Invisible Gammon: Fun with the Ecological Fallacy

A lot of people are really impressed by the insight that survey research among the British reliably suggests that both strongly left-wing and strongly right-wing positions are popular. A more nuanced version refines this to economically left-wing and socially right-wing positions. Nick Barlow’s post here is an example, but I don’t want to single him out particularly – the post is not stupid, it’s just an example that was to hand. Some people take this all the way to thinking that there is a big opportunity out there for a political party that wants to “fund the NHS and hang the paedos”.

There is a serious problem with this. Consider, if you will, a population where 40% of the people support the Red Party, which is strongly redistributionist, socially liberal, and cosmopolitan. Another 40% of the people support the Blue Party, which is economically libertarian, socially conservative/authoritarian, and nationalist. The people’s priorities among these issues are the usual ones – economy first, then society, then foreign policy. If this seems familiar it’s because I have selected parameters that resemble Britain today.

Now let’s ask some people what they think. Let’s draw a representative sample of people and ask them to, say, rank a list of policy positions taken from the Red and Blue election manifestos. Fairly obviously, the Reds will rate the flagship Red policy – say, generous public healthcare – top and the Blues will do the same with theirs – say, hating foreigners. If nothing else happens, both these policies should be ranked top by 40% of the public, plus or minus the sampling error. Exactly which one comes top will be determined by the margin of error. Something similar will happen as we go down the ticket.

In this example, it would probably be sensible to expect that politics in Red and Blue Land would have a lot to do with the two tribes’ favoured top policies. However, it would be really stupid to imagine you could become president by running for election with the list of top-rated policies in your manifesto. In our example, by construction, there is literally nobody who believes in both parties’ top priorities. Rather than a historic compromise, the hypothetical centrist candidate would be more like Steve Bell’s proposal to reconcile Northern Ireland’s conflict by launching the Fuck The Pope And The Queen Party, or the wonderful Yiddish concept of the kosher chazzer. The apparent public consensus is a vacuous artefact of aggregating very different groups.

This is a common enough error to have a name: the ecological fallacy, which occurs when you assume that the average properties of a population are informative about those of individuals within it.

And you can’t save this by appealing to the politics of emotion. The kosher pig candidate would have to project generosity and meanspiritedness at the same time, some trick. If you think political platforms are aesthetic and emotional constructs, this implies that coherency of style, tone, and story is important. If you think they are programmes for action, they need to be administratively workable and physically possible. In case you’ve not noticed, we’re having a test of that one.

6 Comments on "Invisible Gammon: Fun with the Ecological Fallacy"

  1. As an aside, while in fact it was not the case*, one can imagine that UKIP at peak gave us a sense of the actual numbers of people who were up for an “authoritarian left” and it’s not really enough to justify a lot of the electoral frothing from commentators, esp. in a FPTP system.

    *In fact, UKIP at peak was in reality (even in “The North”) composed of factions, and a significant one was “people who stopped voting Labour in 1979.”


  2. Ah but UKIP had gone from 5% to 25%, so the only real question was whether they were about to go from 25% to 45% or to 125% – can’t argue with mathematics! I caricature slightly, but I really think a lot of the “UKIP breakthrough” thinking was the product of projecting a straight-line graph. It was a revelation to me when I looked at the figures and discovered that, in one seat after another, that big chunky gain in UKIP votes consisted of a bit from here plus a bit from there plus the entire vote of the two local Fascist candidates who had been cultivating that seat for the last ten years. Hoovering up the fash vote, as a political tactic, is many things, but it’s not repeatable.


  3. Ecological fallacy is a real problem … but it turns out the pattern Barlow is reporting is not a result of that (it’s not clear from the single factor graphs in the text, but it is clear from links to Paula Surridge’s stuff and the crosstab Excel links – I prefer needlessly overcomplicated kdes).

    There are different ways to think about the UK populace being “surprisingly Left-Authoritarian”:
    – you can look at where it is relative to some evaluation of “elite discourse”
    – you can look at where the mean is relative to the semi-arbitrary range of the questions
    – you can look at the distributions and see which way they skew

    Ultimately, you get the same answer.

    In fact, even if you decide that you want to tear the scales apart and recreate a new set of scales which are perfectly normally/uniformly distributed orthonormal factors … well, you can’t. There’s too much correlation between elements of the two (on average) uncorrelated sets of variables – net result is that there’s always a lot of Left-Authoritarians and hardly any Right-Liberals.

    It’s basically very, very hard to come up with a way of splitting up the UK population that doesn’t lead to an excess of Left-Authoritarians …

    … even if you’re deliberately *forcing* an equal split.

    I’ve had the inklings of partial success … if I remove the “redistribution” variable from the set of the 5 variables that make up the standard Economic Left-Right measure.

    But the reason that works is that Redistribution is heavily associated with Social Liberalism – you’re basically conceding that the fundamental way we think about left-right economics is biased towards Social Liberals … in some ways, that’s a bigger concession than just saying “yep, a whole lot of Left-Authoritarians!”.


  4. Re: The point that the Blue Party and the Red Party are effecting opinions.

    That definitely happens – there’s no question that there are partisanship effects …

    … however, there’s no evidence that those effects change answers on the scales mentioned. In fact, it’s quite hard to establish anything that really shifts them. You can see some effects that appear to be life-stage related (e.g. buy house, become a bit more economically right-wing), but these things are pretty stable over 40 year longitudinal studies.

    You can make a strong case that parties – specifically the party in government – has influence on the values of voters …

    … but only voters aged ~15-24 years old. After that, do not expect a readily identifiable impact on basic (ec. left/right; auth-lib) values.

    Even if that *weren’t* the case, both parties have been pushing redistribution and social liberalism and doing their best not to make immigration salient – “Left-Authoritarian” voters are really surprisingly unkeen on either of those things and very unkeen on immigration.

    FWIW, my suspicion is that the raw number of Left-Authoritarians would not translate into a reliable party turnout because Left-Authoritarians are, due to their other values, never likely to be enthusiastic repeat voters in a liberal democracy.

    But that’s just a suspicion.


  5. A commoner and equally annoying instance of the fallacy is the insistence that “US voters prefer a divided government” because they so often end up with a president from party X and a Congress with a party Y majority, and therefore that must be what they want.


Leave a Reply to ajay Cancel reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.