Instruments, indicators, targets, transmission mechanisms and goal variables (my pathetic attempt at a general theory of monetary policy)
Macroeconomists disagree about everything. We can’t agree on the proper goal variable of monetary policy, or indeed even the criteria you would use to determine the optimal goal variable. We have proposed dozens of transmission mechanisms, with no general consensus emerging. We don’t agree on the proper intermediate targets; should it be interest rates, M2, exchange rates, or NGDP futures? Even worse, we don’t even agree on what the term ‘target’ means. Are interest rates an instrument of policy or a target? Is inflation a target or a goal variable? But it’s even worse than that. Even if we can agree on interest rates being a target, we don’t even agree on whether high interest rates mean easy money or tight money. In other words we completely disagree on what constitutes an indicator of the stance of monetary policy. And we even disagree about policy instruments, the most basic question of all. What does the central bank actually do? Does it directly control short term interest rates or does it control the monetary base? It’s no surprise that people debating monetary policy find it hard to even communicate with each other.
Physicists are always trying to unify the basic forces of nature, to develop a simpler and more elegant model. I’m going to suggest dropping the transmission mechanism, and combining indicators and targets. I’ll end up with a three stage process; the monetary base plus IOR as dual policy instruments (with very different purposes), NGDP futures as a policy target, and actual NGDP as the policy goal. This post will be a sort of comment on some excellent posts by Paul Krugman and Nick Rowe. It will be too long (like my early posts) but will also contain some new ideas (unlike my recent posts.) I’ll work backwards, from goals to instruments.
I’m just going to assume an NGDP policy goal; the post will be too long to waste time defending it. (I defend it in this National Affairs article.) It seems logical that you would want to adopt a policy that is expected to succeed. Thus monetary policy should be set in such a way that the expected growth in NGDP is equal to the desired growth in NGDP. That means (according to Lars Svensson) that policy should target he forecast, i.e. that the optimal policy target is expected NGDP growth. Svensson prefers to target the internal central bank forecast (presumably because he worries about a circularity problem with targeting market forecasts.) However there are at least two NGDP targeting proposals that use market forecasts and avoid the circularity problem. Hence I’ll assume NGDP futures prices are the optimal policy target, as they are much more closely correlated with expected future NGDP than are any of the other proposed targets (interest rates, exchange rates, M2, etc.)
What about transmission mechanisms? Mishkin argues that certain market prices are good indicators of policy because they are important parts of the monetary policy transmission mechanism:
Other asset prices besides those on short-term debt instruments contain important information about the stance of monetary policy because they are important elements in various monetary policy transmission mechanisms.
I slightly disagree, and would like use dimes and M2 as a counterexample. I think almost everyone would agree that dimes are not a very important part of the monetary policy transmission mechanism, certainly less important than M2. After all, if you don’t have any dimes in your pocket, that won’t have much impact on your expenditures, you’ll use nickels or quarters instead. And yet, dimes have recently been a better monetary policy indicator than M2. Why is that?
Coin production is very cyclical. Because coins are durable, the production of new coins is highly correlated with changes in the total stock of coins. Hence during recessions (when transactions fall) the stock of coins rises much more slowly, and new coin production falls sharply. Coins made during recession years tend to be “rare” and valuable to collectors. But if the dime-making machine broke down and the US Mint made fewer dimes during a given year, it wouldn’t cause a recession—people would make do somehow. Indeed something like that happened in 1964-65 when silver coins were hoarded due to the rising price of silver, and there was a coin shortage. And of course 1964-65 was right in the middle of a long boom.
Now consider M2 by comparison. Suppose that M2 velocity is quite erratic, due to financial innovation and changes in interest rates. But also assume that M2 velocity is uncorrelated with changes in the stock of M2. In that case M2 might be a lousy indicator of money policy, but still an important part of the transmission mechanism. That is, more M2 would cause more NGDP, given the level of velocity (which is determined independently.) BTW, I don’t wish to argue that M2 is in fact an important part of the transmission mechanism (nor am I convinced that M2 and M2 velocity are independent) just that one can make a more plausible argument for M2 than for dimes, even if dimes are a more reliable indicator of tight money.
This link shows that coin production plummeted about 80% between 2007 and 2009. And this link shows M2 growth accelerated during the recession. Some might object that the Fed can’t control coin production, that it is “demand determined.” You are not thinking outside the box. The Fed can control NGDP, and thus could have prevented the recession, or at least made it milder. If they had done so, then US Mint coin production would have fallen much less sharply. Coins really are a pretty good indicator that money is too tight (albeit not perfect, it would be less informative during hyperinflation.)
We should abandon attempts to find a monetary policy transmission mechanism. It’s too complicated. We know (intuitively) that 100 to 1 monetary reforms immediately lower all nominal aggregates in proportion; both the price level and NGDP immediately fall by 99%. How is it different if monetary policymakers suddenly double the monetary base, as in Hume’s thought experiment? One difference is wages and prices and debt are sticky during money supply changes, but not during monetary reforms. The other is that monetary reforms are expected to be permanent, whereas a doubling of the money supply might be reversed in the future. Recent NK models treat money as a financial asset, where its real value is very sensitive to changes in its expected future real value. So that makes the monetary transmission mechanism even more complicated. We need to deal with some wages and prices being sticky, some being flexible, and also with asset prices that are strongly impacted by future expected NGDP. There are a million transmission mechanisms; better to keep them inside the black box and focus on what we can measure and control.
