Region, local context, and voting at the 1997 general election in England

There has been considerable debate in recent work on voting patterns in Great Britain regarding the importance of regional effects: are these "real" or are they simply statistical artifacts of decision-making processes at smaller spatial scales which are aggregated up to the regional scale if not incorporated directly into any modeling? Using a multilevel model design, this article reports on analyses of survey data for the 1997 general election in England which allows tests of whether regional variations are no more than aggregation effects. Individual voters are nested within households, neighborhoods, constituencies, and regions and when all of the smaller-scale spatial levels are included in the model, the observed regional effects are statistically insignificant. At the 1997 general election, at least, regional variations within England in support for the three main parties - basically, a north-south divide - are aggregation effects.