This article attempts to estimate the demand for Indonesian behavior money in 1997.1 – 2002.4 using non stationary technique. A new asymptotic theory of regression is introduced for possibly non stationary time series. These techniques are less dependent Johansen’s maximum likelihood of co integration but more depend on the ordinary least squares (OLS) estimation of the equation included in the ECM. The dynamic OLS estimation proposed by Phillips &  Loretan in 1991 is used to estimate cointegration. Meanwhile, Vector Autoregresion (VAR) is used to forecast the model which has an interelation time series. Since it desirable to include national income and exchange rate as regressor in the money demand function. To estimate demand function in the short run is used autoregressive distributed lag ECM/ADL ECM, which known Hendry type ECM. The results have find that there are non stationary condition in the time series data in the research. Meanwhile, the estimation with VAR, DOLS and ADL ECM is suggested that volatility of exchange rate impact to demand for Indonesian money.