Robert Waldmann: Multipliers, Reverse Causation, and the Heroic Assumption That E(εη) = 0: Monday DeLong SmackDown Watch
...@foosion That is a very good question. There are two reasons. First even when looking at a scatter it is best to worry about reverse causation (as Brad does at length in the post). Government revenues are very strongly affected by GDP growth. To consider the effect of tax increases and cuts, one has to look at changes in the tax code not the automatic effect of GDP growth on tax revenues. To do that for all the countries would require a major effort not a blog post.
Second PK is attracted to New Keynesian models (NK) in which deficits don't matter (the standard NK models have Ricardian equivalence). In those models, the varaible which matters is government purchases (G). By the way, G doesn't included transfers such as social security pensions or unemployment insurance. In any case, he used the variable which he should use given the models he likes best.
Even given the old Keynesian IS-LM model, an increase in G financed by an equal increase in taxes should stimulate. I certainly agree with you about the different effects of taxes on the best off and the rest, but that implies that taxes increases on both will have a smaller effect on aggregate demand than an equally large spending increase (this is what the IS-LM model says and this fitst the data).
@Brad It seems to me that the multiplier is too large. Your quick calculations definitely add value to PK's valuable post. But I have to guess that there is something else going on too. The red flag is the heroic assumption that epsilon and eta are independent. I think the question is whether there are shocks wich affect both G and GDP directly (not just through GDP and G respectively). Trying to describe them I go to pick your country casual empiricism and to the PIIGS. To try to be brief I think the causes of austerity includes financial crises. I guess the simplest example is Ireland which spent 30% of a year's GDP rescuing banks. This put pressure on G. I guess the other common cause could be high price levels (the over-valued Greek, Italian, Spanish, Portoguese and Irish Euro). I think trade deficits lead to austerity as governments don't like them and (correctly for once) believe stimulus causes higher trade deficits.
Anyway just arithmetically, something which directly affects G a little and GDP a lot can cause a spuriously high multiplier.
I would say: touché. I said it was a heroic assumption.
But do notice that the G is real government purchases, and hence excludes bailout transfers (I believe--it's supposed to exclude them, and I hope EuroStat did)...
