Mark Thoma asks for thoughts on the following point
Psychologists mock what economists call the microfoundations of consumer behavior…. That this framework is suitable for aggregate systems in a globalized economy simply because the tribe called economics has agreed to adhere to these ad hoc assumptions makes no sense. Increased interactions with disciplines that economists have often mocked as unscientific would greatly improve economists’ understanding of the real world and would be more truly scientific. …
For the core questions that economists are interested in, optimization is always a valid framework. Now before I go on, I want to say that I am not sure I get the compulsion to use optimization in all cases. If you just posit some relationships and they work then more power to you.
However, I don’t think there is something missing in optimization.
Optimization is just picking. You have a bunch of different things. You want to say: show me things like this, not things like that. This is an optimization problem. It need not have any normative flavor even in describing the process. You need not be getting the “best” items by any non-trivial definition of the best.
Now in practice economists often imagine that you are dealing with a situation where you take some range on the number line and then match every single real number in that range to a different item in your bunch of things.
We call this an objective function. Its going to allow us to pick items using calculus.
I am of course going to tell you that spacetime is discrete and so at a basic level this simply cannot describe reality. But, that’s fine because economics doesn’t address problems in which the discreteness of reality becomes apparent.
So, we can act as if we can do this matching problem and then boom you have an objective function.
At the same time though you have to recognize that you just have some pile of crap and a number line. That’s it. You don’t want to go around imbuing objective functions with meaning they don’t have.
For example, take utility maximization. Here we have an objective function that matches possible human behaviors to the number line. We set up the matching so that the if we allow the human in question to engage in a larger set of behaviors he or she will in fact engage in a behavior that corresponds to a larger number.
So ultimately what we have done here is set up this relationship between possible things this human can do and numbers on the number line, mediated by observations about what the human actually does.
Fair enough, but what’s the meta-meaning of higher numbers. Nothing. There is no meta-meaning.
Are higher numbers in any sense fundamentally “better?” Not unless you believe that whatever this person does when the constraints are eased is fundamentally better.
Do higher numbers make the person happier? Maybe. Since happiness is presumably a real thing you need some way to test this. However, there is no underlying reason why higher utility ought to correspond to greater happiness. Not unless you think freedom from constraint is the ne plus ultra of happiness.
I can give you a prime example why this probably isn’t true right here:
The most celebrated technique in this book is a return to swaddling where you wrap a baby like so
Now by all appearances babies seem happier when swaddled.
Maybe someone wants to come along and try to convince me that they have some method of establishing that the baby has higher utility when swaddled. Seeing such an argument is going to be a real treat.
This is a literal constraint on behavior. It only works if the friction of the blanket is enough to simply overwhelm the amount of force the baby can generate. Its hard to generate a scenario that is more oppressive than being completely bound head-to-toe.
Yet, it really, really looks like the baby is happier.
Are the only relevant constraints material? Not if this person acts anything like most people. The line that separates the set of behaviors that do and do not result in being the subject of laughter and ridicule is a pretty strong constraint for most folks.
The line that separates the behaviors that lead to esteem and being the object of sexual desire from the those that do not is also a pretty big deal.
What do I mean by “strong” and “big deal”
I mean this. Take your all material constraints. Look at how much they vary among the population. Now vary them that much on your subject and observe his behavior. How much does it change? How far up and down the number line can we go.
Now, take the material constraint to the least binding of the values you used and move only the ridicule constraint. See if we can sweep out a range of numbers just as large as we got before. I am betting we can.
This means we can do just as much by subjecting this person to various amounts of ridicule as you can by subjecting them to all the various material constraints present among our population.
Put in more human terms it would go like this, even if you are as rich as the wealthiest person in the world or have as many resources at your command as the most powerful person in the world we can make you behave as desperately as any pauper simply by subjecting you to enough ridicule.
That is what it would mean to say, ridicule is a big deal.
But, what does all this have to do with anything?
Three takeaways at least.
One, sticking with optimization doesn’t really limit the phenomena we can model.
Two, despite its name optimization doesn’t tell you anything subjective about what is happening. The proof is always in the pudding. If you like it when folks get to do what they would do if they had fewer constraints then that makes higher utility “better.” If not, then not.
Maybe when people are allowed to do whatever they want they will go out and get a subprime mortgage, move their family into a brand new house, not be able to keep up with the payment and then get foreclosed upon.
If you look at that and say: man that sucks. Then utility maximization is bad in this case.
Three, there is just no reason to suspect that real people will be influenced only by material constraints or that the range of behaviors people display will be tied mostly to changes in material constraints.
10 comments
Comments feed for this article
Wednesday ~ January 25th, 2012 at 8:12 am
anon
“Optimization” is mathematically equivalent to an equivalent iterative approach. So you are right in claiming that optimizing alone is not normative – it’s a mathematical identity. Which one do we use? It’s typically problem dependent: iterators are simpler and more expressive for some things – fields and optimization is simpler for some other things. Mathematically they are one and the same thing, they give the same answers.
