Wednesday, February 06, 2008

Default Rules and Soft Paternalism

Consider two firms, otherwise identical, with different default rules. Firm A tells its employees that it normally diverts 10% of their salary to a pension fund but will be happy to pay the money directly to the employee if he prefers. Firm B tells its employees that it normally pays them all of their salary, but will be happy to divert 10% of the salary to a pension fund for any employee who prefers that option.

On the face of it, one would expect about the same fraction of employees to go with each option in each firm, since the amount at stake is large enough to make the extra cost of filling out a request or stopping by the human resources office to tell them that you want to switch away from the default trivial in comparison. I am told that in fact such default rules have quite a large effect--that many more employees will go with the pension plan in firm A than in firm B. I have not looked into the literature on the subject myself, but for the moment will assume it is true.

If so, that suggests the possibility of "soft paternalism." If the government thinks employees would be better off putting money into a pension plan, they require all employers to act like firm A. It is "soft" because the employee is still free to choose whether or not to go along. I gather that some thinkers on what I think of as the new new left--academics who accept a good deal of the libertarian view of the desirability of markets and individual choice while still looking for ways of altering behavior in what they think desirable ways--have made proposals along these lines. Presumably part of their argument is that if individual choice is being affected in this way by default rules then it isn't entirely rational, so there is nothing wrong with taking advantage of the irrationality to get people to voluntarily choose what they "ought" to choose.

When I discussed this issue with my daughter Rebecca, she offered an interesting explanation of the pattern of behavior--interesting in part because it makes the behavior rational. The cost of switching into or out of the pension plan is negligible, but the cost of getting the information needed to decide whether to switch in or out is not. In this case as in many others, one cheap way of getting information is by observing what other people do. If, as seems plausible, the firm will have chosen as the default the option most of its employees prefer, thus saving trouble for all concerned, the default rule is a signal of the choices of other employees, hence cheap evidence of what an employee who doesn't know which option is better should do. So the employee rationally goes along with the default option unless he has some good reason to think the alternative is better.

If this analysis is correct, it implies that soft paternalism won't work, or at least won't work for very long. Once it becomes clear that default rules are being chosen not by the employer to fit employee preferences but by the government to nudge employees into doing what the government thinks they should do, the argument for going along with the default breaks down.

Of course there might be a new argument--that if the government thinks you should get a pension that's a reason to do so. But if people actually accept that, there is no need to use default rules; the government can simply tell employees what it thinks they should do.


Patri Friedman said...

it's a cute argument, but I believe that in practice, much of the effect would still remain. It seems like enough evidence has accumulated that people really are irrational that we should often accept irrational behavior at face value rather than searching for rational explanations, fun though it may be.

Anonymous said...

Maybe people just can't decide, even with effort. So they do nothing, taking the default. This inaction is rational regardless of how the default was chosen.

Anonymous said...

Not exactly analogous, but my company plays this neat trick when it wants to divert some employee salary to charitable causes (like flood relief, etc.). It sends a mail saying it will deduct 100 bucks from everyone's salary for so and so good cause. There is no obligation to contribute. Anyone not wishing to contribute can send a mail to the HR.

I don't think anyone has ever written to HR asking not to deduct money from their salary, whether they agreed with the cause or not.

Anonymous said...

I agree with Patri. Anopther point is this, even granting human rationality: It may be rational to do as the presumptive authority tells you that you should, unless you have reason to think that you know better, or unless you think that the good general procedure recommended by the presumptive authority does not apply because of the special circumstances in your case.

Thus, it might be a good general rule to put 10% of your income in a 401(k) plan if the government or your employer thinks you should, but not if you urgently need the money, or you think the 401(k) plan is badly managed, or something. That could be a reason not to have this kind of thing actually required; if it's required,it will be imposed on people even when it's a very bad idea.

Anonymous said...

Assuming that we are irrational in the way suggested and that we're better off with employers making thoughtful default choices for us, we may still be rational enough to prefer employers with good default rules over employers with bad default rules. What is the argument for government intervention?

Raphfrk said...

Another advantage of the 'soft' system is that it acts as a safety valve against crazy options.

If the government sets a crazy rule as its default option, then people could shift en mass to the other option. If everyone is saying at lunch that they overrided the default option, then after a critical point, most other employees who haven't done so will shift.

This can be somewhat reduced by requiring that people continuously override the instructions. For example, you might have to tell HR every six months that you still want to use the non-default option. (To make it more annoying, they could require that you have to submit a signed form to HR rather than allowing it as an email and/or maybe prohibit the company from sending out reminders).

Anonymous said...

Another explanation is that we don't like responsibilities. If you stick to the default, it appears someone else is making the choice and therefore that he should take the blame if the choice is wrong.

Sticking to the default is a way to shy out of responsibility. "Obeying orders" falls in the same category.

Troy Camplin said...

I think it's a combination of the two, plus just plain laziness.

Mike Huben said...

A long time ago, David, you pointed out to me that any outcome can be justified in terms of hidden preferences that are supposed, but not measured or shown to exist.

Aren't you doing the same thing here with hidden rationality?

I'm sure that I could make up explanations based on hidden rationality for why Las Vegas should not find gambling profitable, by pointing out that people would observe the loss rates of other people.

