Category Archives: quantitative methods
The past few posts have been quite critical of Neorealism. I thought it might be interesting to say something qualified in the theory’s favour. In Networks of Nations, Zeev Maoz presents what he calls the theory of networked international politics. It hasn’t been discussed very much on IR blogs, apart from by Braumoeller over at the Monkey Cage during a discussion of systemic theory and networked vs. traditional approaches. Maoz’s theory applies familiar mechanisms drawn from existing IR theories to analyse social network patterns of interstate interaction. Drawing on Neorealism he analyses patterns of strategic interaction, Liberalism patterns of mutually beneficial self-interested interaction, Constructivism patterns of homophily and the formation of different cultural cliques, and World Systems Theory patterns of inequality and hierarchy. He also makes some novel claims about prestige and status in the network of nations, developing some original hypotheses that develop concepts within social network theory (but which also echo recent and classic discussions of status in international relations). Interestingly, he finds that each theory makes successful predictions in its own core area of competence. Combining the different mechanisms, Maoz puts forward a complex and detailed account of how the spillovers generated by cooperation amongst security-seeking states results in an increasingly complex, evolving network structure of cooperation within the international system. Democratisation within strategically interacting groups of states tends to reduce competition and accelerate the formation of cooperative relationships. Yet the system remains unequal and hierarchies of status may generate conflict.
This is only a brief summary of a dense, detailed book that examines dozens of hypotheses and a huge range of interactions at different levels of analysis within different areas of world politics. One area of focus is what Maoz calls the strategic reference group (SRG), which he refers to as the basic unit of for the analysis of national security policy within his framework. An SRG is a state’s security environment, it is made up of all the states who are likely to be perceived as threats to a state – those the state has been involved in military disputes with, those the state has been in a rivalry with, and allies of both sets of ‘enemy’ states (the friend of my enemy is my enemy). Neorealists believe that states often engage in ‘internal balancing’, arms build-ups, if they believe that they are vulnerable. The military capabilities of others are dangerous unless they are checked and negated by counterveiling power. Building up a state’s own power resources is one way a state can make itself less vulnerable, forming alliances is an alternative strategy. Do states actually behave this way? Maoz finds that yes, states will tend to build up their military capabilities if they are weak relative to their strategic reference group*. Forming alliances makes a state less likely to engage in a military build-up. So it seems that isolated states with many powerful enemies seek to increase their capabilities. States that are powerful, have powerful allies and have few enemies may feel themselves safe enough to reduce their capabilities – perhaps opting for butter instead of guns.
This at least seems to provide support for a Neorealist perspective, especially the Defensive Neorealism of Waltz, Walt and Glaser, as Maoz’s results are consistent with the hypothesis that states seek security by attempt to achieve ‘enough’ power. But there are one or two problems. First, Maoz’s definition of the strategic reference group isn’t thoroughly Neorealist. In Maoz’s analysis, states are not concerned about all other states, just specific threatening states. States have specific issues and disagreements with past enemies, strategic rivals, and their allies. Anarchy is not necessarily a war of all against all in the theory of networked international politics. Second, I’m slightly cautious about the findings as there could be other factors not included in the model that might lead to states within SRGs to experience a growth in military personal and expenditure at the same time or sequentially. Economic growth and industrialisation are likely to be regional processes and may be associated with a rise in military capabilities. Third, Maoz notes that states that are very strong compared to their SRGs are less likely to get involved in conflict. Does this support or undermine Neorealism? If Neorealists are committed to the hypothesis that unipolar global and regional inter-state systems are very unlikely to form and very unstable because they provoke counter-balancing coalitions, then this seems to undermine the theory. Fourth, although Maoz argues that these results are robust and in line with Neorealist expectations, he notes that they don’t have a great deal of predictive power. So Neorealist theories don’t explain all that much how states respond to potential threats – perhaps (as Maoz suggests) because states have many foreign policy tools available to them apart from military build-ups and alliance formation.
Maoz examines many other processes within the global inter-state system in further detail, again finding that some Neorealist claims are borne out. But so are the claims of other theories, such as democratic peace theory. Even on a charitable interpretation of patterns of conflict and cooperation, Neorealism provides an incomplete account of the phenomena that it was created to make sense of.
Maoz’s Networks of Nations is a fascinating application of familiar theoretical frameworks to a host of levels of analysis within the international system – social network analysis seems to offer a veritable smorgasbord of possible levels of analysis for international relations scholarship. Focusing on indirect as well as direct interactions between states offers a very powerful set of tools for scholars looking for behavioural patterns in world politics. It would be especially interesting to see a network analysis of other middle-range theories in IR – is for example territorial peace and conflict a network phenomena? In my view such research would help us think creatively and rigorously about the nature of international systems and move past exhausted theories.
