Wednesday, 8 February 2012

People who disagree with me are stupid

Or perhaps not. They might not have chosen to consider my argument properly, if at all.  Or I might not have expressed myself clearly enough.  Or they might have a different view of the desirability of trade-offs involved.  Or I might be wrong.

However, there is a line of research purporting to show that right-wing ideology (which disagrees with me) is associated with being a bit thick.  A recent paper added to this corpus has been seized upon by George Monbiot in the Guardian  "conservatism thrives on low intelligence and poor information".  I think that's a fair description of Republican politics in the USA currently - the candidates for the Republican presidential nomination have recently been conducting a series of debates of quite extraordinary stupidity.  But in the UK, the Conservative Party, whose policies are well to the left of the intelligent and thoughtful Barack Obama, strikes me as no less intelligent than the Labour Party.  Its problem is a lack of concern for the problems of the economically unfortunate, not any inferiority in policy analysis, nor a reluctance to argue its case coherently. (Tim Worstall makes the case that in the UK 'conservative' and 'Conservative' may not be the same thing.)

The paper itself finds "that lower general intelligence (g) in childhood predicts greater racism in adulthood, and this effect was largely mediated via conservative ideology".  What the authors did is analyse some existing data from two studies in the UK, each of which tested a group of children then years later asked them questions as adults.  From these data they distilled measures of general intelligence g, social conservativism, racism, socio-economic status, and education level.  They then used a factor analysis to show that social conservatism is quite strongly negatively correlated to g, even taking into account socio-economic status and education level.  Another factor analysis including social conservatism as a factor showed that g is not a predictor of racism once you take social conservatism into account.

There are a lot of criticisms that can be made of the statistical methods: William Briggs outlines them here (but his comments about the tiny correlation along what the paper calls path c' are rather unfair: the authors are clear about it).  But the fundamental problem is that this paper is not setting out to test competing hypotheses, it's just looking for confirmation of one idea, by means of a crude measure of social conservatism.  The paper gives examples of the questions used in this statistic: “Give law breakers stiffer sentences”, “Schools should teach children to obey authority”, and “Family life suffers if mum is working fulltime”.  Similarly, the example questions it gives for the racism statistic are "I wouldn't mind working with people from other races” and "I wouldn’t mind if a family of a different race moved next door”.  What the results tell us is that the same people who answer yes to the 'conservatism' questions tend to answer no to the 'racism' questions, and those people tend to have lower g scores.  But those answers are not just conservative (by one definition) and racist, they are also unnuanced.  Perhaps the same people would have agreed also with some other questions on the questionnaire, such as "Ordinary working people do not get their fair share of the nation’s wealth" or "Private schools should be abolished".  If so, a differently motivated analysis might have found that g is negatively correlated with left-wing views, and left-wing views predict racism.

Perhaps a social scientist would care to conduct a statistical analysis to determine whether social scientists tend to seek confirmation of their own prejudices in the way they choose to analyse data.  Or perhaps it would be safer to get someone else to do it.

1 comment:

  1. People who disagree with you ARE stupid. Remember, the Bardie way is the only way, and you are always wrong.