Previous research and theorizing suggest ethnic discrimination in the labor market should be higher when employers face few search difficulties in the hiring process —i.e., discrimination should increase with the number of applicants competing for a given job vacancy. This literature, which draws on Gary Becker’s seminal work on discrimination, assumes employers have a hidden taste or propensity for discrimination, which they can only realize when they have enough candidates to choose from. This study challenges previous Beckerian accounts by presenting an alternative mechanism, which combines statistical discrimination theories in economics and mainstream theorizing in social psychology, on the one hand, with norm-referenced evaluation theory in educational research, on the other. This alternative argument sees discrimination as the outcome of information deficits (as in statistical discrimination models) and contends employers are rational actors that learn about the state of the world through their own experiences, including crucially their experiences as evaluators of applicants’ résumés. We argue employers learn about each applicant’s quality by comparing them with other candidates in the pool. This process of evaluation by comparison allows employers to fill at least some of the information gaps on which statistical discrimination is based. Thus, the larger the pool of candidates, the more information employers can acquire. This means that, unless employers do have indeed a hidden taste for discrimination, ethnic minorities who are qualified for a given job will have greater employment chances if they can compete in a larger pool. Hence opposite predictions on the relationship between the size of the applicants’ pool and employers’ discrimination propensity follow from these two competing hypotheses, which we call the discrimination as second order motive hypothesis and the learning-by-evaluating hypothesis. We test these competing hypotheses by exploiting the unique features of the data on ethnic discrimination collected in Spain by the D-Lab using an online correspondence test. The software developed at the D-Lab allows us to retrieve information on the number of job applicants competing for each vacancy advertised in our targeted job-search platform –a well-known high-traffic job-search website in Spain. This unique feature helps us overcome some of the limitations of the previous literature, thus providing an extraordinary opportunity to empirically test the two competing hypotheses discussed in this study. Using a Linear Probability model that interacts applicants’ ethnicity with the size of the applicant pool, we find discrimination against minority applicants decreases with the number of competitors in the pool. This finding challenges previous Beckerian accounts of the association between labor-market tightness and discrimination propensity, while being fully consistent with our alternative learning-by-evaluating mechanism. By analyzing the empirical question of the association between labor-market tightness and discrimination using unique data at the job-vacancy level, we claim this study provides a new test for the information-deficit mechanism that lies at the heart of all statistical discrimination models, while at the same time presenting curricula evaluation as a new mechanism for employers’ learning.