*I think that there is a mistake on page 143, which states that the variable State/SRG capability imbalance is generated by ‘subtract[ing] a state’s military capabilities from the sum of the military capabilities of its SRG members’. The name of the variable, the discussion in the text, and the operationalisation of the variable in the case where a state has no SRG all suggest that this is a typo and that the actual operationalisation is the state’s military capabilities less the sum of the capabilities of the states in its SRG. This operationalisation is consistent with Maoz’s interpretation of the negative coefficient for the variable in regression model for military build-ups: ‘As the difference between the capabilities of the focal state and the aggregate capabilities of its SRG increases, the extent of the absolute and rate-of-change in the state’s capabilities declines’. In other words, states with large capabilities and a weak set of states in their SRG (i.e. a positive State/SRG capability imbalance) are likely to reduce their military capabilities.
Two years ago prominent Neorealist authors Mearsheimer and Walt lamented the current inattention to grand theory in IR as an academic field (which they also summarised over at Duck of Minerva). They argued that ‘simplistic hypothesis testing’ had replaced inquiry into the fundamental features of world politics and the debate among rival intellectual frameworks seeking to make sense of international relations. There probably is too much ‘simplistic hypothesis testing’ in IR, the use of positivist methods to answer micro-questions that no-one has every actually cared about abounds in major journals. But Mearsheimer and Walt’s critique was odd and misdirected in lots of respects, as others noted at the time it was published. Notably, M & W are dismissive of any non-positivist approach such as critical theory and feminism – yet their commitment to positivism hardly sits well with their aversion to ‘normal science’.
I’ve got time for Stephen Walt as a theorist, his version of Neorealism is flexible and non-dogmatic. His analysis in his blog is measured and rejects nationalistic claims about US exceptionalism. But along with Mearsheimer, his criticism of ‘simplistic hypothesis testing’ reads like rearguard action in support of a version of Realism that just doesn’t convince any more, doesn’t provide any useful leads for empirical research and doesn’t warrant any further theoretical refinement.
The example that they use to illustrate ‘simplistic hypothesis testing’ is telling, as it does little to support their case. Research by conflict theorists on strategic rivalries, they claim, produced:
an expanding set of empirical findings but did not produce a broader synthesis or a general explanation of the various positive and negative results. Instead, we get generalizations of the following sort: ‘Dyads that contend in territorial disputes have a greater probability of going to war than is expected by chance,’ or ‘[Enduring] rivals have a greater probability of going to war than other dyads’ (Vasquez and Leskiw, 2001: 308–309). But we still have little idea why.
The field hardly needs realism to tell it that states will oppose threats to themselves (if they can) or that revisionist states will seize opportunities to gain re-wards (especially if the risks are low).
Once again, Neorealism offers little more than the banal observation that international politics is a rough-and-tumble world and that states pursue their interests – if we define interests broadly enough to include just about any possible goal that a state might feasibly pursue. Walt has offered some well-judged observations about contemporary international politics, but these are often made in spite of Neorealism – indeed as corrections to the baseline Neorealist model developed by Waltz. The discipline of IR doesn’t need Waltzian Neorealism as a grand theory, certainly not to the exclusion of critical theory, or middle-range research programmes such as steps to war, or the much more convincing rival systemic theories that despite M & W’s protests are actually out there.
My favourite school of realism? You’ve probably never heard of it (well, you might have done).
I’ve thought highly of William Wohlforth and Stephen Brooks for quite a while, their ‘post-classical’ take on international relations – inspired by Gilpin’s War and Change in World Politics – is much more refreshing than the dogmatic, tortured interpretations of contemporary international politics offered by those few who still cling to the neorealist raft. They are still a little bit straight-jacketed by the core assumptions of traditional realism, but their contributions have helped to unpick many of the unthinking assumptions within debates about power politics. But I’ve become even more persuaded as I’ve realised that this approach is much more consistent with the empirical evidence than Waltz’s neorealism. Its proponents and its critics all agree that neorealism is a positivist theory, or an attempt to provide the foundations of one, but I wonder more and more what its positive, falsfiable claims actually are. I don’t even think neorealists have even defined the balance of power, their central concept, in a clear and unambiguous way. It is a distribution of capabilities, a configuration of capabilities, or a process arising from the individual balancing behaviours of states? Here’s Wohlforth offering a damning verdict on attempts a few years ago to defend what Vasquez would call the neotraditional theory of the balance of power:
The recent decline in the United States’ economic fortunes does not vindicate any prediction made by any balance-of-power realist, has no implications for any theoretical proposition about the functioning of a unipolar system, and has not caused a structural shift to bi- or multipolarity. Things can be made to seem otherwise only when scholars use inconsistent measures of capabilities, do not define terms with precision, forward inherently unfalsifiable arguments, and fail to clarify causal mechanisms.
The Gilpin-Wohlforth-Brooks approach, by contrast, views international relations as a set of overlapping global and regional hierarchies. Conflict occurs as states chafe against their place within the hierarchy. This not only allows dialogue with other hegemonic leadership theories (such as that of Giovanni Arrighi, in honour of whom this blogged was named) but also with quantitative analysis of international conflict – which finds that states of similar material capabilities and close proximity are more likely to experience conflict. All that is needed is a bold, clear theory that unites existing empirical evidence across lines of inquiry and provides a compelling set of causal mechanisms…
But for now, the ‘postclassical’ approach provides the best off the shelf take on geopolitics for students and scholars.
