Now that George Selgin has placed his almost legendary “Less Than Zero” essay online, I thought it was a good time to discuss his ideas. I’m going to nitpick a few points, so it might be useful to first lay my cards on the table—do I support the productivity norm or not?
I think there are actually two (or three) distinct ideas in George’s essay, which I see as being logically separate issues. There are two versions of the productivity norm, based on labor productivity and total factor productivity. But the difference between them is not all that important in my view. Much more important is his argument that prices should fall at the rate that productivity is increasing. I see this as a completely separate argument that raises an entirely different set of problems. So here is my preliminary response, before getting into details:
1. The productivity norm as a monetary policy regime? It’s OK with me if it is OK with you (meaning my fellow Americans.)
2. Mild deflation at the rate of productivity increase? Yes, but only if several stringent preconditions are met. And we haven’t yet met them.
George uses many of the same methods that I employ to make arguments. He mixes intuitive thought experiments, economic history, history of thought, pragmatic arguments, fairness arguments, and illuminating analogies/metaphors.
He starts with an intuitive argument. Suppose you had n industries, and there was a sudden increase in productivity in one of those industries, say computers. The increase in productivity in the computer industry would lower the relative price of computers. Under a price level target, you would have to inject enough money to slightly raise the price of all other goods, in order to offset the fall in the price of computers. That seems unnecessarily cumbersome. In contrast, under a productivity norm (assuming unit elastic demand for computers) you could let the price of computers fall and keep all other prices the same. This little thought experiment shows that the “menu cost” argument doesn’t necessarily show that price level targets are superior to the productivity norm. (BTW, he employs a very nice analogy using Baroque fugues and Romantic symphonies. It’s too long to summarize here, but take a look at pages 23-25.)
Selgin actually discusses several versions of this policy. He considers a NGDP rule to be superior to a price level rule, as it would allow productivity increases to depress the overall price level. But he thinks a productivity norm would be even better. He discusses both a labor productivity norm (where prices fall at the rate labor productivity increases) and a total factor productivity norm (where prices fall at the rate which total factor productivity increases.) Although he leans toward the TFP standard, I am going to use the labor standard in my examples, for these reasons:
1. In practice, I don’t think the implications differ very much from the TFP norm.
2. I like it better.
3. I find it easier to understand and explain.
4. It is similar to Earl Thompson’s 1982 proposal.
Those interested in the TFP version should consult his essay.
In practice, with the US trend GDP growth rate around 3% and the labor supply growing around 1% per year, the labor version would result in a trend rate of deflation of about 2%. (And I believe the TFP version would lead to closer to 1.5% deflation, on average, but am not certain.) So aggregate nominal wages would be stable and workers would earn real wage increases through falling prices. Indeed something like this occurred between 1870-1897. This is another reason to discount the “menu cost argument” for a price level rule. Yes, a price level rule would keep the average price of goods stable. But it wouldn’t stabilize factor prices. And wages are an especially sticky factor price, which suggests changes in the aggregate wage level may be very costly.
Another common argument for preventing inflation is that it harms creditors. But suppose prices rise because of an adverse productivity shock; George asks why creditors shouldn’t bear some of the burden. If a big asteroid devastated the US, then under a price level target the Fed would prevent all T-bond holders from suffering any loss in purchasing power. The lifestyle of those government bond coupon clippers would be supported by the back-breaking labor of the remaining citizens, picking through the wreckage of a Cormac McCarthy-type landscape. Is that fair? Maybe, but it’s not obvious why it’s more fair than spreading out the pain.
In principle, a productivity norm does not have to aim for mild deflation. You could target nominal wages to grow at 2% per year. In that case inflation would average about zero percent, but the actually rate of inflation would vary inversely with labor productivity shocks. Since I find most of George’s arguments for the productivity norm to be fairly persuasive, I’d like to nit-pick his arguments for a mild deflation productivity norm, rather than near-zero inflation productivity norm.
First let me start with an issue on which I think we agree. George indicated that he agreed with my argument that 2008 was not a good time to institute a policy of even mild deflation. Many wage and price contracts had been negotiated under higher inflation expectations, and of course the banking system was already heavily stressed by defaults from the housing crash. So let’s consider the long run arguments for mild deflation, assuming the system was phased in during a more stable period.
