Scott Sumner  

Central banks should have listened to Eliezer Yudkowsky

The Peculiar Political Economy... Other People's Money...

Eliezer Yudkowsky has a great post over at Less Wrong, presenting an imaginary dialogue with a central banker:

Frequently Asked Questions for Central Banks Undershooting Their Inflation Target

I recall reading this dialogue years ago, but I am not sure where it was originally posted, or when. The original date of the post would be especially helpful, as it's turned out to be quite prophetic. For instance, at one point the imaginary central banker keeps saying that his (her?) experts insist that inflation is likely to be two percent, and that TIPS market forecasts predicting sub-2% inflation are flawed. Yudkowsky replies that deep and liquid markets provide the most efficient forecast. How'd that dispute work out?

Yudkowsky has a rare ability to look beyond all the noise, all the framing effects, and get to the heart of an issue. The post is rather long, but it's engagingly written and I hope readers will persevere to the end---where they'll find a pleasant surprise (well, pleasant to me.) It's probably the best single introduction to the market monetarist way of thinking in the entire blogosphere.

PS. Off topic, I was intrigued by this post comparing data mining in the social sciences to what might be called "theory mining" in particle physics. But I don't feel qualified to evaluate the claims. I wonder if someone who does (Yudkowsky, David Deutsch, Scott Aaronson, Robin Hanson, etc.) has expressed a view on this issue.

HT: Anon, Razib Khan

Comments and Sharing

COMMENTS (21 to date)
Jim Glass writes:

Larry Summers was pretty critical of the Fed so undershooting its 2% target in an interview with Tyler, leading up to...

"Right now, I have an instinct towards nominal GDP targeting"

... among a lot of other topics, in case not reported here before.

Brian Donohue writes:

Bravo. Yudkowsky nails it.

Jason writes:

That's probably my favorite Yudkowsky. I think he originally posted it on Facebook. The date on that post is February 12, 2016.

On a related note, I've always (and still) have trouble with the fact that the case for market monetarism seems so terribly obvious to me, and yet some very intelligent people don't seem to be persuaded. Your occasional updates on how popular (economic) opinion is moving towards your views are reassuring on that matter.

Alec Fahrin writes:

I believe your first link is from 2014-2015.

The problem here is that we have had four years of evidence that monetary policy has consistently been too tight, but the Fed still wants to raise rates for a third time this year.
2.5% wage growth YoY. 3.7% total compensation growth YoY. CPI is averaging 2.2% this year. Producticity about 2%.
Therefore, real compensation is still not matching the growth in productivity.

The “data-dependent” Yellen Fed was 50% theory, 40% their traumatic 1970s experiences, and 10% data.

Scott Sumner writes:

Thanks everyone.

Matthew Waters writes:

The post is really good, though like MM in general sort of glosses over the zero-bound concrete steppes, especially in a financial crisis.

There are a mix of concrete steppes in a crisis. There is general idea of negative rates and buying assets. Central banks also did not reduce rates quick enough in 2008, they paid IOR and did not purchase government assets available. They left ammunition on the table, for no discernable reason.

And that's all true! But if the ideal policy is no bailouts, then MM policy would have to offset not only the 2008 NGDP drop as it happened in 2008. It would ALSO have to offset the effect of Bear Stearns, AIG, Lehman, others going bankrupt. Bear and Lehman bankruptcy happened under expectations of a much higher NGDP growth path.

Furthermore, the ideal MM policy would not have the discount window, money market insurance and commercial paper purchases. These responses were less stealthy than the bailout, but they put significant liquidity into the market (including making commercial paper liquidity-equivalent).

I have toyed with some concrete steppe ideas, other than OMO's or negative rates. One idea I had was to purchase Treasuries maturing soon for far above par. Say a Treasury maturing for $100 in a week might trade for $99.98. The Fed offers $120.

The Fed could make the offer to anyone on a simple price basis. It shouldn't be to just primary dealers, but to any institution in FedWire Securities Service or individuals with TreasuryDirect.

That's my nutty idea, anyway, as an alternative idea to negative rates or private asset purchases.

Scott Sumner writes:

Matthew, It's a mistake to focus on the concrete steps in isolation. The relevant trade-off is between size of balance sheet and trend rate of NGDP growth. If you want a smaller balance sheet, have a higher NGDP growth target.

SG writes:

Great piece. I think Yudkowsky first posted this on facebook a few years ago. My favorite line:

"Q. Just because creating enough money would make prices rise, doesn't logically imply that if prices don't rise then I haven't created enough money!

