The dormitive quality of rational choice

by John Quiggin on October 10, 2007

This Matt Yglesias post has already made it on to my colleague Andy McLennan’s door. It’s short enough to quote in full

I’m not sure I understand why Greg Mankiw thinks economists “don’t understand tipping.” When I was learning economics, I learned that people are utility-maximers and that whenever you see some behavior that doesn’t seem explicable in purely financial terms that must be because people are deriving utility from the foregone financial advantage. Thus, as any economist could tell you, people tip because of the utility they derive from the tipping in much the way that economists can explain all aspects of human life.

Have I ever mentioned that philosophers tend to think that economics is vacuous? Which isn’t to say that you shouldn’t listen to economists. These days, they tend to know a lot of math, and math is a very useful thing.

Matt omitted the irony alerts, but I tried to spell out the same point here.
Given any data on any observed set of problems involving the selection of one or more choices from a set of alternatives, the observed choices can be represented as the maximisation of an appropriately specified function.
Playing straight man to Matt, that doesn’t mean utility functions are useless – the functional representation lets you do lots of math that is much harder if you try to work directly with preferences. But any competent economist knows that utility isn’t an explanation of observed choices, it’s a way of representing them. The representation is simpler if choices satisfy some minimal consistency requirements, like transitivity (if you prefer A to B and B to C then you should prefer A to C).

The only place where I routinely encounter the confusion between substantive predictions about behavior can be derived and the observation that all choices maximise some kind of utility function is among adherents of the “rational choice” school of political science. As mentioned above “rational choice”, as described here, is essentially a tautology. More importantly, perhaps (and unlike the situation with most market interactions) crucial aspects of political behavior can’t be explained in terms of rational egoistic choice. Most obviously, voluntary voting doesn’t make sense for an egoist, since the probability of being decisive is so low as to ensure that no private benefit from one outcome or the other can be worth the effort. In my experience, rational choice fans often want to explain voting by arguing (correctly) that rational choice can encompass any preferences, including altruistic or expressive preferences, then go back to claiming that rational choice models yield substantive predictions (which can only be obtained by surreptitiously disallowing any preferences except egoistic preferences).

{ 61 comments }

1

Scott Martens 10.10.07 at 12:00 pm

So, in other words, if people do something it is, by definition, something that seemed like a good idea at the time?

2

Kevin Donoghue 10.10.07 at 1:06 pm

Which isn’t to say that you shouldn’t listen to economists.

Indeed one should listen to economists, but not for the reason Matt gives. The real payoff is that you can ensure that your errors are original. I vaguely recall Kaldor explaining why he was unable to help the British government: “I was too late; all the really big mistakes had already been made.”

3

aaron_m 10.10.07 at 1:12 pm

Isn’t a big part of the problem with the utility assumption that economists seem to lack understanding on or guidelines for when it is appropriate to model based on the assumption as you describe it. Much economic thinking/modeling on cooperation seems to me to have this problem. In thinking about the prospects for cooperation on various terms we want to be able to try at least to separate out other regarding reasons for action and self-regarding reasons, and to have some notion of the directions in which they are pulling (i.e. cooperate or not). Mashing it all up into the same metric without any obvious way to measure the intensity of utility satisfaction from different kinds of reasons for action just makes a mess of it. No?

4

Soullite 10.10.07 at 1:17 pm

People Tip out of fear of social ostracization.

This is why economics sucks. It’s basically a gimped form of psychology with a few mathematical formula tossed in. People don’t derive ‘utility’ in the way ‘utility’ is classically defined, Economists have to twist the concept so that it encompasses all non-financial motives a person can experience.

Plus there’s the whole part of them being wrong about 70% of the crap they say, and lying about the purpose of the other 30%, But that doesn’t really matter for that specific example.

5

abb1 10.10.07 at 2:03 pm

I feel that self-interest is something like gravity. Yes, there are all kinds of motivations: social/cultural norms, rules, patterns, indoctrinations, altruistic impulses, etc. Their effects can be significant, even huge, and yet self-interest, survival instinct, is probably the strongest driving force in there.

6

JP Stormcrow 10.10.07 at 2:36 pm

It truly is a comfort to live in this, the most rational of all possible worlds.

7

rvman 10.10.07 at 2:56 pm

There are some usable neoclassical economic theories of tipping involving repeated visits, but yes, for one-time visits, individual tipping isn’t ‘rational’ if we assume ‘rational’ to be ‘maximizing homo economicus’s(Econ’s autistic sociopath base model) bang for buck’. Even Econ has moved away from homo economicus, except as a starting point for further elaboration. We have to look in the utility function from econ’s perspective, which means we borrow from the realm of social and cognitive psychology. There are really two questions – “Why tipping as a convention” and “How is the convention enforced”. Econ has a lot to say about the former, very little about the latter, except to claim that it has to be set up to be in the long-run individual interest to do it, otherwise it will fail.

