Past Event

About the Speaker

Professor Raymond Duch

Raymond Duch is an Official Fellow at Nuffield College, University of Oxford, and the Director of the Nuffield Centre for Experimental Social Sciences (CESS), which currently has centres in Oxford, Santiago (Chile) and Pune (India). Prior to assuming these positions he was the Senator Don Henderson Scholar in Political Science at the University of Houston. He is currently the Long Term Visiting Professor at the Institute for Advanced Studies, Toulouse School of Economics, a Director of the European Political Science Association, and Vice-President of the Midwest Political Science Association. He is a member of the UK Cabinet Office Cross-Whitehall Trial Advice Panel to offer Whitehall departments technical support in designing and implementing controlled experiments to assess policy effectiveness. Professor Duch’s research focuses on responsibility attribution, incorporating elements of theory, experiments and analysis of public opinion. In 2008 he published an award-winning book, The Economic Vote, that demonstrates that citizens hold political parties accountable for economic outcomes. His experiments have identified the information shortcuts that individuals deploy for responsibility attribution. More recently, Professor Duch has conducted experimental research into cheating, exploring its implications for tax compliance, corruption and economic performance. Professor Duch has conducted lab, field and online experiments throughout the world He lectures and also publishes on experimental methods. His research appears in the leading political science and economic journals. He is the founder of Behavioural Analytics that advises public and private clients.

Abstract

Do digital traces accurately reflect individual preferences? If so, can signals from social media be used to measure public opinion about socially relevant quantities, such as elections, protest, consumer behaviour, and viral outbreaks?  In my presentation I will describe a novel MRP estimation strategy that combines samples of digital traces, our CESS Global subject pool, along with a post stratification dataset that includes individual-level socio-economic data, in order to generate area forecasts of the outcome social phenomena of interest.  We demonstrate that a relatively small virtual sample can be quite representative of the overall population.  I will demonstrate how we can use these digital traces along with global convenience samples to accurately predict social, political and economic behaviour.  Included in the presentation will be efforts to predict lying and cheating behaviour in the population, congressional district vote outcomes in the recent 2018 mid-term elections in the U.S. and our forecasted winners for the current 2019 Lok Sabha elections in India.

 

---

---