As I said previously, I’ve changed my mind about what quantitative methods can contribute to international relations research. Becoming more familiar with quantitative research has exposed me to the existence of a more diverse set of viewpoints on the appropriate use of statistical techniques and what they can tell us about the social world. I’ve found the anti-inductivist arguments of scholars of the analytical sociology movement and the creative, innovative positivism of Philip Schrodt particularly useful in their criticism of standard practices in quantitative social science.
Another unorthodox perspective is provided by Salvatore Babones. I first became aware of Babones research on the global income distribution a long time ago when I studying for my MA. His work was one of the influences that led me to gradually take the empirics of global inequality more and more seriously, leading me to my current set of interests. Babones, however, is an anti-positivist – something that he considers to be compatible with the employment of statistical techniques. He argues that quantitative methods should not be put in service of theory-testing, which he regards as an attempt to emulate the natural sciences that is of dubious merit when dealing with observational data. Instead, he advocates the use of statistical techniques as powerful tools to enable the researcher to engage in a dialogue with the data as part of a holistic, reflexive research enterprise. This leads him to a surprising conclusion in a recent article:
The goal of interpretive research is not really to answer research questions. The goal of interpretive research is to develop the expertise of the researcher. The decomposition of new environments into basic building blocks that have already been studied and the assembly of those building blocks into conjectural policy solutions is what human experts do. The practice of interpretive data analysis helps them learn how to do it better.
There seems to be some overlap here with the emphasis on the concatenation of mechanisms by analytical sociologists. Interestingly, Babones notes that he is more sympathetic to the use of traditional statistical technique such as regression than some analytical sociologists. Perhaps the difference arises from the more optimistic and philosophically realist position of analytical sociologists: they believe that sufficiently sophisticated and realistic models can succeed at identifying underlying data-generating processes. Babones seems a bit more sceptical, he offers an interpretative perspective in part because he holds that variables are always at least one remove from the entities we are interested in (I wondered if this might dispose him towards latent factor analysis and it turns out he’s edited a book on the topic). In places, Babones’s account seems a bit too inductivist from the position in the philosophy of science that I occupy – but I intend to read his book on Macro-Comparative Research to engage with his standpoint in more detail, as Babones is an expert researcher who has offered a distinct perspective on quantitative methods.
For the first few years I studied international relations I was of the view that quantitative approaches didn’t have a great deal to add to the subject. This view was not, I should note, based on a general antipathy towards the natural sciences or a fear of numbers – both of which are common in the social sciences and the humanities. Instead, I thought that the most important issues in international relations were theoretical and related to issues in philosophy of social science. It also seemed to me that the quantitative research that I was aware often missed the point of the issue it was intending to address, operationalising concepts in a manner that was unconvincing and thus reducing fairly subtle processes to crude measures. It also seemed to be atheoretical and involved throwing a load of different variables into the hopper to see what would come out the other side.
I realised, however, that many of the issues that I was interested in could not be resolved through theory alone. It might seem strange that I ever believed (implicitly) that they could, but a lot of debate in IR and the humanities more broadly seems to involve the evaluation of a set of claims based on a set of theoretical desiderata alone. This is a rut, and its a rut that a lot of areas of scholarship fall into. Engagement with philosophy of science and social theory is important, but it can’t adjudicate between competing empirical claims about topics such as inequality, conflict, democratisation and state formation.
Reading more statistical research, I realised that there was much more in the way of theoretically sophisticated, historically informed quant research in IR than I had credited. I started to take quantitative contributions much more seriously, but interrogate them in more detail. Every statistical model makes theoretical assumptions and those assumptions can be questioned and problematised just like those of any other theoretical claim. Getting into the nuts and bolts of quantitative research on topics ranging from economic growth and inequality, to democratisation and economic development, to strategic rivalries between states gave me a more informed understanding of the contributions and limits of quantitative research in world politics. The most significant problems are that first, the entities social science works with are not stable over historical time, the social world is never really in equilibrium, and so the relationships between variables cannot be expected to be constant across time and space. The second problem is the issue of causal complexity, the way in which causal factors may interact and combine in specific configurtions to produce certain outcomes. The third problem is the lack of reliable data and measures based on well-operationalised concepts.
Many quantitative scholars and methodologists acknowledge these problems, however, and have attempted to devise ways to address these tricky issues. Braumoeller and Wimmer come to mind scholars who have made major contributions to world politics recently through the intelligent, theoretically informed use of quantitative methods. Paul Schrodt has made some searing criticisms of status quo quantitative practice in the study of politics and attempted to push analytical techniques forward in the discipline. His contributions to efforts to make event data useful for research in politics and international relations are very interesting, and I’ll be watching out for what he and others do with the huge new ICEWS database. It’s possible that with data at this level of granuality, quantitative scholarship can move beyond a focus on broad structural correlates of outcomes and towards a greater focus on political processes. Of course, there are pitfalls and obstacles, but I’m more optimistic about the prospects than I was before I really started to engage with this area of research.