Clark take-down

Standard rationalizing models, those that have agents optimizing their own material outcomes, don’t explain all economic facts, especially those dealing with the very, very long run. This has obliged many economist to augment the standard models to include other motivations or objectives of the agents.

For example, to explain the fact that in rich countries richer people tend to have fewer children (which is opposite of the case poor countries), in addition to their desire to optimize their own income you can add the desire of parents to optimize their children’s income. These parental preferences create a trade-off between the quantity and quality of children. Parents choose to have many low-quality children to work on the farm or few high-quality children they send off to college.

Professor Clark has criticized these sorts of adjustments to the standard model, saying they represent arbitrary additions. He complains: “This should make clear that the references specified over goods and children in all these models have no function other than making a bow towards the form of maximizing over preferences in economic models. They do not somehow better explain the world they are just ways of reproducing, mathematically, observed behavior.”

Personally, in the last month I’ve seen models that include preferences for “fairness”, “equality”, “efficiency” and “reciprocity”. There was a macro seminar a couple weeks ago that used “ambiguity aversion” to explain the home portfolio bias and I’ve seen models where “beliefs” and “norms” determine payoffs. In some cases, I felt these additions to the standard model were arbitrary and ill-supported, but my main issue is with model robustness. If we add one sort of psychological or sociological motivation to the model and we get non-standard results, why shouldn’t we add all such motivations? Perhaps adding more psychology to our models would change the results further, perhaps some psychological quirks cancel out other ones.

I guess I’m saying that if economic-man is unrealistic, why isn’t economic-man plus “reciprocity” (or whatever) equally unrealistic?

In any case, The Economist blogger ((who I bet has a really cool first name)) thinks adding such preferences to the standard model is moving the science forward and that Clark is wrong to object.

I think economists should become more comfortable with non-rationalizing models.

15 thoughts on “Clark take-down”

  1. Hold on, we had non-rationing models, in the Dark Night of vulgar Keynesianism… and it was Bad ™!

    Some things you must take as given, it can’t be “turtles all the way down”! If you want to have people caring for their children endogenously, will you want an endogenous discount factor next? — I don’t think there’s a problem with models in with households are self-interested, rather than individuals.

    Sure, some of the modifications you cite are really ad-hoc-ish and awful and all that, but don’t throw the baby out with the bath water.

    The choice of primary, given elements in much like the choice of state variabiles. There’s some flexibility but you go with plausability.

  2. What’s wrong with endogenous discount factors? They’re endogenous. In the long run. The very very long run. And they’re ln 2. Clark has it in his book.

  3. There’s nothing wrong with them. My point was that while you can make this and that endogenous, i.e. preferences over children number and type, *something* must be given.

    I don’t think rule-of-thumb agents are the answer.

  4. I think what that “something” is depends on the question you’re asking. If you’re looking at business cycle then it makes perfect sense to take discount rates as given. If you’re looking at world economic history from 100,000 BC to present then they’re endogenous. Fertility and preferences over children are somewhere in between.

  5. Everything is endogenous in the long run.

    If in the long run preferences aren’t structural, then rationalizing model of the long run can’t be doing what they claim to be doing. Who’s optimizing over what preferences?

    Besides the very long run, I tend to think rationalizing models are problematic when the choice we’re observing happens once in a lifetime or involves payoffs that might take decades or even generations to realize. Here I’m thinking about things like renewable resource management, fertility decisions and the decision to accumulate human capital. In the last case, its tenuous at best to assume 18 y.o. bright-eyed, bushy tailed agents can predict which skills they’ll need to accumulate to best meet the demands of the labor market over their lifetimes. Personally, I went to college because “that’s what middle class Americans do”, i.e. its the norm. Period.

    A good model of the decision to go to college would have norms (not preferences) as the primitive and it would talk about how those norms spread and evolve in society. BTW, this isn’t abandoning our cherished methodological individualism… norms are transmitted from person to person and we are modeling the individual’s norm learning behavior.

