The Rules vs. Discretion Paradox

Marcus Nunes posted the following report from the Cato Monetary Conference in a comment on Sumner's Money Illusion:

"One young man asked whether the adoption of a nominal GDP target would satisfy Mr Taylor’s desire for the Fed to be governed by rules rather than the whims of policymakers. Mr Taylor had no problem with steady nominal GDP growth as a goal of monetary policy but he did not see how a rule along the lines of “keep NGDP on its trend path” would be useful because it does not address how to achieve this objective. Expectations matter, he noted, but they are nothing without actions that justify those expectations. A policy rule is useless if it does not to relate to the instruments at the disposal of policymakers."

This points to a paradox regarding rules and discretion.    Are the consequences of a rule desirable, or at least, tolerable?   Market monetarists argue that nominal GDP level targeting has consequences are tolerable, and more to the point, better than the alternatives.   Unfortunately, the direct actions of a central bank can only impact nominal GDP--spending on output--by influencing the spending decisions of millions of households and firms.  

While this suggests that the direct actions of the central bank might not be able to influence the choices of those households and firms so that nominal GDP cannot be kept on target, perhaps more to the point, central banks have no confidence in their ability to keep nominal GDP on a target.    They prefer to commit to doing things that they are confident they can accomplish.

And what do central banks feel they can accomplish?  

First, what can they actually do directly?   They can purchase specific amounts of particular assets--what has come to be called quantitative easing.   Or, they can control the total amount of their liabilities--targeting base money. (Market monetarists, like other monetarists, tend to fixate on this)   They can also control the interest rates at which they make loans to banks--the discount rate in the U.S. (or primary and secondary credit rates.)    And they can directly control the interest rate they pay on the reserve balances that banks hold with them.

However, it is clear that some of those things that central banks can do, they have little interest in doing.   As all monetarists know, getting a central bank to make some kind of commitment regarding base money is nearly impossible.   While they have been willing to commit to purchasing certain types and amounts of financial assets, they don't seem very comfortable following that approach.

No, they like to manipulate supply and demand conditions in short term credit markets so that some kind of market interest rate is on target.  They keep short term interest rates steady, and then at periodic meetings, they decide to make adjustments.   In the US, the Fed uses the federal funds rate as a benchmark.   (Interestingly, when other short term interest rates failed to move with the federal funds rate in 2008, the Fed began all sorts of new interventions to directly impact other short term rates.)   Clearly, the Fed believes that it can "control" short term market interest rates, even though this control is indirect for anything other than the interest rate the Fed itself charges on the loans it makes or pays on the deposits it accepts.

Of course, keeping short term interest at some particular level is not a tolerable rule.    If interest rates (either nominal or real) are set too low, the result is a hyperinflationary disaster.    If, on the other hand, interest rates are set too high, the result would be an equally disastrous deflationary depression.

Central bankers appear to have learned that they must adjust the interest rates they "control" to prevent disaster.   Rising consumer prices and rising unemployment both upset voters, and so, indirectly, the politicians.   If consumer prices start rising to much, central banks have learned that they must raise interest rates and if unemployment starts to rise too much, they need to lower interest rates.

If we look at a period where inflation and unemployment both remained acceptably low, like the Great Moderation,  then we can see exactly how much the central bank adjusted interest rates during those period where inflation or unemployment began to rise.   We can then advise central bankers to continue with that approach.   If the "natural" unemployment rate  is steady over the period, then the changes in unemployment would reflect deviations of the unemployment from the natural rate.   So, the central banks actions can be characterized as responding to inflation and deviations of real GDP from potential GDP.

Keeping inflation at a low, steady rate, and real GDP close to potential is a tolerable rule.   Of course, central banks can only influence inflation and real GDP through the spending decisions of millions of households and firms--much like the level of nominal GDP.   (The primary difference between nominal GDP and inflation and unemployment is that voters don't care about nominal GDP statistics, but they do care about inflation and unemployment.)

What Taylor proposes is to take the observed relationship between the interest rates that a central bank "sets" and inflation and the real GDP gap (which is related to the unemployment rate that voters and politicians worry about,) and then tell central bankers to continue to adjust interest rates in that fashion.   Take what was their discretionary response to money market conditions constrained by the need to avoid excessive inflation and unemployment, and make that into a mechanical rule.

If the relationship between inflation, unemployment, and short term interest rates remain similar to those that held over the "good" period, then such an approach seems reasonable.   But what happens when conditions change?

In my view, the notion that economists can develop a simple formula relating something that central bankers are confident they can control and those things that it is at least tolerable to control, is hubris.   Sure, it appears that it is possible to find some useful rules of thumb that work in some times and and some places.   But reifying those rules of thumb into binding constraints on a central bank is an error.   (Of course, in practice, central banks have followed such rules just as long as they felt like it.   When they believe conditions have changed, they quit following them.)

In my view, the binding constraint on the central bank should be something that is tolerable--nominal GDP level targeting.   If central banks can develop rules of  thumb that relate something they are confident they can control--like short term interest rates--to the level of nominal GDP, then that is just fine.   But when the rule of thumb breaks down, the central bankers need to change what they are doing.

For example, when the rule mandates a lower short and safe interest rate, and it is already at zero, then the central bank needs to drop the rule of thumb and  seamlessly shift to raising the level of base money.   If the relationship between short and safe interest rates and other interest rates change, there should be no notion that the central bank needs to get all the other interest rates to move so that they have the "proper" (really the past) relationship with short and safe interest rates.   If the rule of thumb breaks down, then adjust the short and safe interest rates more than ordered by the broken rule.

To sum up, the paradox is that those things that it is sensible, or at least tolerable, for a central bank to control, the central banks have no confidence in their ability to control.   And those things that the central banks have confidence in their ability to control can lead to intolerable macroeconomic consequences.  It is to avoid those disasters, or at least intolerable macroeconomic results, that central banks need to be limited by rules.  

I think the answer is discretion subject to constraint.   Some way needs to be developed so that those actually issuing money can adjust the interest rates they pay and charge and/or the quantities of monetary liabilities they issue, subject to the constraint that the expected value of some relevant nominal quantity remains anchored.

I think the least bad nominal anchor is a steady growth path for nominal GDP.    While index futures targeting (or convertibility) is a promising approach that deserves more research, the least bad option today is to allow a central bank discretion to manipulate those things it directly controls subject to the constraint that it try to keep nominal GDP on target is the least bad approach.   In my view, the first step is for Congress to legislate the target growth path and then hold the central bankers accountable for hitting that target.    If they fail, then they can appear before Congress and explain why.   And they can also explain what they are doing differently to try to get nominal GDP to the target growth path.