Methodological Developments in Modelling Family Dyads Focusing on Gender Attitudes, Women's Control over Decisions, and Autonomy CANCELLED

Thursday, July 17, 2014: 6:00 PM
Room: 303
Wendy OLSEN , University of Manchester, Manchester, United Kingdom
I compare gender in India and Bangladesh, raising issues of method and interpretation. There is an enduring problem of violence against women in spite of India’s growth spurt up to 2007 and Bangladesh’s rapid export growth in recent decades.  The roots of violence against women can be located in cultural expectations associated with gender. This paper uses national survey data from an interdisciplinary socio-economic standpoint.

I compare alternative models that measure couples’ attitudes about women’s and men’s roles, and employment.  The latest DHS surveys for India and Bangladesh allow us to marry up women with their husbands. The methodological lessons are useful for future mother/child research and child labour research. Dyadic panel data could be used, but there is a shortage of panel data for India and Bangladesh on a national scale.  The growth curve models could be adapted for dyads, and this paper highlights two methodological issues that would arise. Here,  dyadic cross-sectional regression allows cross effects and offers innovative findings.

The paper presents findings for women’s and men’s attitudes across wealth categories, social class and educational levels.  The dyadic regression literature (Kenny, Kashy, and Cook, 2006) suggests an actor-partner interdependence model if there is gender asymmetry. A fresh, detailed review of literature underpins my interpretation.

A related  task is to interpret regression slopes given the overlap of coverage of important variables.  Statisticians choose models which are either parsimonious or complex, and then worry about bias via endogeneity.  I sidestep this perennial dilemma through sociological reasoning that allows for context to work through variables as well as being represented by variables.  The logic of considering variables as causes is questioned. Instead, variables are traces of the real underlying structures and institutions. The structural equation modelling methods demonstrated here enable qualitative and survey data to be integrated.