This article attempts to estimate demand for M2 money in Indonesia using time series non-stationary technique in 1997.1 - 2006.4. There are four methods are used in research, first, VAR estimation used to forecast model which have interaction of data time series. Second, function impulse response to see response from every variable to structural innovation of the other variables at the same time. Third, variance decomposition to know dissociating variation change of shock from each variable to other variables in model. Fourth method, ADL ECM to see long-range adjustment in variable, before and after addition of variable. The result, there are non-stationary condition in the time series data in the research. Result of VAR estimation show that there is no causality relation two ways among fifth of variable. From impulse, response known that response of M2 variable to other variable very fluctuative but finally the condition will return to stabilize.