Academic journal article The American Journal of Economics and Sociology

Women's Employment: Joining Explanations Based on Individual Characteristics and on Contextual Factors

Academic journal article The American Journal of Economics and Sociology

Women's Employment: Joining Explanations Based on Individual Characteristics and on Contextual Factors

Article excerpt

Introduction

The dramatic increase in women's activity and employment rates explains a large part of the secular increase in labor supply that took place during the second half of the 20th century in many industrial countries. It constitutes one of the most interesting facts of contemporary economic history and indeed it stimulated an extensive debate. This article takes issue with the mainstream economics explanations of this phenomenon. The mainstream economic theory traditionally approaches the issue of gender differentials in the labor market by means of microeconomic (often game-theoretic) models of the household, ultimately based on gender differences in preferences and/or endowments (sometimes attributed to gender-based discrimination). At the same time, the applied mainstream literature has extensively studied other relevant factors such as the role of education, social policy, macroeconomic conditions, and culture. This literature offers interesting insights on the determinants of women's employment but it suffers from a main limitation; it mostly developed into two related but separate strands. One of these strands deals with individuals' characteristics and how they affect women's labor supply, by usually employing micro-data surveys. The other strand investigates the relevance of the institutional context and public policy (or "macro" factors), by usually analyzing cross-sections or panels of aggregate data. (1)

While each strand of literature recognizes it is linked to the other, they usually ignore or assume as exogenously given the relations, variables, and theories of the other strand. This separation is problematic from an applied point of view because it marks a possibly untenable ceteris paribus hypothesis. This article argues that the ignorance of possible interactions between micro and macro factors may produce sensible biases in the applied estimates, for two reasons. (2) On the one hand, if the impact of individual characteristics depends on the environment, the individual estimates of any regression that do not control for these interactions could at best be considered as valid only on average and for the specific setting in which they are computed. On the other hand, the macro-level variables may not only exert a direct effect on women's employment, but also an indirect one, by modifying the impact of micro variables. The resulting estimates of a regression that ignores these interactions may thus suffer from a missing variable bias and lead to biased estimates of the impact of the micro variables.

In the few papers that explicitly deal with this issue, the problem is usually treated by means of a difference-in-difference approach, that is, by employing dummy-variable interaction terms between micro and macro variables. However, this method is not feasible if the environment affects individual values by means of several factors simultaneously, again for two reasons. First, observations may be clustered by region; therefore, the error term of the regression is not independently and identically distributed. Second, accounting for the simultaneous direct and indirect impact of many environmental factors may require an excessive number of dummy variables. This article proposes the use of multilevel analysis as a more direct and rigorous way to simultaneously estimate the impact of micro and macro factors within a unified regression model. This method allows one to consider a wider set of determinants of women's employment than is possible with a standard regression analysis, for example, the simultaneous role of education and social policy. However, due to utmost computational requirements, the use of multilevel analysis suggests the prior application of a multivariate analysis to summarize a great number of macro variables into a small set of synthetic macro indices.

This article aims to contribute to the literature on understanding women's work in two ways. On the one hand, we criticize the standard applied mainstream literature by showing the existence of significant interactions between the micro and macro levels. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.