The IMF grabbed headlines and upset officials earlier this month when it released an analysis which concluded that, starting in 2009, the fiscal policy multiplier has actually been considerably larger than previously supposed (IMF 2012). The Fund’s new estimates, which range from 0.9 to 1.7, suggest that Europe’s policies of austerity are in fact directly responsible for the fact that the continent’s recessions have been even deeper than initially forecast.
We are shocked – shocked – to find that there’s multiplication going on in here.
Actually, not so shocked. This is of course just what standard theory would suggest: that the fiscal multiplier will be unusually large when there is little monetary response to the fiscal impulse, whether because interest rates are at the zero lower bound or for other reasons. One might object that the ECB has, in fact, reacted to the Eurozone slowdown by easing. To this we would respond "not so much". What matters in this context are not targeted interventions like last year’s Long-Term Refinancing Operations or this (or next?) year’s Outright Monetary Transactions, which are designed to enhance the operation of particular markets and help specific sovereigns. What matters for the multiplier are economy-wide measures like interest rate cuts and quantitative easing. To date, the latter has been non-existent, while the former have been underwhelming.
The problem is that standard theory doesn’t tell us much about the precise magnitude of the multiplier under such conditions. The IMF’s analysis, moreover, relies on observations for only a handful of national experiences. It is limited to the post-2009 period. And it has been criticised for its sensitivity to the inclusion of influential outliers.
Fortunately, history provides more evidence on the relevant magnitudes. In a paper written together with Miguel Almunia, Agustin Bénétrix and Gisela Rua, we considered the experience of 27 countries in the 1930s, the last time when interest rates were at or near the zero lower bound, and when post-2009-like monetary conditions therefore applied (Almunia et al. 2010 ).
Our results depart from the earlier historical literature. Generalising from the experience of the US it is frequently said, echoing E Cary Brown, that fiscal policy didn’t work in the 1930s because it wasn’t tried. In fact it was tried, in Japan, Italy, and Germany, for rearmament- and military-related reasons, and even in the US, where a Veterans’ Bonus amounting to 2% of GDP was paid out in 1936. Fiscal policy could have been used more actively, as Keynes was later to lament, but there was at least enough variation across countries and over time to permit systematic quantitative analysis of its effects.
We analyse the size of fiscal multipliers in several ways. First, we estimate panel vector regressions, relying on recursive ordering to identify shocks and using defence spending as our fiscal policy variable. The idea is that levels of defence spending are typically chosen for reasons unrelated to the current state of the economy, so defence spending can thus be placed before output in the recursive ordering. We also let interest rates and government revenues respond to output fluctuations. We find defence-spending multipliers in this 1930s setting as large as 2.5 on impact and 1.2 after the initial year.
Second, we estimate the response of output to government spending using a panel of annual data and defence spending as an instrument for the fiscal stance.
Here too we control for the level of interest rates, although these were low virtually everywhere, reflecting the prevalence of economic slack and ongoing deflation. Using this approach, our estimate of the multiplier is 1.6 when evaluated at the median values of the independent variables.
These estimates based on 1930s data are at the higher end of those in the literature, consistent with the idea that the multiplier will be greater when interest rates do not respo