![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
Using data from the extended section of the 2010 Mexican census (2.9 million households), we study how school enrolment is associated with wealth inequality and with the educational environment the child is exposed to at the household and municipal levels. We provide robust evidence of wealth inequality as a negative predictor of school enrolment for children in primary, secondary and high school age ranges while a positive role is played by the educational environment. Through the introduction of interaction terms, we account for how economic and educational variables are intertwined at both the household and the municipal level, and we are able to illustrate the considerable heterogeneity in the role of adult education for households at different standards of living.
Acknowledgement
We thank for helpful comments and suggestions Ed Anderson, Luna Bellani, Paul Clist, Ben D’Exelle, Sarah Tustin, two Referees of this Journal, and the participants of the VI Meeting of the Society for the Study of Economic Inequality, the XXV Coloquio Mexicano de Economía Matemática y Econometría and a seminar held at The Institute Of Education (IOE-UCL). We would also like to thank the Mexican Science and Technology Council (CONACYT) for financial assistance.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. See, for example, Knack and Keefer (Citation1997), Takata (Citation2003), Uslaner (Citation2005), Elgar and Aitken (Citation2011), De Vries, Gosling, and Potter (Citation2011), Loughnan et al. (Citation2011), Neville (Citation2012), Piff, Stancato, Côté, Mendoza-Denton, and Keltner (Citation2012), Piff et al. Trautmann, Van De Kuilen, and Zeckhauser (Citation2013) and Piff (Citation2014).
2. See Peugh (Citation2010), Reiter and Raghunathan (Citation2007), and Gutiérrez, Carter, and Drukker (Citation2001) for technical discussions of the log-likelihood ratio test and Valentine, Verdes-Tennant, and Bonsel (Citation2015) and Vu, Le, and Muhajarine (Citation2013) for empirical applications. Results from our log-likelihood ratio tests (Chi2 statistics and the p-values) can be found at the end of our regression tables.
3. Standard Principal Components Analysis (PCA) assumes that the variables are multivariate normal. Following Kolenikov and Angeles (Citation2009), we run PCA using polychoric correlations to better approximate the normality assumption and estimate the amount of variation explained by the first component. Finally, it should be noted that financial assets are not included in our measure of wealth because unavailable in the survey – we do not expect this to have created a relevant bias in our use of wealth as a predictor of school enrolment.
4. The migration index used in the analysis is calculated by the Mexican National Population Council on the basis of the percentage of households that receive remittances, percentage of households with members in the United States, percentage of households with visiting members who live in the United States and percentage of households with returning members who lived in the United States between 2005 and 2010.
5. We remark that this interpretation of mean asset index relates to these specific models – as extensively explained by the body of literature quoted above. The use of mean asset index as explanatory variable in an OLS model where the dependent variable is municipal enrolment rates reveals (as expected) a positive and significant coefficient – results available upon request.
6. The introduction of an interaction term in a logit model allows for heterogeneity in the shape (rather than only in the position) of the curve representing the conditional probability that the dependent variable equals one as a function of the explanatory variable of interest; in other words, it allows this shape to differ at different levels of the interacted variable. This means that if a continuous variable is interacted with a dummy variable, we will have two possible shapes for this curve – one for each value of the dummy variable; if two continuous variables are interacted then we would have many (virtually infinite) shapes.
7. The analysis using age dummies is available upon request.
8. For this analysis, continuous explanatory variables are set to mean values and dummies are set to zero (therefore this child is male, non indigenous, does not have any disability and so forth.)