How does Operation Twist differ from QE?

In a direct sense, almost none at all. Despite the common perception that “Operation Twist” is an ineffectual, conservative move, while further quantitative easing would be a powerful and risky one, the fundamental economic difference between them is quite minor.

But maybe perception itself is the problem.

In its press release two weeks ago, the Fed pledged to:

…purchase, by the end of June 2012, $400 billion of Treasury securities with remaining maturities of 6 years to 30 years and to sell an equal amount of Treasury securities with remaining maturities of 3 years or less.

Is this different from quantitative easing? QE2 was equivalent to the combination of two open market operations:

  • (1) Buying short-term Treasuries with newly created money.
  • (2) Swapping short-term Treasuries for longer-maturity ones.

The Fed’s new policy is just operation (2), disconnected from (1). Operation Twist is less effective than a potential QE3, therefore, to precisely the extent that operation (1) makes a difference.

Does it? First, let’s be even more precise, breaking down (1) into two smaller components:

  • (1A) Buying T-bills (extremely short term Treasuries with duration less than a year) with newly created money.
  • (1B) Swapping T-bills for a broader mix of short-term Treasuries (e.g. those with remaining maturity “3 years or less”).
  • (2) Swapping short-term Treasuries for longer-maturity ones.

Virtually everyone agrees that (1A) is useless: yields on 1-year T-bills hover around 0.1%, which is almost exactly the same as the effective federal funds rate. Yields on shorter T-bills are even lower. For all practical purposes, T-bills and reserves are equivalent assets—they’re both extremely liquid and offer a safe nominal return. Exchanging them does nothing.

If there’s a difference between Operation Twist and QE, then, it really has to be in (1B). This is already a little amusing: all the rhetoric contrasting Twist and QE, if it has any logical interpretation, boils down to an extremely specific statement about the effects of a particular maturity swap. (I’m pretty sure this is not what everybody is thinking!)

Of course, it’s conceptually possible that (1B) matters. Maybe reserves and T-bills do have some special “moneyness” (e.g. liquidity value) that 1-3 year Treasuries do not, and by expanding the relative supply of this moneyness we can make it less valuable—thereby pushing down yields on the securities that lack it. But this is easy to check: let’s just look up the difference in yields! This will tell us the maximum possible effect of the policy.

The yields on 2 and 3-year Treasuries are currently 0.25% and 0.42%, respectively. Needless to say, these aren’t much different from the current 0.1% yield on reserves and 1-year T-bills. But that actually isn’t the right comparison—we should be looking at the rates on 2 or 3-year debt versus the rate expected on reserves/T-bills over the next 2 or 3 years. This will tell us whether slightly longer-term securities are trading at a discount because they lack the features of money. As it turns out, the average rate over the next two years on the Fed Futures market is 0.15%, while the average rate over the next three years (assuming Sep. 2014, which is missing, is the same as Aug. 2014) is 0.32%.

In both cases, the Treasury yields are 0.1% higher than the corresponding average of forward rates. This is the gap that (1B) might address. It’s not zero, but it’s incredibly small relative to the other possible impacts of the policy, like the impact on long rates (even a 0.1% decrease in yield on the 10-year would be vastly more important, since it implies a much larger increase in price than the same change on a 2-year), or signaling.

To sum up: the fundamental difference between Operation Twist and QE, which boils down to the effect of a very specific asset swap, is extraordinarily minor.

Of course, most people disagree with this analysis—they think, for whatever reason, that QE is much stronger stuff. As I’ve explained, this doesn’t make sense from a fundamental perspective, but nevertheless it may be partly self-fulfilling. After all, perceptions matter.

This is possible even if you assume market participants themselves are rational and understand monetary policy. Consider the following simple model of the Federal Reserve: it wants to make monetary policy easier, but it’s constrained by political pressure from the Rick Perrys of the world and internal pressure from the Richard Fishers. Its decision to try Operation Twist, a mostly equivalent but marginally less effective alternative to QE, is a signal that it’s bowing to pressure from politicians and FOMC hawks, who irrationally think that QE is much more dangerous. While the true effect of replacing QE with Twist is minor, this signal about the Fed’s decision process has serious implications for how policy will be set in the future, and that’s extremely important.

