Friday, 28 August 2015


Oh yeah your science we can't reproduce that.

Key horrible/hilarious statistic: "47% of original effect sizes were in the 95% confidence interval of the replication effect size". (For non-experts: that is not a good number.)

Thursday, 20 August 2015

Honesty in fifteen countries

A sneak preview of my honesty research. I'll have a poster at Decepticon in Cambridge next week.

Monday, 17 August 2015


(Mingardi on) John Gray on Hayek.

Oliver Sacks on Robert Aumann (and other things).

Kwasi Kwarteng's new book about Thatcher's toughest time. He's a sitting MP so he should know.

Dodgy charts in general

There is a bigger problem with the chart I previously mentioned (here you go:)

 Similar charts do the rounds at every US election, "showing" how Democratic presidents are better for growth. I've even seen them circulating on political science mailing lists, whose members should know better.

The bigger problem is assuming that there are no long-term policy variables. But it is obvious that one period's policy lays the foundations for growth or failure in future periods.  It is also well known that you can boost GDP by spending too much, and that this will have to be paid for in future.

Maybe Democrats or Labour are in fact better for the economy. But charts like the above don't prove that, because "growth is better under X" does not imply "X is better for growth". I spend big when I go to Swansea Vegas; that doesn't mean that moving to Swansea Vegas will increase my spending power.

Sorry: empirical public economics is harder than that.

Crooked timber, dodgy chart

Crooked Timber comment on a chart which has been doing the rounds.

This is a perfect application of interocular trauma econometrics – it hits you right between the eyes. It’s all up-and-down or left-and-right. The sort of thing that generates the difficult cases for liberal political philosophy – increases in inequality which nevertheless benefited the worst-off, which would have showed up as a southwest-to-northeast upward slope – never happened.
The chart looks at median household income versus inequality. That sounds sensible (we care about the average Joe) but it is an odd way to think about the growth versus inequality tradeoff. (It's a fine way to think about the Rawlsian inequality-versus-helping-the-worst-off tradeoff, but who ever thought Thatcherism was Rawlsian?)

Growth normally means the growth in average wealth, i.e. the mean. If you are only going to count growth of median incomes then you have already decided that any growth which only goes to the richer half of society is purely harmful – it does not change median incomes and it does increase inequality. But that is just the issue we want to discuss when we think about the growth-inequality trade off. Suppose poor, middle and rich incomes change from £20K, £40K and £60K a year to £20K, £40K and £100K. I can think of many legitimate reasons that might be a bad thing. Inequality may be bad for society even if it does not increase (absolute) poverty. But we must at least acknowledge that somebody has got richer and might like it. The median statistic ignores that: it stays fixed at £40K.

Wednesday, 5 August 2015

Robert Conquest RIP

Robert Conquest is dead. He was a poet, too, and wrote funny satirical verse. Off the top of my head:
There was a great Marxist called Lenin
Who did two or three million men in.
That's a lot to have done in
But where he did one in
That grand Marxist Stalin did ten in.


Attacks on freedom of expression in Europe.

My first Wikipedia article: McKelvey-Schofield chaos theorem. Just a stub!