If you’ve read the short description of our project that we put up some time ago, you’ll realize that this project tries to produce good estimates of public opinion in British constituencies, and to use those estimates to assess the link between constituents and their MPs.
We’re not operating in a vacuum here. Others have written on the link between constituency opinion and MP behaviour — but they’ve had to rely on very loose proxies for opinion.
Let’s take a 1987 paper on the link between constituency opinion and a series of free votes on abortion, capital punishment, and homosexuality. This paper, by John Hibbing and David Marsh, has to rely on a proxy that others have used — namely, the proportion of Catholics in a constituency. Hibbing and Marsh find that, after controlling for party, MPs in constituencies with a higher proportion of Catholics are more likely to support more socially conservative positions.
So far, this approach hasn’t been greatly improved upon. A more recent (2004) paper by John Baughman, on abortion, has a very nice way of handling votes on abortion (an item response model with a meaningful scale) — but the measurement of constituency opinion relies on two dummies for Catholic-heavy regions (e.g. Strathclyde, Merseyside).
These kinds of proxies — percentage of Catholics, percentage of homeowners, etc., — are used for want of more direct measures of constituency opinion. They are useful indirect measures of constituency opinion on a particular political issue to the extent that membership of different demographic groups tends to go with different opinions on that issue.
The trouble is that the relationship between demographic groups and opinion is by no means a perfect one: Catholics do not have uniform political views, and neither do homeowners. In other words, demographic proxies are very noisy measures of constituency opinion. As a result, when we use these proxies to assess the link between MP behaviour and constituency opinion, the true picture may be obscured or even misrepresented.
Another problem with using demographic proxies for constituency opinion is that there are only a limited number of political issues where peoples’ opinions are strongly associated with their observable social characteristics. Because of this, researchers have been restricted to assessing the MP-constituency link on this limited subset of issues.
In sum, we need better, more specific, measures of public opinion in Westminster constituencies if we want to better understand the link between MPs and their constituents in Britain.