Alan S. Blinder, "Quantitative Easing: Entrance and Exit Strategies（pdf）"（CEPS Working Paper No. 204, March 2010）
明らかに、それ（＝量的緩和政策）はここアメリカにおいても採用される可能性がある。2008年12月16日、連邦公開市場委員会（FOMC）は、1937年以来最悪の不況に発展しそうな事態（＝大停滞）と闘うために、フェデラルファンド金利（FF金利）の誘導目標をほぼゼロ％に引き下げる決定を行った。その後、この大停滞と闘う中で伝統的な政策手段のすべてを使い尽くしたFedは、それはそれは素晴らしき量的緩和の世界へと勢いよく足を踏み入れることになった。ベン・バーナンキ議長はFedが実施した新しい政策を「信用緩和」（“credit easing” ）と呼ぶよう心がけたが－おそらく、日本銀行がそれ以前の時期において実施した政策と区別しようとしたのであろう－、その呼び方が定着することはなかった。
おおざっぱにいうと、量的緩和というのは、市中への流動性の注入（市場流動性の回復）や信用条件の緩和を目的として、中央銀行のバランスシート上における資産構成（composition）やバランスシートの規模を変化させる政策のことである。 量的緩和を反転させる（reverse）ことは「量的引き締め」（“quantitative tightening”）とでも呼ぶことができるのであろうが、誰一人としてこのように呼ぶ気配はないようである。通常の議論の中では、量的緩和を反転させることは「出口戦略」（“exit strategy”）と呼ばれている－「出口戦略」という言葉が使用される背後には、量的緩和（quantitative easing ；“QE”）は何か異常なものであるとの認識が控えているのであろう－。大勢にしたがって、本論文でも量的緩和を反転させることを「出口戦略」と呼ぶことにしよう。
以下で展開する議論の流れを説明すると、まずはじめに量的緩和の採用を後押しする概念的な根拠のスケッチを行う。そこでは、量的緩和がなぜ適当な政策となり得るのか、量的緩和はどのようなメカニズムを通じて機能することになるのか、といった問題を検討するであろう。次には、Fedの入口戦略（entrance strategy） －おそらくはもうすでに採用済みの過去の戦略ということになるのであろうが－に話題を転じ、その後に出口戦略－まだこの先の将来に問題となる戦略であろう－について議論する。Fedの入口戦略と出口戦略とを議論する際には、2001～2006年における日本の経験との簡単な比較が行われるであろう。そして最後に、量的緩和の採用に伴って引き起こされることになるであろう中央銀行の独立性にまつわる質問をいくつか取り上げたのちに、本論文の内容を簡潔に要約して議論の終わりとする。
量的緩和の採用を後押しする概念的な根拠: 流動性の罠（the liquidity trap）
まず当たり前な点から触れておくと、金融政策について学んでいるどの学生も、中央銀行が保有する伝統的な政策手段－オーバーナイト金利（アメリカでは「フェデラルファンド」金利）－の方が量的緩和よりも効果的で（powerful）頼りになる手段だ、と考えることだろうと思う。そうだとすれば、どうして理性的なセントラルバンカーたちは量的緩和なんかに打って出ようとするのだろうか？ その答えはあまりにも明白である。経済が非常に厳しい逆境に置かれているような状況において、中央銀行が名目金利をゼロ％にまで引き下げてもなお経済を十分に刺激するには至らない、といったケースが生じることがあるからである。名目金利がゼロ％の下限に達してしまっているような状況は、「流動性の罠」（“liquidity trap”）との名称－ケインズのオリジナルの意味とは若干異なった使用法ではあるが－で呼ばれている (Krugman, 1998)。
なぜ経済が「流動性の罠」に陥ることがあるのか、そのロジックを簡単に見ておこう。まず前提として、総需要の決定において重要となるのは、名目金利（i）ではなく実質金利（r）である、と想定される。深刻な不況下では、時に、中央銀行は実質金利（r = i – π；π はインフレ率を表している）をマイナスの水準にまで引き下げる必要に迫られることがあるが、一度名目金利（i）がゼロ％に達してしまうと、それ以上名目金利を引き下げることはできなくなってしまう。結果として、実質金利（r）は–π の水準－低水準ではあるだろうがおそらくはプラスの水準ーにとどまることになる。一度名目金利がゼロ％に達してしまうと（i=0）、伝統的な金融政策は「スッカラカン状態」（“out of bullets”）になってしまう（＝打つ手がなくなる）のである。
Actually, the situation is even worse than that. Recall Milton Friedman’s (1968) warning against fixing the nominal interest rate when inflation is either rising or falling: Doing so invites dynamic instability. Well, once the nominal rate is stuck at zero, it is, of course, fixed. If inflation then falls, the real interest rate will rise further, thereby squeezing the economy even more. This is a recipe for deflationary implosion.