Alternatively, maybe 20% of the market doesn’t understand monetary policy and thinks that QE will be wildly inflationary. Here we hit upon a quirk of monetary policy: expectations of inflation can be partly self-fulfilling, especially at the zero lower bound (where the Fed doesn’t move to counteract them). That 20% will purchase more (after all, their savings are about to be inflated away!), write contracts that embed inflation, hike prices, and so on. This leads to a little more inflation. Anticipating this, the other 80% adjusts its inflation expectations upward. This will lead to even more inflation. Anticipating this, the rational 80% adjusts expectations even more, and so on. Bottom line: under conditions of strategic complementarity, which many economists believe apply to price-setting, a small population with irrational beliefs can make a big difference.

Both these effects probably play a role. And that’s why, even though economics tells us that Twist is essentially the same as QE, it may ultimately have far less of an impact—precisely because so many people think it’s less potent.

(Yes, it’s frustrating to be an economist sometimes.)

About these ads

15 Comments

Filed under macro

15 responses to “How does Operation Twist differ from QE?

  1. JKH

    On the flip side of perception, twist is a post-QE clean up to what otherwise would be an inefficient use of the Fed’s existing balance sheet.

    The overall objective (with QE or twist) is to remove duration from the market.

    Once a threshold of duration extraction has been achieved with QE, it’s logical to examine potential inconsistencies in the entire remaining legacy portfolio.

    And if the continuing objective is to remove duration from the market (increasing duration at the Fed), at some point there’s no good reason to be running fairly short dated asset duration against the short dated interest rate sensitivity of reserves.

    So swap short dated asset duration next, rather than expand the balance sheet.

    • I agree, but only subject to the artificial yet politically realistic condition that the overall size of the balance sheet matters. If, for whatever reason, there is some cost to increasing the size of the Fed’s balance sheet by an additional dollar, then of course (to the extent that it is possible) it is logical to remove duration from the market through an operation like Twist that does not increase that size.

      A more economically meaningful measure of the Fed’s balance sheet, however, would recognize that long-duration assets are vastly different from short-duration assets: when the “size” of the balance sheet expands by printing money to purchase a T-bill, the risk and return characteristics of both public and private balance sheets are left essentially unchanged. When a T-bill is swapped for a 10-year T-note on the Fed’s balance sheet, on the other hand, the effective size of the Fed’s intervention in terms of redistributing risk between private and public hands grows, even if the nominal “size” remains the same.

      In practice, and for no particularly good reason, pundits pay attention to the nominal size of the Fed’s balance sheet. As I tried to hint in the post, the policy implications of this fact can go either way. If you want to achieve the maximal portfolio balance effects subject to some political constraint on nominal size, indeed you should do Twist instead of QE. If you believe that most of the effects will actually be achieved through signaling, and perhaps the manipulation of irrational market participants who form inflation expectations based on the wrong proxy (e.g. the amount of base money), then you want to blow up the nominal size of the balance sheet as much as possible.

      • JKH

        I agree with all that, but I’ll make one additional point.

        As you say, I think, net dollar duration of nominal balance sheet size is one measure of effective balance sheet size for the purpose of gauging interest rate risk (weighted average duration of assets minus weighted average duration of liabilities). The twisted balance sheet can always be replicated with a non-twisted QE’d balance sheet in the sense of such a measure of risk.

        However, there is a difference at play between the two alternatives in terms of the full nature of the interest rate risk exposure for the Fed.

        There are two high level types of interest rate risk at play in commercial banking, and these would also apply to the Fed. The first is net interest margin risk. The second goes by various labels, including economic value risk, marked to market risk, present value risk, duration risk, etc. Volumes can be written about how these play out on a commercial bank balance sheet, and maybe on a Fed balance sheet as well. One of the key differences relates to the effect of time on the two measures of risk. It is also important here that the Fed accounts for its “bottom line” mostly on an accrual accounting basis for interest rate risk (as opposed to marked to market), and this accounting approach corresponds mostly to the net interest margin form of interest rate risk.

        Consider the $ 400 billion nominal twist differential in question.

        Twist will deliver $ 400 billion of net duration addition of some magnitude (quantified for example as the average net duration addition per nominal dollar) to the Fed balance sheet without changing its nominal size. In particular, bank reserves as a Fed “funding” quantity will not be affected, holding other things equal.

