Incentives, belief in incentives, the left, the right and moral hazard

Several old political problems turn out to be based on the same underlying question. Here are some examples.

Should we tax the rich more? The "yes" argument says rich people need the money less (that is, poor people's utility from wealth is higher), and wealth partly arises from luck and therefore is a legitimate target for taxation. The "no" argument says that if people expect to be taxed when they're rich (and receive handouts when they're poor) they have less incentive to earn wealth, and therefore less wealth will be produced by society.

Should we pay higher unemployment benefits? The "yes" argument: people are unemployed through no fault of their own; their spending will support the economy; it's moral to share resources with the poor; if it happened to you, you'd want benefits too. The "no" argument: it gives people an incentive not to work; it requires confiscation of resources from hardworking people to pay those who are (for whatever reason) not working; we don't know whether people are unemployed only through bad luck or whether their behaviour and choices contributed to it.

Should Greece default on its debts? The "yes" argument says its people need the money more than international bond investors, and if a default is inevitable there's no point wasting cash now if the consequences are going to come anyway. The "no" argument says debts should be honoured, and that Greece won't be able to borrow in the future if it defaults now.

Should we target higher inflation? The "yes" argument says inflation is bad for rentiers (who live off past earnings) and good for people currently earning; therefore it increases GDP - and what's more, helps the poor, who need it more, at the expense of the rich. The "no" argument says not all rentiers are wealthy, and not all wage-earners are poor, so we might be hurting the wrong people; and more fundamentally, inflation reduces the incentives to invest and save money which will lead to bad results in the long run.

Should we treat people better in prison? The "yes" argument says it's humane to do so, and it results in better outcomes for prisoners once they're released, and less repeat offending. The "no" argument says that if we let people believe prison isn't so bad, the incentives to commit crime are distorted and there will eventually be more crime.

Should we bail out struggling banks (or General Motors, or the airlines back in the day when they needed it?) The "yes" argument says that if we don't, there will be economic chaos in the short term, a major recession, and lots of people thrown out of work. The "no" argument says if we do, banks will come to believe they will always be rescued, and will take more risks in the future; we must set an example so we are better off in the future.

In all of these debates, the "no" argument is essentially this: if we act expediently now, people will come to believe we will act expediently in the future. And this will reduce their incentive to earn money, stay in a job, lend to Greece, save money, act lawfully, or avoid excessive risks - all forms of "good behaviour". The "yes" people want us to do what's best right now; the "no" people want us to make a sacrifice now in the hope of creating the right incentives for a better future.

The "no" strategy is a risky one. In all cases it involves definite costs today. But will it create strong enough incentives to bring those promised benefits tomorrow? The answer depends on how people's beliefs are formed today, because those beliefs are what influence tomorrow's behaviour.

The "yes" strategy is also risky, because it cashes in benefits today without any certainty about the impact on tomorrow. At least we get some definite benefits today, but the long-term costs might be much greater.

We'd quite like to have it both ways: have people act as if the incentives for good behaviour are powerful and sharp, but when things go wrong, do what's convenient and somehow avoid changing anyone's future behaviour. I guess this dilemma will be familiar to parents who love their children and don't want to cause them pain, but believes punishment is sometimes necessary to help them know how to behave in future.

This lays open an unstated question underlying the old debate between economists and other behavioural scientists: "do people act purely according to incentives?" In fact, even if they do act according to the incentives they think they face, what beliefs do they have about those incentives? We don't know enough about how beliefs are formed to give a clear answer to this question. But the answer is critical. I would argue that many of our beliefs about those incentives are formed at an early age and are very hard to change by the mere application of evidence. Evidence clearly has some impact - we don't really know how much - but most psychological research shows that people put less weight on evidence than they statistically should. So maybe we can act for today with impunity, because people will mostly ignore the bailouts/taxes/whatever and the impact on tomorrow will be minimal? That feels a bit too easy.

