Government Spending Multipliers in Good Times and in Bad: Evidence from US Historical Data

Using a newly constructed quarterly US data set spanning 1889–2015 — built specifically to include the large wars and deep recessions that the post-WWII sample most of the literature relies on simply lacks — Ramey and Zubairy estimate government spending multipliers that are below unity regardless of whether the economy has slack (as measured by the unemployment rate), a finding that holds across two different spending-shock identification schemes (a military-spending “news” series and the Blanchard-Perotti shock) and across two different estimation methods (Jordà local projections and a threshold VAR). They find no evidence that multipliers are higher during high-unemployment states than low-unemployment states; the one case where multipliers differ significantly by state (using the Blanchard-Perotti shock) is driven by unusually low multipliers in low-unemployment states, not by elevated multipliers in high-unemployment ones. For the zero lower bound, the evidence is more mixed: across the full sample there is no evidence of multipliers above one, but once the WWII rationing period (1941q3–1945q4) is excluded, some specifications — particularly using the military news shock — imply multipliers as high as 1.4. The authors also show that a leading contrary finding in the literature, Auerbach and Gorodnichenko’s (2012) estimated 2.24 five-year recession multiplier, rests on an assumption that the economy stays in the recession state for the full horizon used to compute the multiplier, which is a poor approximation since their own data show recessions lasting only a few quarters on average; using a more data-consistent assumption about how the economy transitions between states brings their estimated recession multiplier down to about unity, in line with Ramey and Zubairy’s own numbers.

Sarah Zubairy | Macro Paper Warehouse

Government Spending Multipliers in Good Times and in Bad: Evidence from US Historical Data