They are just simple correlations, but a sample of 34 (split into two groups) beats a sample of the one country the economist just happens to live in. Correlations of data are themselves data, and it is our job to try to interpret (or explain) the data.
This is how I (tentatively) interpret those correlations:
1. Fiscal policy worked in countries ("places" would be a better word) with fixed exchange rates. That didn't surprise me much. It's what I've been taught, and what I've been teaching for decades. I could rig up a model where fiscal policy does not affect NGDP even under fixed exchange rates (e.g. expected future distortionary taxes reduce current investment, or government spending is a close substitute for current private consumption or investment), but I would be surprised if that model I rigged actually worked at all well, except maybe in special cases.
2. Fiscal policy did not work in the other countries (or "places"). That surprised me a bit. I've been arguing that monetary policy can offset fiscal policy, even at the ZLB, and I could have been wrong about that. But the claim that monetary policy would in fact offset fiscal policy is a stronger claim, that I was less sure of. The counterargument, that central banks don't like doing the "unconventional" monetary policies that would be needed, and so won't in fact do them to the amount needed to fully offset fiscal policy, is not a silly argument. But it seems to have been wrong.
It's 6 days now, which is a long time in the blogosphere. I have seen posts about who said what about who said what. What I want to see are posts that interpret those correlations. And other interpretations/explanations are always possible (though econometricians bravely try to minimise the number of plausible interpretations). How would you explain them?
[Update: yes of course there are 1,001 reasons why Mark's simple regressions might bias the estimated coefficients on fiscal policy in a downward direction. But that's the beauty of splitting the sample into two. You need to explain why that bias would be smaller in the fixed exchange rate countries than in the other countries. You have to explain why, despite that bias, we still get a positive coefficient on fiscal policy in the fixed exchange rate countries, but zero in the other countries. And that's a lot harder to do. (If my econometrics were better, I would maybe start talking about difference in differences estimation, but it isn't, so I won't.)]
[Update 2: not that this really changes my policy recommendations (at least for Canada). Because I might be wrong, and it made sense to wear both monetary belt and fiscal braces if you wanted to be really sure your pants don't fall into depression. Sacrifice a goat too, if goats are cheap.]