Uneasy Money

I have been giving John Taylor a lot of attention lately (see here, here, here, and here). I should probably lay off, but I think there is an important point to be made, and I am going to try one more time to get to the bottom of what I find disturbing about Taylor’s advocacy of rule-like behavior by central bankers. Some of what I want to say has already been said by Nick Rowe in his excellent post responding to Taylor’s criticism of NGDP targeting, but I want to address more directly Taylor’s actual statements than Nick did.

Taylor argues that targeting objectives, like NGDP, is not a policy rule at all, but rather a way of giving the central bank the discretion to do what it wants under the guise of what purports to be a policy rule that isn’t a rule at all. Here is how Taylor put it.

NGDP targeting may seem like a policy rule, but it does not give much quantitative operational guidance about what the central bank should do with the instruments. It is highly discretionary. Like the wolf dressed up as a sheep, it is discretion in rules clothing.

For this reason, as Amity Shlaes argues in her recent Bloomberg piece, NGDP targeting is not the kind of policy that Milton Friedman would advocate. In Capitalism and Freedom, he argued that this type of targeting procedure is stated in terms of “objectives that the monetary authorities do not have the clear and direct power to achieve by their own actions.” That is why he preferred instrument rules like keeping constant the growth rate of the money supply. It is also why I have preferred instrument rules, either for the money supply, or for the short term interest rate.

As I pointed out in my previous post, Friedman’s proposed rule for money supply growth is an exceedingly inapt example of an instrument rule, because the Fed has no more control over the money supply than it does over NGDP, as Friedman himself belatedly had to acknowledge. Now it is true that the Fed can control the short-term interest rate, so it is an instrument. The Taylor rule, a recipe for the short-term interest rate, specified in terms of the difference between the actual and the target rates of inflation and the difference between actual and potential GDP, does qualify as an instrument rule (on Taylor’s understanding of the term). But, as Nick Rowe points out, potential GDP being unobservable, it must be estimated. The higher potential GDP, the lower the implied short-term interest rate. So even the Taylor rule does not preclude an exercise of discretion by the Fed.

Perhaps the failure of the Taylor rule to preclude any exercise of discretion is why Taylor in the article by Amity Shlaes to which he refers seems to be offering a pared down version of the Taylor rule. Shlaes writes:

In response to my query about NGDP, Taylor sent a description of the reform he seeks — not widening the Fed’s growth mandate, but rather removing it [my emphasis]. Taylor says he would like to see reform happen in this order: 1) Congress enacts a single mandate for price stability; 2) Congress enacts reporting requirements for the Fed on what its strategy or policy rule is; and 3) the Fed picks a strategy relating to money and interest rates and tells the public what that strategy is.

One can’t really be sure what this means, but wouldn’t enacting “a single mandate for price stability” require the Fed to base its choice of the short-term interest rate solely on the difference between the actual and the target rates of inflation regardless of the difference between actual and potential GDP? After all, allowing the Fed to take into consideration whether actual GDP is less than potential increases the scope for the Fed to exercise discretion.

But there is something else disturbing about Taylor’s fixation on rules. I have been writing about the long-running debate about rules versus discretion in monetary policy on this blog for a while, especially the past couple of weeks, but failing to identify the critical semantic confusion that infects much of the rules versus discretion debate, and especially Taylor’s pronouncements. It was not until I read the following comment by W. Peden on my post Rules v. Discretion that the point suddenly became clear to me.

It’s a shame that there isn’t a widely used term in economics that means something like “precise orders” to distinguish from our usual use of “rule”, which means something like a restriction on possible action. A rule does not give exact and invariable instructions.

So, for example, the Taylor rule is really a kind of varying imperative, since it gives precise instructions on the short-term interest rate. In contrast, targeting NGDP along a trend is a rule because it allows for a huge variety of actions within that period; in a once-in-a-lifetime financial crisis like the early 1930s or 2008-2009, it would require quite radical discretionary actions so that the trend could be maintained.