Selgin has several arguments for mild deflation:
1. Friedman’s optimum quantity of money argument (i.e. mild deflation lowers the inflation tax on cash balances.)
2. Rising factor prices can lead to macroeconomic instability (boom, followed by recession.)
3. Prices would be stable in industries with no productivity variation, thus the price system would more effectively send signals about productivity to consumers and producers.
4. With stable aggregate wage levels, workers would find it easier to respond to labor market signals.
These are aesthetically pleasing arguments, and I’ll add one more ugly pragmatic argument, one that might be even more important. Our tax system discriminates against saving and in favor of current consumption. This is even true when inflation is zero, and becomes worse at high inflation rates. Mild deflation would lower the real effective tax rate on capital, and this would increase the real growth rate in the economy. Cutting the other way is the fact that cash is extremely useful in the underground economy, so mild deflation (and near zero interest rates) might expand the size of the underground economy. Japan has this problem.
Now for the other side:
I am not convinced by the wage rate argument. I worry about two problems. One is money illusion, the fear that workers are irrationally antagonistic to nominal wage cuts. George’s response is that wages don’t have to be cut in declining industries, as any attempt to cut wages will push workers toward growing industries. If this works smoothly, the wages ought to be about the same in each industry (adjusted for skill levels obviously.) But I worry that labor markets don’t work that well. Labor is heterogeneous and it is very costly to move from one industry to another. Thus real wage cuts are often necessary in declining industries. And if aggregate nominal wages are stable, then you are back to the money illusion problem–you must convince workers to take nominal pay cuts, but you can’t “trick them into doing the right thing” with inflation. A very aesthetically ugly argument, but one that I fear is true.
My other major concern relates to the current interest rate-oriented monetary regime, in which central banks see themselves as being powerless in a near-zero interest rate environment. Yes, the US did OK in the late 1800s. But the gold standard somewhat anchored price level expectations during that period, and this helped us avoid a liquidity trap. In addition, real interest rates were higher than today because of the rapid population growth and the industrial revolution.
George makes some theoretical arguments based on the assumption that equilibrium real rates should be at least as high as the rate of productivity growth. But I am not sure that is always the case. Government bonds offer liquidity advantages, and it seems to me that in the US the real interest rate on government bonds may fall below the growth rate on productivity.
My nightmare scenario is Japan. My critics point out that the Japanese should have adjusted to mild deflation by now. True, if the rate was absolutely steady. But what seems to happen is that Japan does adjust for a few years, but then when it falls into recession (like right now) the central bank doesn’t know how to use its policy tools to prevent deflation from intensifying. I think this problem could be eliminated with a futures-oriented NGDP target, level targeting (or indeed the same level targeting approach concept applied to Selgin’s productivity norm.) But we haven’t gotten there yet. So right now I oppose mild deflation. Here is the order in which things need to be fixed:
1. First, we need level targeting of a nominal aggregate.
2. Then let’s target the forecast of that aggregate.
3. Then let’s adopt a near zero inflation productivity norm.
4. Then we can go to a mild deflation productivity norm.
I am also less than convinced by George’s argument that stable wage rates would avoid the boom and bust cycle that can develop when nominal wages trend upward. I don’t see how this argument is consistent with the superneutrality of money. On the other hand you know that I like pragmatic arguments, so I imagine there are all sorts of real world factors that might make rising wage levels more prone to lead to business cycles. So I’ll leave that as an open issue. (As an aside, I completely agree with George’s view that price level targeting can push up factor prices during productivity booms, with possible destabilizing consequences. But I don’t see a problem with a slightly positive trend rate of wage growth, as long as the rate of inflation can still vary inversely with fluctuations in productivity growth.)
A few minor comments on his historical section:
1. There is nothing technically wrong with his comments on WWI, but he might leave some readers with the impression that most of the inflation was merely offsetting the hit to productivity associated with wartime disruption. In fact, the inflation of both goods and factor prices was extremely large, and mostly monetary.