  6. Notsneaky,
    Saying that in very long run models you should always have an endogenous discount factor is like saying that in all macro models you should have an inflation-unemployment trade-off. To which my answer is “maybe”, but I’m not comfortable with the “must”.

    If there’s a case for a variable discount factor as a driving force of development, then fine, make that case, but why have us all assume that it is. Clearly, the Clark book (haven’t read it yet) is a step in this direction, but there are other ideas out there too.

    Pushmedia1,
    Maybe we can compromise. We can keep rationality, at the individual/generation level but have OLG where the discount factor and labor elasticity and so on are different for each generation.

    As for norms… why are people following them and what if they don’t follow them? You don’t absorb norms like a sponge (or maybe I think too highly of the human species). Norms, as suggested here, seem to me rather like the Keynesian consumption function models. Let’s assume people follow norms. Let’s assume a constant propensity to consume.

    Plus, with norm models I’m always having trouble figuring out who’s better off. I’d rather keep revealed preference satisfaction as measure of welfare.

    Overall, I agree with many/most of the ideas you support here (you and Notsneaky), but I fail to see how rationality *must* be a victim.

  7. If you have a choice between a parsimonious model with 1 or 2 explanatory variables and an ad-hoc model with 3, 4, 10 or 17 explanatory variables, both equally positive, you have to choose the former model. St. Occam says so.

    Sometimes the parsimonious model is rationalizing and sometimes its not. The point is that we shouldn’t limit our tool box. Not to get all cargo cultish, but in physics they have different models depending on the scale. Big stuff is modeled by relativity and small stuff by quantum mechanics. Physicists think this is a bug and they’re working to quash it, but the point is that there’s two very different models to help explain physical phenomena. If the physicists can get away with it, I think we should be able to too.

    And the welfare issue: its not the case that you’re always concerned about policy. If you are, just translate the outcomes to the utility equivalents and comparative static your butt off.

  8. But, but… but rationality doesn’t add explanatory variables, it restricts the model with a preference ordering. Can you trust a model if it was a bunch of free, pseudo-structural parameters?

    I can take a multiplier-accelerator model, tweak it, and beat the pants off baseline RBC model at all frequencies. Does that mean that we should take Prescott’s Nobel away?

    But we agree… “Sometimes”. I’m just saying that, all things equal, I’d rather have rationality.

  9. >> “If you have a choice between a parsimonious model with 1 or 2 explanatory variables and an ad-hoc model with 3, 4, 10 or 17 explanatory variables, both equally positive, you have to choose the former model.”

    Were you at the thing, with the thing, that’s now in MP4 format going around?

  10. All else equal, you prefer rationality… Are you determining your preferences via revealed preference? Because economists use rationalizing models, those models must be preferred?

    Shouldn’t you prefer rationality in the case in which its most parsimonious. After all, they don’t have rational quarks in physics, rational molecules in chemistry, rational dna in biology or rational stars in astronomy. Those silly scientist, don’t they know they should be using rationalizing models. :-)

    Rationality is a form of micro-founded discipline (i.e. its a theory of why people make decisions) and that’s why Prescott deserves his Nobel. I agree we shouldn’t just throw reduced form models at problems just to replicate the data, but that’s not the only alternative.

    Economists shouldn’t just fall back on rationality/optimization because that’s what they know. We need micro foundations, but that’s a bigger class of models than the rationalizing models are.

    BTW, beta and rho are just as “free” as parameters in other types of models. Models with “efficiency” preferences or “ambiguity aversion” or whatever add more parameters to the models (gamma, lambda, theta, oh my!). Also, these parameters are “psuedo-structural” in that they’re not structural at all in the long run.

    Oh, yeah, I went to the thing, but MR seems to have removed the link to the recording… Do you have it?

  11. Since this turned out to be Clark central, I have a question… Does taking Clark seriously imply we should take eugenics seriously too? (In a scientifically-serious way, this time.) Dawkins would approve, I think. Axelrod too, maybe?

Comments are closed.