Saturday, February 24, 2018

Slok on QE, and a great paper

DB's Torsten Sløk writes in his regular email analysis:
Yesterday I participated in the annual US Monetary Policy Forum here in Manhattan, and the 96-page paper presented concluded that we don’t really know if QE has worked. This was also the conclusion of the discussion, where several of the FOMC members present actively participated. Nobody in academia or at the Fed is able to show if QE, forward guidance, and negative interest rates are helpful or harmful policies. 
Despite this, everyone agreed yesterday that next time we have a recession, we will just do the same again. Eh, what? If we can’t show that a policy has worked and whether it is helpful or harmful how can we conclude that we will just do more next time? And if it did work, then removing it will have no consequences? There is a big intellectual inconsistency here.

Investors, on the other hand, have a different view. Almost all clients I discuss this topic with believe that QE lowered long rates, inflated stock prices, and narrowed credit spreads. Why? Because when the Fed and ECB buy government bonds, then the sellers of those government bonds take the cash they get and spend it on buying higher-yielding assets such as IG credit and dividend-paying equities. In other words, central bank policies lowered risk premia in financial markets, including in credit and equities. As QE, forward guidance, and negative interest rates come to an end, risk premia, including the term premium, should normalize and move back up again. And this process starts with the risk-free rate, i.e. Treasury yields moving higher, which is what we are observing at the moment.
These lovely paragraphs encapsulate well the academic and industry/policy view, and the tension in the former.

I'm interested by the latter tension: Industry and media commenters are deeply convinced that the zero interest rate and QE period had massive effects on financial markets, in particular lowering risk premiums and inflating price bubbles.

I'm deep in the academic view. The industry view forgets that the Fed does not just suck up bonds, it issues interest-bearing reserves in exchange. For every $1 of bond the market does not hold, the market has to hold $1 of additional reserves.  Industry analysis is very insightful about individual traders and investors and the mechanics of markets but forgets about adding up constraints and equilibrium which are the bread and butter of academia. You personally may sell a bond and put the money in to stocks. But someone else has to sell you that stock and hold the reserves.

The risk premium is the same if you borrow at 2% and lend at 4% than if you borrow at 4% and lend at 6%. So there is no relationship at all in basic economics between the level of interest rates and the risk premium, or between the maturity structure of outstanding government debt (reserves are just overnight government debt) and the risk premium. That one cannot see any movement at all in 10 year rates or inflation with QE is also noteworthy.

But us academics need to listen as well as to lecture. Often industry people know something we don't.    So I find this striking difference interesting. Though I haven't changed my mind yet.

Torsten wrote back:

TS: But that argument only holds in a closed economy, no? In other words, what if the US based seller of Treasuries to the Fed took the proceeds of their sale of US Treasuries and invested it in Indonesian government bonds?

JC: Then the seller of Indonesian government bonds now is sitting on US reserves.

TS: And what it the European insurance company used the cash they get for selling bunds to the ECB to buy US IG credit?

JC: Then the seller of US IG credit is now sitting on reserves. Someone is sitting on reserves. And reserves are now just very short term Treasury debt.

TS: Anyways, you may say the market view is partial equilibrium but almost everyone in the industry saw the portfolio substitution with their own eyes and believe that it is real.

JC: That helps. Yes, but they saw one side of the portfolio susbstitution. They did not see the other side of that substitution! I think in the end it's mostly foreign banks now sitting on the reserves, so those banks took deposits from someone who sold securities to your industry contacts.

A fascinating conversation..

