Civil rights experiments versus enrichment experiments in wage gap analysis

The choice of the non-discriminatory vector of returns in Oaxaca-Blinder wage gap decompositions affects the results. Rather than being arbitrary, that choice should depend on the nature of the intended policy to address wage differentials in the labour market. The effectiveness of policies at the extensive margin, such as those offering greater access to higher education, is better estimated by the explained part of the wage gap, when choosing the lower-wage group’s vector of returns as non-discriminatory (the ‘enrichment experiment’). Alternatively, the effectiveness of affirmative action policies is better estimated by the unexplained part of the wage gap, when choosing the higher-wage group’s vector of returns as non-discriminatory (the ‘civil rights experiment’). We provide an example of applying this methodology for ethnic and gender wage differentials in the Georgian labour market.