2. Informe del Director
4.2. Proyectos de Investigación
Our hypotheses address several issues which we group into the following categories: Monetary incentives, reputation, social in-game influences, objective quality, and personal traits. The main monetary incentive is the cash prize. However, because
the prize has not changed over the years, the monetary motivation (e.g., change of the prize) can be disregarded, and thus, extrinsic motivation does not vary across observations and for this reason cannot be analyzed on basis of our data. To ap- proximate the impact of reputation on our dependent variable, we consider the impact of past voting behavior. This hypothesis is derived the work of Bernheim (1994) and Benabou and Tirole (2006). Bernheim (1994) considers the emergence of customs or why people tend to conform to certain social standards. His model explains conformity by comparing the relative value of popularity to intrinsic utility, and shows that agents with non-extreme preferences choose to conform to a norm in equilibrium. Benabou and Tirole (2006) consider a model in which people also care about appearing pro-social, depending on the visibility of their actions. Since voting behavior is visible to the public once the show is aired, past voting behav- ior can be considered as a social norm current contestants want to conform to. In the literature this behavior is often associated with the term “bandwagon” effect (Aardema et al., 1977). If a high level of evaluations has evolved as a “social” focal point (Schelling, 1960) over time, this might help to explain the voting behavior of participants. (Young, 1996) shows that focal points such as a 50:50 split in bar- gaining situations are not focal right away, but evolve over time and remain stable afterwards. We control for the influence of past voting behavior by including the independent variable “Average evaluation level,” which reflects the average points given over the last 24 weeks. Our first hypothesis is then as follows:
H1: “The higher evaluations have been in the past, the higher evaluations are today.”
Concerning social in-game influences, we first consider potential direction-of- comparison effects caused by the sequential voting structure of the show. Direction- of-comparison effects can influence voting behavior in the following way:
”It appears that judges form an impression of each new option by comparing it to those that preceded it. Using that option’s features as a ‘checklist,’ more weight is given to unique ones than to ones shared with previous options. This unidirectional comparison process produces increasing ratings in options with unique positive fea- tures, and decreasing ratings when options have unique negative features.” (Bruine de Bruin and Keren (2003), p. 91).
We use the starting position of a cook to measure this. Haigner et al. (2010) find that in “The Perfect Dinner,” first day contestants are evaluated significantly more harshly than contestants who cook later. However, the authors do not control for a range of effects that we consider: the effect of objective quality, the effect of being
socially similar to another person, the effect of past voting behavior and the effect of having already performed. We expect to see an overall positive evaluation bias for contestants that perform later, since subsequent cooks can adjust the performance elements of their dinner (e.g., having an artist perform, playing music themselves, etc.) and gather more information about other contestants (e.g., sympathy or pref- erences). Of course, unique negative performances of later cooks will lead to a lower rating, but overall contestants that cook at a later stage in the game have a better chance to adjust to the expectation of the group. Due to this process we expect to see an overall positive effect of cooking later. Our second hypothesis is then the following:
H2: “The later a contestant performs, the higher the evaluations he receives.”
The variable “Already Cooked” indicates whether the evaluator has competed before the contestant to be evaluated. This controls for the effect that once a partici- pant has cooked, he might be more critical of another’s performance since he will not be evaluated again. Put differently, we are controlling for contestants overestimating their own performance. This is a recognized phenomenon in the psychological and economic literature7. Moore and Healy (2008) find that overestimation especially
plays a role when difficult tasks are performed. Arguably, preparing a three course dinner for 5 persons can be categorized as such a task, and might lead a contestant to overestimate his own performance. Our third hypothesis is then the following:
H3: “The earlier a contestant performs, the lower the evaluations the contestant gives to the remaining cooks.”
In order to test for the influence of objective measures, we use the total number of ingredients used in the three dishes prepared on one day, the price and difficulty level. Our fourth hypothesis concerning the effect of the objective characteristics of a dinner is then the following.
H4: “The more sophisticated a performance is, the better the evaluations.”
Concerning personal traits, we separate our analysis into two parts: In the first part we consider an aggregate measure of dissimilarity between contestants, mea- sured by the variable “Overall Dissimilarity”. We expect participants who have dissimilar social characteristics to give each other lower evaluations. This effect has been recognized in the experimental economic and psychological literature (Bra˜nas Garza et al., 2010, 2012; Engel, 2011). Our fifth hypothesis is then the following:
H5: “The less similar two contestants are, the lower their mutual evaluations.”
As a second step, we take a closer look at what specific characteristics exert the strongest influence on the aggregate dissimilarity measure. We will not specify a range of hypotheses for each characteristic. Instead, we will present a table with specific results for some traits and their impact on the aggregate measure and discuss the individual impact each factor has8. The next section presents our econometric approach in order to answer these questions.