Note: The goal of this post is to stimulate a conversation on how to model banks using economic theory. It may be impenetrable to those who are not already aficionados of economic theory.
In this post I am going to reinterpret a model of banking written by David Andolfatto that is available here. Before I reinterpret the model that Andolfatto presents, let me make some basic observations about the type of environment that is being studied here. First, this is a model of normal times: at present no effort is being made to incorporate crises or even the possibility of crises in the model. Second, there is a sense in which this is a model of short-term lending: all loans are one-period loans and no multi-period loans are considered. Indeed the model is structured so that there is no value to longer term lending.
Andolfatto recognizes that one of the cornerstones of heterodox theory is that banks create the money that they lend. When he introduces banks into his model, however, he ignores this principle and instead models banks according to the standard loanable funds approach as more “trustworthy” than non-banks. That is, he models θb > θ, where θ is a trust parameter. Effectively, he assumes the mainstream view that liquidity is a spectrum phenomenon and that banks just sit incrementally higher on the spectrum than other debt issuers.
I would argue that this framing fails to capture the idea that banks create money. When we say that banks “create money” what we mean is that banks issue liabilities that are generally accepted by the public. If I bring a $20,000 cashiers’ check issued by a bank to purchase a car – aside from confirming that the check is not a fake – just about any car dealership in the country will accept as if it were cash. In short, when we say that banks “create money,” we are saying that the trust parameter is so high that the banks’ liabilities are for practical purposes (in normal times) indistinguishable from fiat money issued by the government. For this reason, the assumption that is consistent with the claim that banks create money is not θb > θ, it is θb = infinity.
On the other hand, at the same time that banks can create money with ease, they are constrained because everybody expects them to give it back on demand. The car dealership accepts the large cashiers’ check, because it represents a promise to deliver the funds to the car dealership within a matter of days, if not faster. Thus, banks can create money and can borrow with extraordinary ease, but the loans are always short-term loans that the bank needs to be prepared to repay promptly.
In fact, Andolfatto presents his results under the assumption that θb = infinity, and he structures the model so that all loans are one-period, or short-term, loans. Thus, we can easily interpret Andolfatto’s model as a model of banks that create money. If we interpret Andolfatto’s model in this way, however, it’s not clear how to relate the model to either financial markets or non-banks.
Market-based lending does not function to finance working capital without bank credit and liquidity support (see, e.g., Stigum and Crescenzi 2007 pp. 976-77 on commercial paper), so if we are going to distinguish financial markets from banks we need to model them as long-term lending markets. Just as the short term assets sold on markets depend on bank guarantees, so do non-banks when they invest in these bank supported assets. Thus, non-bank lending, when it is being distinguished from bank lending, also needs to be modeled as long-term lending. Since there is only one-period, short-term debt in this model, there is no way to discuss market-based or non-bank lending as distinct from bank lending in this model.
This interpretation of the model is completely different from Andolfatto who claims:
“In the model, banks and financial markets are competing mechanisms for allocating credit. Banks are “special” only to the extent they are better than markets at funding investment. This specialness is not (in the model) logically rooted in their ability to create money. In particular, bond-finance in the model is “special” if it is the lower cost way to fund investment. Variations in the parameter that governs the willingness/ability of non-bank creditors to extend credit generates business cycles in the exact same way it would in a banking economy.”
But what Andolfatto has done is to reduce the statement that banks create money to a claim that banks can fund their loans ex nihilo: trust makes it possible for banks to finance working capital in this way. This framework underestimates what it means to say that banks create money, which I argue includes not just (i) the ability to fund loans ex nihilo, but also (ii) the “on demand” nature of the bank’s liability when it funds such loans. In short, there is a fundamental category distinction between bank obligations that are inherently monetary because they are payable at par “on demand” and non-bank obligations which do not have this property.
By modelling in detail only the investment financing side of the bank’s activities and not the monetary or “on demand” aspect of the bank’s liabilities, Andolfatto’s interpretation of his model abstracts from the concept of “money” itself. I would argue that the right way to bring the concept of money back into this model is to recognize that each period over which the bank is lending is fundamentally short, such as a week or a month. There is no evidence that capital markets can finance this type of activity without bank support.
In short, Andolfatto’s whole discussion assumes that “banks and financial markets are competing mechanisms for allocating credit,” and it assumes that it is appropriate to model “credit” as entirely homogeneous. In fact, “credit” is an overarching category that embraces more than one distinct form of lending. Bank credit, because it associated with the expansion of the money supply is categorically different from a bond issue, which does not increase the supply of “on demand” liabilities in the economy. Treating a 10 year bond obligation as substantially the same as a one-month advance of workers’ wages, because they are both “credit” fails to draw enough real-world distinctions about the nature of the financial system to be useful.
Thus, in my view if we are to treat the banking section of the Andolfatto model as a model of banking, then we must also recognize that it cannot at the same time be a model of financial markets. In order to introduce financial markets into the model, it will be necessary to introduce longer term debt.