Sunday, June 10, 2012

Monetary Policy: The Naive View

Christina Romer is certainly consistent. In spite of her many years of experience in academia and policy circles, she consistently surprises me with her un-nuanced views on economic theory and empirical evidence and how they inform policymaking. Case in point: this NYT article.

Romer wants to make the case that the state of the world dictates more accommodation by the Fed. First,
By law, the Fed is supposed to aim for maximum employment and stable prices. But the unemployment rate is 8.2 percent — a good two percentage points above what even the most pessimistic members say is its sustainable level.
...right now, the inflation measure that the Fed watches is a bit below its target of 2 percent...
the Fed’s dual mandate doesn’t say it should care about unemployment only so long as inflation is at or below the target. It’s supposed to care about both equally. If inflation is at the target and unemployment is way above, it’s sensible to risk a little inflation to bring down unemployment.
The "dual mandate" specified in the Full Employment and Balanced Growth Act of 1978, or Humphrey Hawkins Act, is in fact quite vague. Under the Act, the Fed is supposed to be promoting maximum employment and price stability. But any creative central banker would find it easy to make an argument that his or her favorite policy fits well within the Act's guidelines. One could argue (I'm not saying this argument would necessarily be correct), for example, that "price stability" means constant prices (0% inflation) and that employment is lower than it was five years ago due to factors which the Fed cannot correct for. That would then imply that policy should be less accommodating. The set of policies consistent with the Act is so large that the dual mandate argument is not going to help in making Romer's case. She's going to have to make the case by appealing to sound economics.

So what about that? The inflation rate, as measured by the twelve-month percentage increase in the pce deflator, is indeed just below 2%, the Fed's inflation target. The unemployment rate is indeed historically high. Given the Fed's past behavior, one might think more accommodation would be appropriate, assuming of course that the Fed's past behavior was optimal. But what is the Fed supposed to do to be more accommodative? Romer says the alternatives are:

1. Nominal GDP targeting. Romer incorrectly refers to NGDP targeting as an "operating procedure." The operating procedure is actually a description of how the FOMC directs the open market desk at the New York Fed to act. The desk can only do one thing: conduct open market operations. The current fiction (in line with what is written in FOMC statements) is that the operating procedure is the same as before the financial crisis, i.e. the desk conducts open market operations to target the fed funds rate at a level specified by the FOMC. In fact, the fed funds rate is currently determined by the interest rate on reserves, as set by the Board of Governors (for the details, see this post). The desk currently just executes whatever the current quantitative easing (QE) program of the Fed is, without regard to the quantity of reserves that are held in the system overnight.

It would not have made any sense even in pre-financial crisis times for the FOMC to direct the desk by just telling it to hit a NGDP target. Nominal GDP is measured on a quarterly basis, and the National Income Accounts numbers are published with a lag. Currently, for example, it is June 10, and the last NGDP number we have is for first quarter of the year. The FOMC meets about every six weeks. If we were practicing NGDP targeting, how exactly would the FOMC translate the difference between actual NGDP and target NGDP in the first quarter into a directive to the desk at the next meeting? If, as in pre-financial crisis times, excess reserves in the system were essentially zero overnight, then presumably the directive to the desk would have to be in terms of a fed funds target. Under current conditions, and given how the Fed thinks about the monetary policy problem, the directive to the desk would have to be in terms of quantitative goals for the Fed's portfolio. Thus, the operating procedure under a NGDP target would necessarily have to be identical to what it is now.

NGDP targeting does not do anything other than specify the Fed's ultimate goals. As such, there are two problems with it. The first is the absence of a sound theory to justify NGDP targeting. It is unclear why an economy in which NGDP grows at a constant rate is an economy in which the central bank is doing what is optimal. Second, one can imagine circumstances under which particular NGDP targets will not be feasible. In fact, I do not think it would be feasible for the Fed to achieve 5% annual nominal GDP growth by the end of this year, for these reasons.

2. More QE. Romer and the FOMC are on the same page on this one, but I don't think QE does anything at all (again, see the last link above). At best, QE can signal future intentions of the Fed with regard to the policy rate (and thus move asset prices), but the Fed can do the same thing with "forward guidance," i.e. announcements about the future path for the policy rate. Like other people, including Miles Kimball, Romer seems to think that QE isn't doing much because the Fed hasn't done it right:
The previous rounds of quantitative easing may have done little to improve expectations because their size and duration were limited in advance. If the Fed does another round, it should leave the overall size and end date unspecified. Or, better yet, the ultimate scale and timing could be tied to the goals the Fed wants to achieve.
First, I'm not sure how you announce a policy without saying what it is. Second, the last sentence in the above quote is interesting. The Fed claims that, for example, purchases of long-maturity Treasuries will lower long bond yields. If they were confident about that, the FOMC would announce targets for long bond yields rather than quantitative goals. They don't announce the targets, therefore they must not be confident that QE does what they claim.