An easy way to see the different uses for the disciplines is by analysing the Kibbutzim in Israel. The founders are personally invested in success (psychology) and so they are willing to forego individualistic behavior by not free riding on the group, to reach their personal goal of community success. (Econ and social psych) As long as the group is small enough that personal zeal and social ostracism can be used as tools to enforce the social norm (social psych and sociology, with econ talking about how strong the incentives have to be to overcome self-interest), it works. When it grows in size, and a new generation appears, the social ostracism weakens through reduction in group coherence,(sociology) the new generation isn’t as personally invested in the group’s success,(social psych) and so the incentive to make it work is reduced.(Econ) At this point economic self-interest takes over, and the kibbutz (or other cooperative community) fails in the absence of market incentives like prices and wages,(Econ) either through out-migration, fragmentation, (sociology) or macroeconomic collapse (econ).

All three disciplines would like more explanation of why ostracism is for tipping used here and not Europe, where ‘tipping’ is enforced by the restaurant by being included in the bill. An economist would look at “Who gains from tipping?”, would start digging around the Principle-Agent problem and in coordination game theory, and think about balancing competing incentives and goals. (How does the restaurant and customer motivate their joint ‘agent’ – the waiter – to choose behaviors they each want? Note that in the US, both pay the waiter – both are principles to the waiter’s agent.) There may be an implicit social contract involved, which needs all three to explain how and why it works.

Behavioral Economics is what the ‘border’ area between econ and psych is called from the econ side – the psych side is Consumer Psychology. People who want to make money off of learning this stuff major in Marketing. They are really the same thing, using the same methods and involving the same research gurus – Daniel Kahneman is probably the researcher closest to the line between them.) Really, one can look at the social sciences as a spectrum – sociology as being intensely focused on ‘irrational’ and group behavior, economics as being focused on rational individual behavior, and social psychology as being the mediator between reason and passion. (Cognitive Psych gets non-rational individual behavior, Economics studies rational group behavior, calling it game theory and I/O Economics.)

They are all attacking the same problems, really using the same techniques – the difference between econometrics and psychometrics is basically that econometrics focuses on ‘field’ techniques, while psychometrics focuses on experimental techniques. Sociology uses mostly psychometric methods, to the extent that it has caught up with the other two in using math. (Harder with groups, and more can be published with literary description there than in econ.)

8

JSE 10.10.07 at 3:23 pm

Even after thinking a lot about the “paradox of voting,” I don’t really understand why economists consider it a paradox. Papers on this subject typically say something at the beginning like “I guess voting might make you feel good about yourself as a citizen, but discounting that, we can’t see rationally how the voter gets something out of it.” This seems to discount the factor which settles the problem completely, no muss. This would be a vacuous explanation if the utility attached to performance of civic duty were a postulate thrown in to explain the behavior. But it isn’t! It’s what one actually observes in oneself when voting!

9

Ted 10.10.07 at 3:38 pm

There are other reasons for voting. Since 1999, I have had such disdain and contempt for George W. Bush that I’ve voted against him three times (an open primary and two general elections). I suspect for a lot of people this is a powerful motivation, in the same way that support for a football team leads them to paint their faces and buy expensive televisions for watching the game. Also, as a political scientist, I have professional reasons for voting, since it’s kind of expected of me, but admittedly this will be a motivation for very few people. Some people vote for reasons of social pressure from friends, spouses, parties trying to turn out the vote, etc.

10

Ted 10.10.07 at 3:39 pm

Oh, and I live in a relatively small city, so when it comes to local ballot issues and elections for local officials there is a small, but non-negligible, chance that my vote really will determine the outcome.

11

Peter 10.10.07 at 4:19 pm

What the rational-choice theorizers of democracy miss, of course, is that citizens are not simply consumers of public policies, but also producers of policies. In other words, they vote not only to influence the outcome, but also to express an opinion (not to mention, of course, to exercise a hard-won right). The deliberative democracy people have been saying all this for close on three decades, but I guess it is too much to ask the rational-choice theorists to actually read the relevant literature.

12

bob howard 10.10.07 at 4:41 pm

Thank-you for this post and comments. I will use ryman’s comment in class sometime this semester

13

aaron_m 10.10.07 at 5:09 pm

If you assume from the start that neither the benefits or importantly the costs in terms of rational self-interest amount to much at all in the case of voting, why would you try to explain behavioral variation between individuals or societies in terms of rational self-interest? Either the assumptions at the start are wrong or the reasoning behind applying the theoretical model are misplaced.

14

Barry 10.10.07 at 5:19 pm

rvman: “(How does the restaurant and customer motivate their joint ‘agent’ – the waiter – to choose behaviors they each want? Note that in the US, both pay the waiter – both are principles to the waiter’s agent.)”

IIRC, the amount paid by the restaurant is somewhere well below minium wage. The waiter makes almost all of their income through tips.

15

Colin Danby 10.10.07 at 7:17 pm

John’s distinction between a representation and an explanation is the vital point here, and indeed a lot of people don’t seem to get it.

Other kinds of representation of the same phenomena are possible, though they may not be as mathematically tractable. You could take the kibbutzim described in ryman’s excellent comment and see the initial stages not in terms of individual interests but the development of densely relational sociality, with emergent properties all its own. From that perspective the second phase is the breakdown of that dense relationality.

To take the other example it’s possible that at least some people tip because they feel the pull of genuine responsibility to another person, and understand the institutional fact that your bill essentially omits payment for service. That’s not to say that you might not get useful insights by modeling this in choice-theoretic terms.

