Over the last few years, the Federal Reserve has conducted a series of large scale asset purchases. Given the Federal Reserve’s dual mandate, the objective of this policy has been to generate an increase in real economic activity while maintaining a low, stable rate of inflation. The effectiveness of large scale asset purchases and the ability of the central bank to achieve a particular target has been subject to debate. The monetary transmission mechanism is of primary importance for understanding the effects of both the recent large scale asset purchases and of monetary policy more generally. The purpose of this paper is to propose a monetary transmission mechanism and to present empirical support for this mechanism. In particular, this paper suggests that monetary policy is transmitted through changes in the growth rate of transaction assets through both a direct and indirect effect. First, an increase in the growth rate of the monetary base, whether through lump sum transfers or open market operations, generates a real balance effect that increases real economic activity. Second, the indirect effect is through bank lending. Since bank loans are often a function of nominal income, expansionary monetary policy increases bank lending. Since economics agents are forward-looking and the the effects of monetary policy are persistent, monetary policy is transmitted through the expected future time path of the growth of transaction assets and nominal income. This characteristic is especially important in light of the policy recommendations of Sumner (2011, 2012) and Woodford (2012), in which the central bank attaches an explicit target for the level of nominal income to large-scale asset purchases.1
This paper examines the validity of this transmission mechanism in two ways. First, the mechanism is formalized through an extension of the monetary search framework of Lagos and Wright (2005). The monetary search framework is modified in three important ways. For example, the types of financial assets are expand to include not only fiat currency, but also riskless government bonds and deposits. In addition, whereas economic agents move sequentially through different markets in the standard monetary search framework, the present model follows Williamson (2006) in assuming that economic agents enter one of two markets each period, that these markets operate simultaneously, and that agents move probabilistically across markets over time. Monetary injections take place in one market. This implies that only a subset of economic agents receive the monetary injection. As a result, this increases demand in this market, but the price level will not fully adjust to the monetary injection even though prices are Flexible. Monetary policy therefore has an effect on real economic activity. This setup therefore draws on the institutional fact that monetary policy is conducted through financial markets via open market operations. In this sense, the setup bears resemblance to the limited participation framework of Lucas (1990) and Fuerst (1992). In contrast with those frameworks, however, the effects of monetary policy in the present context are persistent.
Finally, the framework is extended to include financial intermediaries. The motivation for intermediation is derived from the fact that some agents experience idiosyncratic productivity shocks. Economic agents that wish to smooth consumption will therefore prefer to borrow before the productivity shock is realized. As outlined below, this assumption implies that agents that experience an adverse productivity shock might have an incentive to default on their loan. Banks therefore impose a borrowing constraint that is a function of the nominal asset holdings and expected nominal income of the economic agent. In the event of default, the bank pays a monitoring cost to seize the assets and output of the borrower.2 If coupled with the limited participation assumption outlined above, it can be shown that the collateral constraint is always binding. This is important because when this constraint is binding monetary policy affects banking lending through changes in expected nominal income.
Whether or not the proposed mechanism is operative requires empirical analysis. The validity of this transmission mechanism is tested using a vector autoregressive (VAR) model. Given the recent advocacy of nominal income targets, the empirical analysis focuses on the role of nominal income expectations in the transmission of monetary policy. In particular, the model predicts that with a binding borrowing constraint higher expected nominal income leads to more lending, an increase in the supply of private transaction assets, higher nominal income, and an increase in real economic activity. Impulse response functions are used to determine whether the dynamics observed in the data are consistent with the predictions of the theoretical model. The estimated IRFs provide support for this prediction.
In addition to the innovation accounting described above, the VAR is also used to construct counterfactual forecasts given different paths of nominal income expectations. Given the evidence of the influence of nominal income expectations from innovation accounting, it would be useful to evaluate the predictions of the model if the large scale asset purchases were tied to an explicit target for nominal income. The forecasts provide an indication of the path of transaction assets, nominal income, and real economic activity when monetary policy is successful at influencing nominal income expectations. This provides an additional mechanism through which to evaluate the theoretical model as well as a way in which to evaluate the claims made by Sumner (2011, 2012) and Woodford (2012) regarding the adoption of an explicit nominal income target. The forecasts are shown to be consistent with the predictions of the theoretical model.
1 More on this point below.
2 This assumption is similar to that of Carlstrom and Fuerst (1997).