Recently, there has been considerable interest in identifying the exogenous policy actions of the Fed and a number of identification methods have been proposed. This paper deals with one of these, namely, using nonborrowed reserves in a recursive structural vector autoregression (VAR). A number of researchers [Christiano, Eichenbaum and Evans (1994ab, 1996, 1997), Evans and Marshall (1997), Strongin (1995), Pagan and Robertson (1995) and Brunner (1994) find evidence of a statistically significant liquidity effect using nonborrowed reserves in a VAR. The success in finding the liquidity effect with nonborrowed reserves in the VAR is attributed to innovations to nonborrowed reserves reflecting supply shocks while innovations to total reserves primarily reflect shocks to demand. The purpose of this paper is to demonstrate that the opposite is true. Evidence of the liquidity effect in recursive structural VARs depends critically on the existence of a negative covariance between the federal funds rate and nonborrowed reserves. Under a variety of operating objectives, the Trading Desk of the Federal Reserve Bank of New York has offset changes in bank-initiated discount window borrowing when implementing the Federal Open Market Committee's policy directive. This practice has created a negative contemporaneous covariance between nonborrowed reserves and the funds rate that has been incorrectly attributed to the liquidity effect. Once the Desk's practice is accounted for, there is no evidence of a statistically significant liquidity effect.