Bayesian Investor Blog

I tried to finish this in time for April 1, 2019, but masterpieces take time, and I didn’t find the time to turn this into a masterpiece. Then I kept adding sections that didn’t fit with the April Fools spirit, for a result that, like South Park, is not suitable for any audience. Finally, an economic crisis prompted me to publish whatever incomplete version of it I could manage to write, before people notice that we’ve avoided a depression. I’m definitely not satisfied with the quality of this post, but can’t afford to put more time into it.

In this post, I’ll try to summarize my guesses as to what are the most controversial parts of Scott Sumner’s monetary policy ideas.

In the process, I’ll try to dress them up to look more like the kind of wisdom that requires years of study to master. After all, I wouldn’t want anyone to get the impression that the Fed’s highly educated experts could be replaced by, say, ordinary bloggers.

In particular, I will attempt to correct the serious shortage of equations and graphs that plagues market monetarist writings.

I’ve attempted to make this fancy enough to belong in a prestigious publication, but I’m little more than an ordinary blogger, and I doubt that I’ve been thorough enough, or have enough experience at creating a prestigious style. Please feel free to write a more sophisticated-looking version of this post, and borrow as much as you like, without any need to credit me.

Basics

I’ll start with an equation that summarizes the central advice of market monetarism:

E = F – T

where:
E is the deviation in central bank policy from what is expected to produce stable growth, mild inflation, and low unemployment,
T is a target for some measure that we’re interested in, most likely NGDP, inflation, or nominal wages,
F is the predicted outcome for T.

Predicted in this context means predictions that are inferred from prices in fairly liquid financial markets. Incentives matter. Forecasters employed by central banks generally have little incentive to correct mistakes in their forecasts, whereas traders lose money when they make bad forecasts.

Caution: monetary policy is only loosely approximated by a single variable. Markets actually respond to more than just the central bank’s current interest rate and open market operations; they respond mostly to the actions that the central bank will take in the future. I’m treating that as a single variable here, since that will be good enough for most purposes if the central bank is following a well-thought-out rule. But if the central bank is stumbling about, trying out different policies, in hopes of empirically determining which one works, then it’s hard to predict whether a single variable will come close to capturing what we want to know about monetary policy.

I will now address some objections to market monetarism, trying to focus on the ones that the Fed is most likely to worry about.

Liquidity Traps

Economists who focus too much on interest rates (mainly Keynesians) complain about liquidity traps, where the central bank allegedly become impotent because interest rates are so low that the central bank can’t lower them further.

I gather that they assume monetary policy operates mostly via poorly understood effects of interest rates on behavior (e.g. via influencing investment). It seems likely that effects of this nature are real. I don’t know of strong arguments for these effects being the primary or sole way by which monetary policy works (here’s an economist asserting they’re the sole way), but it sure seems like many influential people believe something like this.

In contrast, market monetarism says that monetary policy operates via changes in the supply of, and demand for, money.

Even if money has some special features, those don’t eliminate the features that it shares with normal goods and services. So we should expect that changes in the supply of, or demand for, money alter the value of money, thereby reducing or magnifying the effects of sticky wages. This process doesn’t require that monetary policy have any particular effect on interest rates (although it can easily alter interest rates as a side effect).

In this graph, the x-axis represents the supply of money, for a given currency. The y-axis represents the value of money (more commonly modeled via its first derivative, the inflation rate, which can interact in unpleasant ways with sticky wages).

One reason that this is controversial is that “money” is hard to define and hard to measure. Also, there are some problems with measuring the value of money (i.e. measuring inflation). Maybe this means that nobody can really say that Venezuela has an excessive money supply. After all, some have argued that the confusion over what we mean by “intelligence” implies that we shouldn’t worry about superintelligent machines, or that the difficulty of measuring “largeness” means we should dismiss speculation about larger than human machines.

Does the Fed have enough power?

An even more controversial claim is that central banks can increase or decrease the money supply by as much as they want.

Market monetarism doesn’t require that it be easy to increase the money supply. It might even require doing things that the central bank has never done before, like buying many more bonds than are required to maintain the central bank’s interest rate target, or changing the interest rate target more frequently than once every 6 weeks. Central banks appear to have the technology and the legal authority to do those things, but their reluctance to employ them suggests that there are hidden obstacles.