16

mq 10.10.07 at 7:47 pm

Rvman’s comment in 7 is one of the best and most interesting I’ve ever seen on CT.

Also, voting is an expressive behavior. Like cheering at a ballgame. Votes are often based on what emotion you’d like to express or identify with.

17

whomever1 10.10.07 at 8:12 pm

Also, the vote (and hence the act of voting) provides resolution of a long (WAY long, in the case of the current election) campaign of propaganda, patriotic ferver, rantings and rallies. Climax is rewarding (though this is more attenuated than some types of climax). You could also find an element of cognitive dissonance. All these people are telling me this is important–I’ve already devoted a bunch of energy to following the campaign; it would seem pathetic if I didn’t vote. Especially if I’ve been doing it for a few years.

18

Hogan 10.10.07 at 8:57 pm

Behavioral Economics is what the ‘border’ area between econ and psych is called from the econ side – the psych side is Consumer Psychology.

Is anyone studying producer psychology? If not, that would seem to be a meaningful imbalance.

19

soru 10.10.07 at 9:26 pm

Econ has a lot to say about the former, very little about the latter, except to claim that it has to be set up to be in the long-run individual interest to do it, otherwise it will fail.

That seems close to valid. By analogy, engineering (as opposed to architecture) can be seen largely an attempt to catalog all the types of bridges that will fall down.

Probably needs a lot more qualifiers to accurately predict failures, though, something like ‘requires a pattern of repeated choices that are obviously and strongly against the interests of those able to make them’, rather than simply ‘in the long-term interest of individuals’.

Tipping and voting both pass that test, simply because they are cheap and easy: if it was a matter of volunteering to clean the restaurant for a week, or leave your crops to walk for a week to a polling station, it would be surprising if observed behaviour was the same.

Similarly for suicide bombers and presidential candidates: both are rare exceptions. A society that required a majority to behave in either way would be in a lot of difficulty.

20

MattXIV 10.10.07 at 9:31 pm

Most obviously, voluntary voting doesn’t make sense for an egoist, since the probability of being decisive is so low as to ensure that no private benefit from one outcome or the other can be worth the effort.

OTOH, this is a good explaination for why the majority of eligible voters in the US don’t vote. In 2006, 59.6% of eligible voters behaved as the simple rational egoist model would predict. The rational egoist model doesn’t cover all the bases, but in almost any scenario a good deal of the actors will indeed be behaving as rational egoists.

On the voting question specifically, what appears to be missing is that the direct benefits of casting the decisive vote are small, but there are massive externalities – everybody gets impacted by the outcome of the election. Political organization seems to serve the purpose of internalizing these externalities by getting people to vote in blocks. The probability of a block being decisive for an election is the sum of the probability of the individuals being decisive, but the individuals in the block all experience the benefits of getting their preferred outcome.

I think coordination to internalize externalities prior to the actual voting does more to explain voting than the expressive utility mechanism does. If expressive utility were the primary motivation, the ability to bring tax dollars back to a district in pork wouldn’t be a significant strength for a politician.

21

loren 10.10.07 at 9:36 pm

peter: “The deliberative democracy people have been saying all this for close on three decades, but I guess it is too much to ask the rational-choice theorists to actually read the relevant literature.”

Oh, they read it sometimes, and they even put together a few papers and the odd panel featuring formal models of deliberation.

“Deliberation,” however, ends up becoming “signalling,” or a bayesian puzzle of why disagreement might persist — perhaps because there doesn’t seem to be any useful (let alone minimally plausible) way to model the philosophical notion that we might recognize, sincerely debate, and occasionally be swayed by, the force of the stronger argument, framed within public reason, and in light of exant evidence.

With due deference (and attendant apologies) to any current or future friends and colleagues I may offend, it reminds me a bit of the old joke about a group of students who manage to get an advance copy of the zoology exam and find that it deals exclusively with snakes. The professor discovers their misdeed and quietly changes the exam to deal entirely with elephants. After the test, several of the answers read: “an elephant is a large mammal with a trunk that resembles a snake. Snakes are …”

Compare: ‘Political philosophers and theorists write a great deal about deliberation. Deliberation looks a lot like signalling. Consider the following signalling game …’

22

roger 10.10.07 at 9:43 pm

Actually, our life is thick with the to the economist puzzling situations like voting. For instance, in an essay entitled Pissed Off in L.A., Jack Katz, the sociologist, surveyed a group of drivers about their driving habits and found that the majority will insult and yell at other drivers while driving, even though the insults and yelling can’t be heard by the other drivers. In fact, they will enact whole dramaturgies of anger, because apparently it is not enough to ‘think’ these dramas – they have to be spoken, even if there is no audience, or even if the audience is other people in the car, who have no control over the other drivers on the road, either. Which, to me, certainly implies that one of the basic human drives is to poetry. It would be nice if economists acknowledged this every once in a while.

23

loren 10.10.07 at 9:49 pm

…and almost no one bothers with the obvious rational choice argument for voting! …

In most modern elections the chances of your vote being decisive are vanishingly small. It seems that you shouldn’t bother to vote (and certainly you shouldn’t bother to invest time and effort to become informed).

Everyone else will reason this way, however, and therefore there is an obvious incentive to cast the only vote, and since you will certainly be decisive, you ought to become informed enough to vote in your interests.

