Quantitative Easing: A Tutorial
On September 13, 2012, the Fed announced a further program of quantitative easing, or QE. The program, which we will all call by its unofficial name, QE3, will consist of purchases of some $40 billion in mortgage-backed securities each month plus continuation of some existing asset purchase programs. We are told that QE is the most powerful weapon left in the Fed’s arsenal—but it is also among the least understood. For the benefit of everyone who wants to understand the mechanics of QE and review its effects on the economy to date, I am posting a QE tutorial in the form of a brief slideshow.
Will QE3 work? All but one of the voting members of the Federal Open Market Commission appear to think it is at least worth a try. In his August 31 speech at Jackson Hole, Fed Chairman Ben Bernanke cited estimates that QE1 and QE2 together have lowered long-term interest rates by 0.8 to 1.2 percentage points. He also suggested that output is 3 percent higher than it otherwise would be and that about 2 million additional jobs have been created as a result of QE1 and QE2.
However, Bernanke also warns that the Fed cannot solve the country’s economic problems alone. “It is critical,” he said, “that fiscal policymakers put in place a credible plan that sets the federal budget on a sustainable trajectory in the medium and longer runs.” At the same time, he warns against excessive short-term fiscal contraction, which could occur if Congress does nothing to mitigate the effects of the so-called fiscal cliff.
The bottom line: “Monetary policy cannot achieve by itself what a broader and more balanced set of economic policies might achieve; in particular, it cannot neutralize the fiscal and financial risks that the country faces. It certainly cannot fine-tune economic outcomes.”
Follow this link to view or download the Quantitative Easing Tutorial
29 Responses to “Quantitative Easing: A Tutorial”
Just to put in a plug for MMT … as I understand it, the focus of the slide show on reserves is rather misplaced. Banks do not need reserves to lend, nor do the current excess reserves encourage lending. The effects of QE are confined to reducing long market interest rates. Which, already being at rock bottom, has also not been terribly helpful. We need fiscal spending.
A question: I know that usually when the Fed credits banks with its ex nihilo money that this goes into reserves and is not M2, but adds to the monetary base unless banks lend, right? Is this going to be the same when the Fed under QEIII buys those mortgage backed securities? Is this the same when the Fed is buying the securities from banks or from 3d parties? Who exactly are they purchasing this mortgage debt from?
Actually, that's several questions
1. The funds used to pay for securities that the Fed purchases are credited to the bank accounts of the seller. The immediate effect is both an increase in reserves (part of the base) and the money supply (which includes bank balances). If we make the typical textbook assumption that the money multiplier (the ratio of M2 to the monetary base) is constant, then later, banks will use the new reserves to make new loans. Doing so adds still more to the money supply, but does so without causing any further change in the monetary base. However, in recent times, banks don't fully use their new reserves for loans; they just leave them on deposit with the Fed and collect the small interest the Fed pays (.25%). In that case, the money multiplier falls, making monetary policy less effective.
2. The effects on M2 are exactly the same no matter what kind of securities the Fed buys. However, the choice of securities may affect the relative price of, say, mortgage-backed securities vs. Treasury bonds.
3. They buy the bonds at auction on the secondary market. The proximate sellers are securities dealers, but the ultimate sellers for whom the dealers act as agents can be anyone who owned them previously, private persons, hedge funds, banks, foreign governments, whatever. The effects are the same no matter who the sellers are.
Informative tutorial. When the economy recovers and it is appropriate for the Fed to sell assets from it's portfolio and raise interest rates, will it be as easy to do as it was to lower
interest rates? By that time there could be a different Fed chairman, and there could be an asymmetry. Who in Washington wants to be responsible for causing the housing industry to contract?
Quantitative easing could have been accomplished by buying government bonds, rather than mortgage-backed securities. Is there a strong case for allocating credit to a particular sector? There is evidence that lowering credit standards by Fannie Mae and other government agencies contributed to the housing bubble that led to the Great Recession from which we have still not recovered.
