Schedule - Parallel Session 7 - Payment Schemes and Incentives

Engineering F105-106 - 09:00 - 10:30

Contingent Payment Schemes: Moral Hazard with Agents Concerned about Sure Wages

Jingyi Meng


We modify the classical principal-agent model with uncertainty and moral hazard by replacing the Expected Utility preferences of the agent with Chance Theory preferences (Schmidt and Zank, 2013). Chance Theory agents are primarily concerned with the sure wage they can obtain, i.e., the certain component in their contract, as they treat increments in bonuses markedly different to similar changes in sure wages. Similar to the classical predictions, our agents’ optimal contracts are contingent payment schemes, however, they differ with respect to the level of the sure wage. We also contrast our predictions to those of the model of (Herweg et al., 2010), who assume agents with loss-averse preferences.

Jingyi Meng

Student, University of Manchester

Goal Setting in the Principal-Agent Model:Weak Incentives for Strong Performance

Joaquín Gómez-Miñambres; Brice Corgnet; Roberto Hernán-Gonzalez


We study a principal-agent framework in which principals can assign wage-irrelevant goals to agents. We find evidence that, when given the possibility to set wage-irrelevant goals, principals select incentive contracts for which pay is less responsive to agents’ performance. We show that average performance of agents is higher with goal setting than in its absence despite weaker incentives. We develop a principal-agent model with reference-dependent utility that illustrates how labor contracts combining weak monetary incentives and wage-irrelevant goals can be optimal. It follows that recognizing the pervasive use of non-monetary incentives in the workplace may help account for previous empirical findings suggesting that firms rely on unexpectedly weak monetary incentives.

Joaquín Gómez-Miñambres

Assistant Professor, Chapman University

Consistency in Choices from Linear Budget Sets: Reconciling Contradicting Experimental Evidence

Dotan Persitz; Dino Levy


Three types of methodologies are used in laboratory experiments to provide individual level data on choices from linear budget sets in settings that invoke preferences over goods, risk preferences or other-regarding preferences. In the Verbal methodology subjects are asked to complete a sentence that describes their preferred bundle. The Graphical methodology requires subjects to choose their preferred bundle from a budget line that is visually presented and describes the set of feasible alternatives. The Discrete methodology (which is not in the focus of this work) asks the subjects to choose from a set of images that represent the available bundles. The data provided by studies that use these methodologies is similar. For every subject, each observation includes a description of the budget set and a chosen bundle. Since these methods study a similar question and produce similar data it is reasonable to expect that they would also provide similar conclusions regarding individual choice behavior. However, the percentage of subjects that satisfy GARP when the verbal methodology is implemented is significantly higher than when the graphical methodology is used. In addition, the inconsistency measure is much closer to consistency when using the verbal methodology compared to the graphical methodology. We find this discrepancy puzzling and this work is an attempt to reconcile these results. We suggest three aspects for the differences between the methodologies. First, each methodology might frame subjects differently. The other two aspects are related to the power of these experiments – the number of choice problems each subject is required to solve and the span of price ratios each subject encounters. We designed a series of experiments, in the context of other-regarding preferences, where in each decision problem, a subject is given a number of tokens and is asked to allocate them between herself and an anonymous other. Each token that she allocates to herself is multiplied by a parameter a, while every token she allocates to the other is multiplied by a parameter b. Each subject is confronted either with an implementation of the verbal methodology or with an implementation of the graphical methodology. Irrespective of the methodology she is facing, the subject is randomly assigned with a number of rounds and with an upper bound to the price ratio. The existing literature predicts that a subject that will be assigned to the verbal methodology with few trials and small span of price ratios will probably demonstrate consistent choices. However, a subject that will be assigned to the graphical methodology with many trials and large span of price ratios will be expected to demonstrate inconsistent choices. This design induces large variation in the three main aspects described above. We intend to run these experiments between January and April 2016 so that by June 2016 the project will be ready for presentation.

Dotan Persitz

Assistant Professor, Tel Aviv University

Preference for Higher Order Moments and the Principal-agent Dilemma

Luc Meunier; Francois Desmoulins-Lebeault


The 2007 financial crisis has highlighted the importance of the impact and management of extreme risks and of the principal-agent dilemma in the investment industry. The fact that the risk and return profile of the agents differs from the one of the principals, and even more so for large risks, has been pointed out as one of the causes of the crisis. To test the importance of preferences about higher order risk and how it is impacted by agency relations, we submitted a questionnaire to a sample of 308 business school students. In it we tested for investment preferences over the first four moments of the distribution of returns in a principal-agent framework. Respondents acted successively as principals and agents, the order of such tasks being randomized. We find evidence of standard deviation aversion, skewness seeking and kurtosis seeking when investing for self. The overall kurtosis seeking of our sample as a whole is mainly driven by standard deviation lovers, standard deviation averse respondents being neutral toward kurtosis on average. These results are similar to the ones obtained by Deck and Schlesinger (2010), and corroborate their 2014 classification of mixed risk aversion. We find that preferences when investing for a firm are not significantly different at an aggregate level. However, there is a significant impact of financial literacy on preferences for standard deviation. Financially illiterate respondents were more likely to take risks when investing for their firm. We also find a marginally significant effect of gender on skewness preferences when investing for the firm, males tending to become more skewness seeking when investing for their firm whereas the opposite holds true for female. This marginal change in preference for skewness is an interesting addition to the body of knowledge on risk preference differences between male and female. Indeed, we are the first to test for such a gender based difference in higher order risk preferences (see Croson and Gneezy 2009 for a literature review on risk aversion, where a consensus is still to be reached). The change in preference for standard deviation has two implications. First, it helps explain the heterogeneity of results observed in the literature concerning risk aversion under responsibility. The fact that respondents are sometimes more risk averse (Reynolds et al 2011 or Charness and Jackson 2009) and sometimes less risk averse (Chakravarty et al 2011 or Pollmann et al 2014) when investing for others could be explained by different levels of financial literacy. Secondly, this can be connected to the 2007 financial crisis. Before the subprime crisis, the securitization process created opacity. This opacity probably lead to a lower level of relative financial literacy given the complex financial products created. In turn, this could explain why a higher level of risk taking was then observed.

Luc Meunier

PhD student, Grenoble Ecole de Management