Path analysis for discrete variables: the role of education in social mobility

• Published in: Journal of the Royal Statistical Society. Series A, Statistics in society Vol. 173; no. 2; pp. 351 - 369
• Main Authors: Kuha, Jouni, Goldthorpe, John H.
• Format: Journal Article
• Language: English
• Published: Oxford, UK Oxford, UK : Blackwell Publishing Ltd 01.04.2010; Blackwell Publishing Ltd; Blackwell Publishing; Blackwell; Royal Statistical Society; Oxford University Press
• Series: Journal of the Royal Statistical Society Series A
• Subjects: Applications; Biology, psychology, social sciences; Categorical data; Coefficients; Direct effects; Educational attainment; Error; Estimation; Estimation methods; Exact sciences and technology; General topics; Impact analysis; Indirect effects; Inference; Intergenerational relations; Level of education; Linear models; Logistics; Mathematics; Measurement; Men; Modeling; Modelling; Multinomial logistic models; Multivariate analysis; Path analysis; Probability and statistics; Sciences and techniques of general use; Social class; Social mobility; socioeconomic status; Standard error; Statistical methods; Statistics; Survey data; United Kingdom; Upward mobility; United Kingdom > UK; Path analysis; Data analysis; Error estimation; Social sciences; Indirect effects; Multinomial logistic models; Decomposition method; Statistical association; Decomposition; Logistic model; Social class; Multivariate analysis; Mean estimation; Sociometry; Logistic distribution; Statistical method; Association measure; Categorical data; Model matching; Direct effects; Multinomial distribution; Application
• ISSN: 0964-1998; 1467-985X
• DOI: 10.1111/j.1467-985X.2009.00620.x

An important open question in sociology with obvious policy implications is how to assess the magnitude of the effect of educational attainment on intergenerational social mobility. To examine this, we propose a general method of path analysis, which can be used to estimate direct and indirect effects even in systems where some of the variables are categorical. It provides an additive decomposition of total effects which is exact when the effects are expressed as mean differences, and approximate but typically quite accurate for other measures of association such as log-odds-ratios. Estimates of the effects and their standard errors can be calculated by using standard output for fitted models. The method is illustrated by an analysis of British survey data on social mobility.