Deep-π*: a neural network estimate of trend inflation
Key takeaways
- Deep‑π shows that employment and financial conditions are key drivers of trend inflation, providing a more transparent and accurate forecast tool than traditional models.
- Employment variables were the largest contributors to the post‑pandemic rise of the current trend
- Forward‑looking trend peaked 8–12 months before the current trend
- In early 2021, professional forecasters expected inflation of 3% or less, but the forward‑looking trend already signalled an increase
What happened
The European Central Bank released a study showing a new neural‑network model, Deep‑π, that estimates trend inflation by linking it to many macroeconomic indicators and attributing each change to specific drivers using Shapley values.
Why it matters
Traditional trend‑inflation models treat the trend as a random walk, making it hard to see which economic factors push it up or down. Deep‑π’s approach lets policymakers and analysts identify the real‑time contributors to inflation trends, improving transparency and potentially guiding policy decisions.
Short-term and longer-term
Short-term. In the next few weeks, watch for how the ECB’s inflation outlook may shift as the Deep‑π model highlights early signals of rising trend inflation.
Longer-term. Over the coming months, investors should consider how the identified drivers—employment, real activity, money and credit—might influence sector performance, particularly in consumer discretionary and financials, as…
Words on this card
- Trend inflation
- The underlying, long‑term path of inflation, excluding short‑term shocks like energy price spikes.
- Shapley values
- A method from game theory that attributes the contribution of each input variable to an outcome.
- UC‑SV model
- A standard unobserved component model with stochastic volatility used for inflation forecasting.
- CPI Learn
- Consumer Price Index — a common measure of how everyday shop and living prices are changing.