Currency in circulation (CIC) is an important variable in monetary policymaking as it affects overall liquidity in the economy and guides the currency issuance operations of central banks. Given the rich information that may be gleaned from CIC, it is in the best interest of monetary authorities, the BSP included, to continuously track and accurately forecast CIC.
This paper proposes an alternative way to model and forecast CIC based on the balance sheet of the central bank. In this framework, an expansion of the BSP’s assets would be offset by an equivalent increase in the BSP’s reserve money liability. For example, foreign exchange inflows, when exchanged into domestic currency, expand overall liquidity in the system. To the extent that expansion affects the inflation target, the excess liquidity is mopped-up through open market operations. The unsterilized portion of the expansion in domestic liquidity are either kept as deposits in banks or withdrawn as cash, thereby, increasing CIC.
The stylized assets and liabilities of the BSP are employed to forecast the levels of CIC using time series models. This approach is nascent and novel as it departs from the existing literature in currency demand forecasting, anchored on the demand-for-cash framework, which treats transaction motives, precautionary motives, and opportunity costs in holding cash as primary factors in predicting CIC.
Several models are estimated, with mean absolute percentage error (MAPE) ranging between one to two percent for in-sample fit. For out-of-sample forecast errors, the MAPE ranges between one to two percent up to twelve months ahead, suggesting promising predictive ability. Furthermore, forecast averaging methods such as the simple mean, mean square error weighting, and least squares weighting produce superior forecasts compared to the baseline models. These CIC models may complement the BSP toolkit in forecasting and analyzing CIC, which may help inform currency policy (and overall monetary policy) of the BSP.