Enter quantitative easing. Suppose that, while the riskless overnight rate is constrained to zero, the central bank has some unconventional policy instruments that it can use to reduce interest rate spreads—such as term premiums and/or risk premiums. If flattening the yield curve and/or shrinking risk premiums can boost aggregate demand, then monetary policy is not powerless at the zero lower bound.6 In that case, a central bank that pursues QE with sufficient vigor can break the potentially vicious downward cycle of deflation, weaker aggregate demand, more deflation, and so on.
What might such an arsenal of unconventional weapons contain? While the following list is hypothetical and conceptual, every item on it has a clear counterpart in something the Federal Reserve has actually done.
First, suppose the objective is to flatten the yield curve, perhaps because long rates have more powerful effects on spending than short rates. There are two main options. One is to utilize “open-mouth policy.” The central bank can commit to keeping the overnight rate at or near zero either for, say, “an extended period” (or some such phrase) or until, say, inflation rises above a certain level. To the extent that the (rational) expectations theory of the term structure is valid, and the commitment is credible, doing so should reduce long rates and thereby stimulate demand.7 But this would not normally be considered quantitative easing because no quantity on the central bank’s balance sheet is affected.
The QE approach to the term structure is straightforward: Use otherwise-conventional open-market purchases to acquire longer-term government securities instead of the short-term bills that central banks normally buy. If arbitrage along the yield curve is imperfect, perhaps because asset-holders have “preferred habitats,” then such operations can push long rates down by shrinking term premiums.8
The second target of QE is risk or liquidity spreads. Every private debt instrument, even bank deposits and AAA bonds, pays a spread over Treasuries for one or both of these reasons.9 Since private borrowing, lending, and spending decisions presumably depend on (risky) non-Treasury rates, reducing their spreads over (riskless) Treasuries will reduce the interest rates that matter for actual transactions even if riskless rates are unchanged.
How might a central bank accomplish that? The most obvious approach is to buy any of a wide variety of risky and/or less liquid assets, paying either by selling some Treasuries out of its portfolio, which would change the composition of its balance sheet, or by creating new base money, which would increase the size of its balance sheet.10 Either variant can be said to constitute QE, and its effectiveness depends on the degree of substitutability across the assets being traded. As we know, buying X and selling Y does nothing if X and Y are perfect substitutes.11 Fortunately, it seems unlikely that, say, mortgage-backed securities (MBS) are perfect substitutes for Treasuries—certainly not in a crisis.
The Fed’s entrance strategy
With this conceptual framework in mind, I turn now to what the Federal Reserve actually did as it embarked on its new strategy of quantitative easing. Because the messy failure of Lehman Brothers in mid-September 2008 was such a watershed, I begin the story a bit earlier.
Reacting somewhat late to the onset of the financial crisis in the summer of 2007, the FOMC began cutting the federal funds rate on September 18, 2007—starting from an initial setting of 5.25%. While it cut rates rapidly by historical standards, the Fed did not signal any great sense of urgency. It was not until April 30, 2008 that the funds rate got down to 2%, where the FOMC decided to keep it while awaiting further developments. (See Chart 1.) Perhaps more germane to the QE story, the Fed was neither expanding its balance sheet (see Chart 2) nor increasing bank reserves (see Chart 3) much over this period.
However, the Fed was already engaging in several forms of quantitative easing, even apart from emergency interventions such as the Bear Stearns rescue. To understand these brands of QE, it is useful to refer to the oversimplified central bank balance sheet just below. Because other balance sheet items are inessential to my story, I omit them.
The first type of QE showed up entirely on the assets side. Early in 2008, the Fed started selling down its holdings of Treasuries and buying other, less liquid, assets instead. (See Chart 2.) This change in the composition of the Fed’s portfolio was clearly intended to provide more liquidity (especially more T-bills) to markets that were thirsting for it. The goal was to reduce liquidity premiums. But, of course, the underlying financial situation was deteriorating all the while, and the markets’ real problems may have been fears of insolvency, not illiquidity—to the extent you can distinguish between the two.12
The second sort of early QE operations began on the liabilities side of the Fed’s balance sheet. To assist the Fed, the Treasury started borrowing in advance of its needs (which were not yet as ample as they would become later) and depositing the excess funds in its accounts at the central bank. While these were fiscal operations, they enabled the Fed to increase its assets—by purchasing more securities and making more discount window loans (e.g., through TAF, the Term Auction Facility)—without increasing bank reserves (see Chart 3). That’s very helpful to a central bank that is a bit timid about stimulating aggregate demand and/or is worried about running out of T-bills to sell, both of which were probably true of the Fed then. But notice that these operations marked the first breaching—however minor—of the wall between fiscal and monetary policy. In addition, the Fed began lending to primary dealers in the immediate aftermath of the Bear Stearns rescue.