A. Actually—"

Matthew Waters writes:

It just seems like circular logic to me: NGDP level targeting works by using NGDP level targeting.

I'm not worried about size of balance sheet per se. Balance sheet with just government assets is fine, whatever the size.

But I personally am uncomfortable with buying large amounts of private assets. I don't mean private asset QE is politically hard. I mean I would be one of those people that make it politically hard.

Compared to other concrete steps, buying private assets raises a host of issues. What price does the Fed buy the assets at? How does it vote in proxy contest? If it holds bonds, does it do workouts or how does it participate in bankruptcy court?

As I said, a lot of concrete steps would ideally be eliminated under MM (discount window, etc.) Other than private asset QE, the concrete steps remaining are negative rates on reserves (not currency) and government asset QE.

In a financial crisis, those concrete steps wouldn't be enough. There were cases in 2008 where textbook EMH stopped holding as well. For example, a lot of risk-free arbitrage profits became available due to lack of liquidity from banks.

So I do think concrete steps are important part of ideal monetary policy. A concrete step which is both theoretically and practically unlimited, such as negative rates on currency, sending checks to citizens, buying private assets or buying Treasuries far above par, is necessary.

Robin Hanson writes:

Since you ask, that physics post is completely correct.

Sam Roberts writes:

Scott, am I wrong that "just print moar money" is equivalent to "just write off moar irresponsible debt and wipe out moar life savings"?

I'm sure Eliezer's correct that by doing those things, a central bank would be able to ensure that its inflation or NGDP target was met. And I'm also pretty sure he's correct that it wouldn't even need to actually do them, as long as it stated an intention to do them with a determination that was utterly credible.

But am I wrong to think that doing those things would be immoral? And that even threatening to do those things would be immoral?

Eliezer imagines the creation of a quadrillion dollars, and the sending of $2000 checks to everyone; and claims that it is so obvious that these would raise prices that a "12 year old girl raised by goldbugs" could understand it.

Is it wrong for me to react: "Yes, of course - it would raise prices by so much that the distinction between me and Jeff Bezos would be negligible, destroying (by swamping) the value of both of our assets"? I know it's a rhetorical exaggeration, but it's being used to justify the same thing on a smaller scale.

Please tell me I'm wrong, or I'm missing something, because market monetarism seems kind of dystopian to me right now.

PDV writes:

The original post was here

Don Geddis writes:

@Matthew Waters: "It just seems like circular logic to me" No, but expectations matter a great deal.

"the concrete steps remaining are ... government asset QE. In a financial crisis, those concrete steps wouldn't be enough." Market Monetarists expect that simply buying Treasuries (ordinary OMOs) would be more than enough. But this comes back to your first confusion: if the economy expects higher NGDP growth, then far fewer bonds would need to be purchased, in order to achieve that NGDP growth. It's the low NGDP expectations themselves (because of the resulting drop in velocity / increase in money demand), which cause the requirement for huge asset purchases.

Don Geddis writes:

@Sam Roberts: "Am I wrong that "just print moar money" is equivalent to "just write off moar irresponsible debt and wipe out moar life savings"?" Yes, you are wrong. The Fed purchasing Treasury bonds is not the same as "writing off debt". Similarly, successfully targeting small positive inflation does not "wipe out life savings".

"I know it's a rhetorical exaggeration, but it's being used to justify the same thing on a smaller scale." No, it's a reductio ad absurdum. Of course the "scale" is critical. That hypothetical thought experiment is a response to people who (falsely) claim that the central bank lacks the power to create inflation (under some economic conditions). The point is to get you to admit that a central bank could cause hyperinflation, if it wanted to. Even if, for example, interest rates are at the Zero Lower Bound.

Once you agree that the central bank possesses the power to cause inflation, then of course it would be silly for the bank to even threaten hyperinflation. That does not accomplish any goal. What the bank does, is credibly promise to meet its inflation target (e.g. 2%). And the promise is credible, because it clearly has the power to create hyperinflation, so therefore it must also have the power to create 2% inflation.

Matthew Waters writes:

Don Geddis,

The NGDP expectations narratives seems detached from how real market participants acted in 2008.

To put it another way, let's say the Fed in March 2008 broadcasted on Times Square that they were doing NGDPLT. But they would not do any more concrete steps past taking the Fed Funds rate to zero. Then Bear Stearns failed, followed by AIG and Lehman.

I don't think anybody thinks NGDP would stay reasonably within the level target with only Fed Funds rate. Why does that necessarily change with adding Treasury purchases alongside Fed Funds rate as tools?

