Under my ERC Starting grant "Opinion Dynamics" (2018-2023) I study how people form and revise beliefs about themselves and the world and how social identity affects this process.  We study opinion dynamics in social networks,  committees or other groups and  ask when opinion dynamics lead to good outcomes and how they can derail to produce "collective delusions".  Our research has shown among others that deliberation in committees can lead to an increase in gender bias, that social identity affects belief formation in a variety of contexts and that failures to update rationally could be behind a substantial share of observed discrimination. Social Identity can be both a boon and a bane for efficient information aggregation. On this site I periodically update about the activities and outcomes from the grant. Support from the ERC is gratefully acknowledged.

Publications: 

• Gender Bias in Opinion Aggregation, International Economic Review 62(3) (2021), 1055-1080.

Abstract: Gender biases have been documented in areas including hiring, promotion or performance evaluations. Many of these decisions are made by committees.  We experimentally investigate whether committee deliberation contributes to gender biases. In our experiments participants perform a real effort task with subjective performance and then rate the task performance of other participants. In a 3 x2 design we vary the extent to which communication among raters is possible and whether or not the experiment is gender-blind. There is substantial evidence of gender bias with open committee deliberation. In this case 60 percent of ratings received by men are revised upwards after deliberation compared to only 25 percent of ratings received by women. As a consequence women are ranked on average three positions lower after deliberation. We explore several mechanisms and test two interventions for open deliberation. Randomizing the order of speaking does not reduce gender bias, but an information intervention where raters are informed of gender bias in prior sessions does.

• Diversity in Committees (joint with N. Hughes and Z. UH Khan), SSRN working paper (2023).

Abstract:  We investigate the potential of diversity to influence committee decision-making. In our model committee members receive private information on a state of the world, deliberate and vote. Committee members belong to one of two groups which may differ along two dimensions: (a) their preferences and (b) their information structures. If groups only differ in their information structures, welfare is increasing in committee diversity. If groups only differ in preferences, welfare is decreasing in diversity. If groups differ along both dimensions then, depending on parameters, welfare may be increasing, decreasing or non-monotonic in diversity. We test the model's key predictions in a laboratory experiment. As predicted, diverse committees outperform non-diverse committees when preferences are aligned. However, when preferences are misaligned diverse and non-diverse perform equally well. The reason is that participants reveal more information than theory predicts and update imperfectly. 

• Non Bayesian Statistical Discrimination (joint with Pol Campos Mercade),  Management Science 70 (4) (2024), 2549-2567.

Abstract: Models of statistical discrimination typically assume that employers make rational inference from (education) signals. However, there is a large amount of evidence showing that most people do not update their beliefs rationally. We use a model and two experiments to show that employers who are conservative, in the sense of signal neglect, discriminate more against disadvantaged groups than Bayesian employers. We find that such irrational statistical discrimination deters high-ability workers from disadvantaged groups from pursuing education, further exacerbating initial group inequalities. Excess discrimination caused by employer conservatism is especially important when signals are very informative. Out of the overall hiring gap in our data, around 40% can be attributed to rational statistical discrimination, a further 40% is due to irrational statistical discrimination, and the remaining 20% is unexplained or potentially taste-based. 

• Match length realization and cooperation in indefinitely repeated games (joint with L. Orlandi and S. Weidenholzer), Journal of Economic Theory 200 (2022), 105416.

Abstract: Experimental studies of infinitely repeated games typically consist of several indefinitely repeated games (“matches”) played in sequence with different partners each time, whereby match length, i.e. the number of stages of each game is randomly determined. Using a large meta data set on indefinitely repeated prisoner's dilemma games (Dal Bó and Fréchette, 2018) we demonstrate that the realized length of early matches has a substantial impact on cooperation rates in subsequent matches. We estimate simple learning models displaying the “power law of practice” and show that participants do learn from match length realization. We then study three cases from the literature where realized match length has a strong impact on treatment comparisons, both in terms of the size and the direction of the treatment effect. These results have important implications for our understanding of how people learn in infinitely repeated games as well as for experimental design. 

• The impact of social identity for social learning and the dynamics of opinions (joint with V. Grimm and  X. Zhou), SSRN working paper (2025).

Abstract: We link labs in China and Germany to study how national identity affects social learning and the dynamics of opinions. Participants have to guess the answer to a series of factual questions. They form a network and then observe the guesses of their network neighbours in repeated simultaneous learning. We find that social identity induces homophily. Furthermore participants are more likely to connect with members who have information advantage. This effect is stronger if in-group members have the advantage. Social identity does not influence accuracy on average, but has a positive effect on consensus if and only if neither identity has information advantage.

Literature Surveys:

• Gender Differences in Networking - A Survey, prepared for EE Handbook on Experimental Economics of Gender (2024).
• Learning in Social Networks , Chapter 82 in Elgar Encyclopedia of Behavioural and Experimental Economics, ed S-H Chuah, R. Hoffmann and A. Neelim, Cheltenham, UK: Edward Elgar Publishing, February 2025.

Presentations:

• Gender Bias in Opinion Aggregation

• Non-Bayesian Statistical Discrimination (joint with Pol Campos Mercade)

• Diversity in Committees (joint with N. Hughes and Z. UH Khan)

• Match Length Realization and cooperation in indefinitely repeated games (joint with L. Orlandi and S. Weidenholzer)

Conference

The interdisciplinary workshop at the University of Essex on Inequality, Identity and Beliefs had to be cancelled due to Covid.