Once you open the door to human irrationality in economics and political science, predictive value of behavioral explanations becomes suspect. Just like any other system with non-linear, ecological-level complexity, it's hard to guess where the equilibrium will be found short of experiment and observation. Thus, we have "laboratories of democracy".

Jonathan said...

I think many people accept defaults because it's easier, and because they have many other things to do. Maybe in some cases there's a significant amount of money at stake and they'd be wiser to pay attention, but they've just come home from work and want to have dinner and collapse on the sofa with a drink...

Sometimes people are not in the mood to be rational. Being rational requires some kind of effort.

Unknown said...

"The status quo bias is a cognitive bias for the status quo; in other words, people like things to stay relatively the same.The finding has been observed in many fields, including political science and economics.Kahneman, Thaler and Knetsch created experiments that could produce this effect reliably. They attribute it to a combination of loss aversion and the endowment effect, two ideas relevant to prospect theory."

Anonymous said...

Well, it only breaks down if you assume that the paternalist direction is incorrect.

Take a different viewpoint: that people are not necessarily irrational, but *are* lazy. In my observation, that's the best argument for paternalistically pushing investment plans such as 401(k). People can invest on their own, and have no compelling reason not to -- they just don't get around to it.

From this viewpoint, the soft paternalism is encouraging them towards a rational outcome, which they simply wouldn't have bothered with on their own. In that case, there is no reason to expect any breakdown over time; on the contrary, one would expect it to be self-reinforcing...

Steve_Roberts said...

Soft paternalism is not coercion, so IMO it's fine at the level of the firm. It becomes problematic when taxes are raised to pay researchers, legislators, and bureaucrats to define and implement soft coercion by government, because the outcome may be 'soft', but the tax-raising and the enforcement is decidedly 'hard'.

David Friedman said...

A number of people in this discussion, including some on a live journal page linked to it which I can't post to, not being a live journal subscriber, bring in the issue of behavioral economics. They might be interested in a published piece of mine, a draft of which is on my web page. It offers evolutionary psychology as a way of explaining patterns of behavior that seem economically irrational--including the endowment effect and non-exponential discounting.

David Friedman said...

The software seems to have truncated the URL in my previous post:

Rue Des Quatre Vents said...

Professor Friedman,

Your daughter's observation reminds me of an experiment Cass Sunstein often uses to illustrate what he (or others) call an information cascade. Let me explain the problem. I'd like to know what decision you think is rational.

Suppose there are ten people in a room and on a table at the far end of the room, there is one urn, either Urn A or Urn B. In Urn A, there are 66 red balls, and 34 white ones. And in Urn B, there are 66 white balls and 34 red ones. No one knows which urn is on the table.

Each one of the 10 is invited, one at a time, to walk up to the table, and pull out a ball. BUT, they cannot reveal to the others what color ball they pulled. Instead, they can only write down on a sheet of paper lying on the table which urn they think it is.

So person #1 walks up, and pulls out a red ball. The rational choice here is to guess urn A. So person #1 writes down red. Then person #2 walks up, and he also pulls out a red ball. Looking down, he sees person #1 thought it was Urn A. So all the evidence supports his writing down Urn A too. And so he does.

Then person #3 walks up and pulls out a white ball. He looks at the paper and sees the two before said Urn A. What's the rational choice here?

David Friedman said...

"Then person #3 walks up and pulls out a white ball. He looks at the paper and sees the two before said Urn A. What's the rational choice here?"

The first question is what person 3 can deduce about the ball person 2 pulled. If person 2 pulled a red ball (as he did) he would write down A. If person 2 pulled a white ball, he would have concluded that one white and one red had so far been pulled, making even odds on A or B. Suppose we assume that in that situation person 2 would flip a coin.

Reasoning in the other direction, the fact that person 2 said A makes it more likely than not that person 2 pulled a red ball. So person 3 concludes that so far there has been one red, one white, and one more likely to be red than white, hence that the urn is more likely A.

Person 4 knows that person 3's answer provides no information about what person 3 drew, as we have just seen, so he is in the same position as person 3. So is everyone thereafter.

Leonard said...

If person 2 pulled a white ball, he would have concluded that one white and one red had so far been pulled, making even odds on A or B. Suppose we assume that in that situation person 2 would flip a coin.

I disagree that this is likely to be how person 2 would act.

For one thing, person 2 is not person 1. Person 2 should reason knowing that there is a chance, even if low, that person 1 wrote down the "wrong" answer. Perhaps person 1 is perverse, or stupid, or simply got confused. Thus the odds are not exactly even. If person 1 acted logically, they are exactly 50/50; so they must be just a bit less, maybe 49/51. So person 2 should guess B, not A, if he pulls white.

Another practical problem with the strategy of guessing randomly is that humans have no built-in RNG. So we don't guess randomly unless we are egghead types and happen to have a coin or die handy to generate numbers. But I see no strong reason in the problem, as described, to think that persons 3 and 2 are both egghead types enough that person 3 would think it likely that person 2 would use a random decision method. Practically, I have a hard time seeing person 2 doing anything but guessing B if he gets white, even if he knew with certainty that player 1 won't mess up his turn.

Thus person 3 should deduce that two red balls have been pulled thus far, and reason from there. And because most of the information in the game is thrown away by design, two red balls pulled initially should cause the entire set of players to guess A, regardless of what they pull.