The development of methods for detecting labour market discrimination is characterized by a shift from observational data to different forms of experimental data. At the same time, there has been a shift from examining differences in treatment of groups of employees to differences in hiring.
Observational data are associated with omitted variable problems. Field experiments in the form of audit and correspondence studies give better control of what is observed by both employers and the researcher. A limitation is that they can typically be employed only for certain types of (low-skill, early career) jobs and the initial (call-back) stage of the hiring process. Another is that employers’ beliefs cannot be controlled for. Natural and lab experiments can address some of these concerns. Most studies are not able to distinguish theories of between taste-based and statistical discrimination.
The book chapter can be read here.