Grasping Reality on TypePad, by Brad DeLong

Evan Soltas:

Larry Summers seems to have a blank slate when it comes to monetary policy. He hasn't written or said much that reveal his own views. But with an assist from Tyler Cowen, I've found that Summers actually did say quite a lot about monetary policy.

Back in 1991.

He participated in a panel discussion for the Journal of Money, Credit and Banking--a top academic publication for monetary economists--on the long-run goals a central bank should have. Since it is gated, I will summarize it. Here are his top 3 quotes:

1. He's practically in favor of NGDP targeting. "What should be the long-term objectives of the monetary authority?… What the monetary authority surely can control in the long run is the growth rate of nominal income."

2. He doesn't like monetary rules, specifically John B. Taylor's or Milton Friedman's. "[I]nstitutions do the work of rules, and monetary rules should be avoided…. Unless it can be demonstrated that the political institutional route to low inflation--to commitment that preserves the discretion to deal with unexpected contingencies and multiple equilibria--is undesirable or cannot work, I don't see any case at all for monetary rules."

3. He is dovish on inflation. "[T]he optimal inflation rate is surely positive, perhaps as high as 2 or 3 percent…. I would support having someone in charge of monetary policy who is more inflation averse than I." His arguments anticipate one Janet Yellen's in 1996.

The full piece:

THESE COMMENTS ADDRESS THE QUESTION of monetary strategy as distinct from monetary tactics. They are not directed at the question of what monetary policy is appropriate at the present time, or of how monetary policy should be determined at any specific point as a function of prevailing conditions. Instead,the issue I take is one tha thas been of increasing interest to economists in recent years: What should be the long-term objectives of the monetary authority.

Increased interest in this issue as opposed to the issue of fine-tuning the economy reflects the general intellectual consensus that has emerged on two points. First, what the monetary authority surely can control in the long run is the growth rate of nominal income. Second, issues of dynamic consistency and commitment are crucial to the design of appropriate monetary policies. Beyond its recognition of these points, my view is that most of what the academic economics community has had to say on the issue of long-run monetary strategy has been tangential at best, and I fear that many of the papers at this conference have not done much to improve the situation.

I wish to advance four propositions about monetary policy and to discuss their implications. First, an optimal tax theory has little or nothing to do with sensible inflation policy. Second, positive rates of inflation are almost certainly desirable. Third, to an important extent, institutions do the work of rules, and monetary rules should be avoided; instead,i nstitutions should be drafted to solve time-inconsistency problems. Fourth, H. J. Res. 409, the proposed zero-inflation amendment, is a good idea if it is not taken too seriously or too literally, but is instead viewed as a device for strengthening the independence of the Federal Reserve System.

My first proposition is that the connection between optimal tax theory and inflation policy is tenuous. It is no accident that the world has many more monetary economists than it has taxation-of-fringe-benefits economists, or that it has many more monetary economists than it has corporate tax specialists. The stakes in monetary policy are much more important than the stakes in employee fringe benefits, even though any sensible estimate of the deadweight loss from not taxing benefits would dwarf the welfare consequences of taxing money balances through inflation.

Optimal tax theory has little to do with optimal monetary policy. Tax revenues from seignorage in the United States were just 0.2 percent of GNP in 1989 and have never exceeded 0.5 percent of GNP in peacetime. Estimates of the area under the money-demand triangle vary, but the 0.5 percent of GNP estimate that could be formed as the product of the money stock and the nominal interest rate is surely a gross overestimate, and the magnitude of this deadweight loss is small relative to any other real consequence of monetary policy.

Furthermore, the standard cash-in-advance or money-in-the-utility-function model that is used when the optimal tax theory is applied is almost completely irrelevant. Outside money, which is at issue in this analysis, is made up of two components. The distortion from taxing th efirst component, required reserves at banks, can be avoided simply by paying interest on bank reserves. There is no reason for reserves to be taxed as a consequence of inflation.

The second component of the outside money stock is currency,which is supplied in the United States at almost $1,000 per every man, woman, and child. More than half of the currency is made up of money that is held in denominations greater than $100. By almost anyone's guess, a very large fraction of that $100-and-up money is held by people who don't want their transactions to be visible. In fact, the inflation tax is primarily a burden on people who are holding illicitly received assets. I can't imagine a sillier aspect of the monetary policy problem than worrying about the Harberger triangle associated with the taxation of those transactions.