Again, a lot of tools would also be eliminated under this NGDPLT regime. I.e. the bank bailouts, discount window, money market insurance, etc.

Finally, while the Fed didn't have NGDPLT in 2008, they did have a 2% inflation target and failed to meet it with all of these actions. The market expectations argument should work the same for simple inflation targeting, but expectations didn't work in 2008.

Don Geddis writes:

@Matthew Waters: "Let's say the Fed in March 2008 broadcasted on Times Square that they were doing NGDPLT. But they would not do any more concrete steps past taking the Fed Funds rate to zero." Your statements are in contradiction. That is simply not NGDPLT. The whole point is to have a sufficient threat that you can force the market to the outcome you want. And then you don't need to use the threat. But if you begin by handcuffing yourself, and "Fed Funds rate to zero" is not sufficient ammunition to achieve the target, then the whole thing falls apart. Then market expectations immediately come to conclude that you will fail to reach your target. And then you do indeed fail.

"Then Bear Stearns failed, followed by AIG and Lehman." These happened after (and because of) NGDP collapsed. You have an odd hypothetical, essentially assuming that the Fed was successful with NGDPLT, but these failures happened anyway. That isn't a world that makes much sense. You would need to give a lot more detail for how, exactly, such a thing could take place.

"Why does that necessarily change with adding Treasury purchases alongside Fed Funds rate as tools?" Because that allows for tremendously greater debasement of the currency (via expanding the money supply), than merely getting the Fed Funds rate to zero. Buying up all Treasuries (and then other OMO-qualified assets) is orders of magnitude greater intervention than merely going to zero rates.

"The market expectations argument should work the same for simple inflation targeting, but expectations didn't work in 2008." Yes, as you guess, the argument works just fine for inflation targeting too. The failure is the Fed's actual actions. They refused to take the necessary actions that were available to them. They did not "do what it takes" to achieve their targets. Instead, they came to the horrible conclusion that there was "an equal chance" of greater, or lesser, inflation. And for years after 2008, they continued to confidently predict that inflation would rise, on their current path ... only to be wrong, quarter after quarter, year after year. Yet no one held them accountable for their failures.

"Expectations" worked fine in 2008. The market did not believe that the Fed's plan would result in 2% inflation going forward. And the market expectations were correct. The Fed had a horrible plan.

Matthew Waters writes:

Bear Stearns failed in March though. The market overall still signalled typical NGDP expectations in March.

Bear Stearns was indirectly bailed out by the Fed with Maiden Lane. I'm assuming that Bear's failure would have knocked on to Lehman and AIG. September would have moved up to March.

I think we're in agreement that for "sufficient threat" of hitting Fed target, the market will usually "believe" the Fed's stated goals.

I disagree that purely buying Treasuries at market value is a sufficient threat in a financial crisis situation. Once Treasuries go to 0%, the tool is exhausted. The quantity bought is immaterial at that point.

Again, this is without bailouts, discount window, etc. It's tough to test such a hypothetical. But the IB's all became Fed members for access to the discount window. What if the discount window is taken away, with dropping Treasury rates to zero as only tool?

Many MM assumptions after Bear's hypothetical failure get *really* strained. Money market funds and hedge funds faced very large withdrawals. So even if investment firms believe the Fed, it could be taken out of their hands if they have redemptions. The people who redeem won't actually look to reinvest for above-market rate. They're worried about their money disappearing.

Since banks have been highly regulated and had de facto 100% deposit insurance since 1930's, it's hard to game out the hypothetical exactly. I do know financial panics happened regularly before 30's and the deposit-holders weren't really looking to reinvest. It's tough for Fed to offset that with just making all Treasury rates zero.

Don Geddis writes:

@Matthew Waters: "Once Treasuries go to 0%, the tool is exhausted." I don't understand what you mean by that phrase. Before you were talking about dropping the Fed Funds Rate to zero. But what does it mean for central bank purchases of Treasuries to make "all Treasury rates zero", as you are claiming?

"dropping Treasury rates to zero as only tool?" The "tool" of OMOs doesn't have anything to do with "dropping Treasury rates". It is the purchase of existing Treasury bonds (and/or other financial assets, such as MBSes or some foreign assets), in order to expand the money supply. The concrete macroeconomic "work" comes from the money supply expansion, not from anything that happens to the Treasuries themselves.

Matthew Waters writes:

I did not phrase the sentence about Treasuries well. The general principle is that Treasuries which are trading at 0% become roughly equivalent to cash.

