27.4.12

Forecasting turn-out on next May Day?

I have tried to do a bit of "forecasting" with my model for turn-out at May Day demonstrations (based in my "Public Choice" article "It's the weather, stupid! Individual participation in collective May Day demonstrations").  A simplified model including

  • Average of min./max. temperatures (according to DMI's weather forecast, as of Friday, for Tuesday min. 6 degrees Celsius, max. 18 degrees = average 12 degrees)
  • Bandwagon (last year's no. of participants = 55,000)
  • Colour of government (not a "bourgeois" government, i.e. a dummy = 0)
  • Post-Soviet collapse (dummy = 1)
Using these four variables, we get a forecast of 81,782 demonstrators in Copenhagen's Fælledparken (standard error of estimate: 42,000).  Using
just the two first will give a forecast of 78,763 demonstrators (standard error of estimate: 43,000).  These particular "simplified" forecasting models only explain 50-53 pct. of the variation in turn-out for the years 1980-2011, so there is a ample room for other factors to play a role and for the forecast to miss ...

1 kommentar:

  1. According to the Copenhagen Police the 2012 turn-out was "only" ca. 35-40,000, which means that my simplified model's out-of-sample forecast was off the mark.

    SvarSlet