Revised estimates of Leave vote share in Westminster constituencies

Shortly after the referendum I published estimates of the share of the vote won by Leave in parliamentary constituencies in England and Wales. These estimates showed that most parliamentary constituencies in England and Wales had voted to Leave (421 of 574 if you're counting), and that most Labour-held seats had also voted to Leave.

Those estimates were based on modelling the Leave vote share in local authority areas using a set of demographic covariates, and using the estimated relationship between these covariates and the Leave share of the vote to project on to parliamentary constituencies, for which information on the same demographic covariates is also available.

When I published those estimates, I tried to convey as clearly as possible the limitations of my analysis. One key limitation was the fact that my estimates did not exactly match the results from local authority areas where those overlapped entirely. In terms used by geographers, my estimates failed to satisfy the pycnophylactic property.

Since that time, I’ve been working on improved estimates which do have this property, and which also include results from Scotland. The details of the analysis can be found in a research note; the estimates themselves can be found in this Google Docs spreadsheet.

The difference between the two estimates are slight. 32 constituencies were formerly estimated to have voted Leave, but are now estimated to have voted Remain; 11 went the other way. This means a net change of +21 for Remain in England and Wales. Including Scotland, this means that 401 of 632 constituencies (63%) are now judged to have voted for Leave.