(Preamble: Riccardo Trezzi had a chat with Nick Timiraos of the WSJ. You can find it here)
The BIS has published a paper titled “Does money growth help explain the recent inflation surge?” by Claudio Borio, Boris Hofmann and Egon Zakrajšek. The paper argues that monetary aggregates help explaining inflation, a long-standing argument in economics that had lost appeal in the 30 years preceding Covid. We argue against a straightforward interpretation of their results: M2 can be a useful indicator but only to assess the risks around a baseline forecast.
What the paper does and results
Money growth is linked to inflation but only at certain conditions. The paper provides cross-country evidence on the link between “excess money growth” –the difference between money growth and real GDP growth– and inflation. Using a large sample of advanced and emerging market economies (1951–2021), the authors argue that there is a one-to-one link between excess money growth and inflation. However, this relationship exists only when inflation is high, while it is non-existent when inflation is low (less than 10% or so). The main result of the paper is shown in Figure 1 below.
Figure 1. Excess money growth and inflation.
As for the most recent period (2020 to present), the authors argue that excess money growth helps explaining the inflation forecast errors of professional forecasters. In other words, the authors (implicitly) argue that if professional forecasters (and central banks) had taken into account monetary aggregates in their forecast, they would have done a better job in forecasting inflation.
Comments and our own research
Time to go back to Mv=PQ? The answer is “careful”. While some people in the profession see “Mv=PQ” as a mantra, it is nothing but an identity: it is true (by definition) at each point in time because “v” is unobservable and implicit given M, P, and Q. Having said so, the fact that “Mv=PQ” is an identity, it does not imply it is necessarily useful. Here is why:
- First, monetary aggregates say little about causality. The debate about the direction of causality between M and inflation has not been fully settled. The observation that money growth today helps to predict inflation tomorrow does not, in and of itself, imply causality (which in fact can run the opposite way).
- Second, in some cases (i.e. demand-driven, as in 2020, when companies drew heavily on their credit lines) the increase in M is largely endogenous. Not only, but agents may adjust their portfolios in advance of their spending based on changes in their income: in this case, it is income, not money, that causes spending to increase, with the evolution of money balances acting as a signal. (these two bullets largely rely on the text from Borio et al. (2023)).
- Third, monetary aggregates are correlated with price increases only at low frequency (see for instance D’Agostino and Surico (2009) or Benati (2021)), especially when inflation is very high or in episodes of hyperinflation. Therefore, it is unclear how one can rely on “M” when inflation remains relatively low (10% or less). We discuss the implications for policy below.
- Finally, monetary aggregates can ultimately have effects on inflation only via aggregate demand. Therefore, a way to “bridge across religions” (monetarism and NK models) is to think what is the appropriate measure of “slack” is in a NK model without relying on Mv=PQ.
Our research suggests that monetary aggregates are not particularly useful for policy but they can/should be used to assess the risks around a forecast. After running a large set of regressions, our research suggests that:
- For both, the US and the euro area (the growth rate of) M2 does not enter significantly in a regression of core inflation. In fact, the “on impact” coefficient of M2 is negative in both regressions and we could not identify a significant relation. The only way to find something statistically significant (for core inflation in the US only) is to include 6-8 lags of M2. The result is that 3 coefficients are positive and significant. The R^2 is generally pretty low (around .15) and it increases only including the Covid period.
- When repeating the exercise with total inflation as a left-hand-side variable we have obtained similar results for both, the US and EA.
Implications for the Fed and ECB staff
The workhorse model in modern central banking is the New Keynesian Phillips curve and it is likely to remain as such for a long time, given that the alternatives have lower in-sample fit properties and forecasting power. Nevertheless, in our view Borio et al. (2023) will generate some discussion as one of the authors (Egon Zakrajšek) is a former Fed Board officer. Having said so, as we discussed above, monetary aggregates can be useful in assessing the risks around a baseline forecast but, given their limitations, they are unlikely to replace general equilibrium models. We expect the Fed/ECB staff to act accordingly.