Saturday, November 22, 2008

Yet More BNP Data Analysis: Does Not Compute!

OK, so I've spent some time getting more data together on the correlates of BNP membership. I've created a table which contains the following metrics: population growth (%), change in population density (%), Gross Value Added(GVA) in 1991, 2006, change in GVA, % GVA growth, unemployment, long-term unemployment as a % of total unemployment, the shares of GDP accounted for by agriculture, industry, and services, total immigration between 1994 and 2002 per capita, total emigration per capita for the same period, total migration per capita, and BNP members per 100 citizens.

And you know what? I was expecting to find a correlation with the economic variables. I had a theory that long-term, Thatcher legacy unemployment, especially, would be a strong correlate of BNP recruitment. But nothing correlates. None of those metrics have any predictive power. Have a look at this.

Immigration per head is up the Y-axis, bigotry on the X-axis, and the data points are scaled by the unemployment rate. It doesn't seem to track any of these variables at all; I urge you to visit the visualisation home page, where you can try the different data series for yourself.

This strongly suggests that some completely different force is at work; perhaps BNP membership is driven by something else entirely. It could be the distribution of social authoritarian tendencies in the population, as Robert Altemeyer theorises. Or alternatively, it could just be that a gratifyingly small percentage of people are completely fucking stupid and pig-ignorant, that this is normally distributed in the population, and it's essentially a matter of chance what pig-ignorant fucking stupidity they get up to.

It's probably worthwhile pointing out that the average concentration of BNP members is 0.0203 per 100 citizens and the standard deviation is 0.0116. So with the sole exception of Northern Ireland, 1.54 standard deviations below the mean and therefore staggering towards the edge of the 90% confidence interval, the variation between regions is entirely explicable by chance - strong backing for the wanker theory.

(For some reason, this post has started to remind me of Donald Crowhurst's campaign leaflet, which bore the headline "YOU MAY THINK YOU ARE LOGICAL - BUT DARE YOU TAKE THIS TEST?" Inside was a sort of flowchart designed to explain logically why everyone should vote Liberal.)


ydue said...

Excellent stuff Alex. What might also be interesting is a more disaggregated version - I might try a sample from it (not really fancying the idea of looking up 13,000 postcodes) and see if there's any correlation with house prices, which would seem to be the only ultra-local rough proxy available for income/wealth.

Some time ago I did a university module in political extremism and thought that the authoritarian personality thing had been pretty much comprehensively shown to be bollocks, though I may be wrong.

cabalamat said...

I wonder what it would look like if each data point was a parliamentary constituency, not a region?

I suspect if you used constituencies, and plotted BNP membership v. proportion of the population who are Muslim, there might be a correlation.

Anonymous said...

The Graun's 'dark red constituencies' map has proved what us East Midlands antifascists knew for decades: concentrations of BNP membership exist in the 99%-white exurbs of Leceicester, not in any of the urban constituencies where teh brown peepul liv, even though the urban constituencies contain the vast majority of the people whom Blears has spent fifteen years screwing over and has now suddenly noticed.

But - this is a graph of membership, not vote. There's a lot of latent support for a well-run Strasserite programme in the UK. The BNP are still _under_-performing in this franchise, I reckon.


kostenloser Counter