Three quirky little papers from the always interesting (though bizarrely difficult to navigate) SSRN website.
1. Lots of good sense from John Bogle about the perils of “getting what you measure”. Mainly about stock market and accounting issues, but the things he’s saying are of far more general relevance, particularly for economists at Warwick or Birkbeck universities.
2. Geomagnetic storms influence equity returns, apparently. People have been laughing at the behavioural finance crowd for this sort of windy, speculative, data-dredging stuff for years now, without obvious effect. Sunspots, etc. I haven’t seen a working paper yet relating trading volumes to menstrual cycles but I’m sure it’s out there.
3. Long term trends in given name frequencies in the UK. No really. Popularity of first names follows a power law, something that I would probably have intuitively predicted given that power laws tend to be generated by models with positive network effects.
oh yeh, and a bonus. Another paper criticising the old Lott ‘n’ Mustard ‘n’ Guns ‘n’ Crime paper. I haven’t done any in depth reading of the econometrics and frankly don’t propose to, but they do make a couple of points which I’ve made myself on the web in the past but haven’t seen in print: a) that Lott’s “demographic” variables included %black and %white, meaning that the model is collinear and thus highly prone to spurious fits (Lott also did this to a disgraceful extent when giving testimony in the Florida voting machines case, proving in my mind that he doesn’t understand the OLS regression model). and b) that *all* of these exercises are doomed to failure in terms of generating robust results, as it is just not possible to gather data on conditions in local crack cocaine markets, and without that data, the regressions are most likely worthless.