I think the view that inflation is costly is correct, but it has nothing to do with optimal tax theory. It has instead to do with other considerations that are much more important, even if they are more difficult to model. First, standards should remain standard. Imagine the following conceptual experiment: Depreciate the pound--not the British currency, but the American weight measure--by 10 percent each year so that what is a pound today would be 14.4 ounces next year and 12.96 ounces two years from now, and so forth. Some might contend that allowing the pound to lose value steadily wouldn't make a difference. Only real things matter,and you could just redenominate the units and all the contracts could be written to be time contingent. But it would be costly to get organized to write those contracts. If I promised delivery of a pound of sugar, it would benefit me to delay as long as I could: the longer I waited, the less sugar I would have to deliver. That's exactly what happens with all payments when there is inflation. Consider cash management services at banks. One of these services provides faster depositing of checks. Think about the social gains versus the private gains that are producedw hen checks are deposited faster. The private gains are significant:you earn more interest. Of course, the person who wrote the check earns less interest.As a first approximation,the social gain is zero. It is a form of rent-seeking.The higher the rate of inflation,the more rent-seeking will be observed. The more that one party seeks to delay paying bills, the more the other party will try to accelerate the payment of bills. This is all a waste of resources that can be avoided if there is little or no inflation.

I do not know how to estimate the size of rent-seeking caused by inflation as firms and individuals try to accelerate collections and delay payments. I suspect this cost may be large. The gross amount outstanding of receivables and payables must be several times the GNP. Certainly experience with very high inflation in Latin America suggests that the financial sector expands at a very rapid rate.

A second issue, which is even more difficult to model, is the issue of stability,the perception that significant rates of inflation are likely to accelerate, that they cannot accelerate forever, and that when they are brought down, the ultimate recession will cost more in output than was gained in the process of going up. That argument is important to consider if you want to think seriously about whether we should have a higher or lower rate of inflation.

If James Tobin was right that it takes a heap of Harberger triangles to fill an Okun gap, it must take at least a heap-squared of Friedman money-demand triangles to fill an Okun gap associated with a recession. A valid case for low inflation must have to do with the inefficiencies caused by allowing the monetary standard to vary and by the instability that results when the inflation trend is changed. Standard optimal tax issues along Ramsey lines are nth-order considerations. Inflation as a Ramsey tax may be the most overstudied issue in macroeconomics.

My second proposition is that the optimal inflation rate is surely positive, perhaps as high as 2 or 3 percent. On one hand,the losses from low positive rates of inflation are likely to be small. The Harberger triangle is negligibly small. At a 3 percent inflationr ate, very few people are going to rush to deposit their checks in order to accumulate a little more interest. I don't see evidence that instability results at low rates of inflation.

On the other hand, the benefit of a positive rate of inflation comes in three places. Thefirst is the avoidance of the zero interest rate trap. The real interest rate in the United States has been negative in about a third of the years since World War II. The real after-tax interest rate, the rate at which corporations, for example, can borrow, has been negative in about three-quarters of the years since World War II. That couldn't happenif we had a zero rate of inflation. The nominali nterest rate cannot be negative. Negative interest rates may be a bad thing and our historical experience may be just an aberration, but perhaps negative real rates are a consequence of the risk associated with risky assets and the return on safe assets. If so, then an option would be lost to the economy if the rate of inflation were zero. I don't need to argue that the real after-tax rate of interest should always, or even most of the time, be negative. But if the real rate of interest should be negative only at certain times, we forgo that opportunity under a zero inflation rate. Potentially, rationing is much more serious than the triangles that are associated with a small tax distortion. If the equilibrium real rate of interest should be negative 1 percent and instead is 0 percent, the demand for capital would be significantly smaller than the supply.The potential consequences for economic stability would be severe at least enough of a risk that one would not want zero inflation.

The preceding argument did not assume that people would be irrational or suffer from money illusion. The next one does, but I think that it's a plausible view. That is, zero inflation in an economy that grows as slowly as the American economy now grows would mean either significant nominal wage cuts or unemployment. For example, the real wage of college professors fell by 30 percent between 1970 and 1980. Had we been in a 2 percent inflation era, real wages would have fallen by only 20 percent. What would have happened? Scholars over age 35 probably would have kept their jobs. I suspect that those under 35 would have fallen victim to lower staffing requirements if real wages had not been able to fall. Would nominal wages have started to fall if the inflation rate had been low? Perhaps, but not as easily as in an inflationary environment. Certainly the standard tenure provision in university contracts draws sharp distinction between nominal declines and real declines. Eventually that could get itself renegotiated,but it would take a long time.

Of course, this is not just an isolated feature of the professors' market. Men in their twenties with no college education, or who dropped out of high school, experienced real wage declines of 22 percent between 1980 and 1989. This says something about the performance of the real economy during that period, but for our purposes the experience raises a different question. Would those wages have fallen by 2 percent a year in nominal terms if inflation had been zero? WouldCongress have legislated the nominal reduction in the minimum wage that would have been necessary for wages to have fallen by 2 percent a year during the 1980s? I severely doubt it. Instead, I think there would have been a significant increase in unemployment.

