The techniques we’re using on this project belong to a long-lasting and broad tradition called small area estimation.
Small-area estimation has been around since the early eighties. It’s extremely useful if you have considerable information on large units but limited information on smaller units. It’s been used in areas from biology and medicine to economics and social policy.
It’s also something that our funders — the Economic and Social Research Council — have put a lot of support behind. The National Centre for Research Methods has a specialist network dedicated to the topic.
Our methods — which we’ll explain in later posts — fall in to what that special network describes as the “statistical regression-based approach”.
Small area estimation in the study of politics has a long and almost entirely American history, some of which is summarized in a recent and general paper by Jeff Lax and Justin Phillips, How Should We Estimate Public Opinion in the States?. Everything that we do here, however, should be intelligible to epidemiologists and human geographers — as well, of course, as the general public.