Modeling and forecasting currency in circulation for liquidity management in Nigeria

This paper presents forecasts of currency in circulation prepared for liquidity management at the Central Bank of Nigeria. Forecasts were produced using ARIMA, ARIMA with structural variables, VAR and VEC models. The performance of the forecasts was then evaluated under a rolling forecast scenario, where the estimation sample is augmented by one observation and the forecast sample is brought forward. The evaluation of the forecasts was based on average performance over a number of rolling forecasts. We found that the most accurate models were mixed models with structural as well as ARIMA components, augmented by seasonal and dummy variables. We also found that the exchange rate, interbank rate, seasonality, holidays and elections were significant in explaining the demand for currency.