以前ここで紹介した論文の掲載が決まったとのことで、著者の一人のエガートソンがツイッターでスレッドを立てて同論文を解説している(H/T タイラー・コーエン)。

Bernanke famously quipped that “The problem with Quantitative Easing (QE) is that it works in practice but not in theory”.
My paper with Bhattarai and Gafarov on how QE can work in theory is forthcoming in the Review of Economic Studies. A thread:
First, on QE not working in theory: I think Bernanke had in mind QE2 and QE3, where Fed bought long term government bonds with money (bank reserves). Why QE1 worked in both in theory and practice is easier to see since it was aimed at market dysfunction.
E.g. my AER paper with Del Negro, Ferrero and Kiyotaki shows how QE1 had a big effect by replacing illiquid private paper with government liquidity. QE2 and QE3 are harder to justify on this ground, since then the Fed was buying very liquid goverment long term bonds.
For QE2 and QE3 Bernanke probably had in mind a classic “irrelevance” result by Wallace (1981) and myself and Woodford (EW) in 2003: Long-rates are largely pinned down by current and expected future short rates and a risk-premia.
In standard models neither is affected by printing money and buying long-term bonds when the zero lower bound on the Federal Fund rate (ZLB) is binding (see discussion in EW:
Our new paper highlights that the irrelevant result in EW critically depends on expectations about future interest rate being unaffected by QE. Is this a reasonable?
There is compelling empirical evidence, e.g. by Vissing-Jorgensen and Kristnamurthy (VJK) in a BPEA from 2011 that shows that when Fed does QE, market participants lower their expectations about future short term interest rate.
VJK call this a “signaling channel“ of QE. In a nutshell, then, QE works partially works by providing “forward guidance” about future interest rates. Our new paper is a theory of how QE can play this role.
Forward guidance is the idea that even if the short term interest rates are at the ZLB, the Fed can still lower long-rates by communicating that it will keep short-rate rates long for a “substantial time”. This increases demand. It is a central idea in EW.
To get perspective it is useful to take a step back. A key problem is that this is not credible, due to classic time-inconsistency, a point I first formalized in my job market paper from Princeton way-way back in time, when dinosaurs roamed the earth (early 2000s) (see paper)
Chairman Jay Powell nicely summarizes the issue: “Part of the problem is that when the time comes to deliver the inflationary stimulus (implied by the forward guidance), that policy is likely to be unpopular, what is known as the time consistency problem in economics.”
Similarly John Williams, President of the NY Fed: “In the jargon of academics, our commitment technology is very limited. It is simply impossible for us to set a predetermined course of policy that will bind future Committees”.
This is where QE enters the picture in our new paper: It makes forward guidance credible. To see why QE is a useful “commitment technology” think for a moment about what QE2 and QE3 correspond to in practice:
The government is printing money (reserves which are equivalent to short term government debt) and buying long term government debt. Because the government is exchanging one form of debt for another, total government debt is unchanged.
So QE is just equivalent to making the duration of government debt shorter. It is as if you change the 30 year mortgage on your house, where interest are fixed, to a floating mortgage, where interest rate changes month to month with the market.
To see how QE can create expectations about lower future interest rates, consider your own position if you were to enter the meeting of the Federal Market Open Committee (FOMC) today with a 1 million dollar mortgage on (a) floating rate or (b) a fixed 30 year rate.
Would you be more reluctant to raise rates with a) or b)? Presumably you would be more reluctant to raise rates if you had (a), floating rate, since then raising the rate immediately increases the cost of your debt payments while they remain unchanged if you had a 30 year one.
The same insight applies when thinking about QE if you model formally a “game” between the government and the private sector in what's called in academic jargon a “Markov Perfect Equilibrium”.
QE solves the time-consistency problem of optimal forward guidance Powell and Williams refer to at the ZLB. It is a "commitment technology”. If the government shortens the duration of its debt via QE, it credibly commits to lower future rates and a future monetary stimulus.
“Wait!” you might say. Why would the Federal Reserve care about the interest cost or fiscal burden of the government?! That’s not part of its famous “dual mandate” of inflation and employment.
One answer is that the Fed should care about the fiscal position of the government because it is part of social welfare. Moreover, while the “dual mandate” catches most of the headlines, the Federal Reserve Act actually does not only talk about inflation and employment.
Instead it says that the Fed should “promote effectively the goals of maximum employment, stable prices, and moderate long-term interest rates.”
The paper, however, stresses another argument. You don’t really need to assume that the Fed cares about the interest payments of the Treasury. All you need is that the Fed is concerned about its own balance sheet gains/losses.
These are ultimately reflected in how much money the Fed can pay back to US taxpayers over time. We show that a model that captures this leads to the same conclusion as a model with a central bank that directly cares about the governments fiscal position.
To show the practical relevance of this point we document several quotes from recently declassified memos of the Feds staff to the FOMC.
These memos show that the Fed regularly monitored potential losses to its balance sheet due to QE, as well as its effect on the overall budget of the government. In other words, these forces played a role in the Fed's decision making, judging from declassified FOMC minutes.
In the paper we show that this role of QE, i.e. the effect it has on the Fed’s incentives in setting its interest rates, can in principle be strong enough to explain what people have found empirically using event-studies.
We also show in simulation that the macroeconomic effect of QE2 and QE3 can be large.
An interesting finding is that the weight on the term associated with the fiscal variable does not need to be quantitatively large. In our baseline simulation. In fact, the weight on fiscal consideration is about 100 smaller than that on inflation in our benchmark example.
Our bottom-line therefore is, paraphrasing Bernanke: QE worked both in practice AND theory.
PS, almost forgot, here is an link to the paper: https://drive.google.com/file/d/1xyHAsj-0kht_Mtx5McqSVd6s3SeH17iz/view. I sometimes tell friends I don't like twitter, for it tweeting is not a natural form of expression for me. I think a 30 tweet marathon thread makes my case!


我々の新たな論文は、EWでの無関係の結果が、将来の金利についての予想がQEに影響されないことに大きく依存していることに焦点を当てている。この前提は合理的だろうか? BPEAの2011年のヴィッシング-ヨルゲンセンクリシュナムルティ*4(VJK*5)のように、FRBQEを実施すると市場参加者は将来の短期金利についての予想を低めるという説得的な実証結果がある。



a)とb)のどちらの場合に金利引き上げをより躊躇するだろうか? おそらく(a)の変動金利の場合に金利引き上げをより躊躇するだろう。その場合、金利引き上げによって直ちに債務支払い費用が増加するからである。30年だった場合にはその費用は不変である。
「ちょっと待った!」と貴兄は言うかもしれない。なぜFRBが政府の金利費用や財務負担を気にするのだ?! インフレと雇用という有名な「二つの使命」にそれは入っていないではないか。

PS 忘れるところだったが、論文へのリンクはこちら:https://drive.google.com/file/d/1xyHAsj-0kht_Mtx5McqSVd6s3SeH17iz/view。友人にはツイッターが好きではないと言うことがある。私にとってツイートは自然な表現形態ではないためだ。30ツイートのマラソンでそのことが立証されたと思う。

*1:cf. 本ブログでは直近にはここでその話を取り上げた。

*2:cf. ここ

*3:cf. ここ