Thursday, May 1, 2014

Micro-foundations don't escape Lucas Critique

The Lucas Critique is summarised by Mark Buchanan as follows:
Any prior regularity that might have existed in a set of data had been present only in the context of the policies prevailing in the past. Change the policies and those changes, by influencing the way people act and anticipate the future, may well strongly change or destroy the regularity on which you had based your plans.
In economics this observation was made famous by Robert Lucas in this 1976 article.

Lucas' critique has justified the micro-foundations approach to macroeconomics for four decades. Put simply, unless you model the macro economy as a result of ‘deep parameters’ of the human psyche, you will never be sure whether your model will apply in a different regulatory or institutional environment. Overcoming the Lucas Critique is apparently achieved by offering a macroeconomic model that stems from a utility function of a representative agent.

But why should we believe that the ‘deep parameters’ of micro-foundations, the utility functions themselves, are somehow independent of the institutional environment?

You can’t escape Lucas’ critique by plucking a utility function out of the air and giving it to a representative agent unless you believe that utility functions are independent of the social environment.

Which raises the question, how can it be possible for an individuals preferences, their utility function, to arise in a social vacuum?

It can’t. The evidence is absolutely clear on this point. Preferences, and even perception, are socially constructed. There simply are no 'deep parameters'.

The whole micro-foundations exercise has been a waste of time for all involved.

While economics has taken seriously this critique from Lucas, they have generally ignored its logical extension of the performativity of economic analysis. Basically performativity says that the use of an economic model in society to guide decisions itself changes behaviour, thus changing the environment in which the analysis applies. Or more simply, utility functions with change in response to the use of models built upon utility functions.

The easy way to see this in action is in sports. As soon as one coach creates a play that exploits a common behaviour in other teams, using that play changes the other team’s response, and thus the environment in which the coach’s original analysis was relevant.

You can’t escape any of this logic.

The lesson is that to understand economic phenomena requires a better understanding of institutional environments, and historical and social context. The micro-foundations approach has merely been an excuse to continue to conceptualise the economy as self-stabilising and in equilibrium in the face of the Lucas Critique, while any rational response would have been to acknowledge the inherent instability of social processes, of which the performativity of economic analysis itself a part of.


  1. Well argued! I do believe that Lucas might have been referring to such fundamental parameters like human instincts that we couldn't change except on the time scale of evolution (say the regularities of Kahneman and Tversky). But it is possible that there are no such economically relevant stable instincts. I've been exploring an alternate take where you assume you don't know anything about microfoundations:

    Entropy and microfoundations

  2. Hi Cameron, very much enjoyed reading this but I have a question. Aren't you missing the difference between the "utility function" and "behaviour"?

    Starting with the example you give at the end of your piece:

    "The easy way to see this in action is in sports. As soon as one coach creates a play that exploits a common behaviour in other teams, using that play changes the other team’s response, and thus the environment in which the coach’s original analysis was relevant.”

    Surely there has been no change in preference here, no alteration to anybody’s utility function: both teams still want to win! All that has changed is the idea of the behaviour which is seen as most likely to deliver the SAME desired outcome.

    We obviously aren't in a world of omniscience here: your entire analysis is predicated upon the existence of uncertainty.

    And the behaviour adopted under conditions of uncertainty depends upon the information available and the way that information is interpreted…in other words the model used by the individual. (BTW there is no need here to examine the rational choice of model, it’s an important question but has no bearing upon my question).

    Neither a change in the information set available, nor in the model used, in anyway implies a change in the utility function.

    My behaviour emerges from two distinct things: what I want (utility): and what I think I need to do to get what I want (my information set plugged into my model).

    Accordingly, when you say: “Or more simply, utility functions with change in response to the use of models built upon utility functions.”,

    Did you actually mean BEHAVIOUR will change in response?

    I agree when you say: "Preferences, and even perception, are socially constructed." And I suppose a case could (in principle) be made for linking economic models to preferences, but a change in and individual's behaviour...which is the logical concomitant, (under the assumption of rationality), to a change in their model, most definitely does not imply a change in their utility function.

    1. Just noticed egregious typo: "anyway" = any way

  3. I would like to second Peter May, at least partially.

    All we ever observe are choices, never preferences. Neuroscientists perhaps some day will be able to tell us how preferences are formed, certainly partly constructed (brain plasticity) and partly innate, our genetic inheritance and gene expression (epigenetics), but cannot do so yet.

    Further, it is inarguable that people do respond to incentives, and changed incentives lead to changed choices -- this is all we can observe. Always we try to do the best we can under the circumstances (or utility maximization subject to a constraint), but we do make mistakes in the absence of perfect information, which leads to changed choices.

    Changed circumstances lead us to explore perhaps previously unexplored parts of our preference map or utility function, and perhaps cause new or changed choices. But for all we know, the preference map itself has not changed, but that our choices are just generated on a difference part of the map.

    This is standard stuff -- changed circumstances such as prices, incomes and new spouses cause us to mentally reside on a different indifference curve.

    Any us of who have raised children know that try as we parents might to mold their utility functions, their choices clearly are most definitely not fully constructed by their most important social environment, the family.

  4. Thanks for the comments Peter and Haynes.

    Sure, you can interpret the point as "Did you actually mean BEHAVIOUR will change in response?"

    That certainly applies to the sport example where the ultimate objective is quite well known.

    But you certainly can interpret it as a a change in utility, or the preference map, or whatever that unobserved concept is that allows us to make choices.

    Let's think about the the sport example again, and this time with the more concrete example of football (soccer). Before a coach's analysis it was generally accepted that getting a corner kick was of a certain value relative to a throw in near the goal. There was a preference (maybe you would say a sub-preference within the larger 'winning the game' preference) of a corner kick over a throw in.

    Now that this new analysis by a coach has started to be exploited, players are choosing the throw-on over the corner kick. To any observer we must say that the utility of the corner kick (or throw in) has changed. Or that the preference map has changed.

    It certainly is possible to say that the meta-utility function is the same but better information about the costs/benefits is available to implement it. But how could it ever be possible in practice to discern any underlying preference if there are merely unpredictable responses to information and uncertainty all the time!

    Which altogether questions the logic of utility functions or underlying preferences at all? If there is no analysis that can prove or disprove it, because we can always say it's new information, new prices, then it actually offers us nothing.

    In terms of the Lucas critique, what you both seem to be saying is that micro-fountation based on utility function would overcome the critique IF the utility function was complete, perfect and stable, and not merely the product of some observed stylised fact, and that included the net complex effects of interactions with other agent. Sure. But that's just saying we know the answer already to the universe, the economy and everything!

    1. Well said - "utility function" is an abstraction and not a well understood one at that. Further most "micro-foundations" are not founded on actual "utility functions" but rather on "behaviours."

  5. One point I've been mulling over is that the things that are least vulnerable to the Lucas Critique are 'mid level' phenomena like social norms and institutions. The macro level is vulnerable for reasons Lucas articulated - over aggregation and policy changes - while the micro level is vulnerable due to reasons you articulate, such as dynamic inconsistency and advertising. But social norms and institutions tend to persist for a bit longer. Perhaps that is where we should look for 'deep parameters' of behaviour.

    1. Yeah I agree. There are intermediate levels of aggregation that have some very regular patterns.