Sunday, 8 February 2009

Behavioural models of pay, and the agency problem

One of the primary criteria in designing any system of pay is to influence the recipients' behaviour.

Traditional pay structures work on the assumption that people behave rationally. Thus, they will pay people for generating profits for their company; on the basis that if you pay them more for higher profits, they will generate more profits. This has the attraction of simplifying the manager's job - assume that the employee can figure out the best way to make profits, and leave it to them.

However, this structure has at least three intrinsic problems. One is asymmetry: you can pay someone for making profits, but you can't penalise them for making losses. Virtually no employee will work for a company if there are circumstances in which they would have to pay for the privilege.

The other is the agency problem: the employee has control over the company's resources but does not get all of the benefits of using them in the most productive way. Thus, a temptation exists for them to use these resources for their own private benefit.

The final problem is that it is based on a false assumption. People do not behave in the traditional "rational" utility-maximising way.

If, instead, we apply some of the results of behavioural economics, we may be able to design a pay structure in which the same behaviour is in the interest of both firm and employee.

Some of the most important results of behavioural studies include:
  • People are more likely to act in their short-term interest than their long-term interest. Many different behavioural experiments have demonstrated this. We can reflect this in pay and management structures which provide immediate reward for doing things that are unambiguously in the interest of the company. This is challenging, because it isn't always obvious in the short term what the company's interests are. But if we have a way to, for example, measure the amount of risk that is being taken in an investment portfolio, and give bankers an immediate reward when that risk is reduced, this may help to override their longer-term tendency to manipulate the system in their own interests.
  • People are loss averse. This means that it has more impact on someone to lose £1000 of money that they already have, than to gain £1000 that they don't have. To take advantage of this, a clawback provision would probably be effective. As suggested in a number of articles recently, bankers could be obliged to return some past bonuses if a company goes on to make losses on positions they took in previous years.
  • People are motivated by social factors. This result is informed more by traditional psychology than by behavioural economics. But a number of social factors - status in a group and adoption of social norms - compete with individual self-interest to influence our behaviour. Companies and their ecology provide two of the strongest social groups for many employees - their colleagues and competitors. By creating measures of status within those groups, and linking those to social norms related to the long-term stability and growth of the company, managers could create influences that might dominate financial incentives.

    Money, in some ways, is just another social status marker anyway. Anecdotally it seems likely that people work hard to increase their bonus not mainly because of the associated material benefits but because it governs their place in the hierarchy of their social group.
  • People are, under some circumstances, risk-seeking. Traditional economic theory says that people will give up some rewards in order to avoid risky situations. Recent behavioural research says that, sometimes, the opposite holds. Thus, particularly in situations where the risk of something happening is very small (perhaps less than 0.1%) people will sometimes make decisions that increase this risk instead of reducing it. Clearly that can cause major problems when they are managing large amounts of money - just look at Bernie Madoff, Jerome Kerviel or Nick Leeson to see the results. So if a third party can monitor traders' behaviour and take a dispassionate view of the risks, they could help to provide a rational discipline which helps reduce them.

    While traders would at first no doubt resist having someone monitoring all the choices they make, it is possible to demonstrate that this is in their own interest. In software development we sometimes use a technique called pair programming which has two programmers working at the same keyboard writing the same piece of code. While most programmers find this distracting and wasteful at first, the resulting improvement in quality and reliability - and often productivity - becomes clear in many situations, and most programmers who regularly work this way are won over.
Lots of modern management theory explores how to motivate people in a non-financial way. However few of the books and articles I've seen base this discussion in real behavioural research; they tend to be based on heuristics and informal experience. While experience is clearly valuable in making management decisions, behavioural economics provides an opportunity to supplement and enhance it with knowledge derived from scientific research.


justeconomics said...

Please do not do sloppy economics!

Economics relies on "rationality" as a postulate. You can't do economics without it. No one is irrational.

Your first claim that "people tend to act in their short-term interest rather than in their long-term interest" is also misleading. What you mean is that individuals appear to have a very high discount factor for future benefits. At any given moment, by postulate, an individual has a single utility function.

Your second discussion on "psychology versus economics" fails to note that economics includes social utility. The individuals you describe are still rational actor, but merely placing value on non-monetary gain.

Lastly, your observation that employees ought to be rewarded for reducing risk and rewarded more for less risky profit than more risky profit is apt. The best (not biggest) banks already do this. The ones that don't should be allowed to fail.

The problem with the TARP as currently utilized is that poor risk management is being allowed to survive. Rumor has it when JP Morgan asked the management of Bear Stearns how much exposure to mortgage assets they had, the BS team didn't know!

Can you imagine being the management of a major financial institution going insolvent due to a housing bubble collapse and not even having found out how much exposure you have?

But now Citi and BofA are being saved helped. Didn't they make the same mistakes?

