Taking Intermediation Seriously

Abstract

Taking Intermediation Seriously Bruce D. Smith (bio) Many modern approaches to macroeconomics attach no significance to financial intermediation. This is true despite the fact that various measures of banking activity are strongly correlated both with long-run real economic activity and with what happens during business cycles. Of course, the failure to attach significance to the macroeconomic consequences of financial intermediation could be rationalized if it had been shown that these correlations arose in a context where intermediary activity changed primarily in response to (or, in other words, was "caused by") changes in real activity. However, this interpretation of the facts does not seem justified for a variety of reasons. For example, in a large literature on financial intermediation and long-run growth, Cameron (1967), Goldsmith (1969), King and Levine (1993a, 1993b), Atje and Jovanovic (1993), Demirguc-Kunt and Levine (1996), and Levine, Loayza, and Beck (2000) have demonstrated that measures of private intermediary lending are strongly positively correlated with real long-run growth (or with the long-run level of real activity). Indeed, King and Levine argue that measures of financial intermediary activity are the only variables that bear a robustly significant relationship to long-run growth experience. Interestingly, Gurley and Shaw (1955) had argued at a very early point that growth in real activity promoted growth in financial activity and conversely—and that both real growth and financial development were endogenously and jointly determined. This seems like a useful conceptual framework. But, perhaps even more interestingly, the only formal empirical tests, of which I am aware, that attempt to discuss whether real growth affects financial intermediation causally and conversely are by Levine, Loayza, and Beck (2000). They argue that financial development causes growth, but that there is no empirical causality in the opposite direction. With respect to business cycles, as a very young economist influenced by real business cycle theory, I once asked Robert Lucas what the evidence was that monetary factors were important for business cycles. He cited Friedman and [End Page 1319] Schwartz (1963).1 Interestingly, Friedman and Schwartz provide very strong evidence that intermediary activity played an enormous role in business cycles in the U.S. between 1867 and World War II. In particular, all but one of the recessions they examine during that period are associated either with a banking crisis, and the 1937 recession is associated with an increase in the reserve-deposit ratio that was strongly responded to by banks. For most of the episodes examined, the strong monetary aspect of the business cycle is a large increase in the currency-deposit ratio. Lucas' argument seems to imply that this should be taken as an important aspect of at least dramatic business cycle episodes. And yet, Lucas and many others have never done this. The purpose of this paper is to discuss some frameworks for thinking seriously about how financial intermediation affects growth and about how banking crises affect major business cycle phenomena. As will be demonstrated, when this is done some fairly embarrassing results that arise in other macroeconomic contexts that ignore banking are easily overturned. One is that most monetary growth models tend to deliver either a Mundell-Tobin effect in which permanently higher inflation promotes long-run real activity, or a superneutrality result in which higher inflation does not affect real activity or real rates of return.2 In fact, there is strong evidence that— at least at high enough rates of inflation—inflation is significantly negatively correlated with long-run real activity.3 Another is the Friedman rule. In many contexts the Friedman rule (maintaining a zero nominal rate of interest) emerges as optimal.4 And yet, the major observed episodes of zero or nearly zero nominal rates of interest have occurred in places like the U.S. during the Great Depression or in Japan recently. No one regards allocations arising during these experiences as approximately optimal. As will be seen, when intermediation is analyzed seriously, the Friedman rule generally will not be optimal and the Mundell-Tobin effect or the superneutrality of money will not generally emerge. Of course, in order to make intermediation matter, it is necessary to consider environments in which the Modigliani-Miller theorem...