Correspondence Tests
Correspondence Tests
ALTERNATIVES TO CORRESPONDENCE TESTS
Correspondence testing is a field-experimental technique used to investigate discrimination in hiring in the labor market.
THE EXPERIMENTAL PROCEDURE
Two carefully matched, fictitious applications are forwarded in response to advertised job vacancies. To avoid detection the applications cannot be identical, but the logic of the technique is to control strictly for all objective factors, such as education, qualifications, and experience, that influence job performance. Consequently, the only distinguishing feature of the two applications is the characteristic—such as race or sex—that is being tested for, with race or sex being identified by the name of the applicant. To safeguard against the possibility that letter style may influence employer response, the letters are regularly reversed and allocated equally between the candidates. In this way, the influence of race, or sex, on selection for interview is isolated.
HISTORY
The social scientists who innovated this technique were Roger Jowell and Patricia Prescott-Clarke. In a test of racial discrimination in England in 1969 they matched applicants with Anglo-Saxon names (John/Mary Robinson) against those with “West Indian” (Errol/Eva Gardiner) and Indian names (Santokh Singh and Ranwi Kaur). The test was at a time when most adult Indians and West Indians in Britain were first-generation immigrants, so that their race was also identified by their primary schooling having occurred in their country of origin, and by specification of their date of arrival in Britain. Applicants with an Anglo-Saxon name were found to be twice as likely to receive an invitation to interview as applicants with an Indian name. No significant discrimination was detected against applicants who were immigrants from the West Indies.
The first economists to use this technique to test for sexual discrimination were Peter Riach and Judith Rich, who applied the test to seven occupations in Melbourne in the 1980s. They found that men were invited to interview 13 percent more frequently than women in the occupation of computer analyst programmer, and 23 percent more frequently than women in the occupation of gardener. No significant discrimination was detected in the other five occupations. Riach and Rich also tested for racial discrimination in Melbourne in the 1980s, matching an Anglo-Celtic name (Baker) against a Greek name (Papadopoulos) and a Vietnamese name (Nguyen). They found that the Anglo-Celtic applicant was invited to interview 10 percent more frequently than the Greek applicant and 41 percent more frequently than the Vietnamese applicant.
Frank Bovenkerk investigated discrimination against Antilleans in France in the 1970s and found that an applicant with a French name was more than three times as likely to be invited to interview as an applicant with an Antillean name. In the 1990s Bovenkerk, in the Netherlands, found that applicants with Dutch names were 23 percent more likely to be invited to interview than those with Surinamese names and 17 percent more likely to be interviewed than those with Moroccan names.
Doris Weichselbaumer, in Austria, matched a man against a “feminine woman” and a “masculine woman,” in two sex-integrated occupations, one female-dominated and one male-dominated occupation. She was able to do this because job applications in Austria include the applicant’s photograph. She found that both types of women encountered some discrimination in the male-dominated occupation, that there was no significant difference in the treatment of the two female types, and that both types of women were more than twice as likely as men to be invited to interview in the female-dominated occupation of secretary.
E. Fry in 1986, and Pauline Graham, Antoinette Jordan, and Brian Lamb in a 1992 follow-up, tested discrimination against applicants with the disability of cerebral palsy, which confined them to a wheelchair. The occupation tested was secretary, and in both experiments the able-bodied applicant was approximately 60 percent more likely to receive an invitation to interview.
Full details of the above studies can be found in the critical survey article by Riach and Rich (2002).
TWENTY-FIRST CENTURY ACTIVITY
Marianne Bertrand and Sendhil Mullainathan (2004) made the first attempt to apply the technique to racial discrimination in the United States. They matched distinctive “African American” names, such as Aisha (female applicants) and Kareem (male applicants) against “white” names, such as Anne and Matthew. They tested four occupations in Boston and Chicago, and found that the white applicant was called back 60 percent more frequently than the African American applicant. Fictitious mailing addresses were used; consequently, any postal responses could not be recorded. Instead of sending carefully matched pairs of applications, which differed only in name, as in the other cited tests, their resumes were randomly generated from a bank of information and sometimes manipulated to match the advertised vacancy. A further problem has been identified by Roland Fryer and Steven Levitt (2004), who have challenged the control provided by the use of distinctive black names. They have demonstrated that since the 1970s distinctive black names have been highly related to lower socioeconomic background. This means that Bertrand and Mullainathan were recording the experience of a subset of African Americans, and employers could well have been responding to socio-economic status rather than race. Their study is not a basis for generalizing about the labor market experience of African Americans; nevertheless, they did detect discriminatory hiring behavior. In view of the control established by the randomly generated resumes, employers are demonstrating either a statistical or animus-based discrimination purely in response to applicants’ first names. An experiment identifying socioeconomic class by address or schooling would be one way of determining its impact relative to racial discrimination.