… which reminds me of another joke: an economist and a professional wrestler are complaining about the long waits at their polling station …

24

loren 10.10.07 at 9:50 pm

roger: Which, to me, certainly implies that one of the basic human drives is to poetry.”

this is a beautiful thought.

25

sherifffruitfly 10.10.07 at 10:04 pm

Are you basically just explaining the concept of “mathematical modelling”?

26

notsneaky 10.10.07 at 10:08 pm

“Everyone else will reason this way, however, and therefore there is an obvious incentive to cast the only vote, and since you will certainly be decisive, you ought to become informed enough to vote in your interests.”

Which is another way of saying that there’s a multitude of equilibria – basically anything.

27

Bruce Baugh 10.10.07 at 10:20 pm

People, or at least some people, also vote to contribute to the total tally, and this is a rational decision. Throughout the next term and beyond, officials, the commentariat, and others will point at the results: Candidate Hesucks got 63% of the vote, but measure Itsucks that he endorsed vigorously got only 51%. Candidate Ourhero lost by 48% to 52% this time, which is a significant improvement over the 36% she got two years previously. And so on. Insofar as anyone takes the percentages as meaningful, and lots of people do, then every vote is in its way decisive.

28

Arr-squared 10.10.07 at 10:28 pm

“Even after thinking a lot about the “paradox of voting,” I don’t really understand why economists consider it a paradox.”

As a “rational choice” guy, I think the idea of the paradox of voting does a number of legitimate and interesting things.

First and foremost, the paradox shows pretty clearly that for most people, the instrumental benefit of voting outweighs the costs – that’s the paradox: why does anybody vote? The benefits of voting (electing the candidate of your choice, maintaining the democratic system) are public goods, and the odds of one’s vote being decisive are small. So nobody should vote for instrumental reasons.

One simple conclusion is that if it’s desirable to stimulate turnout, those efforts should probably focus on non-instrumental benefits. This generally implies civic education and other ways to increase the expressive and solidary benefits of voting.

Maybe a more subtle conclusion is that, using the paradox as a frame, we can investigate how different types of people experience the causes and benefits differentially. For example, older Americans (particularly the elderly) see voting as more beneficial because they experience more directly many of the things that governments do, like run schools, provide social security, etc. Older Americans, particularly retirees, also experience lower opportunity costs to voting because, by definition, they have more discretionary time.

A younger voter, on the other hand, might be juggling college, work, and a social life and thus have little discretionary time AND perceive little benefit to voting.

29

ben saunders 10.10.07 at 10:30 pm

Last time I voted I did so for a minor party, in the hope that they’d keep their deposit. My vote was probably more likely to influence that than the outcome.

30

abb1 10.10.07 at 10:33 pm

When I say “roll over” my dog will roll over, which is obviously irrational behavior as far as the economists of the doggy civilization are concerned. Well, she’s doing this because she has been indoctrinated by me; she was getting treats at first, and now she does it automatically. This seemingly irrational behavior is still a manifestation of an egoistic emotion.

Humans can understand abstract concepts and thus they are much easier to indoctrinate; all it takes is repeating a bunch words. That’s how they learn to tip, vote and love their countries.

31

John Quiggin 10.10.07 at 10:48 pm

Note that there’s no paradox (or at least not the standard one) if you allow a modest amount of altruism. For example, it seems reasonable to say that the US as a whole would have been at least a trillion dollars better off if Gore had won instead of Bush, but that Bush voters believed something like the opposite

Ex ante, any one voter’s chance of being decisive is roughly proportional to the no of voters, say 1 in 100 million, so the expected benefit to the country from casting a vote for Gore was around $10 000, making a trip to the polling booth look pretty good if you care at all about your fellow citizens. The problem is that each individual’s private expected benefit comes out at 0.01 cents, which is not so appealing.

There are still some subsidiary paradoxes – why vote in cases where the outcome is a foregone conclusion and so on, but these aren’t so important.

32

The Fool 10.10.07 at 10:52 pm

Speaking of social scientist jokes and people locked into unwaaranted assumptions, here’s my favorite:

Two behaviorists hook up at the annual convention and engage in hot steamy sex. At the end, the first behaviorist says to the second, “Wow. That was great for you. How was it for me?”

*rimshot*

33

JSE 10.10.07 at 11:20 pm

Ex ante, any one voter’s chance of being decisive is roughly proportional to the no of voters, say 1 in 100 million

I argued against this computation in Slate a few years ago, for what it’s worth.

34

leederick 10.10.07 at 11:24 pm

What’s the difference between rational choice and Newtonian physics?

You say that if you see someone make a choice you can explain it in terms of utility, and (the killer complaint) that as you can do this no matter what choice then using utility as an explanation is vacuous. Well, if you see an object accelerate you can explain that it terms of it being acted upon by a force, and no matter how it accelerates you can conjure up some force which will explain its acceleration.

Would you disallow both explanations? I don’t think anyone would say that backcalculating like this can’t be used to explain what happened in car accidents, and it just a representation to make the maths easy. So what’s the difference? Just that Newton’s right about the way the world works whereas rational choice isn’t?

35

John Quiggin 10.10.07 at 11:51 pm

JSE, I’m unconvinced. Obviously, the Electoral College complicates things, and most actual elections have some such complicating factor.

