Gender Pay Gap Elements and Locating or Avoiding Missing Elements
|Starts:||16:00 28 Mar 2017|
|Ends:||17:15 28 Mar 2017|
|What is it:||Seminar|
|Organiser:||Cathie Marsh Institute for Social Research|
|Who is it for:||University staff, Adults, Alumni, Current University students|
|Speaker:||Professor Wendy Olsen|
Research on gender pay gaps includes reports of the factors contributing to pay gaps via a decomposition equation which is usually based on cross-sectional data and a linear regression model. In this paper we widen this to allow for change over time but, more importantly, we notice that some inverse relationships can be seen in terms of offsetting positive and negative factors, which when brought together may make the outcome appear to be undifferentiated by gender. More generally, the results of our explorations showed us how to discern underlying causal mechanisms both across discrete groups and over time, especially where these are positive and negative in sign. We remark that the world represented by our data on gender pay gaps seems to be more of a dialectical world (one with countervailing pressures) than a deterministic one (where one might imagine every cause to have its impacts visible in the outcomes). We discuss these findings and a bootstrap extension of our ‘discernment’ argument.
Lastly we consider statistical issues and possible extensions to the research. First the causal model can be represented using a directed acyclic graph and estimated using R-JAGS. Second the non-exchangeability of household members in the underlying Understanding Society survey data may affect our estimates of regional pay drivers; and third we may prefer to use a multilevel model with regions as Level 3 and households as Level 2.
Acknowledgements: We acknowledge the inputs of two teams of researchers over a number of years, including David Bayliss, Vanessa Gash, Hein Heuvelman, Pierre Walthery, Leen Vandecasteele, Anna Ritchie, our funders, the Close the Gap organisation based in Glasgow, UK, and the Government Equalities Office, who claim no authorship of these arguments. However their support and engagement with us over a period of several years (since 2015 for CTG and since 2007 for GEO) has influenced our choice of measurement methods in this paper. We also acknowledge the programming skill and inputs of Min Zhang, Wasel bin Shadat, Sook Kim, Vanessa Gash, David Bayliss, Kunal Sen, Giuseppe Maio, and Amaresh Dubey, without whose work this paper helped us in discerning patterns.
No registration needed. All Welcome. Tea & coffee from 15:45.
Professor Wendy Olsen
Organisation: University of Manchester
Travel and Contact Information
Humanities Bridgeford Street