Test design 15 2 Test design 2.1 A correspondence test Of the approaches described in the previous section,the one settled upon for this study was that of a correspondence test.While its focus is more limited than tests involving actors,in the sense that it can only deal with the early part of the recruitment process,it avoids the problems with those approaches.Tests involving actors are particularly difficult to implement successfully due to the requirement to match candidates across all characteristics relevant to an employer (with the exception of race).It is questionable whether this can be achieved and the objection has been raised that it cannot be known whether there are motivational differences between actors of different ethnic backgrounds who are aware of the nature of the test.By contrast,the feature of the random assignment of ethnicity to applications allows us to be confident that the test is valid. The key elements of the approach were as follows: formally advertised job vacancies were identified across a particular set of occupations and areas over a specified period; a matched set of three plausible applications was developed in response to these adverts; ethnic identity was randomly assigned to each application using names widely associated with the required ethnic group(one of the three was white,with the other two from different minority ethnic groups); responses from employers were monitored; discrimination was measured as differential treatment at an aggregate level between the ethnic groups in the study(the fact that applications were sent for the same vacancies provides the control).15 2 Test design 2.1 A correspondence test Of the approaches described in the previous section, the one settled upon for this study was that of a correspondence test. While its focus is more limited than tests involving actors, in the sense that it can only deal with the early part of the recruitment process, it avoids the problems with those approaches. Tests involving actors are particularly difficult to implement successfully due to the requirement to match candidates across all characteristics relevant to an employer (with the exception of race). It is questionable whether this can be achieved and the objection has been raised that it cannot be known whether there are motivational differences between actors of different ethnic backgrounds who are aware of the nature of the test. By contrast, the feature of the random assignment of ethnicity to applications allows us to be confident that the test is valid. The key elements of the approach were as follows: • formally advertised job vacancies were identified across a particular set of occupations and areas over a specified period; • a matched set of three plausible applications was developed in response to these adverts; • ethnic identity was randomly assigned to each application using names widely associated with the required ethnic group (one of the three was white, with the other two from different minority ethnic groups); • responses from employers were monitored; • discrimination was measured as differential treatment at an aggregate level between the ethnic groups in the study (the fact that applications were sent for the same vacancies provides the control). Test design