Monday, February 18, 2013

When Is Reductionism Not Appropriate in Theory?

. Monday, February 18, 2013

In "The Reductionist Gamble" Oatley actually makes a weak argument: we can't know when outcomes we observe are independent from each other (what he calls data-generating process (DGP) "A") and when they are not ("DGP B"). In Oatley's framework assumptions regarding the DGP may be trivial in many cases even if they are not always trivial. Oatley, in other words, is arguing that our inability to directly observe the data-generating process should make us humble and encourage us to investigate international politics under a range of assumptions regarding the DGP. Instead, IPE -- at least as it is represented in the APSR and IO -- has honed in on one: the Open Economy Politics (OEP) paradigm, which he calls "reductionist" because it assumes independence of observations, both theoretically and methodologically, and because it recommends beginning analysis at the lowest unit-level (individuals or firms), while bringing in systemic effects only when necessary. Oatley argues, persuasively, that in some cases the assumption of independence will be false, and when it is false the reported empirical results are not reliable.

In my view this argument should be uncontroversial. Oatley is not saying that all OEP research is wrong, nor that most of it is; he's merely saying that OEP contains a critical assumption which ought to be justified more regularly than it is. This is Research Design 101. Since it is not, there is a strong likelihood that some unknown percentage of OEP research has been done under false premises, and is therefore biased in an unknown direction. Oatley re-estimates several empirical models central to the literature, while adding terms which allow for conditional "systemic" effects, and finds that the reported results are called into question.

I'm going to make a stronger argument, in speculative form (i.e. I'm trying to be a bit sensational): almost all of the theoretical traditions in IR/IPE suggest DGP A is not a trivial assumption, but most of our research designs go on to assume DGP A anyway. My argument is that, in the typical case, our empirical design does not match our theoretical structure.

Regarding Oatley's post (and article), it might be helpful to think of some of the major theoretical orientations in IPE -- within which mid-level theories are ostensibly embedded -- and place them into different bins: those which assume independence of observations (DGP A) and those which assume non-independence of observations (DGP B). For these purposes, "observations" refers to realizations of some outcome at a given unit of analysis. The outcome could be "amount of trade" or "incidence of war" or "occurrence of a financial crisis". The unit of analysis could be states, firms, or advocacy groups. To greater or lesser extent, all of the prominent grand theories involve an assumption of non-independence: DGP B. David Lake (and presumably other OEPers) hate isms so this might not be of interest to them, but it's useful as a starting exercise.

1. Various realist theories suggest that states will modify their behavior in response to the international environment within which they operate. In particular, state behavior is conditional upon the underlying distribution of power in the system. If that is the case, then treating states as if they are realizations of independent and identically distributed observations is not theoretically sound: the behavior of state A is reactive to the behavior of state B, and is contingent upon the underlying distribution of power in the international system. Economic coercion, "redistributive cooperation", the structure of trade, and other outcomes are responsive to interactions between states.

2. Institutionalists argue that the behavior of international actors is modified by the ways in which they interact. Membership in common organizations -- e.g. the WTO -- can have an impact on common outcomes -- e.g. increased dyadic trade -- even after controlling for domestic factors. The US, for example, does not decide how much it should trade based solely on local factors; how much the US trades depends at least partially on its ability to find willing trading partners. This logic is easily extended from dyadic relationships to systemic relationships: the amount that the US trades with France may not be independent of the amount that the US trades with the UK, by virtue of the common membership of the US, UK, and France in organizations which are designed to facilitate trade. In this way, we are starting to understand that even pairs of states are not independent from other pairs of states, much less individual states. Indeed, even the most vigorous findings in a dyadic context -- such as the democratic peace -- often have trouble holding up in a systemic analysis.

3. Varieties of constructivism also think in terms of DGP B. The identities, beliefs, and actions of agents cannot be understood outside of the social structure within which they exist. For constructivists, interests are not assigned according to immutable materialist Laws of the Cosmos, but emerge via processes of interaction. Thus, it is nearly impossible for constructivists to theorize about social actors and entities as if they existed independently from one another.

4. In its most basic form, which has obviously been complicated in thousands of ways, Marxists believe that the material conditions of society lead to exploitation and imperialism, and that this is a global phenomenon. E.g., the exploitation of labor in the global South is a necessary component of the enrichment of capital in the global North. It is thus impossible to explain the outcomes in the "base" of society -- e.g., employer-employee relations -- without reference to the "superstructure" of society -- its organization of political power. Because capitalism is a global phenomenon, the superstructure is itself global in nature. Therefore, employer-employee relations in Britain are not independent of the British imperial exploitation of the Indian subcontinent. Treating the two as if they were same misses the entire point of the theory.

5. First wave IPE was concerned with "complex interdependence". The name itself implies that an assumption of independence of observations is not warranted (even though its authors didn't generally argue in those terms), and that patterns of dependency will not be simple or monotonic. There were varying types of interdependence that Keohane and Nye were concerned with. Among them was the non-independence of the economic system from the political system; the impact of transnational nonstate actors in influencing power politics; the effect of international integration on the formation of states' perception of interests, the effectual nature of international institutions, etc. There is no contemporary research paradigm centered on complex interdependence (yet), but this rhetoric has colored IPE theory ever since the 1960s. In a recent essay, Robert Keohane professed some nostalgia for the "Old IPE", which focused on complex interdependence, over the new IPE which assumes it away (at least methodologically). He expresses hope for the future by way of diffusion and network analyses; methods which do not assume unit independence:
[T]he null hypothesis that national governments make decisions independently is not sufficient to explain the spread of liberalism or of convergence in certain sectors. Competitive pressures seem to play a major role, and there is some evidence, less clear, that coercion, learning and emulation also are significant factors. In other words, the distinctively international and transnational processes studied by IPE have to be brought back into the picture. 