1. There should be no technical difficulty with selling the Fed's holdings of QE securities at the appropriate time in the future. However, there could be two asymmetries, as you put it. One is political; a new Fed chair might either be reluctant to sell soon enough, fearing damage to the economy at a sensitive point in the election cycle, or might sell prematurely, based on fears of inflationary impact (fears that so far have proved ungrounded). The other possible asymmetry is that the securities might be sold at a loss. They are, after all, bonds, so their market price will fall when interest rates begin to rise, and presumably the Fed will not start reducing its portfolio on a large scale until interest rates to start to rise.
2. I don't want to second-guess the Fed as to its choice of MBS rather than Treasuries for QE3. I simply note that your criticism is widely shared by those who fear that when central banks select specific securities for purchase, they risk crossing a red line between monetary and fiscal policy. In a different way, the ECB is facing this same criticism with its plans to buy Spanish and Italian government bonds.
How is this not "pushing on a string"? "In 2007 commercial banking reserves at the Fed totaled $20.8 billion. By the end of 2008, that figure had ballooned to $860 billion. By the middle of 2011, it had reached $1.6 trillion, where it remains." (June 2012) — From Robert Pollin's paper The Great U.S. Liquidity Trap of 2009-2011. "Over this same time period [2007-2009] that the banks built up these massive cash reserves, credit stopped flowing into the non-corporate business sector in the U.S. For these smaller businesses, total borrowing fell from $546 billion in 2007 to negative $346 billion in 2009 — a nearly $900 billion reversal." And in the following 2 years the non-corporate sector "obtained zero net credit." Pollin's paper is a proposal to "pull on a string." His two interventions would be: "1)instituting an excess reserve tax and 2) expanding the federal government's loan-guarantee program to smaller businesses."
Also, a plan from the EPI, "Putting America Back to Work" includes a menu of 11 proposals which enlarges on Obama's Sept. 2011 jobs plan. Together the EPI's menu cost perhaps $500 billion, it's not clear, and about 15 million jobs would be created. I don't think the economics profession adequately explains or understands the parameters of a self-sustaining expansion. It has to do now with redressing the imbalances in wealth and income, but that may be verboten to any discussion, but the verboten is the necessary, IMO.
1. Yes, it is pushing on a string. Notice that in the graphs in the tutorial, the base, M2, and nominal GDP track closely together up to 2008. That is because in normal times, the money multiplier (M2/base) and velocity (NGDP/M2) don't vary much from year to year. When those parameters are constant, or roughly so, the Fed can slow NGDP by pulling the string on the monetary base. The charts show, however, that pushing the monetary base string has done very little. The controversy over QE all revolves around the idea that if the give the economy even more string, it will eventually take off. Michael Woodford has a very good discussion in his Jackson Hole paper. We really don't know if it will work until it is tried.
2. Yes, other unconventional measures like a tax on reserves might also be worth trying. Denmark has already done this by imposing a negative interest rate (i.e. a "tax") on reserves. Maybe we will learn something from that experiment. Some observers fear it will backfire. The fear is based on the idea that the negative interest rate is a new cost for banks, so banks will have to raise (not lower) their loan rates to make up for the new cost.
Good job for pointing out that QE creates bank deposits and reserves in equal amounts. Most explainers about QE fail to mention the bank deposits.
I would word things differently re: the broad/base money ratio. It shrank dramatically because base money was increased by a very large amount, period. Even if banks hadn't been reluctant to lend during QE1 and QE2, that would have made only a slight difference. The quantity of reserves injected would in even a very gung-ho economy have been enough to fund years of credit expansion and deposit growth before using up excess reserves. The average required reserves to deposits ratio is only a bit more than 1%.
