So you admit 1999 is as arbitrary? And look, you should admit to yourself that the labour market is not the climate and labour market data not a temperature record. The thirty or forty years you need in climate science don't make much sense in a view at the trends of the labour market. I could start in november 1971, but would that really gave us any useful trend? Or in 1948 or 1919 or 1871 or 1815. Would you really do this?

Depends on what question I wanted answered. If I want to answer the question "what was the impact of German reunification on West German manufacturing?" I would look at the twenty years since and the twenty years preceding reunification. If I want to answer the question "what is the impact of Hartz IV?" I would look at all data since the reunification. Why the reunification? Because that's such a major confounder that disentangling the confounding effects of the reunification will be a bigger project than analysing Hartz IV.

Is this arbitrary? Yes, to some extent. All statistics involves arbitrary judgment calls in specifying your model. But notice that I give reasons for my picking those particular dates. If you can give other reasons for picking other dates, we can have a discussion about which dates are more reasonable. If you just pull a set of dates out of your ass, then we can't have a meaningful conversation about their relevance.

Alternatively, one could go the other way: Use one of the several statistical methods that finds breaks in the data itself, and use the location of those break points to sanity-check one's model of the economy.

No problem: Hartz IV started January first 2005. correlation, if you ask me, but you wanted that sort of argument.

I also wanted a properly specified model.

But I can actually well believe that the German "success" since 2006 is due to Hartz IV, because it dovetails nicely with what one would expect from a mercantilist wage suppression strategy. So it's intuitively reasonable.

- Jake

Friends come and go. Enemies accumulate.