نرخ بانکی و بهره در تجزیه و تحلیل سیاست های پولی: یک اکتشاف کمی

The paper reconsiders the role of money and banking in monetary policy analysis by including a banking sector and money in an optimizing model otherwise of a standard type. The model is implemented quantitatively, with a calibration based on US data. It is reasonably successful in providing an endogenous explanation for substantial steady-state differentials between the interbank policy rate and (i) the collateralized loan rate, (ii) the uncollateralized loan rate, (iii) the T-bill rate, (iv) the net marginal product of capital, and (v) a pure intertemporal rate. We find a differential of over 3% p.a. between (iii) and (iv), thereby contributing to resolution of the equity premium puzzle. Dynamic impulse response functions imply pro- or counter-cyclical movements in an external finance premium that can be of quantitative significance. In addition, they suggest that a central bank that fails to recognize the distinction between interbank and other short rates could miss its appropriate settings by as much as 4% p.a. Also, shocks to banking productivity or collateral effectiveness call for large responses in the policy rate.

Recent years have seen great changes in monetary policy analysis, as economists in central banks and academia have come together on an analytical approach of the general type discussed by Rotemberg and Woodford (1997), Goodfriend and King (1997), Clarida et al. (1999), Woodford (2003), and many others. This approach is characterized, as argued by McCallum (2002), by investigations of alternative rules for monetary policy conducted in models that are based on private-agent optimizing behavior but with specifications that include features designed to lend empirical veracity, thereby aspiring to be structural and accordingly usable (in principle) for policy analysis. Despite a widespread belief that this approach is fundamentally sound, and that recent work represents a major improvement over the practice typical 15 or 20 years ago, there are some reasons for unease. Prominent among these are the absence from the standard framework of any significant role for monetary aggregates, financial intermediation, or distinctions among various short-term interest rates that play different roles in the transmission mechanism. A recent paper by Goodfriend (2005, p. 277) develops a qualitative framework designed to overcome these particular weaknesses. Specifically, it “… integrates broad money demand, loan production, asset pricing, and arbitrage between banking and asset markets” and illustrates the logical necessity (in principle) for monetary policy to take account of—among other things—the difference between the interbank rate of interest (used as the policy instrument) and other short rates including the government bond rate, the collateralized bank loan rate, the (nominal) net marginal product of capital, and a shadow nominal intertemporal rate—each of which differs from the others. As noted by Hess (2005), however, Goodfriend (2005) provides no evidence or argument concerning the quantitative importance of these features and distinctions. The primary objective of the present paper, accordingly, is to formulate a quantitative version of Goodfriend's model, develop a plausible calibration, and utilize this model to assess the magnitude and policy relevance of the effects and distinctions just mentioned for steady state interest rates and aggregate variables, and for dynamic monetary policy simulations. Among other things, the paper will investigate the role of the “external finance premium—EFP” that is emphasized in the prominent work of Bernanke et al. (1999). It will do so using a model in which the EFP is endogenously determined by no-arbitrage relationships in an environment in which loan production depends upon both collateral and loan-monitoring inputs, with capital serving less efficiently as collateral than bonds, while bank-deposit money is crucial for facilitating transactions. In this setting, the EFP may move either pro-cyclically or counter-cyclically in response to shocks, depending upon parameters of the model. How does the present paper compare with previous efforts to outline and quantify the role of financial intermediation (banking) in monetary policy? Probably the most prominent line of work of this type is that begun by Bernanke and Gertler, 1989 and Bernanke and Gertler, 1995 and continued by Bernanke et al. (1999), but the literature also includes notable contributions by Kiyotaki and Moore (1997), Carlstrom and Fuerst (1997), Kocherlakota (2000), Cooley et al. (2004), and others. It is apparently the case, however, that in all of these studies the models are fundamentally non-monetary—i.e., do not recognize the existence of a demand for money that serves to facilitate transactions.1 This omission could be of first-order importance for the financial accelerator, however, for its mechanism works via increases in the supply of collateral induced by asset price increases. In models with money, however, such increases also increase the demand for collateral as spenders go to the banking system for additional money to facilitate the additional spending induced by the initiating shock. Accordingly, our analysis focuses on the net effect of these offsetting forces. Our model's “banking accelerator” transmission effects work in much the same way as the financial accelerator does in existing models. For instance, monetary policy that stimulates employment and output in the presence of sticky prices raises the marginal product of capital, the price of capital, and the value of collateral in the economy, thereby tending to reduce the EFP for a given quantity of bank deposits demanded. However, our model includes in addition “banking attenuator” effects, which recognize that monetary stimulus to spending also increases the demand for bank deposits, thereby tending to raise the EFP for a given value of collateral-eligible assets in the economy. Also of some indirect relevance to the present exploration are the studies of Ireland (2004), McCallum (2001), and Woodford (2003, pp. 300–311), which seek to quantify the effects of neglecting monetary aggregates in model specifications that do not adopt the usual approximation that is necessary to keep monetary aggregate magnitudes from appearing in the intertemporal optimizing conditions of a standard policy-analysis model. These papers, all of which found only small effects of this approximation, did not include any explicit banking sector, however, and did not recognize distinctions among various short-term interest rates. In effect, ours is a two-sector model with a goods-producing sector and a banking sector. Goods are produced with capital and work effort as usual. The banking sector produces loans (and thus deposits) according to a production function with inputs of monitoring effort and collateral, the latter consisting of government bonds and capital (employed in the goods-producing sector) held by households. The distinctions among various interest rates arise in the model because loans and deposits are costly to produce, in the sense that they require work effort, while collateral services allow an economization of that effort. Hence, the total return on bonds (or capital) has an explicit pecuniary component and also an implicit liquidity service-yield component that reflects the collateral services that this asset provides. Portfolio balance requires risk-adjusted equality between the various assets’ total returns. Thus the sum of pecuniary and service-yield returns must equal a shadow total nominal risk-adjusted interest rate. However, the pecuniary bond rate is less than the net nominal marginal product of capital because bonds are more productive as collateral than capital. In equilibrium the interbank rate, which is the cost of loanable funds for a bank, is below the (uncollateralized) loan rate by the marginal cost of loan production. Finally, the loan market and the asset markets are linked by a no-arbitrage condition between the uncollateralized loan rate and the shadow total nominal rate.2 The strategy of the paper is twofold. First, it is to use observed average historical values of interest rates and rate spreads, together with observations on banking and macroeconomic aggregates, to calibrate parameters by reference to steady-state equilibrium values for the model with realistic trend productivity growth in the production of goods and loans. Our aim in this regard is to determine the extent to which the introduction of money and banking into an otherwise standard growth model can account for observable interest rate differentials, and how much money and banking matters quantitatively (on average) for aggregates like the capital stock, employment, and output. In this regard, we develop the implications of money and banking with reference to the famous equity premium puzzle of Mehra and Prescott (1985), summarized by Campbell (1999) as the puzzle of “why the average real stock return is so high in relation to the short-term interest rate”. The second part of our strategy is to linearize the model around the calibrated steady state to explore how much the inclusion of money and banking matters quantitatively for an otherwise standard “new neoclassical synthesis” (aka, “new Keynesian”, NNS) model of monetary policy. In particular, we wish to investigate quantitatively how much a central bank can be misled by relying on an NNS model without money and banking when managing its interbank-rate policy instrument. We begin by illustrating the presence of the two effects or mechanisms—the accelerator and the attenuator—by which money and banking influence the model economy's dynamics.3 Next, we illustrate the extent to which a central bank could misjudge its interbank rate response to a goods productivity shock by not taking money and banking into account. Finally, we consider shocks emanating from the banking sector itself—a shock to loan monitoring productivity and a shock to effective collateral that is meant to represent widespread financial distress. The paper's outline is as follows. In Section 2, we begin by building upon the specification of the Goodfriend (2005) model and highlighting some of its features. Next, in Section 3 we emphasize the various interest rates that appear in the model and the frictions that make them differ from each other. Then in 4, 5 and 6 we develop the steady-state solution that forms the basis for the linearized version of the model, which will be used in subsequent analysis, and discuss the steady-state calibration and some of the quantitative consequences of incorporating money and banking. In Section 7 we complete the specification and linearization of our dynamic model. Finally, in Section 8 we conduct various policy experiments to see how this model performs in comparison to more standard specifications without any banking sector. Conclusions are briefly outlined in Section 9.

