely its response to a crisis would be violent.
The data on crisis management technique reveal an even sharper escalation of violence–(democratic (37%), civil authoritarian (49%) and military regime (63%), with a considerable higher tendency toward full-scale war as well (18% and 21% for civil authoritarian and democratic, 39% for military regimes). Conversely, democratic regimes which were most likely to perceive non-violent acts as triggers to their crises tended to choose pacific [crisis management techniques], with negotiation the most frequent among them: it was highest among democratic regimes (32%), and dropped to 11% for military regimes.
Dr. Rummel argues that colinearity between economic freedom, other variables, and democracy interfere with the effect of democracy on militarized disputes. As Dr. Rummel almost certainly knows, but did not explain to the reader, multicollinearity is not a severe problem in multivariate analysis until correlations are quite high, on the order of 0.9 (he argues they are 0.7. I find that the two key variables correlate at 0.4135) [RJR: This is wrong. The effect of multicollinearity is linear. The higher the correlation between two independent variables, the more this effects the regression coefficients.]. Similarly, the idea that democracy creates capitalism is, I think, questionable. [RJR: Not creates, but promotes] Few, if any, of the archetypal laissez faire economies of nineteenth century Europe would be considered democratic by contemporary standards, though they became democratic in time. Similarly, in South Korea, Singapore, Taiwan, and elsewhere in recent decades, capitalism and development gave rise to pressures to democratize, not the other way around. [RJR: I do not say that democracy is a necessary condition for capitalism. Authoritarian systems can be capitalist as well.] Rather than treat democracy as a gift of the gods or something that landed from outer space, it seems more reasonable to recognize that democracies formed out of the same soup as did contemporary capitalism and economic development. In any event, the claim that capitalism and democracy are correlated does not obviously lead to the conclusion that democracy should be given preference (or deference) as the key contributor to liberal peace. If the two processes are related, then why treat one as if it is important and the other as if it does not exist? [RJR: The simple reason is that democratic countries, regardless of their internal socialism and level of development are most peaceful and don’t make wars on each other, whereas authoritarian countries that are capitalist, regardless of level of development, are on the average much more violent and do make wars on each other].
Yet, again to be safe, I remove all of the variables from the regression model, except democracy. Democracy is not remotely statistically significant, even with no competitors (P value 0.448). Maybe economic freedom gets “help” from the other variables? I ran the regression model with just democracy and the freedom index, and find that economic freedom is statistically significant (P value 0.001), while democracy is insignificant. [RJR: Irrelevant, since in his data there are no wars between democracies, and as I say, in almost two centuries, only 3 minor cases of violence] The claim about sampling is debatable, and is debated, in the literature [RJR: No, it is not debatable in the case of statistical inference. Read your statistics books]. Whether we observe all possible states of the world, or just the ones that came to pass in this iteration of history hinges on issues outside the realm of the knowable. Democratic peace researchers have consistently used the statistical significance of democracy as evidence of the validity of their claims. How else can I challenge the conventional wisdom? [RJR: Use the three rules of research: (1)You look at your data and get as familiar with its details as with your own body, (2) you fit the method of analysis to the nature of the data, which is to say (3) you study the assumptions, problems, and interpretation of the method before applying it. The appropriate method in your case is contingency analysis].
At several points, Dr. Rummel notes that “there are NO (zero) wars between democracies over almost two-centuries.” This sounds persuasive, but note that the claim treats as a conclusion that which is presumably the subject of this debate. [RJR: Huh? It is a statement of fact. There were no wars between democracies.] Is it democracy that makes peace or something closely associated with democracy? Dr. Rummel emphasizes that capitalism is correlated with democracy, but refuses to treat seriously the possibility that it is capitalism that causes peace. [RJR: I explained this above: capitalist nondemocracies do make war on each other, while capitalist democracies do not.] The “two-centuries” claim is also misleading. Democratization is a recent phenomenon in world affairs. How many two-centuries old democracies are there? Indeed, we can also say that over the same period, no advanced free market economies have gone to war with each other, either. [RJR: Okay, look at the 117 democracies today, or the 50 or so several decades ago. The same thing holds].
Dr. Rummel asks “How could CATO let such a poor study into their prime report?” Clearly, this is a rhetorical question, but let me answer it as honestly as I can. The study conforms as closely as possible to the state of the art in democratic peace research. Rather than being “incompetent,” I adopted the same variables and evaluation standards, and a similar research design to those of the most widely cited research program on the democratic peace. That this happened to be the approach of Oneal and Russett and not Rummel is unfortunately a consequence of the greater popularity of the former among researchers and the wider public. Dr. Rummel does not like the choices I made in my analysis, but he does not like the choices made by other democratic peace researchers either. Differences between Dr. Rummel’s views and those of the larger democratic peace research community were not made clear in his comments, a possible source of confusion.
At the same time, I do not claim that my findings are definitive. They are a cautionary tale that gives some backing to those who are concerned that enthusiasm for the democratic peace has exceeded good judgment. [RJR: that this is a cautionary tale is inconsistent with the flat statements made in the study] No doubt this is not the end of the debate, though I hope Dr. Rummel and other interlocutors will cease from impugning my professional reputation every time I offer evidence that differs from their conclusions. Science is a perpetual learning process, in which we gradually whittle away at uncertainty. The fervor with which researchers on the subject hold to their respective visions of democratic peace should itself lead intelligent observers to caution. [RJR: Gartzke does not even try to justify his and CATO’s claim that “economic freedom is almost 50 times more effective than democracy in restraining nations from going to war.” To make this claim based on data that show no wars between democracies, and to criticize the American attempt to help democratize Iraq based on this, is where the incompetence comes in.]
Let me add in closing that, while the study Dr. Rummel critiques does not directly contradict the dyadic version of the democratic peace, my other research does. I have replicated the major dyadic studies of Oneal and Russett and others, using several indicators of capitalism, including but not limited to, the Index of Economic Freedom. I find that democracy does not sustain a dyadic effect on conflict either (there is not even a special peace among democracies), when appropriate measures of global market integration and economic development are introduced. I have shared these findings with democratic peace researchers (John Oneal, Bruce Russett, Erik Weede, Patrick James, James Lee Ray, to name a few), and expect that they will soon be available in print. Of course, I will also provide copies to Dr. Rummel, if he wishes.
In sum, Gartzke does not provide any persuasive evidence against the democratic peace, and in fact his data, as well as my own analyses of the other data to which he refers, confirms the democratic peace. Moreover, my contingency analyses of the relationship between democracy, development, and violence show the dominance of democratic freedom. Finally, the claim that stimulated my critique of Gartzke’s study — that economic freedom is almost 50 times more effective than democracy in restraining nations from going to war — remains ridiculous, and also undefended by Gartzke.
Yes, it matters, because free economies are not exclusive to democracies. Authoritarian states with free economies do make war on each other and murder their citizens.
This is why you can’t depend on public services to alert you to a global epidemic of bird flu. It’s up to us blogsters.
Typical and predictable.