But before the verbal melee (which I clearly won, btw), some interesting stuff got said that I thought I'd comment on.
Kevin, responding to a Rafe point, wrote:
Any compensation scheme will eventually start to be gamed. So you have to constantly improve them.
By "gaming," we usually mean two things. Either (1) the employee takes an action that improves measured performance (and hence increases pay) but doesn't actually create value, or (2) the employee fails to take an action that would create value, because taking that action doesn't increase measured performance.
Except in a very few cases, it's generally true that explicit pay-for-performance schemes can be gamed. This stems from problems in measuring performance. Unless your way of measuring performance reflects all the things you want done and none of the stuff you don't want done, you're due for some gaming behavior.
So, what can you do about gaming?
Option (1) Live with it. This option works best when the gaming isn't that costly to the firm. For many salespeople, for example, pay comes from sales commissions. This is an explicit pay-for-performance contract that rewards salespeople who sell more, so that's pretty good alignment between the measure and what the firm wants done. However, there are some documented examples of gaming, even here. Many commission contracts feature quotas or other kink points. Suppose for example that I'm a salesperson whose contract calls for a $10,000 bonus if I hit annual sales of $2 million. Suppose it's December 15, and I've already hit my target --- I'm best off pushing all sales for the rest of this year into next year. This probably isn't what the firm wants me to do, since competition could find these customers and try to steal them away as I dawdle to push the sales into next year. Paul Oyer (at Stanford) and Ian Larkin (HBS) have written nice papers documenting this.
Option (2) Don't use pay-for-performance incentives at all. This option works best when the gaming is really bad. If you think about some of the objections to pay-for-performance for public schoolteachers, they usually go along these lines. Some teachers unions argue that paying teachers for raising student test score causes reductions in education quality, because teachers misdirect their effort toward test scores and away from other valuable educational activities. The argument is that teacher gaming would be so bad that we're better off with no incentive pay than with incentive pay that's directed at test scores.
Option (3) Use subjective assessments of employee performance in addition to the explicit pay-for-performance plan. The idea here is that a "boss" subjectively assesses whether employee performance was achieved in the "right" way. Both the subjective assessment and the numerical measure play a role in determining pay. Examples: A salesperson's supervisor might pay attention to whether the saleperson's performance slackens after a target is met, and reduce pay if so. A school principal might tell a teacher to raise test scores as much as possible, but then monitor the teacher to make sure that the teacher doesn't drop art and science to simply do arithmetic drills all day.
A final example: A board of directors might monitor the CEO to make sure that earnings growth is achieved in a sustainable manner. There is evidence that boards do at least some of this: See my paper with Rachel Hayes from the RAND Journal of Economics in 2000.
So, getting back to the Rafe and Kevin conversation, if it's impossible to design a bulletproof pay-for-performance plan, then maybe instead we think about making sure that we're combining the (inherently flawed) explicit pay-for-performance plan with strong board oversight and subjective evaluation.
One of the big problems with subjective evaluation of performance, however, is that we now face the problem of providing incentives for the evaluator to give good evaluations.
And so one response to Rafe's initial idea is this: Let's not think that making a one-size-fits-all change to explicit pay-for-performance plans can solve all our corporate governance problems. Probably this could never work due to the inherent problems of measuring performance. So let's instead continue to give directors the freedom to compensate CEOs in the way that fits the specific needs of the firm. But let's also figure out how to give stronger incentives to directors to do a good job monitoring CEOs.
Easier said than done, but I'll post some thoughts sometime.
Posts
6 comments:
I think it's funny that we staked out essentially the same positions we did 20 years ago on any arbitrary problem. You always think things are equilibrium problems. I always think they're decision problems. Rafe always thinks they're algorithm problems. Of course, if Smith were participating, he'd think regulating executive compensation was a flow problem :-)
Not just any ole flow... Supersonic flow,
Supersonic or otherwise, there would definitely be lots of Reynolds numbers involved.
"One of the big problems with subjective evaluation of performance, however, is that we now face the problem of providing incentives for the evaluator to give good evaluations."
With information technology this becomes easier. Crowdsource it anonymously. Include input from all employees, all Board members, shareholders, etc. Knowing that the evaluation feeds into executive compensation should be enough incentive to take it seriously. Nobody evaluating needs to know the exact formula or weighting of the eval, that would be counterproductive.
Excellent ME reference with the Reynolds number. Shows almost as good of memory as nailing our Rains unit.
Anonymous "crowdsourcing" (where to you techno-entrep-business geeks get these words?) might help, but it surely won't solve the problem.
Gaming can happen any time that (1) there's an action that creates value but doesn't improve the measure, or (2) there's an action that improves the measure, but doesn't create value.
For an example of (1): Can you imagine a case where the CEO knows what the value-maximizing strategy is, but employees either don't like it (say, because the strategy involves layoffs) or disagree (because they don't have access to all the information the CEO does).
For (2): Is there anything a CEO could do to improve his popularity with the firm's employees but that doesn't actually create value? How about excessive glad-handing site visits to the firm's facilities?
Post a Comment