3. Forward guidance. As with QE, Romer likes it, but she does not like how the Fed does it:
Instead, the policy-making committee could adopt the proposal of Charles Evans, the president of the Federal Reserve Bank of Chicago, that the Fed pledge to keep rates near zero until unemployment is down to 7 percent or inflation has risen to 3 percent. Such conditional guidance assures people that the Fed will keep at the job until unemployment is down or the toll on inflation becomes unacceptable.
If you look through the FOMC minutes (a prize to the person who can find this), you'll find the FOMC's rationale for ditching Evans's suggestion, and I think that rationale was good. From what I remember, the reasoning was: (i) There are too many contingencies to worry about. You can't write them all into the FOMC statement. (ii) Making policy explicitly contingent on, for example, the unemployment rate, would be silly. The unemployment rate is determined by many factors, most of which have nothing at all to do with monetary policy. (iii) The relationship between monetary policy and real economic activity is imperfectly understood.

The important fight that is going on is not one involving the weapons of monetary policy against the poorly-performing US economy. The key struggle is in getting policy people, and those who write about policy, to use the best available economic tools and reasoning to address our current problems.


  1. This might be the quote you're looking for from the January minutes:

    Several participants thought it would be helpful to provide more information about the economic conditions that would be likely to warrant maintaining the current target range for the federal funds rate, perhaps by providing numerical thresholds for the unemployment and inflation rates. Different opinions were expressed regarding the appropriate values of such thresholds, reflecting different assessments of the path
    for the federal funds rate that would likely be appropriate to foster the Committee’s longer-run goals. However, some participants worried that such thresholds would not accurately or effectively convey the Committee’s forward-looking approach to monetary policy and thus would pose difficult communications issues, or that movements in the unemployment rate, by themselves,
    would be an unreliable measure of progress toward maximum employment.

    If so, I'd like to claim as my prize an answer to the question I asked currently at the bottom of the comments on your June 4 unconventional OMO post.

    1. Here's your question:

      "Thanks for this post, it was well written. Question about the aribtrage. An SPV holding even T-bonds will have some haircut, i.e. the repos will be less than 100% of the assets. So the Fed's action can be thought of as unwinding the repo, but it also seems to change the nature of the equity sliver of T-bonds by essentially undwinding the leverage. This still changes the overall portfolio of the private sector by reducing its risk, particular its term premium risk.

      Or are you arguing that playing with the term structure won't matter because it simply transfers the term premium to the government? But isn't that the point? That is, if the government increases the risk of safe nominal assets it might be just as effective as increasing the supply of safe assets in raising the price level."

      The "government" is not another agent we can shift risk to. At best it can reallocate risk. If the Fed buys long Treasury bonds, it is intermediating across maturities, and bearing maturity risk. But what happens if short rates go up and the prices of long Treasuries go down? Now the profit the Fed hands over to the Treasury is smaller, and the Treasury has to make it up somehow. The private sector has to take the hit somehow, no matter how the Treasury finances the loss. You can find a nonneutrality in there somewhere, but only if fiscal policy changes.

  2. Ah, I think I missed the part of your post where the hyperbolic and mildly insulting headline and leading paragraph are given any support. You've dealt with talking about NGDP targeting, QE etc. in the past without collapsing into insults (e.g. your posts on Kimball recently).

    Raising the quality of discourse in the medium and all that...

  3. Naivete is probably the least insulting thing I could accuse her of. Distortion is another. For example, she says:

    "The academic literature shows that monetary policy can be very effective at reducing unemployment in situations like ours."

    This seems to imply that there is a consensus on this issue, which is not correct. The example she links to is her own work.

    She also says:

    "Furthermore, most analyses suggest that the main determinant of inflation is the state of the economy."

    That's pretty sneaky. It's actually tautological, as the "state of the economy" includes the whole history - some part of that has to explain inflation, and everything else. What she means the reader to understand is that the Phillips curve view of inflation is widely-accepted and well-founded - output gaps predict inflation. That's not correct. For example, Atkeson and Ohanian show that a monkey can forecast inflation as well as an economist with a Phillips curve.

