ESRA logo

ESRA 2023 Program


All time references are in CEST

Surveying or gaming: How to best measure socio-economic behaviors and attitudes? 2

Session Organisers Mr Jakob Jonathan Kemper (University of Duisburg-Essen)
Dr Jan Karem Höhne (University of Duisburg-Essen)
Professor Achim Goerres (University of Duisburg-Essen)
TimeWednesday 19 July, 16:00 - 17:30
Room U6-05

People’s socio-economic behaviors and attitudes are of key interest in a variety of scientific disciplines, including, but not limited to, social, political, psychological, and economic research. These disciplines partially differ with respect to their measurement methods. On the one hand, there are many researchers that rely on indirect behavioral and attitudinal measures from large-scale sample surveys. On the other hand, there are also many researchers that rely on more direct measures from behavioral games, such as the dictator, solidarity, and trust games, that are commonly conducted in labs with small convenience samples. Considering the eminent literature, there is some conventional wisdom about the merits and limits of both measurement methods. For example, survey measures are frequently criticized for their hypothetical touch and that they do not decently mirror people’s real socio-economic behaviors and attitudes. Behavioral games, in contrast, are frequently criticized for their artificial settings and samples, which impedes drawing conclusions beyond the studies. Despite this conventional wisdom, there is only a small body of research investigating and evaluating the soundness of both measurement methods. In this session, we therefore invite scientific contributions that present experimental and/or non-experimental research on measuring socio-economic behaviors and attitudes in a variety of research settings (e.g., lab or field) and modes (e.g., in-person or online). We also welcome mixed-method contributions that, for example, combine surveys and games to improve the measurement of socio-economic behaviors and attitudes.

Keywords: socio-economic behaviors and attitudes, experiments, observational studies, behavioral games, interdisciplinary research


Living in Novaland: Can we Simulate the Experience of States, Economies and Public Policies in a Virtual Online State?

Professor Achim Goerres (University of Duisburg-Essen)
Dr Jan Karem Höhne (University of Duisburg-Essen) - Presenting Author
Mr Jakob Kemper (University of Duisburg-Essen)

What if we could experimentally manipulate all characteristics of states, economies and public policies and estimate their effects on citizens? This paper puts forward the evidence from Novaland, a virtual liberal democracy that only exists online and that has characteristics realistically drawn from German and Romanian welfare states. The study consists of an experimental online platform based on text and images in which volunteers are surveyed after they have exited the experience, are randomly assigned to different experiences, such as defined by income, corruption or unemployment, interact with each other simultaneously and thereby co-create collective decisions, such as elections or donation pools, that then determine the course of Novaland. The study was conducted in November 2022 with participants that were recruited via social media platforms, such as Facebook and Twitter. Our main results are: (1) The newly programmed Novaland Experience worked technically well in that over 300 participants could simultaneously interact online at the same time. (2) Participants behaved in an internally valid manner, even though there are some signs of inattentiveness. (3) The political ideology of volunteers was reflected in how they behaved in Novaland, whereas socio-demographic variables showed no systematic effects.

Do firms with works councils prefer agreeable applicants? A discrete choice experiment in German firms

Professor Harald Pfeifer (Federal Institute for Vocational Education and Training (BIBB))
Dr Caroline Wehner (Federal Institute for Vocational Education and Training (BIBB)) - Presenting Author
Professor Andries de Grip (ROA/ Maastricht University)
Professor Julia Kensbock (University Bremen)

Firms with work works councils tend to provide more secure jobs and higher wages (Addison et al. 2010). Consequently, works councils provide an incentive to apply for a job in the respective firms, especially for workers at the lower end of the skill distribution (Kugler/Saint Paul, 2004). This suggests that these firms must screen applicants more carefully to prevent future costs of dealing with “difficult” or less co-operative employees afterwards.
Recent literature shows that personality and attitudes are important predictors for an employee’s productivity and behaviour at the work place (see e.g. Deming/Kahn 2018). Agreeableness has been shown to be an important determinant for conflict-, compromising- and obliging-styles of employees (Tehrania/Yaminib, 2020). If works council firms aim to avoid “difficult” employees, recruiting employees with high agreeableness is particularly attractive.
Thus, we ask whether works council firms value agreeable applicants relatively more than recruiters from other firms to prevent counterproductive work behaviours. To answer this question, we use a discrete choice experiment among firms to assess whether recruiters in firms with a works council have a higher preference for more agreeable workers. Both the discrete choice experiment and the presence of employee representation are assessed as part of the BIBB Cost-Benefit Survey 2017/2018. Our analysis relies on 10,466 worker profiles from 5,233 recruitment decisions from 786 respondents. The job applicants differ in the following attributes: professional competence, personality, and gross wage claimed. We analyse the choices made using a mixed logit regression model (e.g., Revelt/Train (1998)).
Our results confirm that recruiters from firms with a works council have a higher preference for agreeable workers than do recruiters from firms without a works council. Our results provide empirical evidence for employer selection mechanisms induced by works councils.

Mass Online Interactions: an oTree Coalition Formation Game in the LISS panel

Mrs Marije Oudejans (Centerdata - Tilburg University)
Mr Stein Jongerius (Centerdata - Tilburg University) - Presenting Author
Dr Seyit Höcük (Centerdata - Tilburg University)
Mr Joris Mulder (Centerdata - Tilburg University)

Social and economic experimental research is commonly performed with non-representative groups, often in a university laboratory setting using students as subjects. Complementing laboratory experiments with field experiments adds external validity and a more diverse subject pool. oTree, an open source Python-based tool for interactive behavioral research and experiments facilitates this in an online laboratory environment (Chen, Schonger and Wickens, 2016).

oTree can be utilized in small-scale controlled laboratory settings or with large-scale online research platforms, such as MTurk or Prolific. By implementing oTree in the Dutch LISS panel, we offer best of both worlds, a large-scale controlled setting. The LISS panel is a probability-based online panel representative for the Dutch household population. The open-access LISS Data Archive comprises an extensive database of background characteristics, which can be merged for more in-depth analyses or used to select specific population groups. In addition, respondents can be specifically instructed before starting the experiment, while they can receive technical support if needed.

We fielded an oTree game in the LISS panel for a large-scale experiment on human behavior regarding coalition formation. The project, Compose 2.0, aims to understand how human behavior affects scaling up horizontal collaboration in supply chains. A game was designed in which 600 LISS panel respondents played the part of a transport company. The oTree game automatically distributed players into 200 groups of three, who collaborated on the transportation of goods and negotiated on profit distribution (actually paid out).

5,000 panel members were invited to participate in the game. A quarter of the respondents committed to participate in the experiment and 82% of those actually joined the game. The study was successfully completed and taught valuable lessons to further optimize controlled real-time online multiplayer games in the LISS panel.