Trading-off indicator accuracy for incentivization – forecast calls for pain October 3, 2007
Posted by Paul Duignan in : Indicators, Outcomes theory & the news , trackback
An article in the London Times today (4 October 2007) – Why a bulging bonus is part of the problem, argues that the current credit crisis following the melt-down of the US mortgage markets has not yet run its full course. In particular, the problem is that in the global banking system, ‘the losers do not know that they are losers, even now’. It will only be when banks start selling off their investment assets that the real price of these assets will be determined and at that time there’ll be lots more pain. In the meantime, it’s likely that a falsely positive picture will be painted because of the bonus systems for employees in the banking industry. This system means that ‘there is tremendous incentive for everyone from the chief executive down to believe the best when millions in personal rewards are at stake’. As a result they’ll overvalue banks assets in the short term. In the run up to the crisis, the first half year results for banks were the best they’ve ever achieved. In order for employees to get an annual bonus they’ll need to also show good results in the second half of the year or else they’ll lose the potential bonus they have on the books from first half of the year.
This is a classic example of the outcomes theory principle of the trade-off between indicator accuracy and using indicators for incentivization where indicators are easy to falsify. And ‘it is the very nature of the bonus system that may end up prolonging the agony’. Perhaps we could go even further than the article and wonder about the roll of the bonus system not only in prolonging the agony, but as a contributing factor in the creation of the problem in the first place.
Incentive systems have to be based on indicators of some sort or another, otherwise you can’t work out when to give a bonus. Where indicators are easy to falsify, there’s a direct trade-off between the potential benefits of using their measurement for incentivizing staff through bonuses and you being able to use the indicator as a true measure of what’s happening in the outside world. This problem can be seen in a variety of instances where indicators are used for incentivization. Carefully considering the possibility of this happening and explicitly deciding on which trade-offs you wish to make is what outcomes theory teaches. In contrast to making careful design decisions about an outcomes system like this, often systems are just put in place without thinking through the implications of using indicators for incentivization.
If the article is right, the result in this case will be that the market will continue under a delusion about some of its investment assets for a lot longer than would have been the case if the indicators in question were just used for getting the best possible estimate of the truth about the current situation. This will prolonging the situation will make the pain worse in the end.
Of course, thinking more deeply, what role has the use of bonuses like this played in the development of the problem? I’ve not looked into it in any detail, but the reported problem is that lenders have been lending to people who won’t be able to make the repayments in the US sub-prime mortgage markets. From the point of economic theory this would appear to be a bizarre aberration from rational economic behavior. What could have driven it to occur? To what extent has this been driven by the individuals in banks being rewarded for achieving a single outcome – writing loans – made very concrete through the bonus system. Presumably they should have been attending to two outcomes at the top of their outcomes model – writing loans and managing risk. Presumably it’s a lot harder in the short term to measure successfully managing risk rather than just writing loans, and so loan writing, because there’s a powerful economic incentive attached to it, has dominated. I’m not sure what role this has played alongside the way the mortage on-selling market works (but this particular market structure may itself have resulted from banks and their staff setting up systems from which they can receive the greatest short term return). If the use of an indicator for incentivization has played a significant role in this, is major reminder of the fact that outcomes measurement and its use is not a passive activity which simply reports on the world. If such systems are not set up with a clear idea of their intended and unintended consequences and appropriate checks and balances put in place, they’re likely to create all sorts of chaos for those involved in them.
Paul Duignan (outcomesblog.org)
Comments»
[...] an earlier blog posting on the banking system, I commented on the problems associated with indicators which can [...]
[...] blogged in 2007 about the problem of bonuses in the financial system and how it was likely the melt-down would be bigger than people were thinking at the time. Thinking [...]