Even More BNP Data...

Various people asked what would happen if I excluded London and Northern Ireland from the BNP analysis. Here's a table showing the R-squared for each factor, first for the whole data set and then excluding these two outliers. (After all, who needs statistical analysis to know those two are weird?)

Factor	R-Squared	R-Squared Excluding NI, London
Immigration	0.0364	0.0810
Emigration	0.0330	0.0460
Migration	0.0343	0.0843
Services % GDP	0.0639	0.0009
Industry % GDP	0.0885	0.0066
Agriculture % GDP	0.0659	0.1697
Long Term Unemployment	0.2078	0.0074
Unemployment %	0.1080	0.0235
Economic Growth %, 1991-2006	0.0782	0.0692
Density Change 1991-2006 %	0.0369	0.0035
Population Change %	0.0008	0.0160

I'm still not convinced there is any rational pattern here at all. Immigration is still astonishingly weak as a predictor of BNP membership; weirdly, economic growth is even weaker, and positive! (I'm feeling so prosperous...I'm going to join the BNP!) In fact, the only factor in the second set of numbers that has an effect measurable without going into three significant figures is the proportion of GDP accounted for by agriculture. Northern Ireland is both surprisingly agricultural (2.3% of GDP - 130% of the UK average) and unsurprisingly low in BNP members (0.0024 per 100), so we wouldn't have seen this earlier on.

The Thatcher legacy - long term unemployment as a percentage of all unemployment - was the strongest correlate with R-squared = 0.2078, but when you drop London and NI, it vanishes, as does unemployment in general.