But your first pass calculation gives a 1 in 5 million probability of a voter being decisive in Florida, and that’s about equal to the number of voters in Florida. Of course, we know ex post that Florida was the decisive state (in the sense that it was the closest, and had enough votes to swing the result), but voters didn’t know that in advance.

I agree that the Bayesian method is about right, but it seems pretty obvious that it’s going to give an answer of the same general of magnitude as the one I did. You could probably get a factor of 10 reduction by assuming that the probability of a vote for one party or the other is between 45 and 55 rather than between 0 and 100, but that only raises the private return to 0.1 cents.

36

John Quiggin 10.10.07 at 11:53 pm

“Well, if you see an object accelerate you can explain that it terms of it being acted upon by a force, and no matter how it accelerates you can conjure up some force which will explain its acceleration.”

The problem would be analogous if, in practice, it was necessary to invoke hypothetical forces lots of the time, but nonetheless physicists claimed that the fact that this can always be done proved the value of standard (no hypothetical forces) Newtonian physics.

37

JSE 10.11.07 at 12:13 am

You could probably get a factor of 10 reduction by assuming that the probability of a vote for one party or the other is between 45 and 55 rather than between 0 and 100

Oh, certainly — I’m saying no more than what you say, that the “reciprocal of the number of voters” heuristic is too low by a factor of 10 or so in this case — though it might be much more drastically too high if you follow polls and know that your state is polling 60-40 in one direction.

And I’m also agreeing with you that the “expected benefit to the country” benefit is not a very big part of people’s reason for voting. Rather it is some part civic duty, some part magical thinking (“Millions of other people are more or less like me, so if I bother to show up, so will millions of other like-minded people.”)

I think arr-squared’s comment is very good, but I think it reflects a more reasonable point of view (“Voting is the result of some kind of non-instrumental self-interest, so if we want to increase turnout we should figure out which kind it is”) than that taken by many papers on the topic, which seem authentically mystified by the fact that people vote.

38

John Quiggin 10.11.07 at 12:16 am

To be clear, I said that “expected benefit to the country” is a sufficient reason for voting (for anyone moderately altruistic), though I don’t imagine it’s the way most people think about it.

39

JanieM 10.11.07 at 12:41 am

John Quiggin wrote: Ex ante, any one voter’s chance of being decisive is roughly proportional to the no of voters, say 1 in 100 million, so the expected benefit to the country from casting a vote for Gore was around $10 000, making a trip to the polling booth look pretty good if you care at all about your fellow citizens.

I’m just a quiet CT reader who has never taken an ecnonomics class, so if this is a stupid question, I apologize. I would love to be raised out of ignorance, if that’s possible in a blog comment thread.

Under the conditions described, if Gore wins, the country is a trillion dollars ahead. That’s $10,000 per voter (using JQ’s 100 million figure) or, perhaps more to the point from an altruistic point of view, roughly $3,300 per citizen. But one vote doesn’t change the trillion dollar result by $10,000; the result is all or nothing.

To put it another way, I don’t see how you can talk about “expected benefit” from 1 vote when the result is all-or-nothing. We’re comparing a trillion dollars (if Gore wins) with zero dollars (if Gore loses). Casting the decisive vote (if that vote could be identified) would trigger the entire trillion. But all the votes up to that point would not have been accumulating $10,000 increments, so it’s hard to buy that calculation (which concerns a number that has nothing to do with any possible real-world outcome) as a motivation.

What makes more sense is the hope that enough people will agree with me to put my side over the top, and that if I vote, I can increase the chances of that outcome. (I’m not saying that’s the only motivation for voting, just that it makes more sense than the “I can generate $10,000 for the country by my 1 vote” formulation. Because you can’t. You can either be part of generating a trillion for the country, or you can lose, and generate nothing.)

40

John Quiggin 10.11.07 at 2:09 am

janiem, I’ve been a bit sloppy and used “expected” in its statistical sense (mean or mathematical expectation) rather than its ordinary language sense. If you buy a lottery ticket with a 1 in 1 million chance of getting $1 million, the (statistical) expected return is $1. Each extra ticket you buy increases your expected return by $1. Of course, that number doesn’t correspond to expecting a real world outcome of $1 – in reality you either get $1mil or nothing.

That said, there are a lot of reasons why a bet with a statistical value of $1 should be worth something fairly close to $1. For example, if you buy a large number of bets like this, and pay more (less) than a dollar a time you will probably lose (win) in the long run.

41

JanieM 10.11.07 at 3:31 am

John — Many thanks for the further explanation.

Without meaning to quibble, and knowing that I probably should read the whole thread again and more carefully, I would say off the top of my head that it seems unlikely that people make the same “calculation” when they vote as when they buy a lottery ticket. I know I don’t. ;-)

Maybe this is an example of why it’s difficult to represent comlex human behaviors in mathematical formulas. I remember seeing the word “vacuous” up there somewhere……

42

notsneaky 10.11.07 at 4:16 am

Actually utility/preferences by themselves are vacuous. They stop being so when you combine them with constraints. Then you get predictions -> when price of something goes up, you do less of it, or in other words the (compensated) demand slopes downward.
So if you make tipping or voting “more expensive”, basically by making it more difficult, you’d see less of it. Recently John wondered why US elections were on Tuesdays which raises the costs of voting. Well, the non-vacuous prediction here would be that countries or regions with elections on off days would have higher turnouts than ones on work days. I’m sure someone somewhere’s looked at it. Estimating this elasticity might even give you an idea of some kind of marginal cost of voting (or cost of voting to marginal voter or something marginal or other).