It might be argued that the American economy performed well during the l950s when inflation rates were very low, and so negligible inflation rates can be consistent with strong economic performance. This may be a misleading comparison. Underlying productivity growth was much more rapid during the l950s than it is today. As a consequence, there is less need for nominal wages to fall than there would be at low inflation rates in the current environment. Moreover, rapid underlying growth may have masked weak performance in keeping output near its potential level. The more dramatic example of weak economic performance during the 1870s corroborates the view that excessively low inflation is potentially quite costly.

In addition, a small positive rate of inflation is likely to be better than a zero rate because it is more likely to be a credible goal. If the marginal cost of inflation around zero is perceived to be low and the short-run marginal benefit from some small inflation is positive, it would be very difficult to persuade people thatthe rate of inflation will remain at zero.

Start at a slightly higher rate of inflation, and the prospect of being credible becomes more likely. The public would recognize that policymakers would be less likely to raise inflation above that amount because everyone would agree that significant marginal costs would be incurred. Hence, the risk of output losses as the public overestimated the inflation rate delivered by the monetary authority would be reduced. I think there is a much greater chance of credibly establishing a 2 or 3 percent inflation rate than in actually establishing a zero percent inflation rate.

My third proposition is that institutions can do the work of money rules. The dynamic consistency problem is real, important,and no doubt the essence of the inflation problem. For example, it always looks good ex post to cancel your exams so that you don't have to grade them; the students have already studied. Likewise, it always looks good to inflate a little more than people expect. One solution to the dynamic consistency problem is to have a rule to which you are absolutely committed, so that cheating ex-post is impossible.

There are two costs of committing to a rule that must be considered. First, you can't respond to contingencies that your rule didn't contemplate.It was probably a good idea to do something in response to the stock market crash of 1987, something that you would not have perceived as desirable simply on the basis of some feedback rule involving the monetary base and nominal GNP. Aside from the crash,I doubt if any of us would have difficulty writing a scenario for the banking system in which it would be desirable to do something that you wouldn't see as desirable simply by looking at the behavior of nominalGNP.

The second cost of fixed monetary rules comes from the possibility of multiple equilibria. In any model that allows the possibility of multiple equilibria and allows monetary policy to have real impact, there is a possibility that monetary policy can be used to jolt the economy from a bad equilibrium to a good equilibrium. The character of all of the standard monetary rules, no matter how the feedback operates, is such that they would stabilize the price level as the economy grows according to its unique equilibrium in the model that is assumed. In a high-unemployment trap,it may well be desirable to use monetary policy to jolt the economy out of that trap, and I don't see any way that a monetary rule is going to do that. Monetary institutions can do the work of monetary rules. What sort of institutions will work? The same kind of institutions we use for the Supreme Court: appointing people in a way to make them not politically responsive. In addition, we can appoint people who may not share the values of the populace, but instead share the values to which we're seeking a commitment. I think it's a fair assumption that Lee Hoskins' attitude toward inflation is much more hostile than that of the median American voter and this is exactly appropriate.

I would support having someone in charge of monetary policy who is more inflation averse than I, precisely because of commitment. Can it work?Just look at the evidence of… a recent paper by Alberto Alesina and me…. Countries like Germany,where the monetary authoritie sare independent, achieve very low rates of inflation without monetary rules. But one also finds there is no systematic difference in real economic performance among those countries that achieve low inflation and political independence for monetary policy and those that don't have independent central banks. Unless it can be demonstrated that the political institutional route to low inflation--to commitment that preserves the discretion to deal with unexpected contingencies and multiple equilibria--is undesirable or cannot work, I don't see any case at all for monetary rules.

What about the zero-inflation amendment, my fourth major point? If it actually means zero inflation, it is almost certainly a disastrous idea. If it means that the Fed is locked into a rule, it is probably a bad idea. If it means that we're strengthening the hands of the Fed, in the name of inflation reduction,to resist political pressure to lower interest rates when the pressure is strong for monetary expansion, then it is probably a good idea.

We already have a Humphrey-Hawkins bill that mandates low unemployment,so why not have a law similar to Humphrey-Hawkins on the other side to mandate price stability,to make it easier for the FederalReserve to carryout its commitment? It seems to be a step in the right direction as long as it's not taken too literally and doesn't mean that we actually take what would be a tremendous risk of moving to zero inflation. If it is a symbolic commitment to Federal Reserve independence, it seems to me that it's all to the better.