A New Approach to Multitask Agency Problems

22 October 2012 at 12:44 am Peter G. Klein 8 comments

| Peter Klein |

The standard approach to multiask agency problems is to recognize that, if the output of some tasks is more easily measured than the output of other tasks, than others, then piece-rate incentive schemes will lead to a distortion of effort toward the more easily monitored tasks. Ask a sales clerk to sell merchandise and keep the store clean and the displays spiffy, and pay on a commission basis, and you’ll get a messy store. A new paper by Omar Al-Ubaydli, Steffen Andersen, Uri Gneezy, and John List challenges this view, arguing that using a piece-rate schemes signals that the principal is a good monitor in general, which can motivate performance even on the not-easily-measured tasks in a multitask setting:

Carrots that Look Like Sticks: Toward an Understanding of Multitasking Incentive Schemes
Omar Al-Ubaydli, Steffen Andersen, Uri Gneezy, John A. List
NBER Working Paper No. 18453, October 2012

Constructing compensation schemes for effort in multi-dimensional tasks is complex, particularly when some dimensions are not easily observable. When incentive schemes contractually reward workers for easily observed measures, such as quantity produced, the standard model predicts that unrewarded dimensions, such as quality, will be neglected. Yet, there remains mixed empirical evidence in favor of this standard principal-agent model prediction. This paper reconciles the literature by using both theory and empirical evidence. The theory outlines conditions under which principals can use a piece rate scheme to induce higher quantity and quality levels than analogous fixed wage schemes. Making use of a series of complementary laboratory and field experiments we show that this effect occurs because the agent is uncertain about the principal’s monitoring ability and the principal’s choice of a piece rate signals to the agent that she is efficient at monitoring.

Entry filed under: - Klein -, Recommended Reading, Strategic Management, Theory of the Firm.

Mormons in Management Social Justice Quote of the Day

8 Comments Add your own

1. RussCoff | 22 October 2012 at 6:45 am

There is also the distinct possibility that the tasks are correlated. In your example, more customers may come into a clean store with spiffy displays. Rewarding based on commissions will also encourage the other behaviors. This is the assumption in my paper with Rich Makadok (Both Market and Hierarchy — http://dl.dropbox.com/u/10984177/PublishedPDFs/BothMarketAndHierarchy-Hybrid-AMR09.pdf)
2. koppl | 22 October 2012 at 8:22 am

I’m not sure I understand the analytics, Peter. If different principals have different monitoring abilities, only those relatively skilled at monitoring quality will pay piece rates because only they are not subject to the multitask problem. Am I missing something or does it boil down to that?

Let’s say I’ve got it. It’s hard to see what one might object to in that analysis, but then I’m not sure I see where it illuminates something in the world. I scanned only the initial bit. Do they give real-world examples for their model? I do not yet understand why this is an interesting result.
3. Peter Klein | 22 October 2012 at 10:09 am

Russ, yes, but even in that case there’s a potential free-rider problem — the marginal impact on my own sales commissions of contributing to a group activity like cleaning may be small.

Roger, that’s the idea. It’s interesting (to some of us) because it contradicts the standard argument that principals facing multitask problems should either separate the tasks or use fixed-payment schemes or subjective evaluation criteria. But de gustibus non est disputandum.
4. koppl | 22 October 2012 at 2:13 pm

Peter,

It could be that I don’t have the right context to appreciate the value of the study. But I don’t see where it’s a multitask problem if the costs of monitoring “quality” are not only finite, but relatively low. By definition (in Holstrom and Milgrom at least) it’s a multitask problem only when those costs are high or infinite. Should I infer that a lot of people have been imputing to Holstrom and Milgrom views they did not express? You know, like people who mistakenly criticize Coase for saying property right don’t matter.
5. Peter Klein | 24 October 2012 at 1:25 pm

Roger, you’re interpreting multitask agency theory very narrowly. There’s a big literature on this class of problems, with many differences in details. M&R’s model isn’t the only one. To make the problem interesting, you need some differences in task characteristics that lead to a potentially inefficient allocation of effort — differences in the costs of measuring quality is one example, but not the only one.
6. Roger Koppl | 24 October 2012 at 1:36 pm

Peter,

I don’t think I understand your comment. You say, “To make the problem interesting, you need some differences in task characteristics that lead to a potentially inefficient allocation of effort.” Yeah, no kidding. That was what Holstrom and Milgrom were, um, talking about. You continue, “differences in the costs of measuring quality is one example, but not the only one.” Um, yes, that true. I lifted the word from, cough cough, the Al-Ubaydli paper you cite. That’s only dimension *they* refer to in the early analytical part of their paper. I put the word in quotes to suggest that it was a stand-in for any dimension of the job that might be hard for a principal to monitor. I assume that what Al-Ubaydli et al. hand in mind. If I’m wrong, well so much the worse for them for being so narrow I guess. Certainly, H&M do not restrict their attention to quality; they were clear about the basic logical structure of their argument. Sorry, Peter, but I really don’t see where I’m somehow being narrow in my interpretation of the multitask problem.

I did say ask whether there mightn’t be a lot of people who misinterpret the multitask problem in something like the way many people misinterpret the Coase theorem. Your reference to “a big literature on this class of problems” strengthens my suspicions in this regard.
7. Peter Klein | 24 October 2012 at 11:34 pm

Sorry, perhaps I misunderstood your last question. In fact, I’m still not sure I see what you’re getting at. But, hey, either you like the paper or you don’t. I found it an interesting twist on the usual story — even when the standard models say to avoid piece rates, using them may serve as a signal of the principal’s ability.
8. Roger Koppl | 25 October 2012 at 10:18 am

Put it this way, Peter. If “the standard models” say to avoid piece rates in the cases in which Al-Ubaydli et al. say piece rates should be paid (or will be in the separating equilibrium), then the original 1991 multi-task paper of Holstrom & Milgorm is not one of the standard models.