Statistical Discrimination and the Implication of Employer-Employee Racial Matches


In this paper, I test the empirical validity of a statistical discrimination model that incorporates employer’s race. I argue that if an employer statistically discriminates less against an employee that shares the same race (matched) than an employee who does not share the same race (mismatched), then the correlation between the employee’s wage and his skill level (AFQT) is stronger for a matched employee than for a mismatched employee. Using data from the NLSY97, which includes information about the racial background of employees and their supervisors, I find evidence that is consistent with a statistical discrimination model for young male employees.

This is a preview of subscription content, access via your institution.


  1. 1.

    See for example the seminal papers of Phelps (1972), and Arrow (1973).

  2. 2.

    Evidence from various disciplines documents difficulties in communications among members of different racial groups. Lang (1986) summarizes a vast literature documenting evidence that miscommunication is more common for members of a different group (race, gender) than for members of the same group. Similarly, Hecht et al. (2003) examine a number of studies on racial differences in communication. These studies find that cultural differences, such as unique linguistic, rhetorical, and relational styles, contribute to lack of communication (reading 3.1 pages 105-107). In the medical literature, Cooper-Patrick et al. (1999) analyze physician-patient interactions. They find that white and black patients who saw physicians of the same race rated their physician’s decision making style as more participatory and were more satisfied than were patients who saw physicians of a different race. In essence, physicians and patients communicate better if both parties share the same race. Finally, Calvo-Armengol and Jackson (2004) and Calvó-Armengol and Jackson (2007) summarize evidence in the social networking literature that workers find jobs through social networks. These networks are often racially stratified (McPherson et al. 2001). If blacks are more likely to be connected through a black social network, then a black social contact is more likely to refer a black worker to a black employer. This referral can include additional information about the worker’s skill.

  3. 3.

    One might be able to generate the same prediction using a model different than statistical discrimination. “Empirical Evidence” provides more intuition and evidence for why statistical discrimination is the theory that fits best the empirical results.

  4. 4.

    In contrast, the earlier NLSY79 does not contain information about the supervisor.

  5. 5.

    About 80 % of the individuals in the sample took the Armed Services Vocational Aptitude Battery (ASVAB). The ASVAB is a set of 10 tests of which four are used to calculate the AFQT: Arithmetic Reasoning, Word Knowledge, Paragraph Comprehension and Mathematics Knowledge. I first sum these four test scores and then standardize the row scores to the responder’s age at the test in cohorts of three months as in Neal and Johnson (1996).

  6. 6.

    See Altonji and Pierret (2001) and Schonberg (2007)

  7. 7.

    When I add Tenure interacted with AFQT, the qualitative results do not change.

  8. 8.

    NLSY97 contains information about employers’ numbers (name) for every week in a year. Therefore, I matched the weekly survey to the annual survey and include the employers who are reported in the week of the interview.


  1. Altonji JG, Pierret CR (2001) Employer learning and statistical discrimination. Q J Econ 116(1):313–350

    Article  Google Scholar 

  2. Arcidiacono P, Bayer P, Hizmo A (2010) Beyond signaling and human capital: education and the revelation of ability. Am Econ J Appl Econ 2(4):76–104

    Article  Google Scholar 

  3. Arrow K (1973) The theory of discrimination. Discrimination in Labor Markets 3(10)

  4. Calvo-Armengol A, Jackson MO (2004) The effects of social networks on employment and inequality. Am Econ Rev 94(3):426–454

    Article  Google Scholar 

  5. Calvó-Armengol A, Jackson MO (2007) Networks in labor markets: wage and employment dynamics and inequality. J Econ Theory 132(1):27–46

    Article  Google Scholar 

  6. Cooper-Patrick L, Gallo JJ, Gonzales JJ, Vu HT, Powe NR, Nelson C, Ford DE (1999) Race, gender, and partnership in the patient-physician relationship. JAMA: J Am Med Assoc 282(6):583–589

    Article  Google Scholar 

  7. Cornell B, Welch I (1996) Culture, information, and screening discrimination. J Polit Econ 104(3):542–571

    Article  Google Scholar 

  8. Dickinson DL, Oaxaca RL (2009) Statistical discrimination in labor markets: an experimental analysis. South Econ J 76(1):16–31

