This study aims to test the stationarity of the velocity of money for Turkey and to determine the variables affecting the velocity of money. The annual data set covering the period from 1970 to 2017 contains the velocity of money, real income, deposit interest rate and real effective exchange rate variables. Although traditional and nonlinear unit root tests, variance ratio tests and fractional integration tests show that the velocity of money is not stationary, it is noteworthy that many of the structural breakpoint unit root tests' results support its stationarity taking into account the economic and structural transformations experienced in the 2000s. In addition, with the nonlinear autoregressive distributed lag model (NARDL), the short-term asymmetry of deposit interest rates and the long and short-term asymmetry of the real effective exchange rate are significant. The positive changes in deposit interest rates and real effective exchange rates over the long-term are more effective on the velocity of money than negative changes. The income, interest rate and exchange rate elasticity of money are respectively positive, negative and positive, and the results coincide with the studies in the literature.