I can’t think of a similar example for tipping … wait, yes I can. Why do you think that when you order a beer at a bar and your change comes to 5 bucks the bartender hands you back five ones rather than a single 5$ bill?

43

RDT 10.11.07 at 8:15 am

Mattxiv says in 20…
Political organization seems to serve the purpose of internalizing these externalities by getting people to vote in blocks. The probability of a block being decisive for an election is the sum of the probability of the individuals being decisive, but the individuals in the block all experience the benefits of getting their preferred outcome.

I’ve always seen voting (and campaign work) as part of the social contract between myself and everyone else hoping to provide a decisive vote in the same direction.

44

abb1 10.11.07 at 8:54 am

…when price of something goes up, you do less of it…

Usually, but not necessarily. People can be indoctrinated to prefer more expensive alternatives of a similar or equivalent quality. Easily. Make voting more difficult/dangerous and people just might start voting in greater numbers. Raise the price of the cheapest DVD player to make it the second least expensive and you might be able to sell more.

45

Tracy W 10.11.07 at 9:14 am

You say that if you see someone make a choice you can explain it in terms of utility, and (the killer complaint) that as you can do this no matter what choice then using utility as an explanation is vacuous.

Generally economists explain choices as the result of the interaction of utility and budget constraints. Economics is often defined as “the study of the allocation of scarce resources”.

This produces predictions that can be disproved. For example, standard utility functions assumptions combined with a fixed budget constraint produce the prediction that if A buys potatoes and then the price of potatoes goes up while A’s budget remains the same, then A will buy fewer potatoes. This can be tested by seeing if A does indeed buy fewer potatoes when the price of potatoes goes up. Under some conditions it has emerged that A buys more potatoes when the price goes up, so economics has added the concept of giffen goods (and luxury goods).

46

Arun 10.11.07 at 11:27 am

You say that if you see someone make a choice you can explain it in terms of utility, and (the killer complaint) that as you can do this no matter what choice then using utility as an explanation is vacuous. Well, if you see an object accelerate you can explain that it terms of it being acted upon by a force, and no matter how it accelerates you can conjure up some force which will explain its acceleration.

There are much stronger experimental constraints on a postulated force than on a utility function.

It might be helpful to look at some recent science:

The Pioneer anomaly – deep space spacecraft deviate from their expected trajectories, and what force is causing this is still a matter of debate.
More here:

http://en.wikipedia.org/wiki/Pioneer_anomaly

Galactic rotation curves – observations indicate that the matter at the periphery of spiral galaxies is moving much faster than it should be, given the visible mass distribution. This is commonly attributed to the gravitational action of matter not visible to our telescopes rather than to a new force.


http://en.wikipedia.org/wiki/Galaxy_rotation_curve

But alternative explanations have been attempted, e.g.,

http://en.wikipedia.org/wiki/MOND

I do not think economics has the means to test postulated utility functions in the same way physics has to test postulated new forces.

47

Arun 10.11.07 at 11:40 am

While my first comment is in moderation, another way of looking at the difference between Newton’s F=ma and utility functions is that the Newtonian law is supposedly universal, in a way that the utility function is not. E.g., to take a previous example, about the utility function for potatoes, suppose an economist has come up with one and it predicts quite well the demand for potatoes as the supply and price vary; then along come organic potatoes and some varieties of genetically-engineered potatoes. Along come some new widely publicized studies on the effect of potatoes on health. That utility function for potatoes is likely to change in unpredictable ways. In the physics world, however, all potatoes accelerate the same way always.

48

Doh 10.11.07 at 1:04 pm

This is why the “Law & Economics” school of jurisprudence (as championed by Posner) has been such a disaster– it takes poor quality economic thought and uses it to justify judicial preferences.

49

notsneaky 10.11.07 at 2:27 pm

43. I’m fine with it working “usually” since that’s about the best you can hope for in social science.

44. There’s a reason I put “compensated” in them parentheses. Still, Giffen goods are rare if they exists at all, so we’re back to “usually”.

50

Tim Worstall 10.11.07 at 3:41 pm

As someone who waited tables and tended bar through several gap years and then college can I point out something from the agent’s side about tipping?

The tax wedge? In both the US and the UK there are significant differences in the amount of money the server receives for a $ spent by the customer, depending upon whether that money is handed over as a tip or in higher prices which feed through into a wage. The two country systems are very different but in the UK, higher prices pay VAT, then two sets of National Insurance then income tax. Tips pay only the last. That’s a difference of 40 pence in the pound in income for the waiter for one pound spent by the customer.

This doesn’t, as above, explain why single visit people should (or do) tip, but it might well explain why good waiters (yes, I am arrogant enough to insist that I was one) preferentially work at places without a built in service charge.

(I doubt that anyone wants a full description of the tax treatment in each country but if anyone is thinking of papers on the subject I would suggest that this issue of differential taxation is important.)