You may have noticed that the "yes" position in all of the above cases is the left-wing or liberal choice; the "no" argument is the right-wing or conservative policy. The particular yes/no framing is simply a matter of how I have asked the question, but the underlying relationship is genuine: the left-wing position is about what's best for today, while the right-wing position is about setting the best example for tomorrow. That is an unusual characterisation of the political spectrum, but it fits with the traditional constituency-based descriptions of left versus right: rich people can more easily afford to defer gratification until tomorrow, so they're more likely to be right wing.

Or let's put it another way. The left believes that if people are going to behave well, they will do so regardless of (or with little regard to) material incentives. The right believes the incentives are really important. (You can see this same division in the debate over competition in the NHS, or privatisation and market discipline in general.) We might draw the conclusion that people project their own self-image onto everyone else - and therefore left-wingers believe they are not influenced by incentives, while right-wingers believe themselves to be guided only by incentives. But that's also too simplistic.

Yet another way to ask this question: Does moral hazard really exist? Do we behave badly when we have an incentive to do so? To some extent, sure. But how much? Crudely, the right believes there is a lot of moral hazard, the left thinks there is only a little. There have been a few experiments on this but I don't think they answer the question clearly. Maybe it is unknowable.

Thaler, annuities, freedom and self-knowledge

Richard Thaler has an oddly limited writeup of the "annuity puzzle" in the New York Times this week. He presents a choice between two people retiring at 65. One gets a guaranteed $4,000/month for the rest of their life (a standard annuity). The other retains a standard investment portfolio and draws down $4,000/month until it runs out (at the age of 85, if he lives that long) - or $3,000/month till he's 100.

Presenting only these three options, he argues that people should really opt for the annuity as it is lower-risk and saves your family the worry of having to support you after the money runs out. The puzzle, in Thaler's view, is that most people don't buy the annuity but instead keep their money in an investment portfolio that they manage themselves.

But I'm not sure this is all that puzzling. In reality, someone who keeps their money in a portfolio is not likely to draw on it at a consistent $4,000/month rate. Why would they happen to withdraw exactly the same monthly income as they would happen to get if they bought an annuity? Anyone who has managed to save this much (about $600-700,000) has many more options.

You might take a lump sum of $20,000 for a cruise, $40,000 for your dream car or $200,000 for a holiday cottage; or you may prefer to make a contribution to your grandchildren's education. Conversely, you might live more frugally some of the time, getting by on $2,000/month when you can, and spend more at Christmas or in the summer.

You could have all sorts of reasons for this. Your needs may change from month to month; you might have been waiting years for the chance to have that car or cruise and want to enjoy it while you can. You might have specific concerns about the future. You might (as Thaler mentions) want to leave some money in your inheritance in case you die before 85. But one powerful rationale could be that you just want to see how it goes - you won't know how much you will want to spend each month until you have figured out your new lifestyle.

Thaler ascribes most of the behaviour to risk aversion and how the annuity choice is framed (and whether it's a default). But the "see how it goes" effect must be a major factor. This particular form of bias (if it is a bias) results from a lack of knowledge of our own preferences, and as such is not accounted for by either rational choice models, nor most behavioural explanations. But I believe it's much more fundamental to the behavioural economy than defaults, risk framing or some of the other Nudge party tricks.

The economics zeitgeist, 5 June 2011

This week's word cloud from the economics blogs. I generate a new one every Sunday, so please subscribe using RSS or the email box on the right and you'll get a message every week with the new cloud.

The words moving up and down the chart are listed here.

I summarise around four hundred blogs through their RSS feeds. Thanks in particular to the Palgrave Econolog who have an excellent database of economics blogs; I have also added a number of blogs that are not on their list. Contact me if you'd like to make sure yours is included too.

I use Wordle to generate the image, the ROME RSS reader to download the RSS feeds, and Java software from Inon to process the data.

You can also see the Java version in the Wordle gallery (please ignore the "5 May" caption, my mistake - it's actually 5 June).

If anyone would like a copy of the underlying data used to generate these clouds, or if you would like to see a version with consistent colour and typeface to make week-to-week comparison easier, please get in touch.