Thursday, June 6, 2019

So Who Should Be Running The Central Bank?


As Stephen Poloz’s term as Governor of the Bank of Canada nears its presumed end, articles have started to appear dealing with the question of who his successor should be.  The betting seems to favour senior Deputy Governor Carolyn Wilkins (and has done for a while) although a  number of other names are also mentioned. 

One issue which comes up occasionally is the background required of the Governor.  Before he went to the the Fed, Ben Bernanke had had a distinguished academic career, and while Janet Yellen was probably best known from the San Francisco Fed, she also had an academic background.  Jerome Powell, on the other hand, spent much of his career in financial markets.  Peter Diamond  was at one point a nominee for a position on the Fed’s board of governors but was opposed (in a decision most economists found hard to believe) on the grounds that his work wasn’t sufficiently money-macro.  Wilkins has a BA from Laurier and a Masters from Western, and has been a career civil servant.

Clearly the Governor of the Bank needs a very specific skill set.  They don’t necessarily have to be a top-level academic researcher, but they do need to be able to consider the recommendations made by the bank’s research staff and to explain decisions made on the basis of that research to politicians and, increasingly these days, to the press and public.  But there’s still a question of the type of analytic skills the governor should have.


As John Cochrane grumbled not too long ago, “Most macroeconomic policy analysis takes place within late 1970s era Keynesian ISLM framework that has not been seen in major peer-reviewed journals since that time.”  While a lot of the research which the Bank of Canada’s staff does is of a sort which would probably win Cochrane’s approval, things like the Monetary Policy Report suggest that the Bank sees more value in old fashioned ISLM, or something like it, than Cochrane does.

A lot of macro theorists share Cochrane’s unfavourable view of ISLM, but it works better than they’re willing to give it credit for.  Part of the issue comes down to the interpretation of the curves themselves.  The IS and LM curves are not behavioural relations, like the Consumption function, for example, rather, they’re equilibrium loci for the goods and money markets respectively.  So the points on the IS curve are combinations of income and the rate of interest at which the goods market will be in equilibrium.  But if that line shows all of the equilibrium points, then every other point on the graph must be a disequilibrium point.  That means that we can also use the diagram to analyze the behavior of economies which are not in equilibrium, (which, arguably, is most economies most of the time) and if we extend the ISLM framework to include equations specifying how income and the rate of interest will adjust when in disequilibrium, we can move from a static to a dynamic analysis of the macro economy. 

The basic ISLM structure is a quasi-closed economy model (quasi, because we generally include exports and imports in the IS curve), but we can extend it to a proper open economy (i.e. one with a financial account as well as a current account) by adding a BP curve.  That’s a locus of (r,Y) points at which inflows and outflows are exactly balanced, so there’s no tendency for the exchange rate to change.  With that addition we can look at whether the ISLM intersection point gives us an equilibrium which is consistent with exchange rate stability or whether the exchange rate is going to change over time, and having put that latter possibility into the model we can then talk about feedbacks between exchange rate changes and the IS and LM curves – if the exchange rate is changing, exports and imports will change so we expect the IS curve to shift.  (I’m partial to Gandolfo’s exposition of this kind of stuff. 

And we can go beyond that and add a government budget balance line, a locus of (r,Y) combinations at which the government’s budget will be in balance.  That one’s important because if the budget’s not in balance then either the stock of money or the supply of government bonds to the market must be changing.  If the government’s continuously running a deficit at the ISLM intersection equilibrium point, it must either be continuously offering new bonds to the market or, if the deficit is financed by printing money, the money supply must be changing.  Either of those options will cause the LM curve to shift.

So long as you accept the validity of the concept of equilibrium curves, then, we can develop what economic dynamics refers to as a phase diagram for the economy – a setting within which we can look at the static and dynamic – short and long run – consequences of policy changes, for example.
The standard ISLM model is a constant price model, but that’s easily enough extended if we add a supply-price curve.  Recognizing that what we’re looking at is an aggregate supply-price curve lets us make testable predictions incorporating both demand pull and cost push price factors.

A lot of the criticisms of ISLM are really criticisms of how it’s used, or of the construction of the behavioural curves (consumption function, investment function, money supply function) which go into the equilibrium loci.  If you’ve got a model of how expectations enter the investment function, you can build that model into the investment function and show how changes in expectations will cause the IS curve to shift.  If you want to incorporate forward-looking lifetime utility maximizing consumers, you can.  The critical question is whether that change will invalidate the notion of goods market equilibrium or simply change the way the IS curve shifts.

In the process of building up the model, of course, we’re adding a fair number of behavioural equations.  That used to be a major problem for modellers, back when seven or eight equations was a lot to manage.  Now, the computing power is available to make that kind of exercise pretty easy to undertake, and we’ve got  long enough data sets to allow us to do a decent job of estimating the behavioural equations, allowing for non-linearities and dynamic adjustment processes.  If we build it up equation by equation, and are careful about specifying the dynamic feedback processes, we can wind up with the kind of models which central banks have traditionally used for policy making.

And while John Cochrane may regard extended ISLM models as hopelessly outdated, you kind of hope that the reason central bankers still use them is not just because today’s central bankers are old fogies who should have retired years ago, but because they still hold up pretty well in the face of the data.  

No comments:

Post a Comment

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