Peden is completely right to say that the Taylor rule is not a rule at all, it is a command to a policy maker to adopt a policy of a certain kind. But policies are not rules. Rules do not prescribe specific actions they impose certain constraints on the manner in which agents can take actions in the pursuit of goals that they, not the author of the rules, have chosen. Introducing the language of rules into a discussion of policy reflects an effort (either conscious or unconscious) to borrow the authority of the political ideal of the rule of law as a support for the particular policy being advocated. The point was obliquely recognized by F. A. Hayek in a passage from the Constitution of Liberty that I quoted in my previous post. And it bears repeating.

It should perhaps be explicitly stated that the case against discretion in monetary policy is not quite the same as that against discretion in the use of the coercive powers of government. Even if the control of money is in the hands of a monopoly, its exercise does not necessarily involve coercion of private individuals.

The political ideal of the rule of law does bear on the use of coercive powers of government, because the political ideal of the rule of law (sometimes called substantive due process by Constitutional lawyers) is meant to constrain the government in exercising coercive powers. But that political ideal has nothing to do with the formulation of policy when (e.g., in the case of setting monetary policy) it involves no exercise of coercion.

So, notwithstanding Friedman’s assertion in Capitalism and Freedom, endorsed by Taylor, there is no principled presumption in favor of formulating policies in terms of specific commands requiring the monetary authorities to set instruments under their direct control according to a recipe or formula defined in terms of an arithmetic formula. The notion that there is any political principle requiring a policy supposed to achieve some desired objective to be so formulated is based on a semantic confusion between rules and policies and on a complete misunderstanding of the political principle requiring governments to follow rules in exercising their coercive powers.

It is still conceivable that monetary policy in terms of a recipe for an instrument of monetary policy might lead to the best possible outcome. It is also conceivable that flying an airplane on automatic pilot would lead to a better outcome than having a trained pilot fly the plane. Indeed, that could be true under some circumstances, but it verges on the preposterous to suppose that it would never be desirable (or indeed imperative) for a live pilot to override the automatic pilot. But that suggests that ultimately policy ought to be formulated in terms of the objectives sought rather than in terms of what is no more than a recipe for an instrument by which the policy objective is to be pursued.

In my previous post about George Selgin’s comment on John Taylor’s critique of NGDP targeting, I observed that Selgin had correctly focused on Taylor’s ambiguous use of the concept of an “inflation shock” without identifying the nature of the shock (violating Scott Sumner’s dictum “never reason from a price change). As Selgin pointed out, if the inflation shock were caused by a shock to aggregate supply, NGDP targeting would do better than a price-level or inflation rule. If the source of the inflation shock were excessive aggregate demand, well, that just means that NGDP had not been targeted. But there was another part of Taylor’s post, not addressed by Selgin, deserving of attention.  Taylor writes:

NGDP targeting may seem like a policy rule, but it does not give much quantitative operational guidance about what the central bank should do with the instruments. It is highly discretionary. Like the wolf dressed up as a sheep, it is discretion in rules clothing.

Taylor goes on to elaborate, invoking the authority of Milton Friedman, on the properties that a rule governing the conduct of a central bank ought to have.

For this reason, as Amity Shlaes argues in her recent Bloomberg piece, [coincidentally mentioning Taylor six times and quoting him twice!] NGDP targeting is not the kind of policy that Milton Friedman would advocate. In Capitalism and Freedom, he argued that this type of targeting procedure is stated in terms of “objectives that the monetary authorities do not have the clear and direct power to achieve by their own actions.” That is why he preferred instrument rules like keeping constant the growth rate of the money supply. It is also why I have preferred instrument rules, either for the money supply, or for the short term interest rate.

The first point to make about this remark is that the money supply rule Friedman advocated, and Taylor endorses — despite its unworkability in theory (Goodhart’s Law) and its demonstrated unworkability in practice when adopted by the Fed under Paul Volcker in the early 1980s — did not – obviously did not! — satisfy Friedman’s own criterion of being stated in terms of “objectives that the monetary authorities . . . have the clear and direct power to achieve by their own actions.” The monetary aggregate that Friedman wanted to grow at a fixed rate was, for the most part, produced by private banks, not by the Fed, so there was obviously no way that the Fed could achieve its objectives for the growth of M1, M2, or Mwhatever by its own actions.