        QE could deliver the same $ 400 billion of net duration with a $ 400 billion nominal expansion of the balance sheet. It just requires calibrating the target duration of acquired assets in accordance with the same net effect as the twisted duration result. (The Fed could calibrate the same effect with other choices for the size of the nominal QE expanded balance sheet, but I use $ 400 billion as an easy comparison against the twist program.)

        So with the QE alternative, the nominal balance sheet is $ 400 billion larger. Other things equal, that means an additional $ 400 billion in reserves. This has an effect on one of the measures of interest rate risk.

        If you look at the net interest margin sensitivity of the Fed balance sheet to an increase in the Fed funds target (which means an increase in interest on reserves), there is more up front gross sensitivity under the QE alternative. This is the result of the fact that the otherwise twisted short term assets are still on the balance sheet, effectively funded by reserves, which are immediately interest rate sensitive to an increase in the target fed funds rate and therefore on the rate paid on reserves. So if the funds rate and interest on reserves do increase, there will in aggregate be a greater compression of net interest margin than would be the case with QE – because there’s an additional $ 400 billion of asset-liability margin to compress than with twist. And this is all notwithstanding that the net duration exposure of the Fed balance sheet may be the same in both cases.

        The flip side of this difference in net interest margin risk is the return situation under the two portfolio alternatives. If the otherwise twisted short term assets are currently yielding a positive carry against the cost of reserves (which they almost certainly are), then the early stage net interest margin performance of the Fed would almost certainly be superior under QE than under twist – again because there is an additional $ 400 billion of asset-liability margin with positive carry (the actual comparison might depend on a slightly shorter duration for assets purchased under QE, in order to equate the duration result of the two alternative approaches). That positive carry represents the return offset to the additional net interest margin risk that is assumed in respect of a Fed tightening scenario. And this again is notwithstanding the fact that the net duration exposure of the balance sheet is the same in both cases.

        Commercial banks tend to look at the two types of risk together, striking some sort of balance between total duration exposure and the distribution of that exposure over future time as reflected in scenarios for net interest margin performance. (It is not an exact correspondence, but it is an important relationship.) I think the Fed would do the same in managing its interest rate risk.

        All that said, I haven’t looked at the details of the program closely enough to judge whether this interest margin risk I’m referring to is very material, given prospects for a prolonged visit with the zero bound. I suspect it’s a quantitatively small consideration in the Fed’s overall assessment of likely interest rate scenarios. Anyway, these short dated assets that are being twisted out would mature at some point under the QE alternative. Nevertheless, this is an issue which the Fed would be looking at and covering off in its interest rate risk management, and covering off in the most efficient way I presume. Again, and maybe partly for this reason, I see no reason for the Fed to leave this piece of short term interest rate risk dangling on its balance sheet by choosing incremental QE over a balance sheet cleansing twist at this point.

      • JKH

        meant:

        “So if the funds rate and interest on reserves do increase, there will in aggregate be a greater compression of net interest margin than would be the case under twist – because there’s an additional $ 400 billion of asset-liability margin to compress than with twist.”

  2. Dave

    Could the real purpose be for the FED to lock in long term low interest rates on the US debt while it can? Interest rate will be higher at some point in the future.

    • If the Fed wants to lock in long term low interest rates, it should be doing exactly the opposite of Twist: it should be selling long-term bonds at their current extremely high prices and buying short-term debt.

      In fact, the Treasury has recently been trying to lock in low long term interest rates by increasing the average maturity of its debt. As James Hamilton points out, this has substantially offset the effect of QE.

  3. Pingback: Economist's View: links for 2011-10-03

  4. Pingback: Monday Morning Reads « Sky Dancing

  5. Matt Waters

    I would say that the new money does make a significant difference and not just because of psychology.

    For example, the Fed targets the Fed funds rate by buying short-dated Treasuries. The Open Market desk in New York would theoretically print an unlimited amount of money until the interest rate goes down, or sell a virtually unlimited number of bonds until interest rates go up. Since both strategies involve increasing or decreasing the Fed’s balance sheet, Fed funds rate targeting is akin to QE even if the “quantity” isn’t set. If it wants to reduce the interest rate and it doesn’t have “credibility,” then it keeps printing money and buying bonds until its liquidity overwhelms any other market actor.

