The 2013 IASSIST conference was held a couple of weeks ago in Cologne, Germany, with a number of CAPDU members in attendance. The Canadian census situation was a theme running through all of last year’s conference in Washington – the opening plenary was devoted to it, and it came up repeatedly in talks, in comments on sessions, during coffee breaks – like a sore point that we data people just couldn’t stop poking. This year talk had of course died down, though I noticed a few pointed comments on governments that interfere with data collection.
However, the closing plenary was of surprising relevance – perhaps more so than the speaker intended. The talk was on record linkage in the social sciences, and the speaker made some comments suggesting that censuses are obsolete (or soon will be), with administrative data collection and linkage being the future of social science analysis. (This was suggested last year as a possible direction for Canada to consider.) He pointed out that our host country Germany relied largely on administrative data – in 2011 they had conducted their first census since reunification. (Privacy issues can be a political minefield in Germany, for reasons which it should not be too difficult to extrapolate.)
Coincidentally, that same day the results of the 2011 German census were released. They found that administrative data had over-counted the country’s population by 1.5 million, or around 1.5 percent. They over-counted the population of foreign residents by 15 percent. The foreign residents accounted for a large share of the population drop, but there were still close to 500,000 mysteriously nonexistant German citizens.
I’ll admit I’m surprised by this myself – I’d assumed that administrative data would be at least as accurate as Census survey data.