Measuring Inequality of Opportunity with Latent Variables

In this paper I show that recently proposed methods to quantify the level of inequality of opportunity are likely to be downward biased when the dependent variable is a proxy for an unobserved concept. Using a multidimensional framework of development, such as the capability approach, or a standard utility maximization framework with heterogeneous preferences permits us to show that such measurement errors are the rule rather than the exception. I propose to estimate the latent variable of interest through appropriate multivariate techniques to circumvent the aforementioned bias. Using a simulation and an empirical illustration, I show that the use of multiple indicator variables and appropriate aggregation techniques can reduce the bias substantially. Using data from Mexico, it is found that inequality of opportunity for the broader concept of economic well-being is more than twice as high as inequality of opportunity in log income, which is commonly used as a proxy of the first.