    1. Oh Steve, for a man of oh-so-few citations, you sure have a big mouth.

    2. That second rude anonymous is not me (first anonymous). I still think your rhetoric is overly heated here though.

    3. Pro tip: Ctrl + c/v allows you to copy and paste key words into a search engine, like, in order to find any relevant information you need:

  4. "The "government" is not another agent we can shift risk to. At best it can reallocate risk. If the Fed buys long Treasury bonds, it is intermediating across maturities, and bearing maturity risk. But what happens if short rates go up and the prices of long Treasuries go down? Now the profit the Fed hands over to the Treasury is smaller, and the Treasury has to make it up somehow. The private sector has to take the hit somehow, no matter how the Treasury finances the loss. You can find a nonneutrality in there somewhere, but only if fiscal policy changes."

    This argument assumes Barro-Ricardian Equivalence, which is as silly an assumption as it gets.

    1. I suspect your go-to assumptions are much worse, such as there exists free stuff that no one has bothered to pick up yet.

    2. Assume it is a "silly assumption". So are you arguing that QE is effective monetary policy or effective fiscal policy? Seems to me its the latter.

    3. It is hard to be sillier than assuming Barro-Ricardian equivalence.

      First, that only holds for models with immortal agents or even sillier, models with agents that behave as if they are immortal.

      Second, it requires that taxes be lump-sum, which is silly.

      Finally, it goes against all, I mean, ALL empirical evidence.

      In summary, to say that "A" does not hold because of Barro equivalence ==> to admit that one failed to make an argument that "A" does not hold.

  5. Stephen, I think your analysis of quantitative easing in the context of excess reserves is quite powerful, but I don't think it prevents quantitative easing and other unconventional monetary policies from affecting the real economy.

    On your first point, I think Nick Rowe and Scott Sumner have done a rather good job at outlining a large range of mechanisms that a central bank can use to depreciate a currency and raise the price level. Even if these are, to a certain extent, neutralized by IOR, this is just a reason to lower IOR. If lowering IOR has the potential to lead to higher inflation, this is why having a stable NGDP target is even more important. That target can help rein in expectations and prevent an inflationary spiral.

    Your argument about lagged data releases doesn't prevent either:

    1) Targeting of the internal forecast
    2) The creation of NGDP futures to gauge market expectations of NGDP

    Market Monetarism bases a lot of its arguments on expectations, and even if these forecasts aren't always accurate, they do a good job of modeling the assumptions under which market participants operate. Thus, even if the national income data is released with a lag, it doesn't prevent effective implementation of NGDP targeting.

    There's also quite a large set of analytical reasons why NGDP targeting would outperform IT. See: Perhaps not all of these arguments are wrapped up in one tight analytical model, but these general concepts should not be difficult to model in chunks.

    On your second point, I think we're in agreement that shaping market expectations is important. However, targeting the interest rate leads to a certain level of indeterminacy. Is the low interest rate an indication of monetary easing, or is it a symptom of lower NGDP expectations? This is one of the key reasons why explaining monetary policy through interest rates is so difficult. Yes, purchases of bonds will raise their price, lowering the interest rate. However, the higher levels of economic activity will move people to invest in other assets, thereby lowering the interest rate. It's not a linear process, so the focus should stay on money supply and demand.

    And her proposal isn't announcing a policy without an idea of what it should do, rather it's a policy that knows what it wants as an endpoint. Current policy defines the mechanism while leaving the end result uncertain, whereas her proposal leaves the mechanism open to interpretation but targets a fixed result.

    1. Your third point is also a little puzzling. I don't think it's controversial to say that there is a short run trade-off between unemployment and inflation. I don't know about the regression data, but if there were no trade-off this would suggest optimal monetary policy would entail raising the interest rate to lower inflation while also leaving unemployment unchanged. This seems like a highly dubious proposition.

      With this trade-off in mind, Romer's proposal to do QE until certain unemployment/inflation thresholds were reached doesn't seem to ridiculous. Unemployment is affected by a lot more than monetary policy, but what are these factors? Additionally, how do you know these factors aren't, in some way, shaped by aggregate demand considerations? If AD were higher, employers would be more wiling to train employees.

      Moreover, Romer's joint condition (UE<7%, Inflation>3%) for terminating the policy clips your risk. Even if unemployment data is sporadic or may not perfectly measure real activity, if monetary policy gets out of hand and inflation rises rapidly, the policy would also be terminated. The flip side of your argument is that you don't know the returns to lowering unemployment. Maybe there are significant hysteresis effects that your model doesn't take into account. Thus, there's no reason to try for the joint target that, at worst, would cut our losses.

