Monday, 12 March 2007

Trying to post more regularly

Today I've been struggling on with the model. At the moment I am trying to get some comparative statics out of it. I seem to be spending too much time focusing on the decision problem of the dictator (who maximizes revenue with respect to a proportional tax and universal flat rate benefit) and not enough on that of the citizens. It may be time to try a different and simpler approach to getting at the intuitions I want.

Recovered enough from my flu to go swimming again for the first time in 10 days. I've been doing 20 lengths a day - the pool here is huge and free, they even give you a free fluffy towel on your way in, and you never have to share with more than one other person per lane. Came out and the spring evening light reminded me of home: even the people walking had an English look.

Now I have to do something for Marco Battaglini's class - a paper proposal. I'm interested in trying to model ideology, something that economists have had less success with than modelling institutions. (There are actually plenty of models out there but none of them have really "caught" and given us a basis to go forward with.) The basic ideas behind analysis of ideology and discourse seem to be as follows:
  1. People describe states of the world in discrete, lumpy categories.
  2. These categories influence how they perceive the world.
  3. Their categories also influence how they act.
  4. Categories are contested - fought over, manipulated for strategic purposes etc.
Most theorists of ideology back these claims up with one of two positions: (A) social reality is created by our ideological categories, and there is no truth beyond them ("social constructivism") or (B) categories inescapably limit our perceptions of reality ("conceptual schemes"). I don't find either of these very appealing. Position (A) is, frankly, mad. Position (B) is not mad, but it has severe difficulties (see Davidson, "On the very idea of a conceptual scheme" - I don't claim to have any thoughts beyond this, so if you still prefer (B) then don't let me stop you). Methodologically, I also dislike the notion that analysing linguistic categories should tie you into any non-trivial metaphysical position. (Science should be unified, of course, but as the programmers say, "think about loose coupling".)

A simpler backup, with less hostages to fortune, would be (C1) finding out about reality is costly; (C2) we find out about reality partly in order to communicate with others; (C3) we can choose what to look for in our perceptions; and (C4) our categories for communication are determined for us beforehand by a general purpose language, rather than worked out ad hoc for every speech act. (C1-4) will, I think, give us what we need to understand that discursive categories influence perception. The rest follows.

A good starting point in modelling terms is cheap talk signalling games. These have 1 and 3 of my four numbered points but not so much 2 and 4. The main problem is that in these games, equilibria are usually conceived of as unique to (and perhaps optimal for, allowing for the different interests of sender and receiver) a particular signalling situation. That's what I want to get away from: political communication is much more stereotyped than that. But it would be nice to find a way to do so simply, without e.g. having to have message categories evolve in some complicated scenario.

My current plan is to let people look at reality (a continuous variable) by specifying a set and in effect asking nature "is it in the set"? That gives you 2. Then they can tell other people what they find, take actions themselves etc. Perhaps it will be enough to find multiple equilibria, which benefit different groups differently. That gives you 4.

Of course, there's a 90% chance that there is nothing interesting to model here, as is so often the way. The key is to find a simple application and show that "categories matter" and that formal analysis can extract some insights. It would be good if, eventually, dialogue between formal theory and applied research could become as productive as it is for the study of institutions. That day is a long way off.