I continue to be confused at how MMTers such as Buck manage to convince themselves of such nonsense. Banks do of course need reserves in order to lend. From the standpoint of the individual bank, it needs to either have excess reserves or be confident it can obtain additional reserves in order to lend. Imagine a small bank has no excess reserves and makes a loan of $1 million, while all its other incoming and outgoing payments balance. When the borrower spends that $1 million, the bank will be $1 million short of reserves, and will need to source them from somewhere. If it wasn't confident it could do that it wouldn't have made the loan. From the standpoint of the banking system, banks can't expand lending if there aren't excess reserves or a predictable flow of central bank injections of reserves, as expanding lending increases deposits and deposits increases required reserves. But from that overall perspective, lending and deposits can expand at up to many times the rate at which reserves expand.
Just so I'm not misunderstood, when I say that if banks weren't reluctant to lend it wouldn't have made much difference, I'm talking only about QE's impact on the broad/base money ratio. QE still would have greatly reduced the broad/base ratio even if banks were gung-ho to lend, because the reserves were simply much greater than banks could possibly need in the short term.
The more important impact of QE in my opinion is through its expansion of deposits. The amount of excess reserves becomes less and less important as it grows. The initial injections of reserves in 2008 mattered very much and are the reason short rates fell to zero. Since 2009 added reserves haven't mattered.
The only thing that restricts banks from lending is capital. Banks never worry about reserves. They are easy to find after a loan is made and if they can't find them, the Fed automatically loans it to them. Commercial banks cannot make private loans using excess reserves. They can only lend them to other banks which means the total amount of excess reserves in the banking system remains the same. Only the Fed can reduce the total amount of excess reserves.
The suggestion that QE has had the effects BB suggests is wrong. All the purported effects arise from assumptions about theoretical transmission routes. There is absolutely no empirical evidence that those transmission mechanisms are working, or that they would have the effect assumed. The only real effect of QE is to lower the cost of govt borrowing and take short-term pressure of the need for fiscal adjustments.
Great tutorial Ed, please allow me two questions (I hope not too basic):
1) if purchase of securities is paied to public with increased bank accounts I suppose that techinically M2 should increase (while M3 should decrease due to QE as securties should belong to M3 and not to M2).
Is that correct ?
If so, does M2 remain unchanged after Q3 mainly due to deleveraging decisions (and related decrease in banks accounts/loans to public) rather than as a consequence of unused excess reserves ?
2) why central bank may face a "problem" in case of a loss after reselling to the market securities purchased during QE ? As a central bank, it may "recover those losses" by increasing monetary base which it already creates.
Thanks in advance for your clarifications.
"There is absolutely no empirical evidence that those transmission mechanisms are working"
This statement needs qualification. There is lots of empirical evidence that the TM of QE are working; see the footnotes in the Jackson Hole papers of Bernanke and Woodford for numerous examples. However, there is also contrary evidence and even the positive evidence often shows the effects to be weak.
1) Depends on what you mean by M3 (a concept not much used in the US). If you mean an M3 includes the kind of short-term securities that the Fed was buying, the answer to your question would be yes. But in fact, this QE3 will purchase mainly longer-term mortgage backed securities that are not elements of any M3 measure I am aware of, so the answer in the present case is no.
2) There are really two problems that would arise if the Fed had to sell large volumes of securities at a loss.
One would be the fiscal effect: Usually the Fed makes a positive contribution to the federal budget, but if it were to operate at a loss, its impact on the budget would be negative, increasing rather than reducing the federal budget deficit.
The other effect is on the Fed's own solvency. The Fed is thinly capitalized by comparison with commercial banks. In principle, it could become insolvent if it sold large volumes of securities at a loss. It could not (as you suggest) fix this solvency problems by expanding the monetary base. The monetary base consists of liabilities (reserves and currency). Increasing assets and liabilities by equal amounts does not change its net worth. If the Fed were to become insolvent, presumably the Treasury would recapitalize it. That would be embarrassing, for one thing, and would also put a further strain on the Federal budget.
