Wednesday, June 26, 2019

Thinking About Policy Making


Larry Schembri’s discussion  of Canada’s experience with flexible exchange rates in the 1950s includes references to a couple of papers from the 1960s.  One is Robert Mundell’s (1963) Canadian Journal of Economics paper on the effectiveness of fiscal and monetary policy under fixed and flexible exchange rates.  Mundell’s paper is a classic in the literature (which means it’s cited by more people than have actually read it).  Another paper which Schembri cites is a 1964 Journal of Political Economy paper by Rudolf Rhomberg, “A Model of the Canadian Economy Under Fixed and Fluctuating Exchange Rates”
  
The Rhomberg paper was one which I hadn’t run into before, but it presents a nice counterpart to Mundell’s paper.


Mundell’s paper is a theoretical analysis of the effects of international capital flows on policy effectiveness under fixed and flexible exchange rates.  Its results are well known: with perfect capital mobility, fiscal policy is completely ineffective under flexible exchange rates and highly effective under fixed exchange rates, while monetary policy is highly effective under flexible exchange rates and ineffective under fixed exchange rates.  It’s an excellent exercise in pure theory: Mundell sets out an ISLM analysis (although he calls the goods market equilibrium curve the XX curve and the money market equilibrium curve the LL curve) to which he adds a balance of payments curve, FF.  He then goes through his now-familiar analysis of the policy cases he’s interested in and demonstrates his theoretical results.

The Rhomberg paper is an empirical one.  He sets out a 17 equation model of the Canadian economy, which he estimates on quarterly data from first quarter 1952 to fourth quarter 1959.  Basically, it’s a Klein-Goldberger type model for the Canadian economy, and its equations are basically the structural ones which underlie the ISLM-type analysis which Mundell uses.  A number of the equations are dynamic, meaning that lags of various explanatory and endogenous variables show up on the right hand side of some of them.

Once he’s got the econometric estimates of his model, Rhomberg moves on to analyzing it.  While it contains the equations which would permit him to find expressions (with numeric coefficient values) for Mundell’s XX, LL and FF curves, which would let him draw the open economy ISLM diagram relevant for Canada in the 1950s, that’s not what Rhomberg does.  What he does do is a type of analysis which was pretty common in the literature of the period, using small empirical macro models.

The first thing which he does is calculate what are termed impact multipliers, meaning multipliers which show the short run effect of a change in each explanatory variable on each of the endogenous variables.  Since he’s using quarterly data, the impact multipliers basically show how the endogenous variables respond to a change in each of the pre-determined variables in the quarter in which the change in the pre-determined variable occurs.  In other words, if government spending is increased, what can we expect see happen to national income during that quarter.  Remember that Rhomberg’s model is a dynamic one, which means that the effect of a one-time change in an explanatory variable can take several quarters to work itself out.

The second thing he does is calculate equilibrium multipliers.  Remember that we said that it would be possible to solve Rhomberg’s model for Mundell-type XX,LL and FF curves, the equilibrium curves for the goods, money and foreign exchange markets.   In his analysis, Mundell discusses the effects of an increase in government spending in a qualitative sense.  That means that he talks about the direction in which his IS counterpart, the XX curve, will shift, and in a general sense where the new equilibrium will be relative to the old one – whether the new equilibrium level of national income will be higher than the old one, for example.  What he can’t do, given the nature of his analysis, is talk about how much the equilibrium level of income will change by (except, of course, in the case where the new equilibrium is the same as the old one so there's no change at all).

What Rhomberg’s equilibrium multipliers show is how much, numerically, equilibrium national income will change in response to a given permanent change in an explanatory variable.  In other words, Rhomberg’s equilibrium multipliers put values to the shifts in the location of equilibria in Mundell’s diagrams.  Again, he doesn’t draw IS and LM curves, but his equilibrium multipliers put numbers on the comparative statics results in Mundell’s theoretical analysis.

Rhomberg then does something which Mundell does not (although it could be done on a theoretical ISLM diagram using phase arrows).  He makes the point that it’s not enough for a policy maker to know the size of the first step in the adjustment process, and to know where the process will eventually settle down.  It’s not unusual in economic models for the process by which the actual level of national income moves from its old equilibrium to its new one to involve over- and under-shooting.  In other words there may be transition periods in which the actual level of national income and/or the actual level of the interest rate might temporarily rise above or fall below the values at which they will eventually settle down.  That’s the cyclical part of what we term the business cycle.  Clearly, the policy maker needs to have a sense of the nature of the transition path that the economy is going to follow.

To get that, Rhomberg does a series of theoretical simulation exercises.  He starts by letting the economy settle into equilibrium with one, unchanging, set of values for the explanatory variables.  That’s relatively easy to do these days using something like Excel – it would have been more time consuming when Rhomberg wrote his paper.  Then he increases one variable – government spending, say – by a given amount and holds it at its new value from there on.  Given that change, he lets the model work its way through, numerically, to its new equilibrium, and graphs the resulting trajectories (something like this is still done in macro – these days it’s referred to as generating impulse response functions).  Among other things he’s able to establish that the Canadian economy in the 1950s did have tendencies to cyclical behavior, so that overshooting would have been quite reasonable to expect.

The type of numerical analysis Rhomberg did is very much what a policy maker would want to see when deciding how much to change one of the policy variables and when to do it.  So why is Mundell’s paper, which was an exercise in pure theory, the famous one and Rhomberg’s pretty much forgotten, especially since Mundell’s was published in The Canadian Journal of Economics and Political Science (as it was then – now it's The Canadian Journal of Economics) whereas Rhomberg’s was in the much more prestigious Journal of Political Economy.

Part of it is precisely because Mundell’s analysis was qualitative – a piece of pure theory.  His qualitative, comparative statics results didn’t depend on precise numerical values, they would apply to any economy facing perfectly elastic international capital flows.  And they were easy to understand – the ISLM apparatus reduces the outcome of the analysis to one graph in two endogenous variables, and even though all of Rhomberg’s equations are contained within Mundell’s XX, LL and FF curves, if what you’re interested in is the overall policy lesson, you can see it much more easily on a diagram that has three lines on it than in a representation in which all seventeen of Rhomberg’s equations are plotted.  Furthermore, Rhomberg's equations apply specifically to the Canadian economy and would have to be re-estimated for another country, whereas Mundell's results, because they're qualitative, are general.

It’s when you go from the theoretical story to actual policy making in the context of a particular country in a particular period that Rhomberg comes into his own.  At that stage you need both papers.  Mundell’s model tells you what you should be looking for; Rhomberg’s analysis lets you put magnitudes to it.  Mundell drew his diagram on the assumption that the FF curve was horizontal, Rhomberg discusses how to estimate the equations underlying that curve.  That means that, among other things, Rhomberg’s analysis lets you judge whether Mundell’s simplifying assumption, that capital flows pretty much instantly between the US and Canada in response to things like interest rate differentials, is a good representation of reality.  (These days advanced textbook presentations of Mundell's model typically discuss cases of less elastic international capital flows as well as Mundell's case of perfectly elastic flows.)

So when you’re actually making policy you need Mundell’s theoretical apparatus to tell you how to set up the analysis and Rhomberg’s equation building and estimation to put numbers on Mundell’s qualitative analysis.  Would-be policy makers need to know Mundell’s results, but they should really read the two papers in tandem if they want to see how to go from broad theoretical discussions to designing policy interventions for specific economies.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.