But Who Will Monitor the Monitor? [Extended Abstract]



David M. Rahman Department of Economics University of Minnesota 4-101 Hanson Hall 1925 Fourth Street South Minneapolis, MN 55455

[email protected] ABSTRACT

if every deviation is detectable, although different actions may be used to detect different deviations. By the Minimax Theorem, there is a payment scheme that discourages all deviations simultaneously if and only if for every deviation there is a payment scheme that discourages it, where different schemes may be used for different deviations. For any given deviation, such payment scheme exists if and only if the deviation is detectable. The use of request-contingent payments allows for different actions to be used to detect different deviations. Indeed, implementability without such payments requires the same action to detect every deviation. I also characterize virtually implementability of a fixed outcome, rather than every outcome, regardless of preferences. Interestingly, this result reconciles the following infinite regress inherent in monitoring. Suppose that providing incentives for a given outcome requires a monitor to detect deviations. What about the monitor’s deviations? I show that the monitor’s deviations are irrelevant. By the previous Minimax argument, we may consider every deviation individually. If such a deviation is detectable then it can be discouraged with a payment scheme. Otherwise, if the monitor’s deviation is undetectable then by definition it does not change the probability of any reported signals, therefore the deviation itself still detects others’ deviations from the given outcome, and so it continues to fulfill the required monitoring role. This argument also applies to the monitor’s deviations from these deviations, and so forth. Under standard conditions (e.g., in a finite game) not every behavior by the monitor can have a profitable, undetectable deviation. Hence, a given outcome is virtually implementable with infrequent monitoring regardless of preferences if and only if every deviation from the outcome is detectable with some monitoring behavior, but deviations away from the monitoring behavior itself need not be detectable. Heuristically, nobody needs to monitor the monitor. I extend this result by also fixing preferences as follows. A given outcome is virtually implementable if and only if all profitable deviations can be discouraged “uniformly” and “credibly.” Intuitively, uniformity assumes that all deviations can be discouraged with payments of the same order of magnitude even if they are only infinitesimally detectable, where still different deviations may be detected with different actions. Credibility means that if deviations are discouraged with others’ actions instead of contingent money payments, then this disciplining behavior is incentive compatible.

Consider a group of individuals whose behavior is subject to moral hazard, and suppose that providing them with incentives requires a monitor to detect deviations. What about the monitor’s deviations? I study mediated contracts and find that the monitor’s deviations are effectively irrelevant. Hence, nobody needs to monitor the monitor. I also characterize exactly when such contracts can provide the right incentives for everyone. In doing so, several new characterizations of virtual implementation are derived.

Categories and Subject Descriptors J.4 [Social and Behavioral Sciences]: Economics; H.1.1 [Models and Principles]: Systems and Information Theory—Value of information

General Terms Economics, Management, Theory

Keywords Mediated contracts, monitoring, virtual implementation A principal asks some agents to take a given action. They then independently choose some behavior and subsequently acquire private information (“signals”), possibly correlated with others’ choices. To provide incentives, the principal can (i) make confidential, verifiable but non-binding requests, and (ii) pay money contingent on both his requests and the observations reported by the agents. What outcomes can this group enforce with such “mediated contracts”? An outcome (correlated strategy) is implementable if for some such payment scheme, every deviation from requested behavior is unprofitable. A deviation is detectable if for some action profile, the probability of any reported signal that it induces differs from that induced by honesty and obedience. First I prove that every outcome is virtually implementable (i.e., there is an implementable outcome arbitrarily close) with mediated contracts regardless of preferences if and only ∗A full version of this paper can be found by following this link: http://www.econ.umn.edu/~dmr/monitor.pdf. Copyright is held by the author/owner(s). EC’09, July 6–10, 2009, Stanford, California, USA. ACM 978-1-60558-458-4/09/07.

157

But who will monitor the monitor?

tives requires a monitor to detect deviations. ... Mediated contracts, monitoring, virtual implementation ... link: http://www.econ.umn.edu/~dmr/monitor.pdf.

299KB Sizes 0 Downloads 57 Views

Recommend Documents

But Who Will Monitor the Monitor?
thereby provides incentives even when the monitor's observations are not only ... Nevertheless, using a version of Bob's contract I show how to make .... Transportation, Medicare, and the US Department of Labor, to name just a few. These.

But Who Will Monitor the Monitor?
I add to the debate by studying a theoretical model that accom- modates ...... to such budget constraints, and a similar general theme prevails. As they show, in.

VOIP voice interaction monitor
Aug 24, 2006 - devices for identifying at least one predetermined parameter by analyzing the ..... Holfelder, Wieland, Tenet Group, International Computer Science. Institute and .... CTI News, Year End Issue, New Products From Amtelco XDS, Tech .....

Monitor Assets, Anytime, Anywhere
business environment, leaders charged ... The good news: In our wireless world, it's never been ... electronic systems to real, live people out in the field every.

Monitor Extn.pdf
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. Monitor Extn.pdf.

pharma science monitor pharmacognostical, preliminary ...
The mass fragmentation studies revealed the presence ... The free hand cut transverse sections of the seed were taken and ... Statistical analysis was performed using GraphPad In Stat software. RESULTS ..... Email: [email protected]

Spartina Treatment Monitor Internships, San Francisco Estuary ...
Interns will also be trained in the use of desktop-based. ArcPad and/or ... Must be willing to conduct field surveys during and directly adjacent to application of.

Heart Rate Monitor Technical Manual.pdf
Connect more apps... Try one of the apps below to open or edit this item. Heart Rate Monitor Technical Manual.pdf. Heart Rate Monitor Technical Manual.pdf.

CPEC Monitor November 2017.pdf
Nov 1, 2017 - Page 3 of 4. CPEC Monitor November 2017.pdf. CPEC Monitor November 2017.pdf. Open. Extract. Open with. Sign In. Main menu. Page 1 of 4.

AEIT 2016 monitor ottico.pdf
PERATURAN DIRJEN DIKTI PEDOMAN OPERASIONAL. Desember 2014. Page 3 of 4. AEIT 2016 monitor ottico.pdf. AEIT 2016 monitor ottico.pdf. Open. Extract.

Who Monitors the Monitor? Effect of Party Observers on ...
Oct 21, 2016 - turnout polling booths and higher shares for one party (the PRI). Larreguy (2012) exploits .... 15However, there might be other reasons for manipulation: a political party could rig elections to win by ... votes that do not contribute