Today (Wed 28th) Pierre-Olivier Gourinchas (IMF Chief Economist) will present a paper titled “Unconventional Fiscal Policy in Times of High Inflation” (with Mai Chi Dao, Allan Dizioli, Chris Jackson, and Daniel Leigh) at the 2023 ECB forum on central banking in Sintra. Please, find below our review of the paper, including why we do not believe in the results.
(At the end of this note a few words on a new IMF paper on EA inflation)
What the paper does
The authors assess the impact of the set of fiscal measures implemented during the energy-crisis in the Euroarea on economic activity and consumers’ prices. The paper estimates counterfactuals (the evolution of output and inflation absent of energy shock and/or absent of the fiscal response to the energy shock), using two approaches: (i) employing the IMF Flexible System of Global Models (FSGM) -a large-scale multi-county general equilibrium model, and (ii) a partial-equilibrium reduced-form Phillips curve model. The authors claim that the fiscal measures reduced euro area inflation by 1 to 2 percentage points in 2022. About half of the reduction in 2022 reflects the direct effects of the measures on headline inflation, with much of the remainder reflecting the lower pass-through to core inflation. The impact on core inflation is therefore very limited (a few tenths) due to the fact that these measures were implemented in an environment with no economic overheating (as opposed to the US).
We discuss the two approaches used by the authors below.
Using the IMF FSGM
We can safely disregard the first half of the paper. The technicalities of the IMF Flexible System of Global Models (FSGM) are explained here. The FSGM is a large-scale multi-country DSGE model. Without entering into the details, the key feature of the model is that the Phillips curve is linear in slack, captured by the output gap (the reader can refer to page 7 of this IMF paper). Fiscal policy operates on consumers’ prices in two channels: (i) through the direct impact on headline inflation via energy subsidies, and (ii) by raising core inflation via the output gap. The main issue is that while (i) can be quite easily estimated, in a linear world with a flat Phillips curve (ii) plays no role. This explains why the authors conclude that fiscal policy had no impact on core prices. In order for fiscal policy to have any sizable effect on core prices, one would need to assume a shock about 10 times larger than the one simulated by the authors (30 percentage points of GDP instead of 3..). The reader can check that the impulse response functions of inflation in FSGM are typically muted (multi-country charts from page 34 here). Indeed, the same exercise in FRB-US would produce the same result: the output gap would need to be pushed to extraordinary levels to produce some inflation, given the flatness of the linear Phillips curve (FRB-US impulse response functions are here).
(Admittedly, the authors do mention the limitation of a linear framework, which bring them and us to the second part of the paper: the reduced-form exercise)
Reduced-form partial equilibrium Phillips curve model
Tell me what measure of slack you use and I will tell you the results. In the second part of the paper the authors run an exercise similar to Ball et al. (2022). The authors first run a set of Phillips curve regressions (including non-linear specifications). Then, they come up with a counterfactual for headline inflation with no energy shocks. Finally, they estimate a counterfactual for core inflation, given past headline which feeds into core. The claim is that while in the US the labor market had a sizable effect on core prices, for the Euroarea much of the increase in core inflation reflects the pass-through of headline-inflation shocks.
We do not discuss the authors’ results in details as we generally do for other papers because it is unnecessary this time. The reason is what the authors write at page 13, which we report below for convenience (yellow highlights):
For a reason not discussed in the paper, the authors capture “slack” in the Euroarea with the unemployment gap, while they use the ratio between vacancies and unemployed (V/U) for the US. (for the Euroarea the authors rely on the IMF estimate of U* which would be itself questionable). Surprise surprise, the claim is that fiscal policy did not have any effect on core inflation in the EA because the old continent was not overheated (in the regressions, slack plays little role and the dynamics of core ends up being explained by past headline inflation shocks, that is energy prices). Without discussing the regressions results, the choice of the authors is (to say the least..) “biased” for the following reasons:
- As well known, in this cycle the choice of the measure of “slack” is crucial because different measures have been sending different (or opposite) signals. For instance, if the econometrician chooses to run a Phillips curve using aggregate output gap as a measure of “slack”, she *knows* in advance to have zero chances to explain core inflation (leaving room for other variables to accomplish the mission, including past inflation itself..).
- A reasonable estimate of the unemployment gap in the Euroarea (sometimes cited in the slides of Isabel Schnable) is around -1%, not far away from the US estimates (CBO, FOMC). The same is true for aggregate output gap (in both cases estimated close to zero). Therefore, it is unclear why through the lens of the Ugap one can conclude that the US economy is way more overheated than the Euroarea. Not only, but it remains unclear why the Ugap should be used in the regressions for the Euroarea as opposed to other measures of “slack”. (For the record, the IMF estimate of the EA Ugap overtime is presented in the figure below)
Figure 1. IMF estimate of Euroarea Ugap
- The authors claim that the labor market in the EA is not overheated because they do not observe a shift in the Beveridge curve. The main issue (net of methodological and measurement issues which we can discuss via email) is that contrary to the US, the EA has registered a boom in the level of employment, in total hours worked, LFPR, and in the employment to population ratio (EPOP). Just for illustrative purposes, below we have plotted a measure of EPOP for the EA together with an estimated trend. Under this view, the EPOPgap in the EA is the largest ever recorded (while in the US EPOP remains below the pre-Covid level.
Figure 2. Euroarea EPOP and estimated trend
Let us conclude with a provocative question. Take two economies with the same Ugap. One has experienced a boom of employment and the number of vacancies did not increase much because firms filled them quickly. The other economy saw a rebound of employment and a very high number of vacancies. Both economies are experiencing high core inflation, high nominal wage growth, and both were subject to large monetary and fiscal stimuli. Which economy is more overheated than the other? In the end, we do not understand why so many people seem to be sure to know the answer.
Conclusion
We are here commenting a paper that, given the authors, will have a big impact in policy circles and beyond. It is very unfortunate that the paper is unconvincing (at best). The first part of the paper is obvious for those working with linear DSGEs. The results in the second part of the paper strictly depends on the measure of slack chosen. The risk is that the audience will only get the message “fiscal policy had no impact on core inflation in the Euroarea”. To us, that remains an unsubstantiated claim.
A new IMF paper on EA inflation
The IMF has released a new paper on EA profits and inflation dynamics (here). We have decided not to review it because it is in line with previous studies. Given the number of questions we have received (especially after the IMF has released this tweet), let us remind that a rise in unit profits does NOT imply neither a rise in the level of profits not a rise in markups. For the readers interested in the details, we have already discussed the algebra here. The bottom line is that the data do not support the “greedflation” view (which to us does not make any sense) and are more consistent with a combination of other shocks.