Puzzled by Gerry Friedman…

A question about the estimable Gerry Friedman:

How can an increase in government spending of $1.4 trillion/year generate a $14 trillion increase in spending in the year 2026? But it really looks to me like he has both:

  • the very dubious assumption that all 10%-points of the shortfall from the trend as of 2007 can be made up relatively easily’, and
  • a multiplier of not the 3-in-and-near-a-liquidity trap I carry around in the back of my head, but 10.

But Friedman’s text claims his multiplier is not even 3, but less than 2, and averaging roughly 1…

In short: In his runs Friedman has government spending higher in 2026 by $1.4 trillion than in baseline. He has real GDP higher in 2026 by $14 trillion. What other components of real spending are higher by how much in order to make that real GDP number in the year 2016 higher than baseline by $14 trillion? And what mechanisms are making those components higher?

It’s fine to propose aspirational policies based on a hope that the world is such that things will break your way. It’s not so good to put the world breaking your way forward as a central-case forecast of what your policies will do. And it’s distressing that I cannot figure out how to make Friedman’s analysis hold together quantitatively even if I do allow the assumption that the entire output relative to the pre-2007 potential-output trend can be closed easily…

Must-Read: Jan Mohlmann and Wim Suyker: Blanchard and Leigh’s Fiscal Multipliers Revisited

Must-Read: Naughty, naughty!

Jan Mohlmann and Wim Suyker: Blanchard and Leigh’s Fiscal Multipliers Revisited: “[We] do not find convincing evidence for stronger-than-expected fiscal multipliers for EU countries…

…during the sovereign debt crisis (2012-2013) or during the tepid recovery thereafter…. As Blanchard and Leigh did, we find a negative and statistically significant coefficient for 2009-2010 and 2010-2011 but not for 2011-2012…. For 2012-2013 we find a larger estimate than Blanchard and Leigh, but due to the higher standard error the estimated coefficient is no longer significant at the 5% level…. In the two periods we added, our estimated coefficients are close to zero…. As Blanchard and Leigh did, we find a statistically significant negative coefficient in the panel forecast for 2009-2013. This result holds for the prolonged period 2009-2015. However, we do not find a statistically significant coefficient when we perform the panel analysis for the period 2011-2015.

Nowhere in their piece do Mohlmann and Suyker report their estimated coefficient and its standard error for the entire period 2009-2015.

Repeat: nowhere in their piece do they report estimates for their entire sample.

Trying to back out estimates from the information they do give, if they had reported it they would have reported a number like -0.60 with a standard error near 0.23. Compare that to the Blanchard-Leigh estimates for 2009-2013 of a number of -0.67 with a standard error of 0.16.

See that a good and true lead is not “[There is no] convincing evidence for stronger-than-expected fiscal multipliers… during… 2012-13 or thereafter…”

See that a good and true lead would be: “There is no statistical power at all over 2012-13, 2013-14, and 2014-15 to test whether excess fiscal multipliers in those years are different than the strong excess fiscal multipliers found by Blanchard and Leigh…”

Seizing the high ground of the null hypothesis for one’s favored position, and then running tests with no power, is undignified…

Blanchard leigh Google Search