The monetary policy forum is here. The paper is "A Skeptical View of the Impact of the Fed’s Balance Sheet'' by David Greenlaw (Morgan Stanley, so not everyone in industry has the industry view!) Jim Hamilton, Ethan S. Harris (Bank of America Merrill Lynch), and Ken West. It's excellent. It takes 96 pages (plus graphs) to put to rest verities that have been passed around unquestioningly for 8 years. Excerpts from the abstract:
Most previous studies have found that quantitative easing (QE) lowered long term yields, with a rough consensus that LSAP purchases reduced yields on 10-year Treasuries by about 100 basis points. We argue that the consensus overstates the effect of LSAPs on 10-year yields...We find that Fed actions and announcements were not a dominant determinant of 10-year yields and that whatever the initial impact of some Fed actions or announcements, the effects tended not to persist.
This is important. Most of the pro-QE evidence was how yields moved on specific QE announcements. We all know there is price pressure, but it usually lasts only a few hours or days. Much commentary has presumed the price pressure was permanent, as if there is a static demand curve or individual bonds. And the first work will naturally pick the events with the biggest announcement effects, then incorrectly generalize.
...the announcements and implementation of the balance-sheet reduction do not seem to have affected rates much.
And implementation... When the Fed actually bought bonds, interest rates went up 2/3 times. See below.
 Going forward, we expect the Federal Reserve’s balance sheet to stay large. This calls for careful consideration of the maturity distribution of assets on the Fed’s balance sheet.
Mild objection. If QE has no effect, then the maturity distribution is irrelevant, as Modigliani and Miller would have predicted, no?

A much-recycled graph showing 10 year rates have been trundling down for 20 years unaffected by QE or much of anything else, and that actual QE purchases - - increases in reserves -- are associated with higher 10 year rates:


  1. Finance was buying the index, not individual stocks.

    The index was correlated with the government budget because congress was running huge deficits. Finance was really saying, we want the government correlation because index betting is cheap and easy. Little tax payers get stuck with the bill later. Why wouldn't finance want government insured gains?

  2. I have a more skeptical view after about decade in investment management. Markets reward behavior, not ideas, and a lot of goofy ideas are "proven" correct by good results. I've met a lot of PMs who believe that buybacks mechanically increase stock prices (and not due to a differential tax argument). Deep in the recession, Shadowstats reached way too many managers with way too much AUM. Phrases like "stimulus induces risk chasing" and "asset inflation" are usually surface level descriptions of recent price changes.

  3. I would say both are wrong.

    I don't understand the academic treatment of QE. As you point out, during the actual implementation of QE, rates tended to rise. My understanding of the academic explanation of OMO is that if you are moving from a policy that is too tight to a more stimulative policy, this should cause long term inflation and real rates to rise, but that the liquidity effect of central bank buying lowers rates in the spot market. But, then, if you measure the effect of QE by doing an event study based on the announcement of the policy, this seems to be based on the idea that the liquidity effect is anticipated. That seems incoherent to me. If the liquidity effect can be anticipated, then there should not be a liquidity effect.

    The industry commenters are correct that risk premiums were lower. That is why long term interest rates were rising during QE. That is the sign that QE was effective! So, at least the industry commenters have the sign right. But, then they treat that as if it is a problem. The reason risk premiums declined is because risk declined. The reason risk declined is because the central bank was doing its job! In fact, I would go so far as to say that the broad notion that developed by 2007 that it was the job of the central bank to inject systematic risk (or at least stand by and let it happen) into financial markets was the fundamental cause of the crisis itself. The tendency of financial managers to equate low risk premiums with financial bubbles, and so to treat systematic risk as some sort of disciplining policy regime is a terrible cultural development.

    1. The industry commenters are correct that risk premiums were lower.... That is the sign that QE was effective!

      Do I have to point out the logical flaw here? I did recover from the flu. That is the sign that the candles to St. Jude were effective?

    2. I don't understand. Are you questioning whether stabilized economic expectations are a causal factor for low risk premiums?

  4. John, thanks for the post. This reminds me of something I wrote over the summer--about the persistence of excess reserves indicating the limits of QE. While I suspect you won't agree with the whole post (here: as I hold out some hope for QE (I see it extending the lender-of-last-resort role to include tamping down volatility in risk spreads), the excerpt below speaks to the thrust of your post, while also pointing to the issue of risk-weighted capital constraints that we face in the industry. QE purchases essentially ask the private sector to sell zero-risk-weight assets and purchase those with non-zero risk weights...

