Most current empirical work finds no evidence that money shocks lower interest rates. We show that these nonresults are mainly due to a failure to model the conditional heteroskedasticity of interest rates. Autoregressive conditional heteroskedasticity (ARCH) models find a significant liquidity effect where ordinary least squares (OLS) models do not. The existence of a liquidity effect is found using different models and sample periods when ARCH models are used in estimation, but never when OLS is employed.