Brendan Halpin, University of  Limerick
Presents:  "Partner choice and parameter estimates: modelling the effect of preferences"

To get a Zoom-invitation, contact the seminar organisers: SWS Seminar



How can we model complex social processes, such as assortative mating, on the basis of simple outcome data, such as tables of existing partnerships? Log-linear analysis of tables of marriages is used extensively to draw inferences about the underlying processes, such as the presence of a female hypergamy preference. We use simulations capturing a range of ideal-typical partner search processes, involving different forms of preferences, and of structural constraints, to understand how well, or otherwise, log-linear models can extract evidence of the underlying preferences. The conclusion is qualified: loglinear models pick up the underlying preferences, but imperfectly. The features of the dynamic process are not wholly recoverable from the static endpoint data. The dynamic operation is better captured with longitudinal data, if available. If not, the problems can be mitigated by restricting analysis to data on recent marriages only.