Monday, 9 March 2009

Should Lloyds executives be sacked?

A fascinating question today on Robert Peston's blog which mixes rationality and game theory.

Do shareholders of the merged Lloyds-HBOS want to retain the management (inherited from Lloyds) which got them to where they are today?

The strictly rational answer is to look only at the future, and the expected value of retaining versus terminating the managers. Rational agents do not consider the past, as you cannot incentivise for past actions - they have already happened and there is no possibility to change them.

In which case, shareholders should make a prediction about these executives' likely future performance. Of course, they do need to use past actions and performance as data points in estimating future performance. Thus, absorbing HBOS which was probably an error of judgment, should be set against the other (generally smart) actions that they took while managing Lloyds.

However, in reality people do consider the past while making these kinds of decisions, as if their current actions could somehow influence history. This is one of the numerous predictable cognitive biases that humans are subject to.

Game theory gives us some justification for why this bias might exist. Sometimes, the only way to win a game is to appear irrational. The classic example is playing chicken with another driver - pointing your cars towards each other from opposite ends of the road and pressing the accelerator. Whichever of the players rips out their steering wheel first is almost certainly going to win, because the other driver realises they are the only one who can swerve.

So rational actions sometimes are not the best winning move.

Repeated games introduce yet another complication: does your action in the last game create an expectation for the next game? In this case, if shareholders ignore the big mistake, will it lead executives to make more mistakes in future because they feel there is no punishment?

This is also the argument for punishing criminals - someone who kills his wife may be judged unlikely to kill again, in which case we may wonder why we should bear the cost of spending hundreds of thousands of pounds - maybe millions - to run the trial and imprison them. But we do it because it sends a signal to other potential criminals and generates an incentive for their future behaviour.

If these people are skilled executives who will make good decisions in the future, Lloyds-HBOS shareholders might be punishing themselves if they decide to send a signal to future executives - including those of other companies - that they should be more careful about making acquisitions. If the other companies are the ones who will benefit, why should Lloyds shareholders pay the cost by losing these management skills from their firm?

The answer might lie in the government shareholding. If, as Peston suggests, the executives offer themselves for re-election while asking the government to abstain, then the private shareholders will be incentivised to reappoint them. The government, on the other hand, has an interest in capturing the externality that is created by sending a message to other directors.

While game theory does actually give us a way to make this decision if we can work out the probabilities of success and failure, and the costs in each case, it is horrendously complex to do so. Is there a rational way to make the decision which does not require the same level of calculation and risk of getting it wrong?

Externality pricing feels like it would be one way to do this. This position would argue that (believe it or not) the government should actually subsidise companies to fire their managers! A simplistic version of this policy would of course be open to abuse, and it could well be impossible to design a workable version. But pure economics suggests that there is a case for it.

Another way to simplify the decision would be to develop a model to predict the management skills of a given executive. If this model were published on the Web, along with whatever data it needs to generate its predictions, it would give shareholders a way to make a more informed decision - though it would not remove the externality problem.

Of course such a tool would be difficult to design and you could never achieve a consensus on one perfect way to predict performance. Undoubtedly though, it would be an interesting tool for investors to use and would provide an extra incentive for managers to behave well and to demonstrate that they are doing so. The political sphere already has a couple of similar websites - They Work For You and Public Whip, for instance - so there is a model for this.

Finally one could try using market prices, by allowing people to buy and sell some kind of derivative on the performance of particular executives. This would help show the skills of the executives (or at least the market's estimate of them), and could also provide a way to manage their compensation.

In the case of Lloyds this might point the way to the best answer: perhaps they should be allowed to keep their jobs but lose this year's salary and bonus in order to demonstrate that the mistake in acquiring HBOS is being taken seriously. Of course they have probably already lost a huge amount of value in shares and options from the collapse in share value over the last year, but there's still a value for shareholders in signalling on top of this.

3 comments:

Matt said...

A very interesting take on it all. Maybe the past data of the executives' performances could be plugged into this system:

http://www.wolframalpha.com/

All you have to decide is what question to ask it.

Leigh Caldwell said...

That does look like an interesting system - at least from what little is visible on Stephen Wolfram's blog posting.

Not sure if investors or regulators would be keen to rely on something that is so non-deterministic (not literally non-deterministic of course, but it looks like it's complex enough to be impossible for anyone to meaningfully understand the path between inputs and output). But it will certainly be intriguing to see what outputs it comes up with. Roll on May.

Matt said...

More info here:

http://www.twine.com/item/122mz8lz9-4c/wolfram-alpha-is-coming-and-it-could-be-as-important-as-google

I think the potential beauty of a system like this is that it could show us things about the data that we didn't realise were there, as well as bringing together a wide range of data sets. In times to come systems like this will trawl the web for their data and tell us all sorts of weird and wonderful things about who and what we are.