51

Katherine 10.11.07 at 3:48 pm

When voting, I tend less to think “what will I gain from it”, and more “what will it cost me” – the answer is that it will cost me a bit of time, which I have to spare, so it really doesn’t cost me so much.

Do economists always assume people ask “what will I gain” rather than “what will it cost”?

52

rvman 10.11.07 at 5:54 pm

43. Ceteris isn’t necessarily paribus in that case. You probably can’t just jack up the price and get more sales, without a fair investment in marketing that good to the new audience. (This is all hidden in the economic model for monopolistic competition. Marketing is the art of manipulating your demand curve, from an econ perspective.)

You have to build your brand image, and if you already have one, you better hope it isn’t already tarnished as a ‘cheap’ product. Packard Bell would have been hard pressed to move ‘upclass’ with their PCs after years of cheap, low quality products in the ’90s, so they were bought out and the brand was killed in favor of less damaged names. Had they jacked prices, they just would have sold fewer units.

30. (I’m not picking on you, abb1, you just happen to be saying the stuff I thought worth responding to.) Most(not all) economists wouldn’t bother explaining your dog, because the Psych guys have this rather thoroughly and famously explained. (Pavlovian conditioning, combined with the pack-animal’s instinctive eagerness to please the Alpha) Econ might take a stab at how big the treat has to be relative to the ‘trick’ to incent cooperation in the first place, though.

14. The fact that tips tend to be a larger part of waiter’s income than wages suggests that restaurants believe they have cheaper ways of monitoring and incenting the waiter. (By, say, firing waiters who cultivate tips in ways which are expensive for the restaurant.) The customer can’t really ‘fire’ the waiter without ‘firing’ the restaurant. There are industries where the proprietor practically charges his workers for the privilege of working there, because access to the customers is so valuable. (Strip clubs and barber shops can be ‘interpreted’ through this lens, and the model can be adjusted for Flea markets and antique malls.)

53

Peter 10.11.07 at 6:20 pm

Loren at #22 wrote:

““Deliberation,” however, ends up becoming “signalling,” or a bayesian puzzle of why disagreement might persist—perhaps because there doesn’t seem to be any useful (let alone minimally plausible) way to model the philosophical notion that we might recognize, sincerely debate, and occasionally be swayed by, the force of the stronger argument, framed within public reason, and in light of exant evidence.”.

Well, again, there’s a relevant literature which belies this statement. Try searching for argumentation theory — there are now many models (some formal and computational) of how intelligent, autonomous entities may represent, formulate, exchange, consider and revise arguments between one another. In addition to the 2300 years of philosophical writing on the topic, there are now regular conferences on computational models of argument (see the ArgMAS and COMMA conferences) as well as regular conference series in philosophy (OSSA and ISSA). See also the journals Argumentation and Informal Logic, and recent special issues of the journals, Artificial Intelligence Journal, the Journal of Logic and Computation, and IEEE Intelligent Systems.

54

rvman 10.11.07 at 6:32 pm

18: I don’t know the field really well, but Psychology has a field called Industrial Organization Psychology(I/O Psych, not to be confused with just I/O, or Input-Output, which is a major topic of study in the cognitive and computational sciences), which studies things like employee motivation and business organization which probably fits part of the bill for a producer psych. Of course, like marketing/consumer psych/behavioral econ you can make money at this one, too – you just get your MBA in Management.

55

Robin 10.11.07 at 6:51 pm

A long time ago I got into a discussion with a friend in the philosophy deptartmemt and Sidney Morgenbesser on rational choice explanations in the social sciences. (I was in poli sci.) My friend suggested that in it structure it explained why he was drinking water then and there. I suggested that it was crap at explaining the Bosnian Civil War (going on at the time). To which he replied, I’m not trying to explain that. Morgenbesser retorted that one would think from my friend’s attitude that why someone drinks water was the most pressing question of the time. What RCT is good at seems to be trivial issues, by an large. At the larger ones’s it’s kinda of boring or crap.

But Sahlins may have said it best.

“Thomas Kuhn and others have wondered whether
the social sciences have paradigms and paradigm
shifts like the natural sciences. Nothing seems to get concluded because some say that the natural sciences don’t even have them, and others that in the social sciences you couldn’t tell a paradigm from a fad.
Still, considering the successive eras of functional explanation of cultural forms—first, by their supposed effects in promoting social solidarity, then, by their economic utility, and lately, as modes of hegemonic power—there does seem to be something like a Kuhnian movement in the social sciences. Though there is at least one important contrast to the natural sciences.
In the social sciences, the pressure to shift from one theoretical regime to another, say from economic benefits to power effects, does not appear to follow from the piling up of anomalies in the waning paradigm, as it does in natural science. In the social sciences, paradigms are not outmoded because they explain less and less, but rather because they explain more and more—until, all too soon, they are explaining just about everything. There is an inflation effect in social science paradigms, which quickly cheapens them. The way that “power” explains everything from Vietnamese second person plural pronouns to Brazilian workers’ architectural bricolage,
African Christianity or Japanese sumo wrestling. But then, if the paradigm begins to seem
less and less attractive, it is not really for the standard logical or methodological reasons. It is not because in thus explaining everything, power explains nothing, or because differences are being attributed to similarities, or because contents are dissolved in their (presumed) effects. It’s because everything turns out to be the same: power. Paradigms change in the social sciences because, their persuasiveness really being more political than empirical, they become commonplace universals. People get tired of them. They get bored.”

