Tuesday, July 30, 2013

To QE or not to QE?

The FOMC meets today and tomorrow. Here's what the committee members will be discussing. Since May, expected inflation has fallen and real bond yields have risen, as you can see in the first chart.
The chart shows the breakeven rate (10-year nominal Treasury yield minus TIPS yield) and the 10-year TIPS yield. The breakeven rate tends to be biased upward as a measure of anticipated inflation (TIPS insure inflation risk, and a TIPS-holder does not have to pay the Treasury if the CPI inflation rate falls below zero), but typically moves in the same direction as anticipated inflation. As well, if we compare the path for the PCE deflator to a 2% growth path from January 2007, we get the next chart.
Since late 2012, the PCE deflator has been south of the 2% growth trend, and the last observation (May 2013) is about 1% on the low side. If you care about year-over-year inflation rates, you'll come to the same conclusion - the inflation rate is too low, relative to what the Fed wants.

Given that the goal of the latest QE (quantitative easing) exercise, which has been proceeding at a pace of $85 billion in asset purchases ($45 billion in long Treasuries, $40 billion in mortgage-backed securities) per month since fall 2012, was to reduce bond yields and presumably to increase current and anticipated inflation, things are not moving in the right direction.

On the other side of the dual mandate, the unemployment rate is still above where the Fed wants it to be.
An important part of the unemployment story is what is going on with the long-term unemployed, so we can look as well at the long-term unemployment rate (those unemployed 27 weeks or more, as a percentage of the labor force). That's in the next chart.
This is interesting, as about 2/3 of the reduction in the unemployment rate since it peaked has come from decreases in the long-term unemployment rate. You can see that the long-term unemployment rate is still very high relative to where it was before the last recession started (about 2.8% vs. something less than 1%). If the unemployment rate fell in the next months at the average rate of decrease from its peak until now, then it would cross the Fed's 6.5% threshold in about March 2015. This is important, as the FOMC has told us that it will not consider increasing the "fed funds rate target" (actually the interest rate on reserves) until the unemployment rate crosses the 6.5% threshold (and as long as the inflation rate is not too high). Further, Bernanke has told us, essentially in a side conversation, that the policy rate is not likely to increase immediately after the threshold is crossed.

Two points here:

1. If the FOMC wants to be consistent with its previous policy statements, and what Bernanke has said in public, it should plan to purchase assets at a higher rate. The news has been on the down side on inflation, anticipated inflation, and long bond yields, with nothing dramatic happening one way or the other in the labor market. So, in order to keep its promises, there should be more QE, and Bernanke should not be talking about "tapering."
2. It seems almost certain that Bernanke will not be in his current job in six months, so why do we care what he says? If we knew the next Chair would be Janet Yellen, then we care what Bernanke says now, as it seems Yellen would just follow through on that. If Larry Summers gets appointed, then things change entirely. Summers is on record as thinking that QE probably doesn't matter much, and that we might have to revise our views about what the capacity of the U.S. economy is. Maybe QE, thresholds, etc. all go out the window.

In an ideal world, what should the Fed be doing? In the past I've argued that QE shouldn't matter, for quantities and prices. But suppose I were working for the Fed full time, and someone asked me to come up with a rationale for QE. What would I do? The answer is this paper, which is the best I can do at the moment, to come up with a model where QE actually makes a difference. You can read the paper to your heart's content, but here's the idea. To take QE seriously we have to be specific about fiscal policy and its relationship to monetary policy, what private financial intermediaries and the central bank can and cannot do, and the roles of assets in exchange and as collateral. The paper does all that, I think. Start with a world where the fiscal authority is doing something stupid. It issues nominal government debt - in short and long maturities - and sets taxes in such a way that the real value of the consolidated government debt is constant forever, at a value V. The central bank can issue reserves, which are perfect substitutes for short-maturity government debt, and currency. The private sector needs currency for some kinds of transactions, and there are other transactions where people trade intermediary liabilities that are backed by short and long-maturity government debt and reserves. Banks serve to allocate liquidity (currency, reserves, government debt) efficiently to the appropriate transactions. The central bank can execute asset swaps - its job is to manage the composition of the outstanding consolidated government debt, consisting of currency, reserves, short maturity bonds, and long maturity bonds.

