I’ve thought of another way of putting one of the key points of the last post, which should tie it in nicely with the next one that will complete these methodological posts. A key thing to take from it is that cost-benefit analyses are fundamentally flawed. A cost-benefit analysis attempting to establish whether the Russian leadership will take a certain action is flawed as it is implicitly calculated from Putin’s subjective point of view and we do not know what this is. A prediction based on my valuation of the costs and benefits is useless as I’m not making the decisions and am unlikely to have an identical perspective to Putin, as these weights are not objective. The decision makers are weighing up the costs and benefits of different actions subjectively, according to their own preferences, so any prediction of what they will do is attempting to mimic this – to replicate their valuations of specific costs and benefits and so provide accurate predictions of their actions.
The problem is we simply don’t have the information. It is irrelevant how important I think the state of budget is, what matters is what weight is placed on it by the Putin circle relative to other issues. How can observers do cost-benefit analysis from the point of view of Putin if we don’t know how he values the costs and benefits? How can we make predictions of Russian actions if we don’t know their preferences? Putin may be twice or half as concerned about an issue as we had thought and this renders the prediction, however well argued and seemingly justifiable, useless.
This may be sounding fatalistic but there is a way out – prediction as a tool can be salvaged but this fundamental flaw is always present. The problem above is that we don’t know Putin’s preferences at a given moment in time and this can never be definitively overcome. So to predict anything we need to make assumptions or estimates about what these preferences are. There is an inevitable trade-off here between detail of assumptions and validity of predictions. Assuming nothing leaves us, as above, being unable to rule any behaviour out, while assuming very precise targets and desires would give us very specific, but entirely useless, predictions. If I think I understand Putin and predict exactly what he will do I can guarantee that I’ll be wrong.
So what do we do? The aim must be to find some sensible midpoint between the theoretical absolutism of being unable to predict anything at all and the error-strewn detailed predictions based on, by definition, the wrong values.There must be a way we can make predictions about the future actions of Russia (I’ve got to have something to write about!) while acknowledging the inherent problems with the task that mean you will almost certainly get it wrong. Well I think there is: ‘modelling Putin’ is possible and useful.
How do you do this sensibly? Find out next week…