Elsewhere I’ve argued that the terms ‘easy money’ and ‘tight money’ have no meaning in absolute terms, but only relative to the policy objectives. This suggests that policy indicators and targets are one and that same. If the NGDP futures price is above the Fed’s target then policy is too expansionary, and vice versa.
Finally we arrive at policy instruments. I’m sympathetic to Nicks Rowe’s argument that changes in the monetary base are what the central bank “really really” controls. But now they also directly control the interest rate on reserves. I hate IOR for two reasons:
1. It makes monetary policy even more complicated.
2. MMTers seem to like interest on reserves.
But if it’s going to be a permanent feature of our system, then I suppose we need to deal with it. I’ll argue that the monetary base can be used to control both the level of NGDP, and its trend growth rate. In contrast, IOR can only be used to change the level of NGDP (which makes it similar to fiscal stimulus–a policy that also is limited to generating one-time changes in base velocity.)
The Fed can stimulate the economy by lowering IOR, but it will soon run up against a “zero lower bound” problem. Not necessarily a zero IOR, that rate could be negative, but rather a zero level of excess reserves. At that point ERs can fall no further, and the monetary base is dominated by currency. Further increases in NGDP then require growth in the monetary base; IOR can only do so much.
We should think of IOR as a tool that allows the Fed to hit two targets, a trend rate of NGDP growth (achieved by long run increases in the monetary base at roughly the desired rate) and an optimal ratio of base money to GDP. For the US, the base to GDP ratio is normally about 6% (when nominal rates are above zero and there is no IOR.) But at zero rates and/or positive IOR you can achieve much higher base to GDP ratios. Milton Friedman dreamed of an economy saturated with liquidity, which is costless to produce. But that was thought to require mild deflation (as in Japan) and runs up against downward wage stickiness near the zero bound of wage increases. More zero bounds! (George Selgin correctly notes that aggregate nominal wage cuts would not be required, but I’m more worried than he is about wage flexibility in declining industries.) With IOR the central bank can have its cake and eat it too. It can have high levels of liquidity (at least within the banking system) and Argentine levels of inflation. Indeed with electronic currency approaching rapidly, we can foresee even greater gains from a policy of paying interest on all base money.
So the Fed could adopt an IOR that saturates the economy with the optimal base/GDP ratio, and then use the monetary base to control the trend growth rate of NGDP. The best way to understand the difference between one-time changes in NGDP and permanent changes in NGDP growth rates is to look at the following random selection of data from an old Robert Barro textbook:
Country Money growth RGDP growth Inflation NGDP growth Time period
Brazil 77.4% 5.6% 77.8% 83.4% 1963-90
Argentina 72.8% 2.1% 76.0% 78.1% 1952-90
Chile 47.3% 3.1% 42.2% 45.3% 1960-90
Israel 31.0% 6.7% 29.4% 36.1% 1950-90
S. Korea 22.1% 7.6% 12.8% 20.4% 1953-90
Iceland 18.4% 4.3% 18.8% 23.1% 1950-90
Portugal 11.5% 4.7% 9.9% 14.6% 1953-86
Britain 6.4% 2.4% 6.5% 8.9% 1951-90
U.S. 5.7% 3.1% 4.2% 7.3% 1950-90
Switzerland 4.6% 3.1% 3.2% 6.3% 1950-90
In my view the sine qua non of monetary economics is to explain these correlations, and I just don’t see how interest rate-oriented monetary economics can do that. The more “old Keynesian” the model, the more it focuses on explaining one-time changes. Any force capable of generating these persistent inflation rates is ipso facto also capable of explaining one-time changes. But alas, the opposite is not true; changes in fiscal policy or IOR can generate one time changes in P or NGDP, but not persistent changes. People get confused about this because at a given point in time a one-time change looks a lot like a growth rate change. (The longest journey begins with a single step.) But they are conceptually very different.
So a policy of IOR might raise the steady state base/GDP ratio from 6% to 60%. But from that point forward you generate your 5% NGDP growth by raising the base by roughly 5% (on average) each year. Peter Ireland made a similar argument in this paper.
So that’s my grand theory of monetary economics. Use the base to target NGDP futures prices, with the hope of maintaining steady growth in actual NGDP. If you also wish to saturate the economy with liquidity, do so with a positive IOR. Choose the optimal NGDP growth rate to minimize the combined losses from excess taxation on capital (when NGDP growth is too high) and excess labor market inefficiency (when NGDP growth is too low.)
Robert Hall liked to publish papers with a “perfect monetary system” back in the 1980s; at least they were perfect until his next iteration. I think that’s way too modest; I’d like to shoot for this blog post representing “the end of macro.” At least until another wacky idea pops into my head.
PS. If you think those high trend inflation rates require fiscal stimulus, you are wrong. The Fed can buy anything, it doesn’t have to be US government bonds. But even if restricted to government bonds, a deficit of just 2% to 3% of GDP would probably be enough for the Fed to run 100% trend inflation, without ever running out of T-securities to buy. (This is based on the very conservative assumption that base demand would fall to no more than 2% or 3% of GDP if we had 100% inflation.)