But we can't trust banks, which need to (implicitly) collateralize their deposit liabilities. The collateral consists of the bank's assets - reserves and government debt. But collateral is imperfect, in the sense that the bank can abscond with some fraction of any asset in its portfolio (a Kiyotaki-Moore problem). To get a term premium - an upward sloping nominal yield curve, it must be easier to run away with long-maturity government debt. This is definitely a short cut, as I discuss in the paper, but a useful starting point. Also, for there to be a term premium, there must be a shortage of liquidity, i.e. V must be small enough that we can't get efficient exchange - of assets for goods.

What does QE do? A QE experiment involves a purchase of long-maturity government debt by the central bank, holding constant the short-term nominal interest rate. This essentially mitigates the asset shortage. The central bank is swapping good collateral for collateral that is less good. This reduces the term premium, the nominal long bond yield falls, and the yield curve flattens. So far so good for Ben Bernanke. But because collateral is now less scarce in the aggregate, real bond yields rise - at both the short and long end of the real yield curve. Further, the inflation rate falls. That's certainly not how Ben Bernanke thinks QE works. The usual story is that QE will cause real bond yields to fall and inflation to rise. In the model, QE is a good thing to do, though.

Another interesting feature of the model is that, if the central bank holds the short-term nominal interest rate constant at zero, and V increases, then inflation falls, and real bond yields rise. Looks like what is going on in the first two charts, right? V is a convenient catchall in the model. Including private assets in the model (loans, equity) should not be a big deal. Better private opportunities is like having a larger V, and we could also think of higher demand for domestic government debt coming from abroad as lowering V. Currently in the real world, if we think that private investment opportunities are looking better and with a better outlook on sovereign debt problems reducing the demand for U.S. consolidated government debt, we can translate this into a reduction in the scarcity of collateral in the aggregate, and a smaller V in the model. So far so good.

So what does this tell us? We're in a world where real interest rates on government debt are endogenous. Those interest rates include liquidity premia that reflect the marginal value of different classes of government debt as collateral and in financial transactions. Currently real interest rates are going up, so it should not be surprising, with the short-term nominal interest rate pegged close to zero, that inflation is falling. Conclusion: People will definitely call me stupid for saying so, but as Cartman says, screw you guys. The Fed can make all the confusing promises about the future it wants, and purchase all the assets it wants, but it may not be able to increase the inflation rate to the 2% it wants unless the nominal interest rate on reserves goes up. That's what the model says.

The same people who will most certainly be calling me stupid will also want to call me inconsistent, a bad inflation forecaster, or some such. You can go back and read my posts and find me worrying about the potential for inflation that exists in the state of the Fed's balance sheet. Here's Krugman's take on bad inflation forecasters:
I’ve spent five years and more watching the inflationphobes, who weren’t particularly sensible to begin with, descend into shrill unholy madness.
Holy crap, who would want to be one of those inflationphobe characters? To save myself from banishment to Krugmaniac hell, I'm going to appeal to multiple equilibria. I haven't written this up yet, but I've worked out some stuff on what happens when a naive central banker, who follows a Taylor rule, and believes erroneously that the real interest rate is exogenous, goes to work in my model. You can get two equilibria: zero lower bound, low real interest rate, low inflation, or positive nominal interest rate, high real interest rate, and high inflation. That's essentially been my point. There's nothing inherently inflationary about a large central bank balance sheet. The potential for inflation comes from how the central bank behaves given the state of its balance sheet. The multiple equilibrium story is somewhat related to this paper by Jim Bullard, with some important differences, I think.