    "The issue? The banking system is incredibly slow at passing on the liquidity of the QE purchases, to the extent it does at all. Thus, criticism of QE programs has arisen out of the fact that the actual government bond purchases have largely resulted in the buildup of excess reserves (deposits with the central bank) at banks as opposed to the intended “portfolio balance” into new long-term lending.

    As banks face capital restrictions against their risk-weighted assets, they are already staying very close to their regulatory minimum capital levels and thus would be unable to “rebalance” their portfolio into riskier long-term assets to replace the safe Treasuries purchased by the central bank without raising costly equity. Thus, the persistent excess reserve problem may be partially explained by the elevated post-crisis risk premiums and binding capital requirements. Widening risk spreads have been a signature phenomenon of serious financial crises. "

  5. I am not sure I agree that QE was ineffective.

    But set that aside. If one says QE is ineffective, should the Fed simply buy bonds directly from the US Treasury? This was legal from World War II through 1981.

    Also, should the Fed simply open up its own market desk and buy bonds on the private sector without fiddling with the primary dealers? That is, do not create reserves. Rather, digitize money and buy bonds.

    I contend in the next recession the US will hit zero bound rather quickly. Ergo having a bigger toolkit available makes sense. If QE is ineffective, then let us try money-financed fiscal programs.

    Interesting side note: the Fed cannot liquidate about half of its balance sheet, as there is so much more cash in circulation. Net net, does this mean QE has been converted into cash in circulation?

  6. Dr. Cochran,

    There are three variables which express change in monetary policy or stance by the Federal Reserve Bank (FRB), the volume of money, narrow money in particular, short term interest rates, the size of the FRB’s portfolio. There does appear to be any correlation between changes in these variables and aggregate stock prices.

    1) Liquidity - Narrow Money

    If one examines the broader and more detailed history of the stock markets in the United States it can be seen that there is actually an inverse or negative correlation between the volume of narrow money, bank reserves, and the NASDAQ. If one plots the year over year percent change (Y/Y%C) in the NASDAQ Composite Index (NASDAQCOM) [1] over time and the Y/Y%C St. Louis Adjusted Reserves (ADJRESN)[2] when there is an upturn in bank reserves there is a downturn in the NASDAQ [3]. Between February 1985 and June 2016 there were 377 months. The median monthly average Y/Y%C in ADJRESN was 3.2% and the median monthly average Y/Y%C in the NASDAQCOM was 4.1%. The Pearson Product Moment Correlation R2 value was -0.28 (p<0.001) a moderate but significant correlation. This was particularly evident following the Crash of 2008 when the volume of bank reserves skyrocketed with the Quantitative Easing Policy (QEP) [4]. In March of of 2009 the ADJRESN rose 894% but the NASDAQCOM fell 31% which was typical for most of 2009. The PPMC for 2009 for these same variables was -0.95 (p<0.001). However during the height of the Housing Bubble in 2004 the exact opposite relationship was observed, the volume of ADJRESN and the value of NASDAQCOM rose together[5]. The PPMC for 2004 for these same variables was 0.63 (p=0.03). So really there is no general correlation at all between changes in the volume of narrow money and changes in the stock market [6].

    2) Short Term Interest Rates

    Were one to plot the Effective Federal Funds Rate against the NASDAQ composite index over time, there would not be any obvious correlation [7]. The later data, after 1995, might suggest a *positive correlation* between the EFFR and NASDAQCOM. if the data is re-analyzed on a Year over Year Percent Change (YOYPC) basis (averaged monthly), the positive nature of the correlations is more obvious, especially in the later data [8]. Between March of 1973 and August of 2017 there were 557 pairs of monthly averages calculated on a YOYPC basis. The median YOYPC for the EFFR was 0.18% and for NASDAQCOM 13.9%. The Pearson Product Moment Correction was 0.02 (p = 0.61), there was no correlation at all. However if one examines all data after 1999 (n = 223) the PPMC becomes 0.20 (p = 0.003), a weak but statistically significant *positive* correlation. As the EFFR increase, the NASDAQ composite index increase (some of the time).

