Guesstimating the US-Euro Area Core Inflation Differential in April

April HICP numbers are out for the Euro Area. The US reports April CPI on Wednesday. Using the Cleveland Fed’s nowcast for April core (0.52% m/m vs. Bloomberg consensus 0.4%), we have the following picture.

Figure 1: Month-on-month annualized core inflation for US (blue) and Euro Area (red), calculated using log differences. Euro Area core HICP seasonally adjusted by author using geometric Census X-12. April US core CPI uses Cleveland Fed Nowcast of 5/8. NBER defined peak-to-trough recession dates shaded gray. Source: BLS, Eurostat via FRED, Cleveland Fed, NBER, and author’s calculations.

In order to see when core US inflation accelerates with respect to Euro Area, I show the differential below.

Figure 2: Month-on-month annualized core inflation differential US minus Euro Area (black), calculated using log differences. Euro Area core HICP seasonally adjusted by author using geometric Census X-12. April US core CPI uses Cleveland Fed Nowcast of 5/8. Teal line is 2018-2020M02 average, red line is 2020M03-2022M04 average, light blue line is 2021M04-2022M04 average. NBER defined peak-to-trough recession dates shaded gray. Source: BLS, Eurostat via FRED, Cleveland Fed, NBER, and author’s calculations.

Clearly, there is a slight acceleration 2020M03 onward. However, this acceleration is not statistically significant treating inflation as stationary (and the price level as nonstationary). There is much better case for a statistically significant acceleration starting at 2021M04 (2.1 ppt acceleration relative to 2018-2021M03, with significance at 15% msl using HAC robust standard errors). 2.1 ppt is economically significant, in my view, even if not statistically significant in this case. This is likely coming from demand pull (or at least compositionally goods demand pull vs. capacity) rather than cost-push shocks, given the timing. This timing is also consistent with a smaller (in absolute value) output gap in 2021 for the US vs. the Euro Area, as discussed here.

(One could treat the log price ratio as a I(0) series with a structural break vs. an I(1) series, using a Bai-Perron test. Then one could reject in favor of a break with statistical significance; see this post.)