Sunday, March 20, 2016

New paradigms in economics assume the outcome of research

Noah Smith summarizes a list of changes requested by Hanauer and Liu for economic theory, so I thought I'd give the information transfer framework (ITF) view for each.

The list turned out to be less interesting than I originally thought, so I've added this preamble.

TL;DR version: Hanauer and Liu try to tell us how economics should be done. But no one knows how economics should be done until someone has figured it out. Do the research, show us success, and then tell us how to do everything.

We didn't know quantum field theory was the proper approach to particle physics until we found out it worked. We didn't know statistical mechanics was the proper approach to thermodynamics until we found out it worked.

There weren't papers in physics that said: we should try this quantum field theory stuff. There were people working on trying to add relativity to quantum mechanics and the things that worked eventually were called quantum field theory.

There are of course always new techniques you should try, but you shouldn't ever say this is how theory should look ... because it basically assumes you know how research will turn out.

This is a common failing not just with various "schools" and heterodox economics, but even some mainstream economic approaches. These are effectively defined by choosing 1) a theoretical paradigm ahead of time (the reasons could be political, a favorite modeling technique, or -- rarely -- deep insight into the problem), and 2) presupposing solutions to things that are unknown.

You may ask: Wait, isn't that what you're doing with the ITF?

No.

I do not know if the ITF is correct. I have found some empirical successes, but maybe those successes will evaporate (which is why I keep testing). I would not say the ITF is the future of economics. Maybe it will work, maybe it won't. Maybe someone out there will see a major flaw.

The ITF reduces to mainstream economics in certain limits, so it's best viewed as an alternate formulation or expansion rather than a major paradigm shift.

Finally, it does not assume what a recession is. I think every approach to economics assumes what a recession is. What's up with that?

The ITF doesn't really assume any particular mechanism or modeling technique (and I've done both top down theory and agent-based simulations on this blog). I have a couple of different answers to the question of what recessions are. Non-ideal information transfer, coordination (like panics), poor monetary policy or avalanches. Some of these are related.

...

Anyway, here's the list:

1) The replacement of reductionist models with ones based on "complex adaptive systems"
I'm with Noah in not understanding what this even means. However, the ITF makes few assumptions about the substrate of economics, only that agents on average fully explore the state space when you have equilibrium. The ITF applies far from equilibrium and without knowledge of constraints.
2) The use of network models
This kind of assumes a specific model approach ... without evidence. Until you figure out that network models get economics basically right, you don't know that it's the proper approach. For example, in physics we didn't know Yang-Mills quantum field theory was the proper approach until we figured everything out with it. Getting something right with method X is both the outcome and rationale for using method X.
But here's a reference to network models and the ITF.
3) The use of disequilibrium models
I like Noah's snarky comment on equilibrium. You really need to define equilibrium before this makes sense. 
In any case, I have some nice graphics for this. Seriously, the ITF is all about the successes and failures of (information) equilibrium. And it defines equilibrium as information equilibrium, so disequilibrium (non-ideal information transfer) means something specific.
4) The use of nonlinear models
Again this kind of assumes a specific model approach ... without evidence. Until you figure out that nonlinear models get economics basically right, you don't know that it's the proper approach.  
My view is let's get some linear models that work first. Then we can try looking at nonlinearity. Here's Krugman's and my takes.

5) The replacement of "mechanistic" theories with "behavioral" ones
Again, I'm with Noah: what does this mean? And again, the ITF doesn't make assumptions about agents. They could be very simple or complex. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that behavioral models get economics basically right, you don't know that it's the proper approach. 
6) The replacement of optimization with something resembling "satisficing"
Well, the ITF doesn't do this. It replaces finding the optimum with simply finding the most likely state. Since it doesn't make many assumptions about agents or their financial histories, it avoids questions of "what do agents think?" ... since these are basically unanswerable/unobservable anyway, I'd say the ITF is a better approach. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that satisficing models get economics basically right, you don't know that it's the proper approach. 
7) The replacement of forward-looking agents with adaptive agents in econ theories
Again, this is a specific model of agents that the ITF avoids. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that adaptive agent models get economics basically right, you don't know that it's the proper approach. 
8) Modeling people as interdependent instead of independent
Again, this is a specific model of agents that the ITF avoids.
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that interdependent agent models get economics basically right, you don't know that it's the proper approach. 
9) Modeling people as irrational approximators instead of rational calculators
Again, this is a specific model of agents that the ITF avoids. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that irrational agent models get economics basically right, you don't know that it's the proper approach. 
10) Modeling people as caring about reciprocity
Again, this is a specific model of agents that the ITF avoids. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that reciprocity gets economics basically right, you don't know that it's the proper approach. 
11) Modeling win-win situations instead of environments characterized by rivalry
Again, this is a specific model of agents that the ITF avoids. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that win-win situations get economics basically right, you don't know that it's the proper approach. 
12) Replacing models of competition with models of cooperation.
Again, this is a specific model of agents that the ITF avoids. 
And again this kind of assumes a specific model approach ... without evidence. Until you figure out that cooperation gets economics basically right, you don't know that it's the proper approach. 

3 comments:

  1. I noticed a pattern toward the end of this one. ;)

    Cut and paste must have been helpful.

    ReplyDelete
  2. "There are of course always new techniques you should try..."

    It seems like some of these points that Hanauer and Liu bring up could be reformulated as "Here's some things it might make sense to try and here's why we think that."

    ReplyDelete
    Replies
    1. Sorry I missed these comments.

      But yes, that would be the frame. But as yet, none of the things in the list have shown to be successful ... otherwise they probably would have used those successes as evidence!

      "Econ doesn't work; let's replace it with things that we don't know work because they sound cool and/or interesting to us."

      That's not how you do science.

      How about -- let's figure out what a recession is in terms of signatures in the data, and then proceed from there?

      Delete

Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.