The visible hand in economics

Some Canadian developers are making a computer game which simulates driving home drunk. They and police think it will be a great educational tool, showing teenagers how dangerous it is to drive while intoxicated. I’m not sure if I completely agree.

Assume that this game will realistically represent driving while drunk. As some people do get home when driving drunk, there must be some probability of getting home without crashing. Since teenagers can play this game over and over again, they can get better at not crashing in the game, which may make them think that they are less likely to crash when they actually drink and drive. Now it might do this, or it might just give teenagers a false belief of being good at drunk driving (given that the road and obstacles will be different on your home road). However, if our teenagers are rational this shouldn’t be the problem (as they will update their beliefs appropriately), the problem is that drunk driving is an experience good with negative externalities.

People who haven’t gone drunk driving don’t know how likely it would be that they would crash, they are uncertain (they don’t know the probability density function and so base probabilities on arbitrary beliefs). Once someone has consumed drunk driving, they gain information, and they know what the risks are when they drive. Now if current social advertising his mis-led teenagers, to believe that the risks are greater than they truly, a situation with no computer game may be preferable to a situation where kids have played the game.

Although full information is usually preferable, teenagers decision to drink drive has a negative externality which is the damage they cause when they crash. By showing kids the true probability of crashing, we increase their consumption of drink driving (assuming that their prior belief was that it was more likely they would crash) to the point where the social cost outweighs the social benefit of their driving activity.

However, there might still be scope for the game and full information. If we can ‘tax’ the negative externality, we can bring the quantity of drink driving down to the socially optimal level. This would require having police fining people when they catch them drink driving. The fine would have to equal [‘cost of outcomes’ x ‘probability of outcomes’]/[probability of being caught and fined]. In this case the driver takes on the full social cost of their drunken activity, and so will only consume the socially optimal amount. The problem with apply this rule come from quantifying the costs. If a drunk driver kills someone, what is the cost of that in monetary terms?

Ultimately, given the difficulty of quantifying outcomes, I think this may be the case where mis-information (at least a focus on the negatives) may be the best way to improve social outcomes. Discuss 🙂

Matt posted earlier today about someone who gains utility from thinking about buying something even if they never actually buy it. There’s actually quite a lot of work that’s been done on utility gained from anticipation. George Loewenstein’s 1987 paper reports a study in which people were willing to pay more for a kiss they received in three days than they would be willing to pay for the same kiss today. They gained pleasure simply from thinking about the kiss and anticipating how good it would be. I don’t think the students in the study ever actually got the kiss so he doesn’t mention whether it lived up to expectations. Interestingly, this idea of utility from anticipation only seems to hold over non-monetary items. Nobody values receiving $1000 in three days more than $1000 today: monetary gains are discounted as one would normally expect.

Of course, if you get benefits from anticipation then that might affect your consumption of ‘real’ items. Botond Koszegi has a working paper in which he models a ‘personal equilibrium’ which describes the interaction between anticipated and actual consumption. Two phenomena which he identifies as arising from his model are self-fulfilling expectations and behaviours which depends upon unchosen alternatives. Both of these are commonly observed: the second in particular is often referred to by psychologists as a criticism of the standard economic model of consumer choice. We value consumption differently depending upon the proximate reference points: for instance, if you go on a fantastic holiday then even if the next holiday is good it is a disappointment. If incorporation of anticipation into our models of behaviour can make them that much more representative of reality then it’s certainly worth thinking about.

So according to a recent study, Chimpanzees play the ultimatum game more ‘rationally’ than humans (hat tip Marginal Revolution).

For those who don’t know, the ultimatum game goes like this. There are two players, and a sum of money that can be split between them (say $1). The first player gives decides on how this dollar should be divided between the two of them. Given this division the second player decides whether to accept this division or reject it. If the second player accepts they divide the dollar, if the second player rejects the offer they both get nothing.

If both players only value the amount of money they get then the first player will set up the division so they give player two only an infinitesimally small amount. However, when humans play the game we find that they divide the dollar up quite evenly. Furthermore, we find that when people divide the dollar up very unevenly, the offer is rejected, even though that means player two misses out on some money.

When they say that Chimps play a more rational game than humans, they mean that chimps behave in the way closer to what we would expect if the agents involved only valued money. All this really tells us is that humans value concepts like fairness at a higher rate than chimps do. Hardly a surprising result.

However, this does make a good point for economists to take on board. Humans obviously do value fairness,. Part of this is instinct, and part of this is institutional. By institutional, I mean that it is a preference we have developed as a result of the society we live in. Although economists are happy to abstract from ideas like this, when we apply economic theory it is important not to forget about the social norms that people also value.

But more importantly, the social norms that are created through the application of economic ideas may eventually change the preferences of the individuals in society. Fairness is useful as it helps reinforce co-operation in situations where a prisoners dilemma occurs. If analysts introduce policy that undermines fairness, or in some way degrades the social norm of fairness, then the socially optimal co-operative result becomes more difficult to achieve. Even more fundamentally, how does the change in preferences influence the way individuals value things, could the loss of fairness as a social norm leave people feeling more upset ceteris paribus? I think this is what sociologists have been telling economists for a while.

Ultimately, I accept the idea that social norms are important in determining preferences, but academic economists have good reason for not looking at them. Academic economists want to focus on thing they feel that they can objectively measure, so that their work does not become value laden. Defining preferences is not value neutral, and so is steered away from in academic work (except maybe in Evolutionary game theory? Not that I would know 😉 ). However, economists that want to apply their ideas to reality must realise that societies affect on preferences is non-trival. This makes the questions of how a given policy will impact on preferences more difficult.