ISET Economist Blog

Are Giorgi and Mariam More Employable than Arthur and Zeynab?
Friday, 14 December, 2018

It all started with a simple exercise for my Master’s project in which I tried to understand the underlying causes of the observed wage gap between ethnic Georgians and ethnic minorities in the country. After more than a decade, a reputable international journal has published a paper reporting on the experimental evidence my colleagues and I collected and analyzed on labor market outcomes for ethnic minority and female citizens of Georgia.

Back in 2008, using the Integrated Household Survey collected by the statistics office of Georgia for 2007 and 2008, I tried to understand the causes of the significant overall wage gap between ethnic Georgians and other ethnicities as reported by the survey. However what I found was that if Georgians and other ethnic minorities work in the same industry performing the same activities on the same seniority level, then it is highly likely that both earn the same salary; therefore the observed overall wage gap must be explained by the fact that ethnic Georgians and ethnic minorities do not, in general, work in the same industry performing the same tasks; hence segregation of ethnic minorities into lower-paying jobs should have been the explanation for the observed statistics, but I still did not know why this would be the case. Was it because ethnic minorities chose career tracks that paid less or because they were not hired in higher-level jobs despite their willingness and qualifications?

While working on my Master’s thesis I came across a very interesting academic publication (also the source of inspiration for the title of this article), by two University of Chicago-based Professors, Marianne Bertrand and Sendhil Mullainathan. They provided a novel concept in understanding racial discrimination in the United States. Using a set of two almost identical professional resumes, they applied to job advertisements announced in two US cities, Chicago and Boston. The only real difference between the two CVs was the name and surname attached to them, one of them sounding very “white” and the other very “black”; however, both of the CVs were fictitious.  The most striking result of the experiment was that white-sounding CVs received 50% more invitations for interviews compared to black-sounding CVs, despite the apparent similarity of qualifications. Therefore, the argument that the observed wage gap between white and black employees was due to their professional or personal choices could not stand up to criticism anymore.

I was so inspired by their simple yet illuminating paper, I set the goal of running a similar experiment in Georgia. Fortunately, I was supported by a number of people and organizations to carry out this exercise, and so in March of 2009, the study was officially launched. The design of the experiment was replicated from the Bertrand and Mullainathan (2004) study; however, we also added a gender dimension to it, which made the research even more compelling in the Georgian reality.

As a preliminary step, we created a bank of CVs based on the resumes of real applicants by altering them slightly to rule out identifying real applicants.  After a vacancy advertisement was placed on job posting websites, we prepared a relevant set of four resumes complying with the requirements stated in the announcement. Finally, we randomly assigned ethnic Georgian and ethnic minority-sounding last names and female and male-specific names to the resumes. We limited ourselves to two groups of ethnic minorities, Azerbaijanis and Armenians, which are the biggest ethnic minority groups in Georgia and have easily-recognizable last names. We also included a real email address and mobile number, different for the two ethnic groups and genders, and sent the applications to the designated email address. In case we received callbacks on either phone number, we made sure to identify the vacancy and company we were called back for, and politely refused the offer.

When applying for the announced job openings, we created a detailed database of the vacancies we were applying for, including: type of company, industry, ownership of the company (foreign or domestic), position, and detailed requirements of the opening on the one hand and on the other hand, we also created a database of the applicants and their characteristics and qualifications including age, gender, type and years of education, experience, and computer skills. Over 12 months we replied to 552 job announcements and sent 2200 resumes.

Table 1 below presents the key findings from the experiment. Out of 1100 total, CVs sent for ethnic Georgian candidates, 13.36% of the CVs received an invitation for the interview; that is, they got around 147 callbacks from potential employers. The same indicator for ethnic minorities stands at 6.27% out of 1100 total CVs sent. The 7.09 percentage point difference is both economically and statistically significant as reported in column 4. Overall, ethnic Georgian job applicants received about 113% more callbacks than their non-ethnic-Georgian counterparts. Looking at the gender dimension of the experiment, we cannot observe any signs of gender discrimination.  The slight difference (in favor of females) in the callback rates between female and male applications is not statistically significant; however, we can see that the ethnic gap is most pronounced for male applicants. For example, to receive one callback, a non-ethnic-Georgian male applicant needed to send more than 20 job applications, whereas an ethnically Georgian male applicant needed to send fewer than eight. No significant differences between groups of ethnic minorities can be observed.

Table 1. Callback rates

  Ethnic Georgian Ethnic Minority % Difference (p-value)





















Source: Asali, M., Pignatti, N., & Skhirtladze, S. (2018). Employment discrimination in the former Soviet Union Republic: Evidence from a field experiment. Journal of Comparative Economics.

While these kinds of experiments have been helpful in identifying the existence of discrimination in the labor market, they are not as successful in diagnosing the underlying reasons for the behavior they reveal. There are two main groups of economic theories that try to explain the discriminatory actions of “Homo Economicus”. One of the oldest groups of theories is the taste-based discrimination theories, which were initiated by Nobel Laureate Gary Becker (1957), where employers, employees, and/or consumers have preferences against certain groups of people such as women, immigrants, or ethnic minorities. The economic agents prefer to incur additional costs in order to abstain from a relationship with these groups. Therefore, the dislike against some groups results in revealed discrimination by some economic agents. The key prediction from these theories is that differential in wages or employment will be abolished in the long run because of competition. Employers maximize their profits by hiring low-cost labor. Eventually, the demand for low-cost labor will increase the price, and therefore the earnings and employment of minority workers.

Statistical discrimination models (e.g. Edmund S. Phelps, 1972; Kenneth J. Arrow, 1973) are the primary alternatives to the taste-based models in the economics literature. According to these models, employers base their decision to hire on real or imagined statistical information since they do not have complete information about the skills and qualifications of applicants. In other words, employers’ decisions are affected by real or imagined/outdated stereotypes which leads to wage or employment discrimination. Statistical discrimination of this type should eventually decline because of the increased possibility over time of collecting information and distinguishing “high productivity” workers from “low productivity” ones.

Our findings can, to some extent, be explained by both groups of theories; while in the paper we make an effort to link the findings with a particular theoretical model, no credible connection can be established and therefore no underlying reasons for the observed labor market behavior could be justified.  We find that gender has no effect on the probability of a callback, but a job applicant who is ethnically Georgian is more than twice as likely to be called for a job interview as an equally skilled ethnic non-Georgian (Azerbaijani or Armenian), but more research needs to done to understand why. These findings should provide the impetus for further research and policy analysis on how to better integrate more than one-eighth of the Georgian population into economic and civil life.

The views and analysis in this article belong solely to the author(s) and do not necessarily reflect the views of the international School of Economics at TSU (ISET) or ISET Policty Institute.