    Let’s say, hypothetically, that the Fed never expanded or contracted its balance sheet and instead kept a balanced portfolio of long-term and short-term bonds. When it wanted to reduce the Fed funds rate, it would buy short-term bonds and sell its long-term bonds, and vice versa for raising the rate. What would happen then? The Fed would need the cooperation of the market going in the same direction because there would ultimately be a bound on the bonds it buys and sells. Sometimes the market would be long Treasuries the same time as the Fed, for whatever reason, and the Fed could actually meet the interest rate target. But other times, the market would be short Treasuries, again for whatever reason, and the Fed could not meet the interest rate target.

    Extrapolating Fed Funds rate targeting to QE vs. Twist, let’s say that the Fed came out and said that they would keep buying bonds with new money until NGDP went up. What if the market is short NGDP because of bad macro news, if you will? That does NOT mean that the Fed is powerless. It just means the Fed has to buy more bonds. That’s it. Yes, buying short-term T-bills will do very little, but buying longer-dated bonds does have an effect on NGDP. If long-term bond holders truly wanted to just hold on to cash and take zero risk, they would not take on interest rate risk by buying Treasuries at historic highs. If the Fed was willing to buy all Treasuries up to monetizing the national debt, I do not see how much of that money would not be spent and NGDP would not go up. Regardless of interest rates, the new money would be spent one way or another.

    However, if we have “Operation Twist” and the Fed refuses to expand its balance sheet, then the Fed acknowledges a bound to its policy. NGDP may still go up if the market feels bullish on NGDP for whatever reason, but the market can just as easily feel bearish for whatever reason. If the market feels bearish, then “Twist” would have little-to-no efficacy in decreasing unemployment.

    But the market being bearish is ONLY an issue if the Fed bounds itself into a “self-induced paralysis” as Bernanke said on Japan. It would also be an issue if the Fed took that same approach to the Fed Funds rate. If the market does not help the Fed raise NGDP through QE, then so be it. Do more QE with new money (i.e. unbounded by short-term bond holdings) until NGDP does go up.

    • I think this is mainly true, but that it’s conceptually a little different from the point I was trying to make. QE has more potential than Twist in the sense that its scale can be unlimited, while Twist’s scale is limited by the size of the Fed’s current portfolio. I’m not disputing this; instead, I’m saying that conditional on the size of the intervention, there is virtually no difference between QE and Twist. This is consistent with the observation that QE is theoretically capable of much larger interventions. (I should have been more explicit about this in the post, of course.)

      I am more skeptical than most market monetarists about the direct NGDP-boosting effects of long-duration asset purchases. There’s a lot of uncertainty about the scale of the effect, but it seems completely plausible to me that the Fed could, say, buy up basically the entire marketable supply of Treasuries and Agencies and still fail to hit its NGDP target. (Though I’m sure something would happen!)

      Unless some agents have very strong preferences for safe nominal returns of a particular long maturity, it’s unlikely that the yield on longer-maturity bonds will fall much below market expectations of the compounded level of short rates—people like liquidity, and they’ll be extremely happy to hold liquid zero-maturity assets over the next 10 years if those assets are expected to have a higher return than the 10-year bond. At that point, they’ll abandon 10-years en masse, and it’ll be tough to push the yield much lower. And even though some agents (e.g. pension funds) do indeed have “very strong” preferences for safe nominal returns with long maturities, if the Fed buys up so many 10-years that the remaining public supply is limited to these agents it’s not clear what the effect would be. It depends on the these agents’ willingness to substitute, say, long-duration Treasuries and Agencies with similar duration corporate debt. If substitutability is low, then the Fed’s asset purchases will simply push up the price of 10-year Treasuries without much spillover to the actual economy. If substitutability is high, then the Fed’s purchases may make a difference.

      Of course, the Fed can guarantee an impact by buying corporate debt directly, or maybe by being even more extreme and buying equities. But again, to hit an NGDP target the required scale of the intervention could be extremely, extremely large: the Fed might have to become larger than the ~$10 trillion US commercial banking sector overnight! This is logistically nontrivial, and although I’m pretty radical even I become a little worried at the notion of the Fed dominating private asset markets.

      In practice, I don’t think this is so much of a concern, because if the Fed is radical enough to do $10, $20, $30 trillion in QE it’s surely radical enough to make strong conditional commitments to easier policy in the future (when it has the ability to move interest rates). I suppose this is the essence of my practical dispute with Market Monetarists: whereas they view unconventional asset purchases as an effective way to implement an NGDP target, I think that those purchases will sometimes be extremely ineffective when compared to the scale of the necessary intervention (even if they do some good), and that commitments to future interest rate policy are vastly more effective.