24 comments:

  1. Stephen,

    "The breakeven rate tends to be biased upward as a measure of anticipated inflation"

    I think that's debatable (and worth debating). The sign of the inflation market risk premium depends on which states of inflation are most
    correlated with welfare. We all have our biases about which future states are most correlated with suffering (and therefore with the marginal value of consumption), but for a more objective take, we can extrapolate the future as a continuation of the current mean-variance conditions. Since inflation break-evens are highly positively correlated with the market portfolio, this suggests that the inflation risk premium ought to be negative (notwithstanding the opinions of the Cleveland Fed). The market fears deflation far more than inflation.

    "you can get two equilibria: zero lower bound, low real interest rate, low inflation, or positive nominal interest rate, high real interest rate, and high inflation."

    In the real world, the real rate is exogenous. If the Fed hikes to 10% right now spot inflation will not go up! So the the real short rate will go up *a lot*. There! Exogenous rate hike. Getting from there to your medium run rbc equilibrium depends on the pricing dynamics and competitive inefficiencies, but that process is critical and plays out over the medium run (i.e. several years). So maybe someday, after a long depression, we end up in your high inflation equilibrium. But if we want to get there painlessly instead, the solution is to *cut* the real rate now, thereby causing a rise in inflation and a steepening of the real rate curve. (Yes, easier said than done - fiscal policy could help).

    I think your model is interesting though, and I did read the paper (well the first and last sections at least) awhile ago and it gave me lots of food for thought about liquidity effects. I'm skeptical though, both because I think exogenous real rates (and the liquidity trap) are critical but also because I think QE affects the economy through non-liquidity channels (if at all). I prefer to imagine an economy that is non-Ricardian by virtue of OLG effects (rates asymptotically below growth). So Wallace irrelevance fails. The Fed can permanently remove bonds from the economy (flush them down the toilet forever - there are free lunches all over the place). But since bonds have negative beta, removing them from the market *increases* portfolio risk and therefore the market risk premium. So QE is *bad*.

    Final thought: is the positive effect of QE baked in by assuming a positive bond risk premium? If you assume that and then add *whatever* friction to make QE do *something*, will you get a positive effect?

    ReplyDelete
    Replies
    1. "I think that's debatable (and worth debating)."

      I guess the issue is whether inflation risk is nondiversifiable and how the CPI comoves with, say, the intertemporal marginal rate of substitution. I'm assuming this works out so that I'm willing to pay something for inflation insurance.

      "In the real world, the real rate is exogenous. If the Fed hikes to 10% right now spot inflation will not go up! So the the real short rate will go up *a lot*. There!"

      Wrong. I don't know the real rate ex ante, as I don't know in advance the inflation rate over the holding period of the asset. I know the Fed likes to think that expectations are "anchored," i.e. anticipated inflation is exogenous, but generally when the Fed does something, anticipated inflation and the nominal interest rate both change.

      " is the positive effect of QE baked in by assuming a positive bond risk premium?"

      No.

      Delete
  2. The main assets on Treasury's balance sheet are long term: future tax receipts, Federal land, and future seigniorage profits.

    QE reduces the tenor of the government's liabilities from 7+ years to overnight. The resulting asset/liability mismatch creates duration risk.

    In an inflation targeting regime, QE just transfers duration risk from private investors to the taxpayer. Alternatively, the central bank abandons inflation targeting and targets duration losses instead.

    ReplyDelete
    Replies
    1. I'm not sure I buy the analogy. Seignorage profits are a long term asset for the Treasury? It would be more appropriate to look at the consolidated government's balance sheet. But what are currency and reserves then? They enter as liabilities, but they're not really obligations in the usual sense, and they could be outstanding forever if the Fed wants them to be. I've never thought that what you're talking about was an avenue for QE to do anything, at least in the current circumstances.

      Delete
    2. I am thinking of the consolidated statements. Seigniorage are profits to Treasury from holding bonds financed by currency. Currency is a zero coupon perpetual. Reserves are overnight FRN's. If the Fed bought all Treasuries, this would reduce the duration of consolidated liabilities to overnight, against long term assets (land and tax receipts).