    3) Large Scale Asset Purchases (LSAP)

    One might note that the FRB stopped expanding its LSAP programme, the most visible part of the Quantitative Easing Policy (QEP), in October of 2014, three years ago. If LSAP and QEP were the reason for the increase in stock prices, stock prices should have stopped increasing over three years ago.

    If there is no correlation between changes in FRB monetary policy and the aggregate prices of stocks, it is hard to a causal connection.


  7. Valter Buffo, Recce&#39;d, MilanFebruary 25, 2018 at 9:56 AM

    Sorry about NOT sharing the enthusiasm. Worrisome to read that we all do NOT understand if this huge (wild) policy response to the GFC was useful or the opposite. Why do it then? Still convinced we need some "unconventional" analysis (including theories) to solve the inconsistency described by Slok. The enclosed paper is nice, collects a number of ideas and data properly end neatly. Excellent would be when useful to improve our understanding. How comes that reserves are debt, so total debt increases, and total financial wealth increase, and in turn market prices of financial wealth do increase more, a nothing comes as a consquence of it? Just double it all again, then! Still thinking that there are a few pieces missing from the mainstream approach, as history may show shortly.

  8. The question is phrased incorrectly....Who did it help or hurt? Savers....unmitigated disaster. The idea of saving is dead! Gambler banks and traders on the verge of bankruptcy....worked like magic...bonuses and wealth for gambling is record. Billionaire stock holders...never before in history has something so greatly increased their wealth over others in society! For solvency of the world...allowed government to indebted themselves beyond imagination.

    The further question...who is putting Humpty back now the debt is beyond repair and the masses are going to be unhappy as more and more is taken from them.

    1. You must be kidding. The last 10 years have been spectacular for savers -- so long as they invested in long-term bonds or stocks.

  9. With the coming of QE and most importantly forward guidance, the entire mechanism for investing and "betting" on fixed income changed.

    For those that have access as all do now... the FWCM page on Bloomberg is the market's benchmark.

    Tell me where OIS (the market's expectation literally these days of how many "hikes" the FED will engage in before they stop") is five years forwad....add the now tradeable and traded spread between the Ten Year Note and matched maturity OIS... and one "gets" the Ten Year US Treasury yield...

    Not by coincidence the current forward OIS (1 month maturity) Five years forward is within a bp or two of the FED "long term" median 2.75%...

    Given that currently OIS is running at 1.67% (the March FOMC is already priced as a fait accompli)...for those that Do rather than ruminate... this simply means that the "market consensus" today... is that the FED moves "just" 100 basis points over the next Five years.

    Being a fixed income investor boils down to evaluating how likely a proposition that is... given that the market is also.. simultaneously and arbitragibly pricing close 4 "hikes"... by March of 2019.

    This forward path, which now trades directly predominantly in the mid curve Eurodollar options which are nothing but bets on the forwards... as in what will the Herd think rates are going to 5 years forward, come this December... is and should be all that anyone "cares about... for this translates directly in to positive or negative returns on your "bet".

    Modern technology, new trading instruments...a compliant FED... have made this the ultimate Keynesian beauty contest. Having an opinion about the outcome is irrelevant... in order to "win"... your opinion has to be different from others, and had to converge not in 10 years.. but all components of the "bet" have now entered the world of immediate mark to the market.

    As always your posts are well worth the read....