Then came the failure of Lehman Brothers, and everything changed, including the Fed’s monetary policy.
The FOMC resumed cutting interest rates at its October 10, 2008 meeting, eventually pushing the funds rate all the way down to virtually zero by December 16th. (See Chart 1 again.) More germane to the QE story, the Fed started expanding its balance sheet, its lending operations, and bank reserves immediately and dramatically. (See Charts 2 and 3.)13 By the last quarter of 2008, any reservations or hesitation at the Fed about boosting aggregate demand were gone. It was “battle stations.”
Total Federal Reserve assets skyrocketed from $907 billion on September 3, 2008 to $2.214 trillion on November 12, 2008. By the last quarter of 2008, any reservations or hesitation at the Fed about boosting aggregate demand were gone. It was “battle stations.”14 (Chart 2.) As this was happening, the Fed was acquiring a wide variety of securities that it had not owned before (e.g., commercial paper) and making types of loans that it had not made before (e.g., to nonbanks). On the liabilities side, bank reserves ballooned from about $11 billion to an astounding $594 billion over that same period—and then to $860 billion on the last day of 2008 (Chart 3). Almost all of this expansion signified increased excess reserves, which were a negligible $2 billion in the month before Lehman collapsed (August) but soared to $767 billion by December.15 Since the Fed’s capital barely changed over this short period, its balance sheet became extremely leveraged in the process. Specifically, the Fed’s leverage (assets divided by capital) soared from about 22:1 to about 53:1. It was a new world, Tevye.16
The early stages of the quantitative easing policy were extremely ad hoc, reactive, and institution-based. The Fed was making things up on the fly, often acquiring assets in the context of rescue operations for specific companies on very short notice, e.g., the Maiden Lane facilities for Bear Stearns and AIG. But, starting with the Commercial Paper Funding Facility (CPFF) in September 2008, and continuing through the MBS purchase program (announced in March 2009), the TALF (Term Asset-Backed Securities Loan Facility, started in March 2009), and others, the Fed’s parade of innovative purchase, lending, and guarantee programs took on a more systematic, thoughtful, and market-based flavor. The idea now was not so much to save faltering institutions, although that potential need remained, but rather to push down risk premiums, which had soared to dizzying heights during the panic-stricken months of September-November 2008 (and then did so again in February-March 2009).17
This change in focus was notable. It was also smart, in my view. As mentioned earlier, riskless rates per se are almost irrelevant to economic activity. The traditional power of the funds rate derives from the fact that risk premiums between it and the (risky) rates that actually matter—rates on business and consumer loans, mortgages, corporate bonds, and so on—do not change much in normal times. Think of the interest rate on instrument j, say Ri, as being composed of the corresponding riskless rate, r, plus a risk premium specific to that instrument, say ρi. Thus Ri = r + ρi. If the ρi change little, then control of r is a powerful tool for manipulating the interest rates that matter—and hence aggregate demand. But when the ρi move around a lot—in this case rising—the funds rate becomes a weak and unreliable policy instrument. During the panicky periods, in fact, most of the Ri were rising even though r was either constant or falling.
While I will have more to say about the Japanese experience later, one sharp contrast between QE in the U.S. and QE in Japan is worth pointing out right now. The Bank of Japan concentrated its QE on bringing down term premiums, mainly by buying long-term government bonds (JGBs). By contrast, until it started buying long-term Treasuries in March 2009, the Fed’s QE efforts concentrated on bringing down risk premiums, which involved a potpourri of market-by-market policies. It was far more complicated, to be sure, but in my view, also far more effective.
In fact, the one aspect of the Fed’s QE campaign of which I have been critical is its purchases of Treasury bonds. The problem in many markets was that the sum r + ρi, was too high--but mainly because of sky-high risk premiums, not high risk-free rates. Thus the real target of opportunity was clearly ρi, not r, which was already low. Furthermore, a steep yield curve provides profitable opportunities for banks to recapitalize themselves without taxpayer assistance. Why undermine that?
