In this note we show that the assumption that core HICP inflation is expected to slowdown significantly because energy inflation has moderated is built on thin ice. A set of analyses using different models/techniques suggests core inflation is more persistent than the ECB staff thinks and markets seem to assume.
Philip Lane’s speech
In a recent speech, Philip Lane argues that “over the last eighteen months, the correlation of energy and core inflation has become increasingly contemporaneous – reflecting that energy has passed through very quickly to core inflation and been an important factor in the recent increase in core inflation in the euro area. This could imply that the recent decline in energy prices could also be reflected much more quickly than usual in core inflation in the euro area.” This statement, which is at the heart of the ECB staff macroeconomic projections as well as sell side analysts forecast, is based on the YoY correlation between core HICP and energy HICP (see Figure 3 of Lane’s speech). In what follows, we show that the statement does not survive to robustness checks.
(Side note: in the meanwhile, we have received the data for all measures of nominal wage and total compensation growth in the Euroarea. Wages are now growing at 7-8% annual rate (QoQ chart here), with the exception of advertised wages and negotiated wages, but because they are published only as YoY. It is very hard from our point of view to argue against signs of wage-price persistency in the Euroarea at this point. We are very puzzled most dismiss a 7-8% nominal wage growth).
Our research
We investigate three research questions using different techniques. The research questions are: (i) has the pass-through from energy prices to core HICP inflation increased over time and it is asymmetric, (ii) how much of the current core HICP readings can be explained by gas/energy prices, and (iii) given the natural gas shock, what should we expect going forward?
1. Has the pass-through from energy prices to core HICP inflation increased over time and it is asymmetric?
The results in Lane’s speech are problematic for three reasons. First, they imply causality using a correlation. Second, they are based on correlations of YoY variables, which tend to cumulate the errors. Third, the results show the correlation between HICP energy and HICP core inflation which are, at least some extent, endogenous to each other.
To prove our point, we first run basic rolling regressions; but instead of using YoY changes we use MoM (ar) series. Figure 1 shows the rolling correlation coefficient between HICP energy (MoM, ar) and core HICP (MoM, ar) using a 12-month window (results using a 24-month window can be seen here). The takeaway is that the correlation has increased last year but it is now close to zero. Not only, but once the rolling correlation is estimated in a regression as in Figure 1, the coefficient is almost never significant (put it differently, there are no confidence bands in Lane’s chart).
Figure 1. Rolling correlation coefficient HICPX – HICP energy
Note: the figure shows the estimated coefficient of a rolling regression (12month window) of HICP energy on core HICP prices. Series are MoM at annual rate.
A set of alternative specifications reinforces the message. For instance, adding controls such as measures of labor market tightness or fx movements brings similar results (see here). Also, considering the correlation between core HICP and oil (here) or natural gas (here), which removes part of the endogeneity issue, results in correlation that have not increased over time. Finally, when adding dummies to estimate whether the relation is asymmetric (that is whether we should expect a faster drop compared to the runup), we could not identify a statistically significant relation.
Therefore, according to our research the answer to the first question is: “It is hard to conclude that the pass-through from energy prices to core inflation is higher now than in the past, and the pass-through seems symmetric”.
2. How much of the current core HICP readings can be explained by gas/energy prices
We investigate this question by setting up a simple model. We first show that a model containing natural gas, its square and cubes and the lags of them explains as much as half of the recent runup of core HICP, with the model predicting the latest MoM of core HICP at around 3½ percent. These results are shown in Figure 2. Having said so, the model fit before the current episode is quite poor so the reader should take these estimates with caution. Nevertheless, this model results are in line with the recent literature suggesting that about half of the current core inflation in the Euroarea is driven by supply-side factors, and the rest by demand-factors.
Figure 2. Core HICP (MoM, saar) and model fitted values
Note: the figure shows core HICP MoM saar (the blue line) and the fitted values (orange line) of a model regressing TTF, its square, cube, and 8 lags of them on core HICP.
Furthermore, a model that includes only the lags of the dependent variable (results here) produces a better fit of the data, especially in the last part of the sample. This is not to say that energy prices are not playing a role. Instead, this suggests the obvious: inflation is a process, and core inflation is very persistent (for the record, the sum of the autoregressive coefficients in the model is close to .95). Unsurprisingly, a full model including lags of the dependent variable, energy prices (squares and cubes) and their lags, fx, and labor market variables produces a forecast (see here) that remains close to 5% (MoM saar) at the end of 2023, way above what the ECB staff is expecting (re-estimating the model from 2016 ends up in an even higher forecast).
We conclude that the answer to the second question is: “energy prices can explain up to half of the runup in core HICP, which is a very persistent process and it is expected to disinflate only gradually over time”.
3. Given the natural gas shock, what should we expect going forward?
While the answers to the first two questions show why the ECB staff and Philip Lane are skating on thin ice, the most interesting question is another one: if natural gas increases by 1% at time “t”, what happens to core inflation at time “t+k”? We answer this question using local projections techniques (Jordà (2005)). Figure 3 shows the local projection of natural gas (TTF) on core HICP inflation. As a reminder, Figure 3 shows the response at horizon “m” (in months) of core HICP (MoM, saar) to a 1% shock to the MoM of natural gas. As with the simple correlation, the confidence bands are quite large. Nevertheless, this exercise seems to suggest a positive pass-through (CI are skewed to the upside), protracted in time. If we take the estimated coefficients and multiply by the size of the TTF shocks in the last 24 months, the cumulative impact on core inflation is estimated around 3 percent, close to the exercise in question #2 of this note. In any case, even by using this technique the answer remains broadly the same: the energy shock explains only half of the current runup. Not only, but it can be that the pass-through from TTF to core is still ongoing and might remain positive for another few months before putting some downward pressure on core HICP.
Therefore, the answer to the third research question is: we should expect (if any) a very gradual disinflation of core HICP, either because half of it is demand-driven or because in any case the passthrough from energy prices to core takes time.
Figure 3. Local projection of natural gas (TTF) on core HICP.
Note: the figure shows the local projection à la Jordà (2005) of natural gas (TTF) on core HICP. Shock is 1% of the MoM at annual rate of TTF.
Conclusion
Our research has shown that the ECB staff and Philip Lane are skating on thin ice. The shock to natural gas seems to explain about half of the elevated core HICP. It is possible to explain a larger portion of core inflation with energy prices but the downside of this result is that core would be expected to disinflate only very gradually. No matter what technique/model we have tried, we have been unable to find a quick (and significant) passthrough from energy prices to core inflation that might justify the ECB staff and Philip Lane’s optimism.