The question of human behavior in particular depends purely on how the human brain works. Not more and not less. Take physical environment, put human brains into that environment, let the laws of physics iterare both for a given amount of time and observe the results – they will be the same, within quantum randomness error bars.
Wednesday ~ January 25th, 2012 at 8:57 am
Terry Mahoney
I thought I understood the question until I read your answer. Thanks for the challenge, though. It is good exercise for my brain.
Wednesday ~ January 25th, 2012 at 11:39 am
Axel
“Optimization is just picking. You have a bunch of different things. You want to say: show me things like this, not things like that. This is an optimization problem. It need not have any normative flavor even in describing the process. You need not be getting the “best” items by any non-trivial definition of the best.”
I’m not sure to understand what this means. To me picking something and optimizing are not equivalent as optimizing supposes that the way you pick induces relative pricing (ie. you pick things like this, not like that in a consistent way). If there is no picking pattern/consistency I can’t get how you could transpose it as an optimization pb.
This consistency assumption is quite important, as is the underlying assumption of people being rational, ie. able to get a good enough representation of their picking pattern. (actually you only need a representation of a picking pattern, not a picking pattern itself – which is more demanding).
Wednesday ~ January 25th, 2012 at 11:46 am
Becky Hargrove
You were right to question the word optimization: taken too literally, it can become as suspect as many other concepts that are now being taken and reviewed, and yes I am guilty as (so it seems) charged for using the word too much. Those numbers on a line are highly subjective, even in the obvious processes of logical economic flows. However, perhaps the subjectivity of optimization would not such an issue now, were there better ways to measure emerging wealth. Clearly, much that we do for ourselves in a productive sense is not possible to measure through money, so how do we compensate? Should we compensate at the level of our time and record how it is used compared to the time of others, the issue is not really about optimization. It is about recording things important to ourselves, for the benefit of others. People can do many productive things for themselves that enrich their lives, but if others do not know, sometimes people do not realize the world is actually moving to a better, more productive place. That is the real purpose of measuring economic activity: to make people feel better about what is possible, rather than some supposed optimization perfection.
Wednesday ~ January 25th, 2012 at 12:48 pm
The Utility of Getting Swaddled
[...] Karl Smith: Now by all appearances babies seem happier when swaddled. [...]
Wednesday ~ January 25th, 2012 at 1:18 pm
Owen
Of course, you can optimize over discrete sets… (you just can’t rely so much on derivatives)
Wednesday ~ January 25th, 2012 at 7:44 pm
Jeff
This entry is quite odd, given what often is written here. Bloggers on Modeled Behavior have many times said things like markets allow people to chose what they want and giving them more freedom to chose makes their lives better, or that ‘revealed preferences’ are real, or any of a number of other related things. Things are said in a manner (and in contexts) that suggests it is believed that these statements have content, or could bear truth values. However, if this is all just ‘picking’, and there is no reason to believe this _actually_ enhances utility (except in the trivial sense where utility is defined as this), or maps onto happiness in a coherent way, then it is hard to know what to make of such statements. Moreover, it is hard to know what to make of (so called) ‘normative’ economics (n.b., descriptive economics appears to be unaffected). But this is just what the quoted material in the linked post, and seemingly the baby swaddling example, was getting at. These ideas seem to undermine much of the foundations of the idea that we can move from discoveries in economics to ideas about what policies we ‘should’ enact.
Tuesday ~ January 31st, 2012 at 2:10 am
Jeff
Here’s a quote from an econ blog posting that does a better job of capturing what I’m getting at:
“Economists’ policy recommendations – our ideas about which policies enhance economic efficiency and which ones detract from efficiency – are all based on the idea that individuals know what’s best for themselves. That we can draw demand functions and use them to infer the marginal benefits to consumers of goods and services. … But if people’s demands are just a product of framing, salience, and the public prominence of hurty-elbow syndrome, how can we use them to infer the marginal benefits to consumers… if a consumer’s marginal benefit from consumption is something other than what is revealed by his or her demand I could just make up anything as a marginal benefit curve, and economics would become a completely ad hoc exercise.”
source: http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/09/the-behavioural-economists-dilemma-induction-versus-deduction.html
This strikes me as a real problem, for economics, but also for everyone, as it seems to undermine the possibility of finding optimal policies.
Thursday ~ January 26th, 2012 at 7:11 am
Anon
“what economists call the microfoundations of consumer behavior….”
They should start by reading the first few chapters of http://tinyurl.com/6tyvzuy
Thursday ~ January 26th, 2012 at 12:20 pm
Luke Lea
“For the core questions that economists are interested in, optimization is always a valid framework.”
Except that you don’t allow that, other things being equal, a dollar is worth more to a poor man than a rich one. What kind of optimization can you get without that? Remember: Pareto was a fascist.