The Liquidity Effect: Testing Identification Conditions Under Time-Varying Conditional Volatility

In the recent SVAR literature, the liquidity effect has been studied by imposing a variety of identifying restrictions required under the assumption that the SVAR fundamental disturbances are homoscedastic. Using typical SVAR processes, we first show that this assumption is not supported by the data and that the SVAR residuals are not characterized by common conditional scedastic structures. Under time-varying conditional volatility of residuals, we are then able to formally test typical sets of restrictions that have been imposed in previous studies. Our results indicate that to obtain a well characterized liquidity effect, one must measure monetary policy shocks as innovations in the Federal funds rate.