Historians, Avoid the Mistakes We Economists Made!

war

A friend of mine recently drew my attention to something he thought would be of interest to this blog.  Apparently something of a controversy has arisen regarding a recent study in the Proceedings of the National Academy of the Sciences that was discussed on Andrew Brown’s blog at The Guardian. Brown interpreted the study to suggest that war has been the key driving force behind human society. Some on the political left have disputed this as it has rather dire implications for any left-wing political project.

First it should be said that this is not a new idea. Nietzsche, for example, saw war as a constructive force. In his On the Genealogy of Morals he writes,

The beginnings of everything great on earth [are] soaked in blood thoroughly and for a long time.

While elsewhere he openly calls for war to wipe out weaker and less productive humans. This idea has been with us for a rather long time and it is unlikely to ever disappear.

The study though makes a somewhat novel case. It claims that war leads to greater social cohesion — or, what the authors call ‘ultrasocial norms’. The authors of the paper give an example,

As an example of an ultrasocial norm, consider generalized trust. Propensity to trust and help individuals outside of one’s ethnic group has a clear benefit for multiethnic societies, but ethnic groups among whom this ultrasocial norm is wide-spread are vulnerable to free-riding by ethnic groups that restrict cooperation to coethnics (e.g., ethnic mafias). An example of an ultrasocial institution, much discussed by historians and political scientists, is government by professional bureaucracies. Other examples include systems of formal education, with the Mandarin educational system in China as the most famous example, and universalizing religions. (p1)
The idea lying behind the paper is then that war breeds higher degrees of social cohesion which leads to the formation of states which in turn generates greater civilisation. The authors lay out the causal chain as such,
The conceptual core of the model invokes the following causal chain: spread of military technologies→intensification of warfare→evolution of ultrasocial traits→ rise of large-scale societies. (p2)
The authors then use modelling that will be familiar to readers of this blog; that is, they effectively use a variety of regression techniques that are familiar to economists. As readers of this blog know, I am extremely skeptical of these techniques. The above mentioned study indicates that such techniques are now being applied to history under the guise of something called ‘cliodynamics’ and which is often associated with ecology (which I think is a somewhat dubious ‘science’…). The author of a Wired article discussing the study describes cliodynamics as such,
Cliodynamics is a field of study created by Peter Turchin in the early 2000s. The idea is to use data as a means of predicting the future, but also as a way of testing theories about what happened in the past. You build a model that seeks to explain history, and then you test this model using real historical data.
I assume that this will sound more than a little familiar to economists in general and those that read this blog in particular.
There are so many problems with applying these techniques to historical data — which, in my opinion is identical to economic data — that it would take me far too long to enumerate them. But the core problem is that of causality. As can be seen from the causal chain laid out by the authors above the study definitely assumes a rather fixed causality. Yet mathematicians that developed these techniques have long pointed out that they cannot really identify causality.
David Freedman, for example, notes that without being able to do real experiments — which we cannot generally do with historical or economic data — any causal inferences drawn from such models are useless. He writes,

The usual point of running regressions is to make causal inferences without doing real experiments. On the other hand, without experiments, the assumptions behind the models are going to be iffy. Inferences get made by ignoring the iffiness of the assumptions. That is the paradox of causal inference… Path models do not infer causation from association. Instead, path models assume causation through response schedules, and – using additional statistical assumptions – estimate causal effects from observational data … The problems are built into the assumptions behind the statistical models … If the assumptions don’t hold, the conclusions don’t follow from the statistics.

The complex nature of historical data, like that of economic data, means that we cannot draw causal inferences from the examination of the statistics alone. In order to do so we need a theory. And if this theory is flawed in any way — which it likely will be given the heterogeneous nature of historical data — then any results we get from applying it are also meaningless.

Take the recent examples of war in the 20th century. I’m no historian but I think we can make some pretty general claims about three major wars in this century with regards to their leading to greater or lesser social cohesion and progress. First, World War I. Did this war lead to greater social cohesion and development? Absolutely not. In fact, it precipitated two decades marked by economic and political strife across the Western world which eventually led to another war. This was in part due to the war but it was in part due to instabilities arising in the economies of Western nations which were only partly due to the war.

World War II, on the other hand, undoubtedly ushered in an age of economic progress and large-scale social cohesion in Western nations. Why was this? Again, the reasons are complex, but a good deal of it probably had to do with the new economic system that emerged during the wartime planning which increased social equality and ensured adequate state-supported infrastructure for stable economic and social development. What about Vietnam? Well, after Vietnam economic and social cohesion began once more to break down. This was not due to the Vietnam war, although the latter certainly contributed in that it was the final nail in the coffin of the Bretton Woods economic system; rather it was due to the falling apart of the post-war socio-economic model.

Why all these differences? Because history is not homogenous. It is irreducibly heterogeneous. In history there are no perfect causal processes that explain events.

History, properly conceptualised, is a series of constellations. These constellations change through time and are not subject to any overarching meta-narrative. They are contingent and particular and in their contingency and particularity cannot be reduced to a set of ‘laws’. The study of history is never final. It is rather a meditation. Techniques like cliodynamics are the phrenology of our day. They have a veneer of scientificity about them. But they are pseudo-science of the highest order.

Nevertheless, they tend to draw attention — even if this attention is fleeting — and so they tend to draw funding. I would council historians, however, that to go down this path is extremely dangerous. If they would care to look at the contemporary state of economics — a subject now held in public ridicule — they can see the end result of applying pseudo-scientific principles to what is ultimately the study of historical processes.

Advertisements

About pilkingtonphil

Philip Pilkington is a London-based economist and member of the Political Economy Research Group (PERG) at Kingston University. You can follow him on Twitter at @pilkingtonphil.
This entry was posted in Philosophy, Statistics and Probability. Bookmark the permalink.

4 Responses to Historians, Avoid the Mistakes We Economists Made!

  1. retardedpolyticks says:

    Ernst Jünger had something similar in mind when he advocated war and the glorious civilisational rise after it. Your initial passage is very reminiscent of Ernst’s take.

  2. B says:

    I really do prefer your (short) example of counterexample analysis than a statistical regression. It emphasizes possible causal mechanisms pushing in different directions, in words, concise and clearly stated, easy to understand and so to refute. Not that statistical inferences are fundamentally difficult to read, but I have been lost so many times with so many of them because of all the different tests, techniques and implicit assumptions. Without mentioning their shaky ontological foundations.

  3. Tim Cooper says:

    This is a great critique of the limits of applyingof these techniques to historical questions. But I also wonder if clio-dynamics doesn’t point to something else too. Perhaps we historians have gone too far in abandoning meta-narrative and causation, leaving that space to others to fill. Perhaps it is time to think again about the scale and manner in which we seek to address historical problems.

    • I’m not sure that this is the case. There are many meta-narratives that are popular in history right now — from Marxism to feminist theory. I think the reason for the clio-dynamic creep is because it is easy to get funding for something that appears scientific than it is for something that does not. It’s a sign of the times; and of our worship of something vaguely called Science.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s