Some people suspect that past episodes of hyperinflation were caused by a central bank doing something new, so doing new things might be more dangerous than continuing familiar policies. It’s much easier for central bankers to deflect blame for the results of familiar policies than it would be to deflect blame for new, untested, policies. Maybe there’s a serious shortage of central bankers who are willing to accept those risks.

Fiscal Policy

Market monetarists deny that fiscal policy has important effects on business cycles.

Maybe it could have important effects under conditions that are rather different from the current US, but central banks have the power to neutralize those effects, and have policy goals which normally lead them to do so.

This idea can be colloquially approximated by the claim that central banks react to perturbations to the monetary system, in ways that are designed to achieve a pre-specified NGDP or inflation target. Or more formally,

MP1 = MP0 + A*(Id – If)

where:
MP0 = monetary policy (planned open market purchases) at time 0,
MP1 = monetary policy (planned open market purchases) at time 1,
Id = desired NGDP growth rate,
If = forecast NGDP growth rate,
A = a number which is estimated through trial and error; it can be considered a constant over small enough time periods (days to weeks?), but will likely change over months to years.

In order to matter, this controversial assertion needs to be combined with the controversial assertions that I mentioned in regards to liquidity traps.

I expect that many politicians and political activists will reject this part of the market monetarist package, due to its implications for fiscal policy. Yet the Fed’s sources of bias seem likely to dispose it to favor this suggestion, and I care more about the Fed’s monetary beliefs than about the beliefs of other doubters, so I won’t devote much time to arguing for this.

Wild Policy Fluctuations

Some people seem to be concerned that Sumner’s policies would cause the Fed’s policy to fluctuate chaotically, causing confusion.

The reasonableness of that objection is strongly dependent on what we care about. Let me make an analogy with a grocery store’s policy toward stocking toilet paper. Demand for toilet paper soared unexpectedly in March. What policies should grocers prioritize for handling this?

1. they could prioritize predictable pricing
2. they could prioritize being predictable about how much they order from suppliers
3. they could prioritize managing the available quantity to match the willingness of customers to buy

Some people worry about the Fed causing interest rates to fluctuate more than current policy does. That resembles the people who want toilet paper prices to stay constant, even if that means that a fair number of people run out of toilet paper. I can see how that might be 20% of what we should care about when making toilet paper policy.

Some people worry about wild swings in the Fed’s open market purchases and sales (or equivalently, wild changes in the money supply). That’s somewhat like worrying about grocers making wild changes in how much toilet paper they order from producers. There are certainly people who would be better off if grocers focused on stabilizing the available quantity of toilet paper, but it’s hard to see how those effects could make up more than 5% of what grocers should care about when making toilet paper policy.

Option (3) corresponds most closely to what Sumner proposes. It would have produced wilder swings in the quantity of toilet paper traded (money supply) than the alternatives, and more frequent changes in toilet paper prices (interest rates), in order to focus on stabilizing the ability of consumers to buy toilet paper (have their wages denominated in a currency whose value is stable, thereby making employment more stable).

So, yes, if you think that the most important parts of central bank policy are reducing frequent changes in interest rates, or money supply growth, you should expect Sumner’s policies to be destabilizing. You should only support market monetarism if you evaluate policies by outcomes such as NGDP growth, inflation, and/or unemployment.

I apologize for not turning this insight into an equation.

Circularity

I’m a bit confused about the market monetarist position on the circularity problem. The circularity problem seems to be something like this: in the limit where the central bank policy becomes perfect, because it relies on perfect forecasts from markets, the market prices will cease to convey any information, because they’ll always predict that the central bank is doing the right thing.

This sounds to me a lot like a concern that the stock market will become so efficient that all smart investors will realize they can’t beat the market, after which there will be nobody around to cause it to be efficient.

I’m having a bit of trouble tackling this problem, because it sounds like a really great problem to have, compared to what we’ve had the past century. It’s sort of like asking what happens if we suppress COVID-19 to the point where all test reports of it are false positives.

If a country is used to having its central bank make a ~$10 trillion mistake about once per decade, then it’s easy to confuse a $10 billion mistake with an absence of mistakes.

Markets are fairly efficient. Efficient enough that, even with decades of experience, I ought to believe that the S&P500 is priced sensibly nine years out of 10. Yet it just takes one little virus to create some obviously misguided prices. I’m confident that even with much better monetary and pandemic policies, markets would still give the Fed a bunch of mildly mistaken signals, for which the markets would then need to signal corrections.

