©Reuters. FILE PHOTO: A view of the European Central Bank headquarters in Frankfurt, Germany, March 16, 2023. REUTERS/Heiko Becker/File Photo
FRANKFURT (Reuters) – The European Central Bank’s artificial intelligence model shows that inflation in the euro area could fall faster than the ECB itself predicted, but with much uncertainty, the bank’s chief economist Philip Lane said on Thursday.
It is the first time that the ECB has commented on the forecasts generated by its new machine learning model, which uses around 60 variables to capture changes that traditional algorithms fail to detect.
A slide accompanying Lane’s lecture showed that the ECB’s artificial intelligence forecasts put inflation by June much lower than predicted by the bank’s official macroeconomic projections – and closer to its 2% target – albeit within a very wide range of possible outcomes.
“The center of distribution is below December projections,” Lane said as he illustrated the slide at an event in Rome.
The ECB forecast inflation at 2.7% in the second quarter of the year, higher than the 2.3% expected by economists polled by Reuters and forecasts generated by the ECB’s artificial intelligence just below that mark.
Investors betting that falling inflation will force the ECB to start cutting interest rates as early as April may feel encouraged by these new computer-generated forecasts from the central bank itself.
But Lane argued that the “wide distribution” of possible outcomes implied by the AI model requires caution.
“The models say: ‘there are a lot of possibilities here, wait until the (data) tells you because you don’t want to place all your bets on the center of this distribution’.”
The model Lane was referring to was developed by ECB economists Michele Lenza, Inès Moutachaker and Joan Paredes, who presented it in a working paper last year.
In it, they said the model’s predictions closely tracked the ECB’s official projections, suggesting that the staff’s expert judgment allowed them to spot the same unexpected changes – known by academics as “non-linearities” – that the machine captured .
During the conference, Lane also said that the disruption of shipping in the Red Sea has so far had no material impact on Europe’s inflation outlook, given the limited weight of shipping costs in the overall price index.