The correspondence technique is seriously challenged by racial groups whose mainstream members do not have names that clearly distinguish them from other racial groups. The name Patel identifies an Indian, Nguyen identifies a Vietnamese, and Papadopoulos identifies a Greek. They are all very common names, are not associated with any subset of those races and they are readily recognized by the inhabitants of the countries where they are present in substantial numbers. This is not the case with African Americans and Afro-Caribbeans, whose ancestors lost their original names when they were forcibly removed from Africa in the eighteenth century.
Riach and Rich (2006a) conducted the first experiment examining sexual discrimination in England. They found that in the male-dominated occupation of engineer, men had a 40 percent higher likelihood of invitation to interview; that in the female-dominated occupation of secretary, women were twice as likely to be invited to interview; and, without precedent, that women were preferred to men in two sex-integrated occupations: They were 67 percent more likely to be interviewed as computer analyst programmers and twice as likely to be interviewed as trainee chartered accountants. Riach and Rich also conducted the first correspondence test of age discrimination, matching two English women who had graduated simultaneously, but were different ages—one twenty-one, the other thirty-nine. The younger graduate was two-and-a-half times more likely to be interviewed than the woman of thirty-nine (Riach and Rich, 2006b).
ALTERNATIVES TO CORRESPONDENCE TESTS
One alternative approach is to train members of different racial groups to attend interviews and present themselves as having the same skills and level of motivation. British social scientists first used this approach in the 1960s and hired professional actors for their skill in role-playing. The Urban Institute in Washington subsequently adopted this approach, but eschewed the use of actors. A principal advantage of this personal attendance at interview is that it solves the problem of groups that are difficult to identify by name, such as Afro-Caribbeans and African Americans. A second advantage of this approach is that it tests discrimination at the point of job offer, rather than selection for interview. The disadvantage of this personal approach is that you do not have, and cannot demonstrate, the same strict control as exists with correspondence testing. Regardless of the level of training, it is always possible that candidates will present themselves in ways that highlight differences in personality and levels of motivation. This point was made by Robin Ward in the 1960s and again by James Heckman in the 1990s (full details are to be found in Riach and Rich 2002). Heckman has recommended that the candidates should be kept ignorant of the fact that they are operating in pairs and testing for employment discrimination. Ian Ayres did exactly this in a test of racial and sexual discrimination in car sales negotiations, but he also conducted equivalent tests in which the testers were aware that they were testing discrimination. He recorded similar results for both sets of tests (see Riach and Rich 2004 for full details).
Surveying employers about employment practices involves the possibility that there will be a discrepancy between what they report and what they actually practice; whereas correspondence testing unequivocally captures practice.
Inferring discrimination from racial or sexual wage differentials that are deemed inconsistent with productivity-determining variables, such as education and experience, begs the question of what the appropriate independent variables should be.
SEE ALSO Discrimination; Racism
BIBLIOGRAPHY
Bertrand, Marianne, and Sendhil Mullainathan. 2004. Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination. American Economic Review 94 (4): 991–1013.
Fryer, Roland G., and Steven D. Levitt. 2004. The Causes and Consequences of Distinctively Black Names. Quarterly Journal of Economics 119 (3): 767–805.
Riach, P. A., and J. Rich. 2002. Field Experiments of Discrimination in the Market Place. Economic Journal 112 (483): F480–F518.
Riach, Peter A., and Judith Rich. 2004. Deceptive Field Experiments of Discrimination: Are They Ethical? KYKLOS 57 (3): 457–470.
Riach, Peter A., and Judith Rich. 2006a. An Experimental Investigation of Sexual Discrimination in Hiring in the English Labor Market. Advances in Economic Analysis and Policy 6 (2). http://www.bepress.com/bejeap/advances/vol6/iss2/art1.
Riach, Peter A., and Judith Rich. 2006b. How Age Discrimination Works. http://news.bbc.co.uk/2/hi/programmes/panorama/4879938.stm.
Peter A. Riach
Judith Rich