Actually, the whole solvency issue is an interesting one. A couple years ago, I wrote a post called "Could QE2 Cause the Fed to Go Broke?" Check it out here: http://dolanecon.blogspot.com/2010/11/could-qe2-c…
I haven't read all the papers mentioned in the footnotes but I have read carefully the BOE paper on the macroeconomic effects of QE. Although it it often cited as doing so- including in the JH papers – the BOE paper contains no empirical evidence of any macroeconomic effects of QE – see my blog from last October "MPC – Time for a Rethink" at http://drderrick.wordpress.com/2011/10/ While I remain deeply sceptical, it may be the case that some real evidence has been found in the US experience. I will review the papers to which you refer.
I'm sorry but you MMTers are just spouting misinformed dogma. Of course reserves matter to the amount of lending. The Fed tightens or loosens conditions by subtracting or adding reserves. Subtracting reserves raises the cost of borrowing reserves and thus restricts banks from lending. Adding reserves lowers the cost of borrowing reserves and thus encourages banks to lend, unless as since 2009 reserves are already so plentiful that banks rarely need to borrow them and the cost to do is minimal.
The Fed's lending to banks normally costs more than interbank lending and carries a stigma, and it isn't completely "automatic", banks do sometimes get liquidated.
Banks can of course make private loans with excess reserves, why on earth couldn't they.
I believe derrick is right that all of the academic studies that have concluded that QE lowered interest rates did so by estimating/guessing what rates would have been without QE. It's not really possible to do anything else.
I think the most important easing programs were those of fall 2008 to spring 2010, which included QE and others and are hard to disentangle from each other since they overlapped. The initial stages obviously lowered interest rates by lowering short rates to zero. The later stages dominated by so-called QE1 avoided a deflationary contraction of broad money that would have occurred as bank credit was shrinking and extinguishing bank deposits. I would say it certainly lowered rates on riskier bonds and lending, I think without QE there would have been a much longer and deeper sell-off of risk assets. But whether QE lowered rates on Treasurys is very hard to say. On one hand Treasury would have been selling into the market and not to investor groups that were switching out of agencies by selling to the Fed. On the other hand there would have been strong haven demand for Treasurys. I don't see what empirical evidence one could possibly find to settle the question.
I don't think anyone suggests QE hasn't had any effect on the interest rates paid on govt securities; a few hundred billion dollars of additional demand will have an effect on the price of anything! A benefit of that price effect is that financing govt debt has been cheaper than it would otherwise have been, and pressure to make needed fiscal adjustments has been lessened. What is at dispute is whether or not QE has had any further effects on the economy. Has it lowered the price of retail loans or increased their availability, and has it had any effect on aggregate demand? It is these latter effects that proponents of QE claim without adducing any hard evidence.
This whole discussion is interesting. Here is how I would sum it up so far:
Bank lending is subject to several constraints, including
1. A liquidity constraint, which can take the form of on-hand excess reserves (as in the traditional textbook model) or access to liquid funds through overnight fed funds, repos, etc. This constraint is not always binding, including right now.
2. A capital constraint, which is not enforced on a short-term basis and is not usually binding, but may be the strongest binding constraint under stress conditions.
3. A demand constraint, by which I mean the demand for loans represented as a risk-return frontier. Depending on the cost of funds, this constraint may become the binding one, in that there are no borrowers who meet the bank's minimum risk-return cutoffs.
Monetary policy can affect the economy only to the extent that it relaxes whatever constraint is currently binding.
In the traditional textbook model the liquidity constraint is assumed to be the binding one, true enough, for example, when the Fed is "pulling the string" to restrain a credit boom, but not, as now, when the Fed is "pushing the string."
2008 is a classic example of the situation when the capital constraint becomes binding. Then policymakers can only ease by injecting capital into the banking system. The Fed alone cannot do that; it takes something like TARP.
When the demand constraint is binding, policy can encourage lending only by shifting the demand frontier, for example, with tax cuts that encourage borrowing, or by some kind of regulatory forbearance that causes banks to lower their risk-return thresholds.
It is an open question whether QE3 will succeed in relaxing any relevant constraint to the extent needed to stimulate the larger economy.
There seems to be an inconsistency between the contents of slides 10 and 12, as follows:
"The first three steps in the open market operation result in an increase in bank reserves and an equal increase in the money stock, in the form of bank deposits."
