Request Paper

Abstract
The predicting power of ARIMA model was profoundly explored in this paper using Nigerian external debt. The study adopted ex-post-facto research design with sourced data from Nigeri an Debt Management Office Bulletin (2005-2019) editions with Box–Jenkins and forward ste pwise approach. Non-seasonal ARIMA models generally denoted by ARIMA(p,d,q) was also a dopted. ARIMA(0,2,0) was produced by auto.arima function from R software while two model s; ARIMA(1,1,0) and ARIMA(1,1,1) were produced via forward stepwise approach. Findings from the study showed that models with 1 differencing are optimistic about the future unlike t hose with 2 differencing like ARIMA(0,2,0) produced by auto.arima. ARIMA(1,1,0) which is a fine-tuned random walk was shown to be optimistic about the future as seen from the look of t he forecast. This was in agreement with (Robert, 2020) findings. The Akaike Information Crit eria (AIC) of ARIMA(1,1,0) model = 284.03 was smaller than that of ARIMA(1,1,1) model = 285.79 and the smaller the AIC, the better. In vein of these findings, ARIMA(1,1,0) was chose n as the best fitted model. Also, the prediction of external debt from ARIMA(1,1,0) model depi cted a slight upward trend. Researchers were advised among others to stick to the forward ste pwise approach in choosing ARIMA models to avoid wrong predictions/forecasts. Governmen t of the day was advised to perceive the dangers of borrowing and misappropriation of funds t o the economy. Keywords: ARIMA models, External Debt, Prediction, Time series.