The market monetarist position seems to be aiming for the most feasible way of getting an order of magnitude reduction in harm caused by monetary problems. This leaves plenty of room for $100 billion mistakes, such as targeting 4% NGDP growth when the optimal target is 4.5%, and for $10 billion mistakes on a typical day, such as guessing wrong about how the NGDP futures market will react to a particular open market purchase.

Actually, there’s some risk that I’m not describing the strongest version of the circularity problem. Maybe the Fed needs to use a forecast of the monetary base instead of, or in addition to, the NGDP (or inflation) forecast. Sumner describes here how relying on NGDP (or inflation) forecasts would provide less than optimal stability. See here for a longer explanation of the better approach.

Another potential problem that is closely related to the circularity problem is the claim that there’s a long lag between changes in monetary policy and market reactions to those changes. Sumner presents evidence in The Midas Paradox of cases where the markets reacted almost instantly to policy changes, and he argues that times when people see lags have been times when the Fed seemed uncertain about whether it would stick with the new policy.

Dangling loose ends

The Fed seems still somewhat anchored on an approach that approximates a gold standard, and is making valiant attempts to incrementally improve on the gold standard. That still leaves as its default behavior to assume that the money supply should grow at a modest and steady pace.

The Fed could react more agilely to economic downturns if, instead, it focused directly on forecasts of inflation (or NGDP, or wages) from markets or from other forecasters whose rewards are based mainly on their accuracy.

Here’s a COVID-19 test supply analogy: “we’ve made a record number of tests” is not a very satisfying claim if there’s also a record demand for tests. Similarly, the Fed expanded the money supply by a record amount near the beginning of the pandemic, but that doesn’t tell us much.

I haven’t said much about level targeting versus rate targeting, as I can’t find much controversy there. The Fed seems to be slowly inching toward the market monetarist position, and hesitating mainly due to concerns that it will make it easier for people to hold the Fed accountable for mistakes.

I haven’t discussed NGDP targeting versus inflation targeting here – I have an old post here on that topic.

Where Are We Now?

[This section does not attempt to reflect Sumner’s views. He believes in a stronger version of the Efficient Market Hypothesis than do I.]

We’re in a situation where a pandemic is causing large fluctuations in the economy that the Fed can’t control. But the Fed still has a good deal of influence over the inflation rate. Stabilizing the inflation rate is still an important way to stabilize the economy.

The TIPS spread seems to indicate that expected Fed policy will be too tight, but by a much smaller margin than in early spring. It’s outlook is pretty similar to what it was in late 2019, so it’s maybe close enough to the right policy to enable decent growth.

The Fed did an impressive job of expanding the money supply for a few weeks in early spring, but it has been shrinking the money supply recently, in order to avoid meeting their inflation target keep interest rates from falling below the Fed’s target.

Several possible explanations for this year’s fluctuations in inflation forecasts (both of which assume a bit of Fed reluctance to cause unusual increases in the money supply):

Markets in March saw a significant risk that COVID-19 would cause people to stay home for years, and have gradually become more optimistic that vaccines will enable GDP to mostly recover by mid 2021.

There was an unusual demand for liquidity: starting in March, the increased uncertainty that was associated with the pandemic caused increased demand for low-risk assets such as money. Businesses were raising enough cash to handle a worst case scenario, often maxing out their credit lines.

These patterns are equivalent to deflationary forces. Normally, deflation is moderately stable over periods of a few years, so past experience led people to worry that it would be hard to recover from this downturn. Yet better information has gradually helped rule out the worst case scenarios, so the businesses that were able to stockpile a fair amount of money, are now deciding their risk of going broke is receding, and they’re reducing their demand for money, thus reversing some of the deflationary pressure.

The TIPS spread is the best available indicator of whether the Fed will meet its goals, but it’s also helpful to look at Hypermind, superforecasters, and Metaculus. Note that Hypermind’s NGDP prices don’t seem to be publicly available on Hypermind’s site, but are displayed on themoneyillusion.com.

I’m slightly more concerned about deflation than inflation over the next year or two, but the current economic situation is unusual enough that I still consider the risks of inflation to be higher that normal. Gold, cryptocurrencies, and the possible bubble in large tech stocks are all hinting at somewhat inflationary conditions (but maybe somewhere other than the US?). So it’s more important than usual to be willing to quickly switch between loose and tight monetary policies.

P.S. – this post was weakly designed to compete with Eliezer’s post on monetary policy.