Social Capital and Women’s Labor Force Participation in Chile

ABSTRACT

Using data from Chile, this study analyzes the relationship between different forms of social capital and women’s labor force participation, accounting for both endogeneity problems and differences among women of different economic strata. First, the results suggest that only some types of social capital are relevant for labor force participation: namely, networks with weaker yet far-reaching connections, including higher-status individuals. There are neither empirical nor theoretical reasons to believe that women have better access to such networks than men. Second, this type of social capital is only relevant for the economic integration of the richest women, failing to increase labor force participation among women of the other 80 percent of households. Thus, this study concludes that policies targeted at women’s economic integration based on the presumption that women have more social capital than men are deeply flawed.

KEYWORDS:

• Social capital
• labor force participation
• labor market inequality
• Chile

JEL Codes:

• B54

ACKNOWLEDGMENTS

The authors thank Prof. Dante Contreras for the excellent feedback provided during the development of this research. This article was prepared with the generous support of the Centre for Social Conflict and Cohesion Studies (CONICYT/FONDAP/15130009). The contents of this paper are the authors’ sole responsibility.

Notes

1 A common concern regarding official numbers on economic participation and employment is the role of the informal economy, as informal jobs might be invisible to some measurements. This may be one reason that women from poor households show a lower labor force participation rate than other women, for example. Three clarifications are in order regarding the informal economy. First, our econometric analyses are based on survey data from a survey that uses detailed measures to capture unpaid and informal economic participation. Second, official numbers on labor participation in Chile are based on survey data, in this case the National Socioeconomic Characterization (CASEN) survey, which also uses measurements aimed at capturing informal and unpaid labor. Finally, while no source can claim to capture all informal economic activity, the informal economy is comparatively small in Chile (Loayza 1996; Maurizio 2016), and thus the comparative scenario shown here should not change due to this problem.

2 In this analysis, and all others in our work, household income refers to total income after taxes and subsidies.

3 We also tested alternatives using a simple sum of these ordinal variables, as well as a Principal Component Analysis factor, and the first two factors of the MCA model. The MCA solution showed the highest relevance in the models, and the second factor of this solution did not add any relevant information. The results of the MCA analysis are in Appendix Table A2.

4 The occupations included in a “position generator” vary to capture positions with varying degrees of status and types of resources in the contexts where they are applied. The ENES includes positions that are commonly used for this purpose and are relatively common in Chile: an engineer, a real estate agent, a secondary education teacher, a policeman, a civil servant, a nurse, a cook, an agricultural worker, a medic, a storehouse clerk, a janitor, and an unskilled construction worker.

5 The tie-strength factors were “1” for acquaintances, “2” for friends, and “3” for relatives. While this weighting is not common in the literature, it is theoretically sound as a way to account for the “ease of access” to different positions in the final score. We tested all our models without using these weighting factors, and all the results reported here held without meaningful changes.

6 In all regression tables, we report marginal probability effects (and their standard errors) instead of beta coefficients. Marginal probability effects are easier to understand than probit coefficients in terms of their econometric relevance, as they represent the partial effects of each explanatory variable on the probability that the dependent variable Y = 1.

7 The effect of age is not linear; it decreases as women age.

8 It is key to clarify, however, that no statistical test can prove that an instrument in particular is a good instrument. The Wald test, reported in our IV-probit models, simply allows us to test the exogeneity of the instrumented variables. The test’s null hypothesis is no endogeneity: if the results allow rejecting the null hypothesis, then it is adequate to use an instrument to deal with the problem of endogeneity. Otherwise, it might be more appropriate to use a simple probit regression.

Additional information

Notes on contributors

Ismael Puga

Ismael Puga received his doctorate in sociology at the Humboldt Universität zu Berlin in 2013. Since 2014, he has been Adjunct Researcher at the Centre for Social Conflict and Cohesion Studies (COES), and, since 2016, Assistant Professor at the School of Sociology in the Universidad Central de Chile. His research interests include social and economic inequality, ideology, social stratification, legitimation of inequality, feminist studies, and Marxist theory.

Daniela Soto

Daniela Soto studied social anthropology at the Universidad de Chile, and obtained her MA in sociology at the Pontificia Universidad Católica de Chile. She is currently working at the Center for Intercultural and Indigenous Research (CIIR). Her research interests include development studies, inequalities, and feminist studies.