  10. Work the problem backwards.
    Between Treasury and member banks we can net out the residual expenses after payments in and out. The net residual is Fed staff expenses, plus a miniscule carry forward risk. So the Fed, being economic, always finds a spot on the curve where the funds flow keeps the staff just busy keeping competitive with private banks, the Fed chooses their best spot to bet, just like any other large TBTF bank. But their major client is Treasury and the Fed operates so as to marginally favor government debt, help them face the TBTFers.

  11. "I'm deep in the academic view. The industry view forgets that the Fed does not just suck up bonds, it issues interest-bearing reserves in exchange. For every $1 of bond the market does not hold, the market has to hold $1 of additional reserves."

    No one in the rest of the world is holding the net additional cash nor cash equivalent reserves the Fed introduced, the US Treasury is.

    US Treasury vs Rest of World

    In the regular course of business, the US Treasury has to fund the government largesse, by issuing debt.
    Call it $500 to quantify it. You can change the first digit and add a bunch of zeroes later.
    The UST is sucking $500 out of the Rest of the World in regular intervals.
    UST issues -----> Bonds ($500) -----> Rest of the World
    UST receives <----- Cash ($500) <------- Rest of the World
    It indeed is a closed loop between the UST and the Rest of the World.

    Next add the new participant.
    The Fed decides to intervene with QE.
    The right hand of the government is buying debt from the left hand of the government issuing the debt.

    In the regular course of business, the US Treasury has to fund the government largesse, by issuing debt.
    However, instead of the Rest of the World buying all the debt, the Fed jumps in as a new participant.
    Again to quantify it, call it $300.

    UST issues -------> Bonds ($300) -----> Fed
    UST receives <----- Cash ($300) <------ Fed

    UST issues -----> Bonds ($200) -----> Rest of World
    UST receives <----- Cash ($200) <----- Rest of World

    The US Treasury was able to sell $500, receive all its $500 cash, and the Rest of the World only parted with net $200 cash,
    leaving $300 cash for the Rest of the World that didn't go to purchasing net $300 in UST bonds, to bid up whatever is the flavor of the day.

    It does not matter if the Fed buys bonds in the open market from the Rest of the World and the cash go around the world or if the Fed purchased directly from the US Treasury,
    the net effect is the Fed's $300 cash ended up back at the US Treasury for $300 in bonds, as the Rest of the World only spent a net $200 to purchase a net $200 in UST Bonds.

  12. Well put. I wouldn't be suprised if the different views come from different perception of time. Industry tends to look on a shorter time frame of quarters to years, while academia takes the longer view, 3 to 5 years.

    If that were the case I could see how both are right - we just have yet to see the long term effects of qe, and will need another decade to research it's impact.

  13. In this discussion between JC and TS, I'd like to know how swaplines interfere with the picture of international flows. Can they be a sneaky way to close the circle? Or are they mostly a method for the powerful to build and maintain monetary power over others? What is the power of the FED over Euroland? Certainly Denmark can only survive in its stance thanks to its Euro swapline.

  14. Dear Professor Cochrane,

    It must seem obvious to you but I would guess to many non-economists and certainly to me it is not clear why the constraint that someone hold the reserves created when the fed buys assets means a priori there is no impact to rates?

    I suppose the answer is either simple and can be explained in a paragraph or so, or else complex and relies on some theory which you could name and refer us to.

    Thank you!

    1. Good clarification. No, that point does not imply no effect. It just implies that the usual thinking which leads to a huge effect is wrong. We are not asking the private sector, all together, to hold a $3 trillion less long-term government bonds. We are asking them to hold $3 trillion less long-term government bonds but in exchange $3 trillion more overnight government bonds (reserves). It's almost like a change operation. The Fed takes back a $20 bill, but gives you two $5 and a $10 in exchange.

      That might indeed do something. If people overall really care a lot about holding 10 year government bonds rather than overnight bonds, there could be small differences in interest rates. (There are other theorems that say no effect at all, but that needs a few more assumptions). Just as changing your $20 for two $5 and a $10 could cause you to go out and spend more. But an argument that ignores the fact that we are holding just as many reserves as we gave up government bonds is false.