2. I disagree with his view that the post-1929 deflation cannot explain the 1929 crash. I think the crash occurred when the market correctly saw the way things were going. I have mixed feelings about the Austrian argument that the problem was the Fed’s refusal to allow mild deflation in the 1920s. I think a case can be made for that argument, but only as part of a broader argument for discretionary monetary policy under a gold standard regime. Under a fiat regime, the actual 1920s policy could have continued indefinitely.
Another way of making that point is that the stable prices of the 1920s might have been destabilizing in retrospect, but not so much because productivity gains were not allowed to depress prices, but rather because there wasn’t enough gold in the world to maintain this policy indefinitely. If I went back in a time machine to 1922, I would suggest to the Fed that they aim for 1-2% deflation per year as the best way of staving off a collapse after 1929.
[BTW, that’s an idea! Perhaps I should tell the Fed that I came in a time machine from the future, where all central banks target NGDP futures contracts, level targeting. I could tell them that because Bernanke implemented the policy in 2010, he went down in history as the greatest Fed chairman ever.]
3. Again, I can’t find any specific error in his discussion of negative productivity shocks and inflation during the 1970s. However, it is important to recognize that RGDP increased at a fairly normal rate in the 1970s. So George’s argument probably only applies to brief periods when there were severe productivity shocks, such as 1974. The reader should certainly not infer that George favors the high inflation that we actually observed during the 1970s.
A few final comments. I really liked this pragmatic argument on page 66:
The real choice we face is therefore, not really a choice between a true productivity norm or a truly constant price level, but between some crude approximation of a productivity norm and some equally crude approximation of a constant price level.
As you may guess, I’d go further and say; “and some even cruder approximation of a constant price level.”
There is a discussion of free banking at the end of the essay. His proposal to freeze the monetary base and allow free banking might or might not work. It depends how stable the demand for bank reserves is. He assumes the demand is proportional to transactions, i.e. to NGDP. But that is not at all clear (as he admits in footnote 57.) I’d be terrified to put my reputation on the line advocating something like this, especially given the long and sad history of Goodhart’s Law. I’d feel much safer getting to free banking via Bill Woolsey’s index futures convertibility approach.
To summarize, I am mostly convinced by the theoretical arguments for the labor productivity standard. Indeed I reached this conclusion on my own from a different direction, publishing a paper favoring a wage index target in 1995. And Earl Thompson proposed this idea in 1982. My half facetious comment “if it’s OK with you” in the intro referred to the strange politics of this idea. What would workers think about the Fed freezing aggregate nominal wage rates? You and I know that this policy does not impact on real wages over the long run, but would the public perceive it that way? When minimum wages (nominal) were recently increased by 40%, above minimum wages (nominal) would have had to fall. Of course this makes obvious what we would hope voters already understand about the costs of minimum wages, but for now I think the politics of a NGDP target are less iffy. Still, you have to start somewhere, and so I am glad to see George trying to educate people about this idea. I would add that there are cases where you could easily sell the benefits to workers, for instance the sharp reduction in wage growth seen in the current recession would not have occurred under a productivity norm (where nominal wage growth stays constant at zero.)
Bill Woolsey has several interesting posts on the Selgin paper. I would also recommend this recent Cato paper by David Beckworth. Interestingly, David’s paper was published in the fall of 2008. So he did not realize that deflation was about to become a problem. Although I completely accept George and David’s distinction between good deflation (from productivity increases) and bad deflation (from tight money), I still worry that many people may draw the wrong implications from this argument.
I was on a panel at the Southern Economics meetings in San Antonio with fellow bloggers David Beckworth, Josh Hendrickson, and also George Selgin, who would be a better blogger than any of us three if he decided to throw his life away and blog full time. A year earlier at the SEA meeting in late November, 2008, I attended a panel with five right-leaning monetary economists. Several of them seemed a bit too dismissive of the looming deflation threat for my taste. And this is my fear. Playing with deflation is like playing with fire. We need to sort out much more fundamental problems with monetary policy before we risk an intentional policy of deflation, despite the fact that many of George and David’s arguments are quite persuasive. If we don’t, we could accidently burn down the house entitled “free market economics.”
Tags: George Selgin