"The money stock increased only slightly, because banks were reluctant to lend out all of their new liquid reserves."
Surely the only way that the sentence from slide 12 could be correct is if the bonds which the Fed bought were owned be the commercial banks themselves, rather than being owned by the banks' customers?
Good question. There are two parts to the answer.
One part has to do with the way the chart on slide 12 is scaled. As the legend notes, the lines do not represent absolute values, but instead, all variables are scaled so that Jan 2008 equals 100. So, even if M2 increased by the same amount as the monetary base (as the example on slide 10 shows), the M2 line would fall far below the base line. In order for the two to rise equally, M2 would have to increase by the ratio of M2 to the base that prevailed in Jan 2008, which was about 9:1.
The second part of the answer is that the effect shown in slide 10 holds precisely only as the immediate impact of the open market operation, with no time for banks to adjust their holdings of loans. In normal times, banks increase loans after an increase in their reserves, holding the ratio of loans to reserves approximately constant. However, in the period since 2008, banks have lent so little that M2 has actually increased by less than the amount of reserves. To oversimplify, imagine that rather than making new loans when QE increased their reserves, they made no new loans at all, instead letting old loans be repaid without rolling them over. The actuality is not that extreme, but you can see how the arithmetic would work.
Your suggestion that it has something to do with whether the Fed buys the bonds from the public or from banks is not quite on the right track. Who it buys from doesn't really make any difference.
Well, I think right now you're claiming that QE lowered rates on Treasurys without "hard evidence". You're assuming based on sound theory that tells you a new huge buyer lowers prices. Usually so, but in this case QE also might decrease haven demand for Treasurys and so raise rates. There have been periods during QE when Treasury rates were going up. It's impossible to say with certainty what they would have done without QE.
I don't fault anyone for taking positions and making arguments as to what QE's effects were or will be without hard evidence. We need to have the debate, and all evidence available is at least somewhat soft. The hardest evidence I can think of is to compare the periods during QE1 and QE2 with the periods between and after, but even that's muddled by all else that was going on. I would fault somebody if he claimed his evidence was harder than it really is.
Oops, I meant of course that a new huge buyer increase prices, lowering rates.
Thanks again Ed for making that point. The idea that it matters from whom the Fed buys is quite widespread.
What I think Jeffrey is getting at is the idea that if the Fed buys assets from banks, that doesn't increase M2, only M0. The Fed would just credit the banks' reserves account and no bank deposits would be directly created. However as soon as banks replaced the assets in their portfolios, they would create bank deposits for the sellers. The only way QE could avoid creating M2 would be if banks were net sellers using the opportunity of the Fed's bid to dispose of assets. Another way of putting that is that the Fed's purchases would have boosted M2 but the banks' shrank bank credit by an equal amount so no M2 was created. During QE1 & QE2 bank credit shrank, but not by as much as the Fed's purchases, and not because of any sales of assets by banks to the Fed.
Though it's not really important from whom the Fed buys, the fact is it buys from security brokers, who buy from original issuers and all varieties of investors.
Ed and Tom,
Thanks very much for your prompt and detailed replies. Both replies have helped.
So in slide 12, when it says "the money stock increased only slightly etc etc", is it really saying the following: "The money stock increased only by an amount approximately equal to the increase in bank reserves (i.e. there was virtually no multiplier effect) because banks were reluctant to undertake new lending" ?
Sorry, let me try to remove a bit more uncertainty.
"The money stock increased only by an amount approximately equal to the value of the bonds bought by the Fed (i.e. there was virtually no multiplier effect) because banks were reluctant to undertake new lending." ?
+++ Well put, I think you know this stuff better than I do
Right. Or "by even less than reserves increased"
Interesting question. I never looked it up before. You can find the data in the Fed's H-8 releases all on line. The latest release shows "loans and leases" as 3.8 times "cash items" (reserve deposits + vault cash). In Jan 2008 it was 20.9 times. Quite a change!