Session 1_ Lecture 1_ Studying Discrimination: Discrimination and the limitations of observational data
The wage decomposition method, explained
Political and media discussions on gender and ethnic inequality often assume differences in outcomes must reflect unequal treatment (discrimination). Yet this need not be the case –and often it isn’t. For example, gender differences in labour-market outcomes might respond to multiple supply-side factors, as we reviewed last year (see Social Stratification I), while differences in labour-market outcomes between natives and migrants (and their descendants) might reflect differences in unobserved characteristics (e.g. language and other cultural skills, networks, parental background, etc). Estimates of discrimination based on observational data (i.e. surveys) typically suffer from omitted variable bias because we cannot possibly account for all potentially confounding factors (e.g. differences in unobserved preferences and traits potentially affecting people’s choices, differences in unobserved resources and skills, etc). In other words, observed gaps in outcomes should not be interpreted as necessarily -nor primarily- reflecting discrimination. Hence the title of this course, Mind the Gap, should be read as a warning. In this introductory session, I will lecture on the confounding bias problem and introduce students to the logic of field-experimental research for the analysis of discrimination (to which we will go back on Session 2). This lecture will be followed by a discussion of the role of cognitive skills in explaining racial gaps in the US, for which students are expected to read at least one of the two starred readings.