HUNTLEY SCHALLER (*)
By building on the Hamilton (1989) Markov switching model, we examine questions like: Does monetary policy have the same effect in expansions and recessions? Given that the economy is currently in a recession, does a fall in interest rates increase the probability of an expansion? Does monetary policy have an incremental effect on the growth rate within a given state, or does it only affect the economy if it is sufficiently strong to induce a state change (e.g., from recession to expansion)? As suggested by models with sticky prices or finance constraints, interest rate changes have larger effects during recessions. (JEL E52, E32)
Much of the recent work [in macroeconomics] has proceeded ... under the assumption that variables follow linear stochastic processes with constant coefficients. ... [As a result] some of the richness of the Burns-Mitchell analysis, such as its focus on asymmetries between recessions and expansions ... may well have been lost.
--Blanchard and Fischer (1989, 7)
For decades macroeconomists have debated whether monetary policy has the same effect on real output in expansions and recessions. As far back as the 1930s, Keynes and Pigou debated whether monetary policy would have less effect on output during a severe economic downturn. In the 1960s, there were active debates on a very different proposition, namely, whether the rightward portion of the aggregate supply curve was vertical, so that monetary policy would have less effect on real output during expansions. In this article, we provide a new type of evidence on whether monetary policy has different effects depending on whether the economy is in an expansion or recession.
Empirical evidence on this issue is particularly relevant in light of new theoretical work in macroeconomics that predicts asymmetric effects of demand shocks conditional on the state of the economy. Two examples of this work are S-s-type models of price adjustment and models in which there are agency costs of financial intermediation. (1)
The intuition for the latter class of models is simple. (2) When there is information asymmetry in financial markets, agents may behave as if they were constrained. For a variety of reasons, these finance constraints are more likely to bind during recessions when the net worth of agents is low. An increase in interest rates will then have two effects on investment: the standard effect of increasing the cost of capital and therefore reducing investment demand and an additional effect of reducing liquidity (e.g., by increasing debt service obligations) and thus reducing investment demand for constrained agents. As a result, monetary policy actions that change interest rates will have greater effects during a recession. (3) The S-s-type price adjustment models of Ball and Mankiw (1994), Caballero and Engel (1992), and Tsiddon (1993) lead to a convex aggregate supply curve and therefore also imply that monetary policy will have stronger effects during recessions.
One of the more frequently cited empirical papers on the potentially asymmetric effects of monetary policy is Cover (1992), which finds evidence that positive monetary shocks have different effects from negative monetary shocks. We are looking at a different type of asymmetry--namely, between booms and recessions.
We study asymmetries using an extension of the Markov switching model developed by Hamilton (1989), estimated over the period 1955-93. In Hamilton's econometric specification, the growth rate of output depends on a state variable that corresponds to an expansion or recession. This approach has several advantages. First, unlike linear projections, it allows for nonlinearities and asymmetries. Second, in estimating the recession coefficients, it gives greater relative weight to observations that most clearly correspond to recessions (and similarly for the expansion coefficients). …