      Does QE do anything? Yes, it transfers duration risk from private investors to taxpayers. Both are in the "private sector", they are just different groups. So QE has distributional effects that the taxpayer doesn't get to vote on. That's all (except for possible financial frictions).

      Delete
  3. "But collateral is imperfect, in the sense that the bank can abscond with some fraction of any asset in its portfolio (a Kiyotaki-Moore problem). To get a term premium - an upward sloping nominal yield curve, it must be easier to run away with long-maturity government debt."

    Can you go into more detail on this? How is it easier to abscond with a 30 year treasury bond than a 5 year t-note? Is a 1 year note really more pledgable than a 10 year note? I thought the 10 year treasury was considered to be the most liquid financial instrument in the world.

    ReplyDelete
    Replies
    1. As I said, it's a shortcut. A deeper model would have aggregate risk, and fluctuations in bond prices. Long bond prices are more volatile than short, so implicitly a bank won't be able to borrow as much against 10-years as against T-bills. When you say 10-years are liquid, I think you mean the market is thick and they're easy to sell.

      Delete
    2. So given your shortcut, Operation Twist -- which involved the Fed selling short term notes and buying long term ones -- would have increased the quantity of "good collateral" in the economy, driving down liquidity premia and increasing real bond yields? And would have reduced the inflation rate?

      Delete
  4. "But because collateral is now less scarce in the aggregate, real bond yields rise - at both the short and long end of the real yield curve. Further, the inflation rate falls."

    Ok, so an increase in the amount of collateral increases liquidity and lowers existing bond liquidity premia. Bonds must now compensate investors with a larger real return since the liquidity bonus they once provided is no more. Is that right? But I don't get the inflation falling bit. How does that follow?

    ReplyDelete
    Replies
    1. The intuition would be that, holding the short-term nominal interest rate constant (that's the nature of the experiment), all real returns rise in equilibrium, including the real return on currency.

      Delete
    2. I like how you explain it in your paper better:

      "Infl‡ation falls because one of the effects of QE is to increase the real stock of currency held by the private sector, and agents require an increase in currency’'s rate of return (a fall in the in‡flation rate) to induce them to hold more currency."

      In order to get to a state in which inflation is falling, the currency's value will fall at first. So we see an increase in inflation until the currency is sufficiently cheap to induce agents to hold it again. Only then do we hit the situation you describe: falling inflation.

      Delete
  5. Stephen,

    "I guess the issue is whether inflation risk is nondiversifiable..."

    Ignoring, for the moment, that nominal assets are in net-zero supply in the economy, and the representative agent therefore doesn't actually have any, I agree. My answer is that there are relatively broad, multi-currency inflation shocks in the historical record, so probably not diversifiable.

    "and how the CPI comoves with, say, the intertemporal marginal rate of substitution."

    Yes, except the word "intertemporal" confuses me, when we talk about it as "moving".

    "I'm assuming this works out so that I'm willing to pay something for inflation insurance."

    Makes sense. And I think the market price is negative.

    "I don't know the real rate ex ante, as I don't know in advance the inflation rate over the holding period of the asset."

    I don't know the *nominal* rate ex ante over the holding period of the asset either. To me there is no difference, in this sense, between the nominal, real, and inflation rate. All are risk neutral values implied from asset prices (nominal and real return bonds). Term rates are always risky rates (under the usual choices of numeraire), even if we sometimes call the treasury curve a "risk free" yield curve. Real rates are every bit as "real" as nominal rates, and you can borrow or invest at those rates whenever you want.

    "I know the Fed likes to think that expectations are "anchored," i.e. anticipated inflation is exogenous, but generally when the Fed does something, anticipated inflation and the nominal interest rate both change."

    Yes. And if the Fed surprise-hikes the short rate, the following happens:

    1) inflation expectations drop, generally across the curve, but by declining amounts going out the curve.