      Even there I have overstated. The treasury sold more bonds than the Fed bought. I don't have exact numbers handy but roughly speaking imagine the treasury sold an extra $5 trillion long term bonds and $5 trillion short term bonds. The Fed bought back $3 trillion of the long term bonds and gave us $3 trillion overnight government bonds instead. So it's just as if the Treasury had sold us $3 trillion overnight, $5 trillion short-term and $2 trillion long term bonds to finance deficits. Does the Treasury's choice of what kind of bonds to sell us really have a huge impact on interest rates? Was the recession largely due to the Treasury's fault in selling us the wrong length bonds, and good thing the Fed was there to undo it? Does this actually not so large rearrangement of the maturity structure of a given amount of government debt in private hands really spark massive bubbles and stimulus and so on?

      Well, per your question, it's possible. But less likely when you realize it's just a rearrangement not a change in the overall amount.

  15. Thanks for the clarification—it Is very helpful to know I was not just missing some obvious point.

    Your observation about the Fed “correcting” the Treasury’s Term Structure decisions is interesting—I suppose the answers to your questions depend on what you mean by “huge,” “massive,” “bubble,” etc. On the question of who was at “fault” for the recession I suppose you could equally blame the fed for not correcting the distribution of available investments earlier or regulators or right minded stewards of capital for the same failure—I think, if you believe that the recession could have been avoided, then you must believe it could have been avoided by some adjustment to what investments were being made and contracts being written?

    Anyway it certainly seems as though adjusting the available aggregate duration, buying mortgage pass throughs, or even as in Europe buying credits is more than just trading a 20 for two 5s and a 10. Definitely one ought not ignore the fact that someone is holding the new reserves—but isn’t the 20/5,5, 10 analogy tantamount to saying that there is no value in having a Term Structure or in offering investments with specific risk characteristics to those better equipped to bear those is hard to believe there is no value to aligning risks with economic agents.

  16. Professor:

    This is largely excellent, and as someone who works at DB, of particular professional relevance. However, I either disagree with, or don’t understand what I think is an important point on risk premia. Credit spreads (and I believe, implicitly, the equity risk premium) are time varying and correlated to the level of interest rates. It’s one reason why the simplified version of the CapM is so deceiving using a fixed spread for the equity risk premium. To adjust your example: if the ten year treasury is at 1%, investors might require 15bps for BBB credit. If the rate is 10%, they will certainly not accept 15bps, but perhaps something closer to 150 or more. I work in debt capital markets, helping banks issue term bonds. For over five years I helped banks issue in Japan. As JGBs approached zero, the credo spreads banks paid fell from the 30s to a handful of basis points...that is, Rates matter to risk premia.

    I asked Veronesi this question a couple of years ago when at Booth, particularly because my markets experience led me to think their was a flaw in the simple representation of the CapM. He agreed, but pointed me in your direction as the expert on time varying risk premia...unfortunately you’d already moved on!

    My larger point is that, to the degree Fed action substantially (artificially) changes the level of Rates, they absolutely change risk premia, and to the degree that there is stickiness in investor return targets, this will change their purchase decisions by necessity.

  17. JC: I think I may have provided this link while catching up on your blog last year, but here it is again - it supports your view and seems relevant. The idea is that the gaps between the three QEs give us a natural experiment for testing how banks & BDs responded to QE - all we need to do is group the relevant Z.1 data into QE and non-QE periods and plop the results onto a chart, which is surprisingly (to me) clear.

    And here's an annotated version by Focus Economics (loses some meaning):


Comments are welcome. Keep it short, polite, and on topic.

Thanks to a few abusers I am now moderating comments. I welcome thoughtful disagreement. I will block comments with insulting or abusive language. I'm also blocking totally inane comments. Try to make some sense. I am much more likely to allow critical comments if you have the honesty and courage to use your real name.