Forecasting Nigeria's economic outlook: an ARIMA-based model using GDP data (1990–2023)

Abstract

This study develops a univariate Box–Jenkins autoregressive integrated moving-average (ARIMA (1,0,0)) model with drift to forecast Nigeria's annual GDP index using World Bank data spanning 1990–2023. We apply Augmented Dickey–Fuller unit‐root tests and first differencing to achieve stationarity, then determine model orders via autocorrelation function (ACF) and partial‐autocorrelation function (PACF) analyses. Maximum‐likelihood estimation yields an AR (1) coefficient of 0.4821 and a drift of 4.4052 index points, while Ljung–Box Q‐tests, residual ACF diagnostics, and normality checks confirm white‐noise innovations. One‐step‐ahead forecast accuracy is strong: root‐mean‐square error (RMSE) = 3.40, mean absolute error (MAE) = 2.54, and mean absolute scaled error (MASE) = 0.93, outperforming a persistence benchmark. Ten‐year projections (2024–2033) converge toward the long‐run mean within widening 95% confidence bands, and scenario‐based drift adjustments illustrate plausible growth paths under commodity and exchange‐rate shocks. We recommend anchoring fiscal forecasts to the central projection with ±3–4 index-point buffers, institutionalising rule-based stabilisers, integrating leading indicators via ARIMA-X or GARCH extensions, and deploying an interactive forecasting dashboard. This transparent, empirically validated framework enhances Nigeria's capacity to smooth revenue volatility, calibrate countercyclical measures, and reinforce macroeconomic resilience.