    2) The nominal forward rate curve flattens, i.e. moves wider in the short end, and then, in anticipation of future rate cuts to counteract the resulting disinflation, goes lower in the medium term.

    Therefore the real curve rises in the short run, by *at least* the amount of the nominal curve. (Inflation, contrary to the Fed (and Scott Sumner), is *endogenous* but goes in the opposite direction of what you are saying.) We've observed this behaviour with every conventional Fed action as long as I can remember (pretty long). For particularly ugly examples, look at the effects of attempted rate hikes in both in Europe and Japan. I haven't read the relevant empirical term structure literature in a long time, but I am quite certain that it bears out my understanding.

    ReplyDelete
  6. Steve,

    I agree with Cartman and you on this one. I think the paper is in the right direction, and I like it better than the one you recently published in AER, with which it shares of course certain features.

    A quick thought I had while skimming over it is whether the model can provide a link between the size of the government debt and the value of private assets that can serve as collateral. Could an increase in the scarcity of government debt, depending also on the regulatory environment, raise the price for private assets, like homes, whose derivatives can serve as collateral? Could something like this have taken place in the late 1990s, thereby fueling the buble?

    ReplyDelete
    Replies
    1. I left the private assets out of the model, as it was easier that way, and I could get to the main points without them. The answer to your question is yes, except perhaps for the "fueling the bubble" part. I'm not sure about that yet. This is what I'm working on now. It's neat, actually, as with the private assets I can get endogenous "pledgeability."

      Delete
    2. Thanks, looking forward to the results!

      Delete
  7. About the multiple equilibria story, high real rate plus high inflation implies a very high nominal rate.

    I have no idea how you are supposed to get high inflation and high nominal rates. In the real world is the other way around, tight money and high nominal rates lead to low inflation.

    ReplyDelete
    Replies
    1. It follows from a Taylor rule, which is definitely something that some "real world" policymakers live by.

      Delete
    2. Would you mind to point out one historic example in which real rates and inflation were high and nominal rates were thus even higher?

      Delete
    3. At the beginning of the Volcker disinflation. What's your point?

      Delete
  8. Whether your multiple equilibria story that serves as a refinement of your wrong inflation prediction has anything to do with the real world.
    Unless you wanna claim that the seventies, i.e. a price shock of an important commodity plus too loose money / a lack of inflation anchoring, will soon be back it does not.

    I think that taking asset price inflation seriously and incorporating it into a new inflation measure would make monetary agencies more conservative concerning stuff like QE.
    So I disagree with Old Keynesians like Krugman who want more aggressive monetary policy (Leijonhufvud is an exception) and I disagree with monetarists like you who see consumer price inflation around the corner.

    ReplyDelete
    Replies
    1. "I disagree with monetarists like you who see consumer price inflation around the corner."

      What do you see around the corner?

      Delete
  9. "But because collateral is now less scarce in the aggregate, real bond yields rise - at both the short and long end of the real yield curve. Further, the inflation rate falls. "

    Why would you suppose this?
    First of all, why should collateral be so relevant in such situation? With the debt purchase, and being the Market aware of it and trusting authorities, yields would go down. Obviously in the sense of how trade occurs, collateral would never offset this trend. Inflation might not be high but it certainly does not make for such a rise in real yields neither. And so future expectations would make inflation go up, perhaps not sufficiently but is not even the main thing here.

    Another point: even if we believe and agree with what you say, what makes you say that collateral is less scarce for long-term operations? Reserves and currency are not even near being perfect substitutes for that.

    So, I wonder how do you back up your assumption?

    ReplyDelete
  10. 1. If the FOMC wants to be consistent with its previous policy statements, and what Bernanke has said in public, it should plan to purchase assets at a higher rate--Stephen Williamson.

    I just want to go on the record that I did agree with Mr. Williamson on something one time. This is it.

    We should be tapering up to hit macroeconomic targets, such as rates of economic growth, inflation or unemployment. This Fed policy should be transparent, public, clear.

    ReplyDelete