Forecasting currency in circulation with the central bank balance sheet

Currency in circulation (CIC) is an important variable in monetary policy as it affects liquidity and guides the currency issuance operations of central banks. This paper proposes a novel approach to forecast CIC using central bank balance sheet variables, namely assets and liabilities other than currency issued. The balance sheet approach is able to generate monthly CIC forecasts as opposed to demand-for-currency models anchored on quarterly Gross Domestic Product (GDP). This allows for more responsive currency policy, particularly during crisis periods when precautionary motives intensify—reflected in a decoupling of GDP and CIC—or when spikes in currency demand arise due to heightened transaction motives. Dynamic time series regression models are estimated to operationalize the balance sheet approach and are compared to baseline predictive methods such as Error-Trend-Seasonality (ETS) models, Autoregressive Integrated Moving Average (ARIMA), and seasonal naïve methods. Results show that including balance sheet variables significantly improves the predictive ability of CIC models in terms of mean absolute percentage error (MAPE) and root mean squared scaled error (RMSSE). These findings hold across multiple training and test sets through time series cross-validation, suggesting stability of forecast accuracy results.