We conclude with a brief summary of the paper's scope and findings. Our objective was to reconsider the role of money and banking in monetary policy analysis by means of an analytical framework that includes both a banking sector and transaction-facilitating money in an optimizing model that is otherwise of the standard new neoclassical synthesis type. The addition of the banking sector leads to several unusual features, including a number of distinct interest rates. The model is implemented quantitatively, based on a calibration that attempts (within specificational constraints) to be realistic for an economy such as that of the United States. Results obtained are of two types, pertaining to steady-state and dynamic properties of the model. In the former case we are reasonably successful in providing an endogenous explanation for substantial steady-state differentials between the short-term interbank interest rate, typically employed as the central bank's policy instrument, and the following: (i) the collateralized loan rate, (ii) the uncollateralized loan rate, (iii) the one-period government bond rate, (iv) the net marginal product of capital, and (v) a shadow nominal pure intertemporal rate, i.e., our “benchmark” rate. One steady-state experiment involves a counterfactual calibration that makes banking services almost costless to produce; this scenario results in interbank and bond rates that are 4% points per annum higher than with our baseline calibration, thereby indicating a major quantitative effect of a banking sector. Effects on the steady-state capital stock are also sizeable; with costless banking services, the capital stock is over 5% lower than in the baseline calibration. Among other results, in the baseline calibration we find a differential of over 3% points per annum between short-term interest rates and the return to capital, a magnitude that may contribute significantly to resolving the famous “equity premium puzzle”. Finally, we report experiments with the baseline calibration that indicate a sizable sensitivity of the steady-state “neutral” interbank rate to the debt to GDP ratio and to the velocity of aggregate bank deposits. Dynamic results are based on impulse response functions implied by a linearized version of the model. Here we demonstrate the quantitative significance of a “banking attenuator” effect that works in the opposite direction from the “financial accelerator” effects emphasized by Bernanke et al. (1999)—although the latter effects are also present in our model. One of the more significant findings in this dynamic context is that, according to the model, a central bank that based its rate-setting policy on analysis that failed to recognize the distinction between the interbank rate and the benchmark intertemporal rate could miss its appropriate settings by as much as 4% points. For instance, we show that a central bank that utilized an interbank interest rate instrument, with parameters chosen to represent a moderate version of the Taylor (1993) rule, would produce a persistent 2% per annum deflation in response to a 1%, highly persistent positive shock to productivity in goods production. Finally, we demonstrate the quantitative consequences of two shocks emanating from the banking sector itself: a shock to banking productivity and a shock to effective collateral reflecting financial distress. For instance, we indicate that the central bank would need to cut the real interbank rate initially by nearly 5% points per annum to fully stabilize inflation and output against a moderately persistent, 1% decline in effective collateral. Moreover, we show that a more realistic policy response reflecting instead a Taylor-style rule with interest rate smoothing fails to offset the contractionary consequences of financial distress, instead permitting a recession with 2% per annum deflation and a 2% contraction in employment. In short, in our calibrated model, the effects of money and banking are quantitatively of major importance.