Does Monetary Policy Improve Bank Lending- Rate Forecasts in Nigeria? Evidence from Econometric and Machine-Learning Models, 2006-2026

Abstract: Nigerian borrowers do not pay the monetary policy rate; they pay a bank lending rate that is set by commercial banks with one eye on the Central Bank of Nigeria (CBN) and another on their own funding costs. This paper asks whether the policy rate and related money-market indicators help to forecast the prime lending rate one month ahead, over and above what the lending rate’s own history already tells us. Using monthly CBN data from

autoregressive distributed-lag (ARDL) regression with the policy rate, Treasury bill rate, interbank rate and inflation, a ridge regression and a random forest. Each model is re-estimated every month and evaluated on 51 out-of-sample forecasts from January 2022 to March 2026. The prime lending rate is extremely persistent, and the linear models are statistically indistinguishable from one another (root mean squared errors of 0.454 to 0.460 percentage points), while the random forest is significantly less accurate (RMSE 0.618; Diebold–Mariano p = 0.004). At a three-month horizon, however, the macro-financial model reduces forecast error by about 13 per cent and is significantly more accurate than the autoregression (Clark–West p < 0.001). The macro-financial variables are individually insignificant in the forecasting equation but jointly significant (F = 3.32, p = 0.011), and policy-rate changes pass through to the prime rate only partially, with a cumulative three-month coefficient of 0.27. A bounds test finds no stable long-run relationship between the prime rate, the policy rate, Treasury bills and inflation. I also show that a single mis-recorded observation in the source file (September 2025) roughly triples forecast errors and reverses the model ranking if left uncorrected. The results point to slow, partial repricing of bank loans in Nigeria and caution against treating flexible machine- learning models as a shortcut to better forecasts in short, regime-prone samples.

Keywords: monetary policy transmission; prime lending rate; interest-rate pass-through; forecasting; random forest; Nigeria

JEL classification: E43; E52; G21; C53