In the last 20 years, multilevel modeling (MLM, Bryk & Raudenbush, 1992) has widely spread as a method of data analysis in the social sciences, notably in education studies, sociology, psychology, and management. However, economic research using MLM has been rather scarce. Besides, popular textbooks in econometrics, aimed at the students of economics, have mostly ignored MLM. In this paper we seek to explain this situation. We propose and critically investigate two main hypotheses. First, since MLM as a method originated not in economics, economists might have been taking their time to integrate it. This could partly be due to the fact that the research questions MLM is usually used for are of minor interest to economists. Second, from their perspective, economists might be seeing substantial flaws in MLM. Our general preliminary conclusion is that both explanations are relevant. Above and beyond the extant mismatch in the research questions, even in similar cases where sociologists or social psychologists would be using MLM by default, economists would prefer not to do it. In our view, the main feature of MLM that scares off economists is the underlying assumption that unobserved heterogeneity is random (the random effects model), while in economics the ‘gold standard’ has been the fixed effects model. We emphasize, however, that in light of the ongoing discussion on comparative advantages and disadvantages of these types of models, this MLM feature should not be treated as a crucial issue. Given the higher flexibility of MLM as compared with the OLS regression in the analysis of various social effects, we think that MLM has been unduly neglected in economics.

Working language is English.

Guests from St. Petersburg are invited to make a request for a pass to the building to Violetta Korsunova (vikorsunova@hse.ru) by 12 am of the seminar’s day. Guests from Moscow are invited to make a request for a pass to the building to Ilya Lomakin (ilomakin@hse.ru) by 12 am of the seminar’s day.