    Article  Google Scholar 

  9. Dustmann C, Pereira SC (2008) Wage growth and job mobility in the united kingdom and germany. Ind Labor Relat Rev 61(3):374–393

    Google Scholar 

  10. Farber HS, Gibbons R (1996) Learning and wage dynamics. Q J Econ 111 (4):1007–1047

    Article  Google Scholar 

  11. Giuliano L, Levine DI, Leonard J (2009) Manager race and the race of new hires. J Labor Econ 27(4):589–631

    Article  Google Scholar 

  12. Hecht ML, Jackson RL, Ribeau SA (2003) African American communication: exploring identity and culture. Routledge, Evanston

    Google Scholar 

  13. Heckman JJ, Rubinstein Y (2001) The importance of noncognitive skills: lessons from the ged testing program. Am Econ Rev 91(2):145–149

    Article  Google Scholar 

  14. Kahn LB (2013) Asymmetric information between employers. Am Econ J Appl Econ 5(4):165–205

    Article  Google Scholar 

  15. Lang K (1986) A language theory of discrimination. Q J Econ 101(2):363–382

    Article  Google Scholar 

  16. McPherson M, Smith-Lovin L, Cook JM (2001) Birds of a feather: homophily in social networks. Annu Rev Sociol 27:415–444

    Article  Google Scholar 

  17. Neal DA, Johnson WR (1996) The role of premarket factors in black-white wage differences. J Polit Econ 104(5):869–895

    Article  Google Scholar 

  18. Oettinger GS (1996) Statistical discrimination and the early career evolution of the black-white wage gap. J Labor Econ 14(1):52–78

    Article  Google Scholar 

  19. Phelps ES (1972) The statistical theory of racism and sexism. Am Econ Rev 62(4):659–661

    Google Scholar 

  20. Pinkston JC (2005) Test of screening discrimination with employer learning. A Indus & Lab Rel Rev 59:267

    Google Scholar 

  21. Pinkston JC (2006) A test of screening discrimination with employer learning. Ind Labor Relat Rev 59(2):267–284

    Google Scholar 

  22. Schonberg U (2007) Testing for asymmetric employer learning. J Labor Econ 25(4):651–691

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Yariv D. Fadlon.

Additional information

All views expressed in this paper are those of the author and do not necessarily reflect the views or policies of the U.S. Bureau of Labor Statistics.

Appendix: Data Collection

Appendix: Data Collection

NLSY97 consists of 4,599 male responders over twelve years (1997–2008). To be included in the sample used in this study, responders must be either white or black (1,144 fewer out of which 36 are biracial that are not Hispanic). I include responders who completed the ASVAB tests, which are used to calculate the AFQT score (687 fewer). In addition, the sample includes only responders who are not enrolled in school (289 fewer), who made the transition to the labor market and who worked at least 30 hours per week (482 fewer). I lose many responders in the last criterion because about sixteen percent of the responders that were interviewed in year 1997 were not interviewed in 2008. Therefore, many of the responders were dropped from the survey. Using these criteria, I am left with a longitudinal sample of 1,997 responders over 12 years.

I consider the main job that the responders reported in the week of the interview.Footnote 8 To avoid bias due to outliers, I do not include observations with hourly wages smaller than $2 or greater than $100.

Since the sample includes young employees, it is expected that most employees (88 %) are employed in blue collar occupations: service, sales, construction, and transportation. But the distribution of occupations across matched and mismatched employees are very similar. This can be shown in Fig. 1. Specifically, about 14.77 percent of the matched employees are employed in white-collar occupation and about 12.55 percent of mismatched employees are in white-collar occupations (Fig. 2).

Fig. 1

Histogram of occupations by matched

Fig. 2

Histogram of number of Employees

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Fadlon, Y.D. Statistical Discrimination and the Implication of Employer-Employee Racial Matches. J Labor Res 36, 232–248 (2015).

Download citation


  • Statistical discrimination
  • Employer-Employee data
  • NLSY97
  • Wage differentials J31