If the Fed has a standing offer to buy all <3 month T-bills at maturity value, then a $100 T-bill, $100 in Fed reserves and a $100 bill are all exact substitutes. The T-bill is now a zero-interest-bearing asset, same as cash. Substiting an asset that's same as cash with cash itself should not have an effect.

In extreme, the Fed does not have to use OMO's to purchase Treasuries. This would really sound like heresy to market participants such as the Primary Dealers. But both Fed reserves and Treasury ownership are entries in some database. Most Treasuries move their ownership on the Fed's ledger. In extreme, it could offer 0% to all Treasuries, to be settled for Fed reserves. Any participant in the FedWire Securities Service or TreasuryDirect could accept. Same could be done for Agency MBS.

It's just tough to test this hypothetical Fed action against the hypothetical of a financial crisis.

[html fixed. You cannot use the keyboard less-than symbol. Use &lt; instead--Econlib Ed.]

Don Geddis writes:

@Matthew Waters: "Substituting an asset that's same as cash with cash itself should not have an effect." They may be "the same" as far as a target for investment, but that's not the only function of cash; nor is it the important macroeconomic function.

"Money" has a number of important functions. It appears to me that you are only considering the "store of value" function. But, for example, as a "medium of exchange", cash is clearly much more useful than Treasury bonds, in a huge variety of transactions.

Meanwhile, the important macroeconomic function is as a "unit of account" (i.e., the thing referred to in wage and debt contracts over time). In that function, even if T-bills and cash appeared to be equivalent at some particular moment in time, they wouldn't at all be expected to remain equivalent over time. So it matters a great deal what expectations are about the path of the money supply of cash in the future, but much less so about the path of the supply of T-bills.

The key is what happens (and what is expected to happen, in the future) to the value of the Unit of Account. Regardless of their current prices, cash and T-bills are not at all "exact substitutes", as you claim.

Matt Waters writes:


Below is the post I wrote, which I wouldn't blame you for not reading.

In general, I like the broad strokes of MM thought. Strong MM (with EMH and expectations) could hold under very strong "homo economicus" assumptions. I just disagree with those assumptions, especially in a financial crisis.

To my engineer, non-economist mind, these concrete steps all matter. Sorry that it gets expounded to such a long post. I understand if you don't read it.


I don't entirely disagree.

My hypothetical has the strongest "concrete step" possible with only government asset purchases and without negative rates.

The Fed gives an open tender offer to all gov. security holders to buy their asset at 0%, say up to some maturity. The Fed could go up to any maturity at the absolute extreme. A 30-year bond issued in 2007 for 4%, with 58 coupon payments left, could be bought for 216% of the principle. For a $100 bond with 29 years remaining at 4%, that's $100 + $2*58. Remember there is a 0% discount rate applied to all cash flows.

It would not be an OMO in the traditional sense because it's not "Open Market." OMO's have the ~20 primary dealers, on behalf of the Fed, buy at the ask prices in the markets. It would be more radical. It is the most extreme concrete step under gov. asset purchases only.

You would think nearly all Treasury holders would in fact turn their Treasuries in for cash. Then, the thinking goes, the cash would be spent.

However, a practical financial crisis has the existing intermediaries all become suspect. For "depositors" in the shadow banking system pre-2008, there existed a long chain of title to the ultimate beneficial owner of banking liabilities.

Money Market Fund owners may have been governments, institutions, corporations, etc. Without the discount window, there would also not exist a backstop for non-FDIC insured liabilities at Fed member banks.

As Gary Gorton said, the depositors go from information-insensitive to information-sensitive. EMH and expectations just don't really hold in such a scenario. Those with ultimate title to banking liabilities used intermediaries for a reason. They also wanted immediate availability of money for a reason. Such a decline in sophistication had the effect, in 2008, of dramatically reducing the natural rate.

The reduction in the natural rate goes back to the T-bill versus cash argument. I did not mean exact substitute to mean you can take a Treasury to the store and buy gum with it. My point is that if the market is in such a state for Treasuries to trade at 0%, then when they're sold for cash, that cash would not be spent. NGDP ends up the same.

Going to the absolute extreme QE (purchase at 0% of all gov securities, up to 30 years) should find some owners willing to spend the new money. But in a financial crisis, it would have to offset lack of lender-of-last-resort and other facilities that existed. I am not convinced it could without even more extreme facilities (helicopter money, large negative rates, etc.)

Comments for this entry have been closed
Return to top