Self-exciting threshold autoregressive model and endogenous structural breaks of exchange rate dynamics in Nigeria: a Bai and Perron sequential method

Abstract

This paper empirically examines structural breaks in Nigeria’s exchange rates of the Naira to GBP and USD datasets using average monthly frequency covering from January, 2004 to January, 2020. The study adopts a Bai and Perron (1998) sequential procedure. The study revealed that both series followed a nonlinear process. The Augmented Dickey-Fuller test showed that the conventional unit root test is biased towards the non-rejection of the null hypothesis of unit root presence in the series. While, the unit root breakpoint test by Vogelsang and Perron (1998) shows a contrary view as it rejected the null hypothesis. The NGN/GBP series generated a three (3) regime process with two significant threshold values of 259.96 and 388.93 for the period under review. While, NGN/USD series generated a two (2) process with a single threshold value of 359. The generated SETAR models for NGN/GBP and NGN/USD are SETAR(3;5,3) and SETAR(2;5,4), respectively. The diagnostic test (i.e. Breusch-Godfrey Serial Correlation LM Test and ARCH Heteroscedasticity Test) conducted on both generated models showed that, the models are free from serial correlation and heteroscedasticity which implies that these models are adequate for forecasting NGN/GBP and NGN/USD exchange rates.