Moving to theoretical orientations which are often thought of as being "middle range", and thus in the wheelhouse of OEP, we can consider:

6. Theories of globalization posit non-independence in many ways. "Golden straightjacket" and "race to the bottom" models believe that global forces coerce national governments to behave differently than they otherwise would, largely in response to market structures and the nature of strategic interactions with other states as well as market actors. Even careful studies which have complicated these accounts have concluded that policymaking at the national level responds to international developments in strong, if narrow, ways. Other theories of globalization focused on cultural hegemony, homogeneity/"Americanization", and a diffusion of norms and practices across countries all argue that what happens in Country B is not independent of what happens in Country A. That assumption is a necessary component of DGP A.

7. The long tradition of dependency theory obviously requires non-independence. Some read Oatley's "Reductionist Gamble" (and a subsequent article forthcoming in Perspectives on Politics) and thought he was advocating for a return to dependency theory, because of language like "core" and "periphery". That's not unreasonable, although I don't know if Oatley would claim the mantle. Dependency theory has been making a comeback in other corners of mainstream IPE as well.

8. Many developments in trade theory suggest that a focus on domestic factor endowments and political institutions (as exogenously-given) are missing a lot of the picture. New trade theory in economics suggests that scale economies play a major role in determining the type and size of trade, quantum trade theory argues that "beachhead costs" and other factors influence how and why governments might invest in trade-generating policies. We've employed gravity models which explicitly assume that the probability of trade relationships forming is not equal across all observations (while only allowing for several forms of dependence, especially proximity and the size of a state's internal market. We're beginning to look at how the structure of trade networks impacts the establishment of future trade. All of the major advancements in economic and political theories of trade over the past three decades have been at the expense of an assumption of independent and identically distributed observations. All of our theoretical improvements, in other words, make it harder and harder to assume DGP A.

9. Transnational financial flows also involve complex processes which make independence a questionable assumption. How do countries get "lumped in" to categories by investors, so that what happens in Thailand has an effect on Indonesia? How does petrodollar recycling lead to a crisis in Brazil? How might US Federal Reserve policy have influenced the Arab Spring? How could a collapse in home prices in the US threaten the existence of the European Union? All of these questions require a theoretical structure which assumes non-independence: DGP B.

I could go on, but I'll stop here. The point is that it's hard to think of a major theoretical orientation -- whether "grand theory" or "mid-range theory" -- in IPE that establishes DGP A. And yet the dominate research paradigm -- open economy politics -- explicitly argues in favor of such an assumption. There is a disjuncture here.

I believe this is why does an article like "The Reductionist Gamble" resonates with some and infuriates (or confuses) others. Why should such an article exist in the first place? How did we get to this point? That will be the subject of my next post.


Jay Ulfelder said...

Terrific post. What you say here of IPE is also true of work in comparative politics and political sociology on civil unrest and regime survival and change. What's methodologically feasible or convenient is often theoretically problematic, and we haven't been paying enough attention to that fact.

LFC said...

I confess I scrolled through this fairly quickly, but somewhat by accident I clicked on the link to the first Wibbels article. It's 6 yrs old, but I don't really keep up w IO, so it was new to me. I looked through the opening pp.

A translation into less academic language of the argument would yield something like "the structure of the world economy and capital markets leads poorer countries to cut social spending in recessions" -- exactly the opposite of what they shd do, and presumably w bad consequences for the poorest segments of their populations. This wd seem to support, and perhaps from another angle, what Thomas Pogge (for one) has argued about the way in which the richer countries deliberately pursue policies on the int'l level that favor themselves and injure the poorest people in the world.

Anonymous said...

It's funny that this is perhaps the one area where the bombs-and-rockets people have it all over the IPE and domestic conflict types. You can find "DGP B"-type theories of international conflict going back at least 50 years, probably farther (Harvey Starr's work, for one prominent example), and -- through the efforts of people like Mike Ward, Kristian Gleditsch, and others, it's still an important part of that subfield.

Kindred Winecoff said...


Sorry I neglected this previously. I'm not familiar enough with Pogge to comment on that. I think there are some issues with Wibbels' account, but in general I like that he's thinking dynamically.

For example, a lot of people said that the reaction of the US and UK to the financial crisis -- bail out the banks, pass stimulus, increase regulation -- makes them hypocrites, because they didn't recommend the same things during the Latin American and Asian crises. But that completely ignores the fact that because of the difference in systemic position of the those countries there was a completely different set of opportunity costs. It's not hypocrisy; it's different responses to different crises in different places.

Ironically, folks seem to get this in the context of Europe: Germany's response is different than Greece's. Why? Because Germany has money and Greece doesn't.


My point is not that there is no work in IR/IPE that doesn't conform to the general pattern; just that they are not the typical mode of research. So I like Ward and Gleditsch a lot, but the typical empirical design in security studies/conflict studies/peace science is still to run logits on COW dyads, treating those dyads as if they were independent even though (from theory) we believe they are not.

When Is Reductionism Not Appropriate in Theory?




Add to Technorati Favorites