So what should one conclude from this? That Friedman was a hypocrite? I don’t think so. But he did have a propensity for getting carried away by his enthusiasms, and he took his enthusiasm for rules to an extreme, supposing that all problems of monetary policy could be solved by prescribing a rule for a fixed rate of growth in the money supply. Even in 1960, that was a remarkably simplistic, actually simple-minded, position to have taken, but in his obsession for simple rules, he thought he had found the Holy Grail of monetary policy. That John Taylor, half a century later, could approvingly cite Friedman’s rule for the rate of growth in the money supply as a benchmark by which to judge other monetary policy rules speaks volumes about Taylor’s grasp of what constitutes good monetary policy.

Actually to gain some obviously needed insight into monetary policy rules, Professor Taylor could do a lot worse than to start with Chapter 21 of Hayek’s Constitution of Liberty. Friedman could have profited from reading it as well, but Friedman, obviously disdaining Hayek’s abilities as an economist, probably did not take it seriously. Let’s have a look at some of what Hayek had to say on the subject of rules and monetary policy.

The case for “rules versus authorities in monetary policy” has been persuasively argued by the late Henry Simons [one of Friedman’s teachers at Chicago] in a well-known essay. The arguments advanced there in favor of strict rules are so strong that the issue is now largely one of how far it is practically possible to tie down monetary authority by appropriate rules. It may still be true that if there were full agreement as to what monetary policy ought to aim for, an independent monetary authority, fully protected against political pressure and free to achieve the ends it has been assigned, might be the best arrangement. . . . But the fact that the responsibility for monetary policy today inevitably rests in part with agencies whose main concern is with government finance probably strengthens the case against allowing much discretion and for making decisions on monetary policy as predictable as possible.

It should perhaps be explicitly stated that the case against discretion in monetary policy is not quite the same as that against discretion in the use of the coercive powers of government. Even if the control of money is in the hands of a monopoly, its exercise does not necessarily involve coercion of private individuals. The argument against discretion in monetary policy rests on the view that monetary policy and its effects should be as predictable as possible. The validity of the argument depends, therefore, on whether we can devise an automatic mechanism which will make the effective supply of money change in a more predictable and less disturbing manner than will any discretionary measures likely to be adopted. The answer is not certain. No automatic mechanism is known which will make the total supply of money adapt itself exactly as we would wish, and the most we can say in favor of any mechanism (or action determined by rigid rules) is that it is doubtful whether in practice any deliberate control would be better. The reason for this doubt is partly that the conditions in which monetary authorities have to make their decisions are usually not favorable to the prevailing of long views, partly that we are not too certain what they should do in particular circumstances and that, therefore, uncertainty about what they will do is necessarily greater when they do not act according to fixed rules. (p. 334)

And a bit later:

There is one basic dilemma, which all central banks face, which makes it inevitable that their policy must involve much discretion. A central bank can exercise only an indirect and therefore limited control over all the circulating media. Its power is based chiefly on the threat of not supplying cash when it is needed. Yet at the same time it is considered to be its duty never to refuse to supply this case at a price when needed. It is this problem, rather than the general effects of policy on prices or the value of money, that necessarily preoccupies the central banker in his day-to-day actions. It is a task which makes it necessary for the central bank constantly to forestall or counteract development in the realm of credit, for which no simple rules can provide sufficient guidance.

The same is nearly as true of the measure intended to affect prices and employment. They must be directed more at forestalling changes before they occur than at correcting them after they have occurred. If a central bank always waited until rule or mechanism forced it to take action, the resulting fluctuations would be much greater than they need be. . . .

[U]nder present conditions we have little choice but to limit monetary policy by prescribing its goals rather than its specific actions. The concrete issue today is whether it ought to kepp stable some level of employment or some level of prices. Reasonably interpreted and with due allowance made for the inevitability of minor fluctuations around a given level, these two aims are not necessarily in conflict, provided that the requirements for monetary stability are given first place and the rest of economic policy is adapted to them. A conflict arises, however, if “full employment” is made the chief objective and this is interpreted, as it sometimes is, as that maximum of employment which can be produced by monetary means in the short run. That way lies progressive inflation. (pp. 336-37)

Professor Taylor, forget Friedman, and study Hayek.