In any case, the Fed’s QE attack on interest rate spreads appears to have been successful, at least in part. Charts 4 and 5 display two different interest rate spreads, one short term and the other long term. Chart 4 shows the spread between the interest rates on three-month financial commercial paper and three month Treasury bills; Chart 5 shows the spread between Moody’s Baa corporate bonds and ten-year Treasury notes. The diagrams differ in details, with, e.g., short rates much more volatile than long rates. But both convey the same basic message: Once the Fed embarked on QE in a major way, spreads tumbled dramatically. Admittedly, other things were changing in markets at the same time; so this was hardly a controlled experiment. Still, the “coincidence” in timing is quite suggestive.
The Fed’s exit strategy
The Fed’s exit is still in its infancy. Chairman Bernanke first outlined the major components of its strategy in July 2009 Congressional testimony, and then followed that up with a speech in October 2009 and further testimonies in February and March 2010.18 So by now we have a pretty good picture of the Fed’s planned exit strategy. Here are the key elements, listed in what may or may not prove to be the correct temporal order: 19
1. “In designing its [extraordinary liquidity] facilities, [the Fed] incorporated features… aimed at encouraging borrowers to reduce their use of the facilities as financial conditions returned to normal.” (p. 4n)
2. “normalizing the terms of regular discount window loans” (p. 4)
3. “passively redeeming agency debt and MBS as they mature or are repaid” (p. 9)
4. “increasing the interest on reserves” (p. 7)20
5. “offer depository institutions term deposits, which… could not be counted as reserves.” (p. 8)
6. “reducing the quantity of reserves” via “reverse repurchase agreements” (p. 7)
7. “redeeming or selling securities” (p. 8) in conventional open-market operations.
Notice that this list deftly omits any mention of raising the federal funds rate. But the funds rate will presumably not wait until all the other steps have been completed. Indeed, Bernanke (2010a) noted that “the federal funds rate could for a time become a less reliable indicator than usual of conditions in short-term money markets,” so that instead “it is possible that the Federal Reserve could for a time use the interest rate paid on reserves… as a guide to its policy stance” (p. 10). I will return to this not-so-subtle hint shortly.
The first and third items on this list are the parts of “quantitative tightening” that the Fed gets for free, analogous to letting assets run off naturally. As the Fed has noted repeatedly, its special liquidity facilities were designed to be unattractive in normal times, and Item 1 is by now pretty close to complete. The Fed’s two commercial paper facilities (one designed to save the money market mutual funds) outlived their usefulness, saw their usage drop to zero, and were officially closed on February 1, 2010. The same was true of the lending facility for primary dealers, the Term Securities Lending Facility, and the extraordinary swap arrangements with foreign central banks. The TAF and the MBS purchase program were just completed, and the TALF is slated to follow suit at the end of June.
Item 2 on this list (raising the discount rate) is necessary in order to supplement Item 1 (making borrowing less attractive), and the Fed began doing so with a surprise inter-meeting announcement on February 18, 2010. A higher discount rate is also needed to enable the Fed to shift to the “corridor” system discussed below.
Note, however, that all these adjustments in liquidity facilities will still leave the Fed’s balance sheet with the Bear Stearns and AIG assets and huge volumes of MBS and GSE debt. Now that new purchases have stopped, the stocks of these two asset classes will gradually dwindle over time (Item 3 on the list). But unless there are aggressive open-market sales, it will be a long time before the Fed’s balance sheet resembles the status quo ante.
That brings me to Items 6 and 7 on Bernanke’s list, which are two types of conventional contractionary open-market operations, either done via reverse repo (and thus temporary) or outright sales (and thus permanent). Transactions like these have long been familiar to anyone who pays any attention to monetary policy, as are their normal effects on interest rates.
However, there is a key distinction between Items 1 and 3 (lending facilities), on the one hand, and Items 6 and 7 (open market operations), on the other, when it comes to degree of difficulty. Quantitative easing under Item 1, in particular, wears off naturally, on the markets’ own rhythm: These special liquidity facilities fall into disuse as and when the market no longer needs them. From the point of view of the central bank, this is ideal because the exit is perfectly timed, almost by definition.
Items 6 and 7 are different. The FOMC will have to decide on the pace of its open-market sales, just as it does in any tightening cycle. But both the volume and the variety of assets to be sold will probably be huge this time around. Of course, the FOMC will get the usual market and macro signals: movements in asset prices and interest rates, the changing macro outlook, inflation and inflationary expectations, etc. But its decisionmaking will be more difficult, and more consequential than usual, because of the enormous scale of the tightening. If the Fed tightens too quickly, it may stunt or even abort the recovery. If it waits too long, inflation may gather steam. Once the Fed’s policy rates get lifted off zero, short-term interest rates will presumably be the Fed’s main guidepost once again—more or less as in the past.