      Yes, sound reasoning and the correct economic tools are important, and the monetary policy ideas brought on by the Market Monetarist revolution have an important role to play in that discussion.


      You realize that was written by a high school student?

      "Perhaps not all of these arguments are wrapped up in one tight analytical model, but these general concepts should not be difficult to model in chunks."

      How do I model in chunks?

    3. But Scott Sumner has predicted that this high school student will be the winner of the 2030 John Bates Clark award! So buck up and start learning how to 'model in chunks.'

    4. I think Yichuan Wang ought to learn a little economics before he opens his mouth again. By the way Steve, he's a high school student too. Do we listen to high school physics students when we build bridges?

    5. You should listen to them if their arguments have merit. If you don't actually care about changing anyone's mind, here's a hint: don't laugh off people's reasoning because they are young, or inexperienced, or lack tenure, or because you think they smell bad. Just engage with their reasoning.

      As a disinterested observer, I find the arguments Evan Soltas espouses are usually just as convincing as the ones Steve Williamson puts forward. Why should I care if he's in high school?

    6. I admit, I'm no expert in economics, but I don't think my substantive points on joint conditions and forward looking policies are that out of line.

      Perhaps "chunks" was not the best noun to use (partially? components?), but I was talking about looking at the aspects that Evan mentions piece by piece. There was the Bernanke paper on oil price shocks that showed a lot of the harm from supply shocks came from the IT monetary policy. On his point about growth there was the recent NBER working paper

      No, you shouldn't let high school physics students design bridges. But it doesn't take a physics whiz to understand that suspension bridges are good for wide rivers. Our lack of technical genius shouldn't prevent us from understanding the fundamentals.

    7. "Our lack of technical genius shouldn't prevent us from understanding the fundamentals."

      But you don't understand the fundamentals, and therein lies the problem. Can you compare and contrast the Bernanke paper you cited with Jim Hamilton's arguments about the importance of oil shocks? No? Then why are you asserting Bernanke is right and Hamilton is wrong?

      You may well be smart, and you may well be interested, but so are a lot of other people whose opinions on economics should also be ignored. Get yourself educated, stay interested, and come back armed with tools. Otherwise, you're just contributing to the noise.

    8. Hamilton and Bernanke can both be right -- it's not contradictory to state that oil shocks have been significant contributors to past downturns AND a different monetary regime would have mitigated these shocks.

      But I'm sure you know that already, and that's beside the point. If you spent more time discussing the facts and less time trying to brand Yichuan as unfit for discussion, maybe you would be contributing to a productive conversation, rather than 'contributing to the noise'.

    9. Of course I knew that, and I also know the identification schemes used to extract these facts from the data, as well as how to interpret the confidence intervals and so on and so forth. But my point is that Wang doesn't, so why should we accord his opinion -- yes, it is an opinion -- any weight at all? He might as well be advocating we bleed the humors, for all the scientific evidence behind his statements, and that goes for the other kid too.

    10. If you're evaluating what weight to place on people's opinions based on their age and academic credentials, then you don't have a good method of evaluation.

      I could turn your question around and ask why I should give your opinion any weight. I'm not sure who you are or what your credentials are (you don't seem interested in attaching your name to these posts). You could be Bob Lucas posting anonymously, but ultimately I just don't care.

      Welcome to the internet, where your years of research, your many advanced degrees from prestigious institutions, your long track record of academic publication, and your nobel prize -- all put together -- mean exactly nothing. The only importance your many accolades have, in this context, are in how they inform your arguments. Wang has staked out views based on analysis in this comment thread and in cited blog posts and papers. If Wang's opinions deserve no weight, it should be apparent from simple logical flaws in that analysis. Kindly highlight those flaws with analysis of your own and support your analysis with empirical evidence. I'm still waiting for you to say something more substantial than "you don't deserve to argue with me because I'm older than you".

    11. Keep waiting, I'll get back to you when you show you know something.

    12. Thanks for your time anyways. Posting all these comments got me to think a bit more about interest rate forward guidance, which has helped in my writing (

      I hope I haven't put too many leeches on you (aspirin is probably better for pain anyways), and I'll be sure to be back the next time a monetary policy op-ed pops up.