56

loren 10.11.07 at 9:08 pm

Peter, yes, fair enough: I’d forgotten all that stuff in formal argumentation theory. And of course that work, and indeed arguably any system of formal logic, amounts to a formal representation of reasoning.

But perhaps you can make your case about the relevance of this work as models of, or useful for, deliberative democracy?

As a fledgling grad student I fell under the sway of Hayward Alker for a time, and endured (too) many an evening in the MIT AI lab hearing about the promise of work by McCarthy and Shank and Abelson, and how RELATUS (or some such thing) would eventually interpret political arguments in real time, and somehow do this in ways that had something to do with a Hegelian ontology, and reflected the beauty of LISP and SCHEME …

… guess I blocked it all from my mind — not that LISP isn’t beautiful; I just never saw the software doing anything remotely like what was being pitched.

For all that work in argumentation theory and computational models of human understanding and argumentation, how are things working out so far?

You guys have any systems yet, based on all those varied formalisms, that can reliably interpret complex political and philosophical arguments, and yield a recognizably cogent evaluation and judgement based on their interpretation?

More pointedly, do any formal models of argumentation help us clarify, challenge, and extend (or perhaps simply reject) extant work on deliberative democracy? Obviously there are general ways to make a claim of relevance (but that’s open to the game theorists too: deliberation may often involve signaling). I’m looking for formal models of argumentation that speak directly to the sorts of concepts and intuitions at the heart of deliberative democracy.

I’m more than happy to be educated on this, and that isn’t a rhetorical flourish: I really mean it. A lot of water must have gone under the bridge since I last bumped into some of the literatures you mention, and if there is recent work in formal argumentation theory that speaks directly and fruitfully to the deliberative democracy literature, then I definitely want to know about it.

But I wonder if for now I’ll stand by my a somewhat weaker verson of my original quip? There’s not (yet) any useful way to formally model the most important concepts and intuitions lurking in the deliberative democracy literature — certainly not in the game-theoretic literature on signaling, which really was the point of my original quip.

57

John Emerson 10.12.07 at 1:28 pm

Robin: Sahlins link or citation?

58

Julian Elson 10.12.07 at 3:26 pm

Well, if you’re going to say that utility functions are no more or less than a monotonic transformation of revealed preferences, then it seems to me that cardinal utility cannot exist except as any of infinitely many possible arbitrary cardinal utility functions which represent the preferences appropriately.

If you don’t think cardinal utility matters in a meaningful way, except as a way of turning stuff that’s hard to use calculus on into stuff that’s easy to use calculus on, then it seems to me that you can’t really accept the expected utility model of risk-aversion, lotteries, etc.

If you don’t accept the expected utility model, then it seems to me that the reduction of compound lotteries axiom isn’t really meaningful.

If you don’t think that the reduction of compound lotteries axiom means anything, then it seems that the Allais paradox is no paradox at all, since the Allais paradox depends on the expected utility model of risk or somesuch.

If the Allais paradox isn’t meaningful, then all of your work on rank-dependent expected utility seems to be a waste of time.

You fool! Don’t you realize your entire life has been a waste? You cannot even follow your own assumptions in deciding what is even a meaningful problem to attempt to tackle!

Just kidding. Maybe.

Sorry. I’m a little medicated at the moment.

59

John Quiggin 10.12.07 at 10:44 pm

There are a few little gaps in your argument, Julian, but the main point is this. If preferences display certain consistency properties (like RCLA) they can be represented by a cardinal utility function, unique up to an affine transform. If in addition they display an independence property (which Allais shows they don’t) they can be represented as EU.

The kind of reasoning I object to would be analogous to the following. An EU fan might argue that (i) even preferences that don’t satisfy EU can be represented by some kind of utility model (for example rank-dependent). (ii) Hence, such observations support utility-based analysis of choice under uncertainty (iii)Hence, it’s OK to use EU whenever you feel like it.

Adjust your meds and see if this makes sense :-)

60

Bruce Webb 10.13.07 at 6:10 am

I am sure this point has been covered but it is too late in the evening to review the whole thread.

At my local bar I have an acknowleded table, certain control over the TV channel, service speed and pour rate of booze that fully compensates for my level of tipping. Now I have friends that may not keep up on tipping but still keep up on service levels,then again they are here everyday. Money may not buy you love, but consistently high levels of tipping can pay significant dividends going forwards. My score on the stud muffin scale may be distressingly close to ‘2’ on the other hand my drink doesn’t go dry and menu substitutions don’t end up a problem. Does my tipping rate actually maximize my efficiency outcome? Not always. Does the equity equation get better? Well mostly yes.

Ask anyone in the hospitality business, rich people are crappy tippers, mostly because they simply don’t relate to the concept of ‘There for the Grace of God Go I” For them tipping just tranactional friction and not a recognition that societal distribution of the gains from productivity is not distributed in equitable manners to actual contribution

61

Robin 10.13.07 at 5:58 pm

Comments on this entry are closed.