There is a principal/agent problem as you note but the principals are the ones who have the most to gain if they can figure out how to solve it. Punish shareholders by letting banks fail and they will find a way to get better management.

Leigh Caldwell said...

Thanks for your comment.

I agree with much of what you say, and it would indeed be too simplistic to claim that "people aren't rational". I do believe that people respond to incentives and - within certain constraints - act in their own interest. That was not clear from my post.

However, the classical homo economicus model of rationality departs substantially from how people really behave. Sophisticated economics cannot be done if we maintain that postulate in its simple form.

One of the major challenges of behavioural economics is that there is not yet any coherent, analytic, alternative model of this "realistic rationality". In a later posting I'll be saying more about this.

So what we are looking at is the difference between the classical economic conception of rationality, and the kind of rationality to which people really adhere.

For instance, the notable feature of the discount rate is not that it is high, but that it is inconsistent. The discount rate people apply in the present is generally higher than the discount rate they apply to income deferred to the future. The classical utility model says otherwise.

I certainly agree with the statement that social utility is a component of an individual's utility function. Perhaps this point is not about rationality but simply about understanding non-financial incentives.

Finally I'm in agreement that shareholders should be the ones to impose these pay deals on banks. If there is a role for the public sector here, the obvious one is that the British state, now a major shareholder in some of the UK's banks, should act as an activist shareholder in proposing this kind of structure.

OldNick666 said...

Hertzberg's Hygiene Motivation theorey postulates considerable non linearity between motivation and rewards. He postulates that money is only one of many rewards. I agree that people tend to act rationally (unless in a highly stressed situation) but the motivation/reward relationships vary from person to person and with time.

In my experience, money is a hygiene factor rather than a motivator for all of the people I have employed. It is a big mistake to make money into a status motivator as it can seriously damage profits.

Mark said...

"No one is irrational"

I'm surprised that this kind of stuff is still being suggested. Maybe you chaps need to catch up? The consensus emerging from across the broader behavioural and cognitive sciences goes something like this

i. we are social creatures first, foremost and last (e.g. Dunbar etc)
ii. rational independent thinking is much less important that we'd like to believe (e.g. Kahnemann's point about cats being able to swim if they have to)
iii. Most of our mental abilities have arisen and are adapted for life in complex social world of other people (Dunbar & Schelling)
iv. most of the "rational" thinking that goes on, happens after the fact (e.g. Eliot Aronson's more rationalizing than rational)
v. Much of what we do, we do because those around us are doing it (Watts, Bentley et al)

That changes quite a bit of the argument, doesn't it?

Leigh Caldwell said...

Thanks Mark - good to hear from you as I am a regular reader of your blog and agree with much of what you propose in it. I haven't read Herd yet, but certainly intend to.

You're spot on in suggesting that economists need to catch up with results from cognitive and social sciences in general. Behavioural economics as a field is trying to do this, but in my experience its ambitions are quite narrow.

What I'm working towards (slowly, but then it is a multi-year project) is a way to synthesise the insights you've mentioned into a positive and tractable (by economists' standards) model of individual behaviour - which can then be extrapolated to give results on the macro level.

Undoubtedly the homo economicus model does not work. The results you mention do need to be incorporated into economics. To some extent this can be done by simply positing utility functions or preferences for social goods (as justeconomics suggests). But that is too general an approach to make any useful predictions from, and it does not capture the universality or the essential structure of these phenomena. I have the impression that your work on Herd gives some useful illumination on that.

Separate from the modelling of individual behaviour is the emergence of group phenomena which are inconsistent with individual maximisation. The prisoner's dilemma is probably the most famous, along with unpriced externalities, information asymmetry and agency problems - but Schelling has many other examples and I suspect you may touch on this in Herd too.

Broadly speaking, I believe the main gaps in classical economic theory are to do with individual departures from rationality, and market failures related to group phenomena. So far as I know, no consistent theory exists to address both of these.

I will probably pull some of this discussion out into a separate post, because it is pertinent to my longer-term research.

Mark said...

Thanks, Leigh

I hope the synethesis is useful.

Perhaps the best place to start is to put aside the (largely) anglo-saxon view of our curious species as one in which the individual agent is primary. Other cultures (indian, latin, African, asian) tend to conceptualise us as socially interconnected creatures.

In doing so, one useful area for consideration is cognition: whilst we can conceptualise it as an individual capacity, the truth is - as socially embedded creatures - we 'outsource' this kind of cognitive load all the time: the "wisdom of crowds" phenomenon and many of the claimed benefits of markets for distributing knowledge are versions of this outsourcing. Ditto, the distributed intelligence celebrated by the mass-collaborationists like Leadebetter, Shirky and co.

Once you go from "I" to "We", the map starts to shift radically.

Hope that helps kick start things