  6. Thanks, that was helpful. Is a decrease in the NPV of future surpluses (or an increase in their risk) a "non-neutrality" in this case in the sense that it could raise the price level? If so, might it not be just as fair to assume the NPVs of primary surpluses are exogenous such that a shorter duration of government bonds doesn't change taxing and spending plans? This seems like a decent argument if the Fed is in a liquidity trap and might passively accept an inflation increase that it otherwise could not easily orchestrate, while the fiscal authority is neutral within some range of inflation scenarios.

  7. Are you thinking of a fiscal-theory-of-the-price-level equation that uses the present value of primary surpluses to determine the price level?

  8. Yes, exactly. I was thinking of a Christopher Sims story but with the idea that even a mere increase in the duration risk retained by the government (as opposed to a change in taxing or spending plans) could be enough to reduce the budgetary NPV if you were willing to assume the primary surplus was exogenous or passive within some range. Is something more needed for this to make sense?

  9. "... if you were willing to assume the primary surplus was exogenous or passive within some range."

    Why do you think that matters?

    Not related to what we're discussing exactly: I was thinking at one time that the fiscal theory equation that, for example, Cochrane is fond of writing down, is not model-free. I can think of simple monetary models where it does not hold - e.g. a pure currency model with a fixed money supply. I think it's because valued money is essentially a bubble.

  10. I agree with you on FTPL, which is why if there can be said to be a strong form of FTPL that proclaims not to be model free it gets accused of all kinds of problems in various models suchs as an overdetermined price level (I think in CIA models) or turning money into phlogiston in a supposedly cashless model (this is what Buiter said I think). Back in I think 2009ish Nick Rowe on WCI wrote a post about Cochrane's money as stock piece that I think used almost the same words you just did to describe its limitations. But as far as I can tell (which may not be all that far), at least people like Sims and Woodford mostly talk and write like they are aware of these limitations.

    The reason I thought it was important to assume the primary surplus was exogeneous is because otherwise the duration risk just looks like a temporary fiscal action that might get reversed through larger primary surpluses (higher T or lower G) if instead the net deficit and not the primary surplus was exogenous. Maybe even then would turn more on whether that reversal was expected to come while the Fed was still in a liquidity trap (so the effect is truly nullifed) or once it was out (then just as good as a non-ricardian FTPL policy bc the Fed is able to offset the late reversal once again with normal monetary policy). I am in no way saying this effect is likely to be important at the levels of QE or twist I am more just trying to talk through it because I feel shaky on it.

  11. All of this gets into issues of what is fiscal policy and what is monetary policy. In the models we write down, there is a sequence of consolidated government budget constraints, and then we have to start making assumptions about what piece of policy is passive. The exogenous-primary-surplus assumption is what I used to get an interesting notion of what monetary policy does in one of the papers I wrote recently.

  12. Steve: Your killing me with the monkey and inflation forecasting. My eyes are so watered up I can't see the screen when typing. Wasn't Bonzo in a movie with Ronald Reagan called "Bedtime for the Phillips Curve"?

  13. Hello Stephen Williamson,

    Here is the blog post you are looking for, where Scott Sumner details how to create an NGDP futures market that guides the daily Treasuries trading activities of the Fed.

    All best,
    Jason Odegaard

  14. here are 600 Bn reasons why you should believe in the wealth effect.

    1. state and local tax revenues have mostly gone up and are slightly higher than 2008.
    2. state employment has gone down.

    huh? why is that?

    the answer: state and local pensions lost 621 Bn dollars in 2009 from 2008 (see below, you can check the tax receipts which i pasted below). contributions in 2010 were about 125 Bn.

    this is just a microcosm. govts and corporations have had to cut back employment and such to fund pensions, and many people are nervous about their 401k. Even if the only effect of QE is for the stock market to go up to the point where retirement and other obligations make people nervous, that's a huge real effect.

    Total State and local tax receipts
    2011 1,345,339
    2010 1,287,025
    2009 1,265,020
    2008 1,319,551
    2007 1,274,824


    1. " make people LESS nervous, that's a huge real effect"

  15. Naive

    I think that is about the nicest think one could say, under the circumstances.

    Ms. Romer had the wheel, so to speak, in her own hands. It seems to me that her bias justifies paying no attention to her remarks now, whatsoever.

    That SW is willing to consider the merits of her arguments shows something.

    But, the old SW emerges at the end. "The relationship between monetary policy and real economic activity is imperfectly understood."

    Wouldn't it be more accurate to say, "The relationship between either macroeconomic or monetary policy and real economic activity is not understood at all and thus no consensus is possible."

    Doesn't this follow from the fact that we have no consensus?