Summary
We built a model that produces one-year ahead forecasts for the year-on-year monthly growth rate of euro area HICP.
We combine 15 forecasting models that range from state-of-the-art deep-learning specifications to simple random walks.
We expect headline inflation in February to be between 4.6% and 5.8%. Our model suggests that in a year’s time headline inflation may still be significantly above the ECB’s target.
Our point forecast for 2022 is 3.7% compared to the December ECB staff macroeconomic projections looking for 3.2%.
To have full access to our research and content visit www.arkesresearch.com
Forecasting Inflation in the Euro Area
One of the consequences of the pandemic is the return of inflation in developed countries. After many years of subdued or missing inflation, policymakers have been surprised by the upward pressure on prices. In fact, the world’s major central banks first overlooked the rise in prices and then rushed into accelerating the withdrawal of monetary stimulus.
After the Fed and the BoE, the ECB re-calibrated its inflation narrative and consequently its intended path of monetary policy adjustment. The ECB is likely to announce on March 10 that it will taper asset purchases through September and lift rates by 25 bps by the end of 2022. At the next press conference, the ECB will also disclose its March staff macroeconomic projections and we expect the ECB to raise its inflation forecast for 2022 at least, which is now 3.2%.
Given the importance of forthcoming inflation prints for the market, we decided to build a model that tracks inflation rates in the euro area. In particular, we forecast the year-on-year monthly growth rate of the Harmonized Inflation Consumer Price (HICP). The forecast horizon is 12 months. Our model combines 15 models according to weights based on real-time forecasting performance. We will publish an update to our forecast every Monday.
The model
Forecasting combination is one of the most powerful techniques in time-series econometrics since it allows averaging across models, thereby reducing the root mean square forecast error (RMSFE). Therefore, we constructed 15 different forecasting models with the aim of enhancing the accuracy of our forecasts. The models we considered range from state-of-the-art deep learning models to simple random walk specifications. We backtested each model for a period starting in 2010 until today. We derived monthly weights from the backtesting exercise that we then used to aggregate the models into a final forecast combination. Put simply, models that appeared to have a lower RMSFE for month j have a greater weight in our aggregation function than models with higher RMSFE.
Results
The February release will be crucial since it will precede the Governing Council meeting. However, ECB policymakers will only observe the flash estimate of inflation (2 March) since the final release will only be published on March 17 after the meeting. Assuming our forecasts are reliable, the ECB will be deciding on the withdrawal of asset purchases with February headline inflation reaching a new peak at around 5.2%. Besides, in one year time, the ECB may still be well above the 2% target. In fact, our model suggests a point forecast of 3.7% for the end of 2022, against the December ECB’s forecasts for 2022 at 3.2%.
Looking at our data, it seems likely that inflation in the euro area peaked in January or February. However, while the declining path broadly reflects the ECB staff macroeconomic projections, we are still far from the ECB’s objectives. Most of the decline in our specification comes from lower input prices, an easing of supply bottlenecks in the second half of the year, tighter financial conditions and a slowdown of the economy. In other words, the transitory components of the well-known “temporary” narrative begin to fade but not as much as its supporters would like. We can conclude that inflation risks are definitely tilted to the upside. The ECB must be convincing in laying down a tightening cycle that anchors inflation expectations to the target. Two consequences will arise from beginning a robust tightening cycle: the growth rate in the euro area is set to soften and rates fragmentation is set to increase unless the ECB intends to calibrate an ad-hoc tool to tame the market’s peripheral sell-off (a very unlikely solution as of now).
Given the massive attention to inflation prints around the world, we built our own forecasting model to predict the evolution of headline inflation in the euro area. We combined 15 models ranging from state-of-the-art deep learning frameworks to simple random walk specifications. Our forecasts suggest February headline inflation will be between 4.6% and 5.8% with a point forecast of 5.2%, a new peak. Besides, while headline inflation displays a downward path mainly due to a softening of temporary factors such as energy prices, at the end of 2022 HICP is still projected to be well above the inflation target as well as the ECB staff macroeconomic projections.
Armando Marozzi
armando.marozzi@arkesresearch.com