This discussion leads naturally to Item 5 on Bernanke’s list, the novel plan to offer banks new types of accounts “which are roughly analogous to certificates of deposit” (p. 8). That is, instead of just having a “checking account” at the Fed, banks will be offered the option of buying various “CDs.” But here’s the wrinkle: Unlike their checking account balances at the Fed, the CDs will not count as official reserves. Thus, when a bank transfers money from its checking account to its saving account, as individuals do all the time, bank reserves will simply vanish.
The potential utility of this new instrument to a central bank wanting to drain reserves is evident. The Fed has announced its intention to auction off fixed volumes of CDs of various maturities, probably ranging from one to six months. Such auctions would give it perfect control over the quantities but leave the corresponding interest rates to be determined by the market. Frankly, I wonder why these new fixed-income instruments would be attractive to banks since they cannot be withdrawn prior to maturity, do not constitute reserves, and cannot serve as clearing balances. In consequence, they may have to bear interest rates higher than those on Treasury bills. We’ll see.
I come, finally, to the instrument that Bernanke and the Fed seem to view as most central to their exit strategy: the interest rate paid on bank reserves. Fed officials seem to view paying interest on reserves as something akin to the magic bullet. I hope they are right, but confess to being a bit worried. Everyone recognizes that the Fed’s QE operations have created a veritable mountain of excess reserves (shown in Chart 3), which U.S. banks are currently holding voluntarily, despite the paltry rates paid by the Fed. The question is: How urgent is it—or will it become—to whittle this mountain down to size?
One view sees all those excess reserves as potential financial kindling that will prove inflationary unless withdrawn from the system as financial conditions normalize.21 We know that under normal circumstances, and before interest was paid on reserves, banks’ demand for excess reserves was virtually zero. But now that reserves earn interest, say at a rate z which the Fed controls, banks probably won’t want to reduce their reserves all the way back to zero. Instead, excess reserves now compete with other very short-term safe assets, such as T-bills, in banks’ asset portfolios.22 Indeed, one can argue that, for banks, reserves are now almost perfect substitutes for T-bills. So excess reserve holdings won’t have to fall all the way back to zero. Rather, the Fed’s looming task will be to reduce the supply of excess reserves at the same pace that banks reduce their demands for them. The questions are how fast that will be and how far the process will go. Notice that, as the Fed’s liabilities shrink, so must its assets. So as it reduces bank reserves, the Fed must also reduce some of the loans and/or less liquid assets now on its balance sheet.
There is, however, an alternative view that argues that the large apparent “overhang” of excess reserves is nothing to worry about. Specifically, once the relevant market interest rate (r) falls to the level of the interest rate paid on reserves (z), the demand for excess reserves becomes infinitely elastic (horizontal) at the opportunity cost of zero (r-z=0), making the effective demand curve DKM rather than DD in Chart 6.23 Another way to state the point is to note that banks will not supply federal funds to the marketplace at any rate below z because they can always earn z by depositing the funds with the Fed.
As Chart 6 shows, as long as the (vertical) supply curve of reserves, SS, which the Fed controls, cuts the demand curve in its horizontal segment, KM, the quantity of reserves should have no effect on the market interest rate, which is stuck at z. Therefore, the quantity of reserves should presumably have no effects on anything else, either. Infinitely elastic demand presumably means that any volume of reserves can remain on banks’ balance sheets indefinitely without kindling inflation. It also means that the Fed’s exit decisions should concentrate on how quickly to shrink the assets side of its balance sheet. The liabilities side, in this view, is the passive partner and matters little per se.
The idea of establishing either an interest rate floor, as depicted in Chart 6, or an interest corridor, as depicted in Chart 7 below, may become the Fed’s new operating procedure.24 The corridor system starts with the floor just explained and adds a ceiling above which the funds rate cannot go. That ceiling is the Fed’s discount rate, d, because no bank will pay more than d to borrow federal funds in the marketplace if it can borrow at rate d from the Fed.
25 The Fed’s policymakers can then set the upper and lower bounds of the corridor (d and z) and let the funds rate float--whether freely or managed--between these two limits. Under such a system, the lower bound—the rate paid on reserves, z—could become the Fed’s active policy instrument, with the discount rate set mechanically, say, 100 basis points or so higher.26
If the federal funds rate is free to float within the corridor, rather than stuck at the floor or ceiling, the Fed would be able to use it as a valuable information variable. If the funds rate traded up too rapidly, that would indicate the Fed was withdrawing reserves too quickly, creating more scarcity than it wants. If funds traded down too far, that would indicate that reserves were too abundant, that is, the Fed was withdrawing them too slowly. Such information should help the Fed time its exit.