Market rates enhance accuracy of bank lending rate forecasts

The Banque de France study presents an effective method for forecasting new lending rates for households and businesses in France and the euro area.

This approach augments an autoregressive model with market interest rates, improving forecast accuracy by an average of 36% compared to purely autoregressive models.

Crucially, the model demonstrates a superior ability to forecast trend reversals, a key limitation of traditional methods.

Commercial banks generally set their interest rates based on market rates, which reflect both their financing and opportunity costs.

The study uses overnight index swap (OIS) rates at 3-month, 2-year, and 10-year horizons as market indicators.

These OIS rates often precede movements in bank rates, acting as leading indicators.

For instance, household loan rates are particularly linked to 10-year OIS rates, while business loan rates correlate more with 3-month and 2-year OIS rates.

The study employs an auto-regressive distribution lag (ARDL) model, augmented with market interest rate expectations from OIS rates, to overcome the limitation of traditional autoregressive models in forecasting cycle reversals.

This methodology simulates a realistic forecasting environment, using an expanding window approach for estimation and making out-of-sample forecasts for horizons up to twelve months.

Predictive quality is assessed using the root mean squared error (RMSE).

Results show that the optimal market rate for forecasting varies by loan type and geography.

For household loans in both France and the euro area, the 10-year OIS rate provides the best predictive power.

For business loans, the 2-year OIS rate is most effective in France, while the 3-month OIS rate is superior for the euro area.

This aligns with the typical maturities of these loan categories.