  6. David Pearson

    I think you raise a broader issue. The Fed can extinguish liquidity risk by virtue of its irredeemable deposit base (reserves). Thus, as you point out, the Fed can remove the liquidity risk embedded in shorter-dated Treasuries. Unfortunately, this liquidity premium, at the moment, appears quite small.

    What the Fed cannot do is remove duration risk. When the Fed buys long-dated Treasuries, it shifts duration risk from private actors to its balance sheet. The risk of losses, from there, shifts (through remittances) to the Treasury, which in turn shifts it back to private actors in the form of higher taxes/Treasury supply. These actors will react by shortening duration.

    The reason why QE2 and Twist rely on the same market inefficiency to work: that private actors will not absorb the information cost of discovering the risk to their tax liability.

    • I agree that the risk ultimately doesn’t disappear—it’s merely transmitted to the public sector and ultimately must find its way back to private actors. To me, the key question is whether the government is more efficient at distributing that risk than the private sector. It seems possible to me that even if private actors know the precise risk to their tax liability, QE2 and Twist can be (somewhat) effective, simply because the government has power to spread risk in a way that no private entity ever can. (Admittedly, this argument as a justification for QE/Twist as a policy makes me a little uncomfortable.)

      An alternative approach is to note that the duration risk is endogenous to the government sector, since it depends on the Fed’s decisions about interest rates. By increasing the amount of interest rate exposure on the government’s balance sheet, QE/Twist may make the Fed less willing to raise rates, since larger public financing costs imply a substantial deadweight loss from taxation. I talked about this a while ago; ultimately it depends on whether (1) the Fed cares about the government’s balance sheet very much or (2) maybe Fed’s own interest rate exposure is high enough that it runs the risk of undercapitalization if interest rates are too high. Gauti Eggertsson’s work on “committing to be irresponsible” is the key piece in the academic literature on this topic.

  7. Matt: “Here we hit upon a quirk of monetary policy: expectations of inflation can be partly self-fulfilling, especially at the zero lower bound (where the Fed doesn’t move to counteract them).”

    Yes, but only “partly”? Even holding the real interest rate constant, in a standard NK model an exogenous 1% sustained increase in expected inflation would lead to a 1% increase in actual inflation. Add in the fact that the ex ante real interest rate would be lower, and the increase in actual inflation would exceed 1%. Then add in the fact that a lower ex ante real interest rate would cause expected investment demand and output demand to increase, and this would make investment more profitable and raise the (temporary equilibrium) natural rate of interest too….and so on.

    I would say that expectations of inflation would be *more* than self-fulfilling. A credible announcement about a higher future NGDP target by the Fed would require the Fed to raise not just nominal but real interest rates too, to prevent overshooting its target.

    • You’re right—widely held expectations of inflation are almost completely self-fulfilling. (At least, the reaction of actual inflation to expected inflation is basically one-to-one, holding the output gap constant; this is particularly clear in the New Keynesian model but true to an extent in other models as well.) There is an additional kick at the zero lower bound, when the Taylor Rule is inoperative. So “more than self-fulfilling” seems like a pretty good characterization, at least under popular models.

      “Partly” is definitely not the wisest choice of words, but I threw it in because I wanted to emphasize that irrational expectations of hyperinflation from a fraction of the population can’t themselves make that hyperinflation come true—both for (A) the obvious reason that it’s only a fraction of the population and (B) the fact that the Fed is ruling out sufficiently inflationary outcomes with its policy rule, as rational market participants know. There is also (C) the problem that I’m not sure how to apply models like the New Keynesian Phillips Curve when we’re assuming that some market participants are straight-up irrational, at least as far as the model is concerned. The NKPC emerges in pristine form from a somewhat ugly Calvo optimal price-setting problem because the contribution to optimal prices of expectations from time t+1 onward can be summarized by expectations of pi_{t+1}; any modification of the model or failure of agents to understand it will break this results. (In fact, this is probably a big part of why other models of price setting are less popular.)

  8. Pingback: Joseph Gagnon explains monetary policy | Kajulew

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s