In this doctoral dissertation, my co-authors and I offer a comprehensive exploration of labour market discrimination, integrating different perspectives, methods and contexts to measure and explain the complex dynamics of the phenomenon. In primary order, the thesis examines the extent of labour market discrimination and its underlying mechanisms. In secondary order, we evaluate some possible strategies to combat discrimination.

First, we present the results of a global meta-analysis of recent correspondence experiments measuring discrimination in hiring. Worldwide, we find significant and persistent discrimination against job applicants based on race, ethnicity, national origin, employment disability, age, appearance, religion, wealth and marital status. The meta-study reveals apparent differences in discrimination across discrimination grounds, minority groups and regions. While ethnic minority candidates are strongly disadvantaged in the application process, this disadvantage is even more significant for persons with disabilities, unusual appearance, or older people. Notably, on average, the European member states tested have stronger age selection bias than the United States.

Second, we compile and evaluate the empirical evidence for the economic theories, taste-based discrimination and statistical discrimination, which attempt to explain ethnic discrimination in the labour market. Although neither theory can fully explain ethnic labour market discrimination, we find that ethnic discrimination in hiring appears to be primarily motivated by personal preference. However, we identify inconsistencies in the methods used to distinguish the mechanisms of discrimination, which behoves us to be categorical about the true mechanism behind ethnic labour market discrimination. Incidentally, we believe it pays to integrate psychological and sociological perspectives with economic perspectives to obtain a more nuanced understanding of labour market discrimination. For example, social psychology can explain how certain groups oppose each other, and sociology can explain how organizational characteristics, such as company policies, may perpetuate discrimination.

We then switch from a global to a more local focus and examine hiring discrimination in Flanders through a contextual analysis of the results of a correspondence experiment. The experiment’s findings reveal significant discrimination against candidates with non-Flemish-sounding names. Furthermore, we find that less discrimination occurs in non-profit organisations and large organizations.

We conclude with findings from two lab experiments. In the first experiment, we examine differences in the effects of rewarding non-discrimination versus punishing discrimination, primarily motivated by personal preference. In line with the behavioural economic principles of "loss aversion"—i.e. losing a sum of money has a higher (negative) perceived value than the (positive) perceived value of winning an equivalent sum of money—we find that sanctions manage to reduce discriminatory behavioural intentions to a greater extent. In the second experiment, I investigate whether technological applications based on large language models, such as ChatGPT, can banish human biases from the selection process. However, using a simulated resume screening task, I find systematic ethnic biases in ChatGPT's candidate judgments similar to those of human recruiters in correspondence research.

The research in this dissertation leads me to the following policy recommendations. First, I argue for a broad interpretation of diversity in labour market policy that goes beyond race and ethnicity to encompass all marginalized groups, particularly those most disadvantaged in the selection process. Second, I argue for more consistent enforcement of anti-discrimination laws, supported by empirical evidence that punishing discriminatory behaviour motivated by personal preferences and prejudices can reduce them. Third, I emphasize the importance of creating a guiding organizational and legal framework around applications of artificial intelligence-based technologies that can be used in selection processes involving humans. Such technological developments should be accompanied by ethical innovation.

Toward future research, I would like to underscore two more elements. First, there is a need to move from merely quantifying discrimination to (even) better understanding how discrimination comes about and identifying effective solutions. Second, sometimes it pays to look beyond disciplinary boundaries; psychological and sociological theories can help us explain the discrimination phenomenon, too. However, within economics, there are numerous opportunities for further research into the mechanisms of and solutions to discrimination. An excellent example of the latter is that competition should drive discrimination out of the market; discrimination would be low when the balance of power between firms (within a given sector) is proportionate or when job openings are difficult to fill. Future research should consequently shed more light on the validity of this proposition.