Identifying the Liquidity Effect: the Case of Nonborrowed Reserves

Despite the fact that efforts to identify it empirically have largely been futile, the liquidity effect plays a central role in conventional monetary theory and policy. Recently, however, an increasing volume of empirical work [Christiano and Eichenbaum (1992a,b), Christiano,, Eichenbaum and Evans (1994a,b) and Strongin (1995)] has supported the existence of a statistically significant and economically important liquidity effect when nonborrowed reserves is used as the indicator of monetary policy. This paper shows that there is an identification problem associated with using nonborrowed reserves. Specifically, the strong negative relationship between nonborrowed reserves and the funds rate can stem from the presence or absence of a liquidity effect. The paper points out how changes in the demand for borrowed reserves can be used to identify whether the relationship between nonborrowed reserves and the funds rate is due to liquidity effect. The evidence presented suggests that the “liquidity effect” that Christiano, Eichenbaum and Evans and others have identified is actually due to the interest sensitivity of the demand for borrowed reserves and the definition linking nonborrowed and borrowed reserves. Consequently, the evidence suggests that the liquidity effect is nil. JEL CLASSIFICATION: E40, E52 Daniel L. Thornton Assistant Vice President Federal Reserve Bank of St. Louis 411 Locust St. Louis, MO 63102 I would like to thank Adrian Pagan and John Robertson for providing me with their computer programs and data. This paper was begun when the author was a visiting scholar at The Business School, City University, LONDON. The liquidity effect—the transient yet persistent declines in real and nominal short-term interest rates associated with unanticipated expansionary monetary policy shocks—plays a central role in conventional monetary theory and policy.1 Despite its prominent role, the liquidity effect has received scant empirical support [Cagan and Gandolfi (1969), Melvin (1983), Thornton (1988b), Reichenstein (1987) and Leeper and Gordon (1992)]. A number of analysts [Bernanke and Blinder (1992), Christiano and Eichenbaum (1991, 1992a,b) and Goodfriend (1991)] have argued, however, that the lack of empirical support is a manifestation of the Fed’s attachment to interest rate targeting in one form or another [Goodfriend (1991)]. They argue that innovations to monetary aggregates, such as Ml, the adjusted monetary base or total reserves, reflect shocks to money demand rather than to money supply. Consequently, the inability of researchers to isolate a statistically significant and economically relevant liquidity effect stems from their failure to correctly identify the exogenous policy actions of the Fed. Recently, Christiano and Eichenbaum (1991, 1992b), Christiano, Eichenbaum and Evans (1994a,b) and Strongin (1995) have argued that nonborrowed reserves reflect the exogenous policy actions of the Fed.2 Using nonborrowed reserves and Vector autoregression (VAR), they find a liquidity effect that is both statistically significant and economically important. Recently, however, Pagan and Robertson (1995) and Christiano (1995) have shov:n that the liquidity effect identified in this way vanishes after the early 1980s. This paper shows that there is an identification problem associated with using nonborrowed reserves as an indicator ofmonetary policy and shows how changes in the demand See Thornton (l988b), Reichenstein (1987) and Pagan and Robertson (1995), 2 Since Christiano and Eichenbaum use a VAR methodology, it is more precisely correct to say that they use exogenous shocks tononborrowed reserves. Most of their empirical work, however, is motivated by a simple statistical analysis of the relationship between nonborrowed reserves and the funds rate. I take the same liberty later and initially focus on the relationship between nonborrowed reserves and the funds rate. Later the analysis focuses on shocks to nonborrowed reserves using a VAR model similar to that ofChristiano and Eichenbaum (1991, 1 992b) and Christiano, Eichenbaum and Evans (I 994a,b). Identifying the Liquidity Effect: The Case ofNonborrowed Reserves Page 2 for borrowed reserves can be used to identify the liquidity effect using nonborrowed reserves. Specifically, it shows how the negative relationship between nonborrowed reserves and the federal funds rate may result from either the presence or al5sence of a liquidity effect.3 Furthermore, it shows how changes in the demand for borrowed reserves can be used to identify which of these alternatives account for the relationship between nonborrowed reserves and the funds rate. The evidence presented here suggests that the negative association between nonborrowed reserves and the funds rate that Thornton (l988a), Christiano and Eichenbaum (1991, 1992a,b), Christiano, Eichenbaum and Evans (l994a,b), Strongin (1995) and Pagan and Robertson (1995) report, stems from the absence, rather than the presence, of a statistically significant and economically important liquidity effect. Also, the evidence suggests that the liquidity effect did not change in the early 1 980s, as Pagan and Robertson (1995) and Christiano (1995) suggest, but rather that it never existed. I. Nonborrowed Reserves, Borrowing and the Liquidity Effect The liquidity effect associated with nonborrowed reserves is motivated by the market for reserves. The demand for reserves is derived from reserve requirements imposed on certain deposit liabilities of banks.4 Banks’ demand for such deposits is assumed to be inversely related to a short-term interest rate, i, representing the opportunity cost of holding such deposits. Hence, so too is banks’ demand for reserves. That is, 3 This point has also been made by Coleman. Gilles and Labadie (1995). 4 Note that the demand for something called reserves likely would arise endogenously ifreserve requirements were not imposed by the Fed. In this case, however, the nature of reserves and their relationship to bank liabilities might be quite different from those imposed by the central bank. Identifying the Liquidity Effect: The Case of Nonborrowed Reserves Page 3 (I) Rd = tf(i), f’ 0, where f(i) denotes the demand for reservable bank liabilities and ‘r denotes the marginal reserve requirement. For convenience assume that reserves are congruent with the monetary base and arise from two sources, the Federal Reserve’s holdings of government debt, Bf, and borrowed reserves, BR.5 Hence, (2) R5 = BR+B1. Borrowed reserves are supplied when the Fed makes loans to banks at the discount window. The demand for borrowed reserves depends on the spread between the federal funds rate, i~,and the discount rate, ~d’ and on other factors, ~3•6That is, (3) BR = + h(i1. — ‘d)’ h ‘ 0; f3 0. Equations 1, 2 and 3 are combined to obtain the reserve market equilibrium condition, (4) ‘rJ(i) = B1 + + h(i~. To close what can be thought of as the reserve block of the credit market it is necessary to have a relationship that links the federal funds rate and the short-term interest rate. For this SIn this model, changes in nonborrowed reserves are congruent with open market operations; however, this is not true in a model that allows for other sources ofreserves. ~Thedegree ofadministration of the discount window has changed over time. See Goodfriend (1983), Thornton (1986) and Cosimano and Sheehan (1994) for discussions ofthe discount window. Identifying the Liquidity Effect: The Case of Nonborrowed Reserves Page 4 purpose, we assume the following condition holds,