Bulletin of Monetary Economics and Banking

Document Type



This paper examines the demand for currency and quasi-money in Indonesia with linear and neural network models. The goal is to predict better the recent financial distress, reflected by the flight into currency and decline of quasi-money.The results show that neural network approaches, much more than linear models, are capable of accurate out-of-sample predictions for both monetary aggregates. However, for the very turbulent period of November and December of 1997, even the neural network models show large out-of-sample forecast errors. When a proxy for exchange-rate uncertainty supplements the network models, the out-of-sample currency demand becomes quite accurate, even for the last month of 1997. The quasi-money demand forecast also improve, although not as dramatically as those of currency demand. The analysis shows that a credible program, which reduces uncertainty in exchange-rate expectations, may mitigate the flight into currency from broad money, and the ensuing demonetization of the financial sector.

First Page


Last Page