Modelling valid population estimates using non-probability samples
This project undertakes an experimental approach to improving the accuracy of public opinion surveys conducted online. By leveraging unique datasets from the partner organization, the project team will endeavour to develop a model that produces more reliable estimates of public opinion than conventional polling methods. This model will be applied in the first instance to the forecasting of elections as a proof of concept and then extended to the measurement of public opinion more generally. As the partner organization is in the business of forecasting election outcomes and conducting public opinion research the findings of this project will help to inform its methodology on a go-forward basis.