Wednesday, October 19, 2011

Nominal GDP Targeting

Nominal GDP (NGDP) targeting seems to be getting a lot of attention. The idea seems to go back at least to the 1980s, when Bennett McCallum talked about it. Scott Sumner and David Beckworth have taken this up as a cause, and Charles Evans has discussed NGDP targeting in a speech. Some of the business media think it matters.

To make sense out of NGDP targeting, start with the original Taylor rule, as specified in Taylor's 1993 paper. Taylor proposed that monetary policy should be conducted according to the following rule:

R(t) = p(t)- p(t-1) + a[y(t) - y*] + b[p(t) - p(t-1) - i*] + r*,

where R is the fed funds rate target, p is the log of the price level, y is the log of real GDP, y* is the target level of real GDP, i* is the target inflation rate, r* is the long-run real interest rate, a > 0 and b > 0. Taylor did not derive his rule using theory, but instead argued that this rule worked well according to some loss criterion in some macroeconometric models. The Taylor rule found its way into New Keynesian (NK) models, and into monetary policy discussions. Taylor rules have been derived in NK models, though the arguments are a little slippery. Generally, an optimal policy rule in a NK model would be some relationship between the policy instrument(s) and exogenous variables, but of course real GDP and the price level are endogenous, so one has to go through some contortions to coax the Taylor rule out of any model. The argument would seem to rely on what is observable to the central bank and what is not.

So, suppose that the central bank adopts a NGDP target. Then, the central bank must also have an approach to implementing such a target. Presumably the advocates of NGDP targeting think that standard central banking practice works, i.e. that a sensible approach to policy over the very short term is to specify an intermediate target for the fed funds rate, with the target set according to the current state of the economy relative to the NGDP target. Thus, we could specify the implementation of the NGDP target as a rule

R(t) = p(t) - p(t-1) + c[y(t) + p(t) - y* - p*] + r*,

where y*p* is the log of the nominal GDP target and c > 0. We can then rewrite this rule as

R(t) = p(t) - p(t-1) + c[y(t) - y*] + c[p(t) - p(t-1) - p* + p(t-1)] + r*

What's the difference between this and the basic Taylor rule? Not much. (i) The coefficients on the terms governing the response of the fed funds rate to the "output gap" and the deviation of the inflation rate from its target are constrained to be the same. (ii) The interpretation of y* may be different. In the NK literature y* is the efficient level of aggregate output ground out in the underlying real business cycle model. The NGDP targeters seem to think of y* as the trend level of output. For practical purposes it does not make much difference, as the people who measure output gaps tend to think of trend GDP as potential GDP. (iii) The target inflation rate is not a constant, but the percentage deviation of the target price level from last period's inflation rate.

In sum, the NGDP rule fits well within the set of Taylor rules that people have considered, which deviate in various ways from the basic rule that Taylor wrote down in 1993. So what's new? Could it be that there is something different about what happens at the zero lower bound, which I have not accounted for thus far? Suppose we are at the zero lower bound, which is essentially the case currently, and the Fed announces, say, a target path for NGDP of 5% per year indefinitely. Could the Fed actually achieve such a target, even if it wanted to? No. Under current circumstances, there are no actions the Fed can take that could necessarily achieve such an outcome. Indeed, it is possible that the Fed could promise to keep the policy rate at 0.25% for five years in the future, and NGDP growth could fall below the target.

There is no magic in a NGDP target. I know people look at the state of the economy, and think that the Fed should keep trying things. Maybe something will work? Well, I'm afraid not. Even the FOMC dissenters, and their supporters are not quite ready to say that there is nothing the Fed can do under the current circumstances that could increase employment. But they should.

60 comments:

  1. Taylor Rule is backward looking, NGDP target is forward looking. It's also important that it be level targeting rather than rate targetting. Shortfalls in one year will be made up for in another year.

    "Presumably the advocates of NGDP targeting think that standard central banking practice works, i.e. that a sensible approach to policy over the very short term is to specify an intermediate target for the fed funds rate, with the target set according to the current state of the economy relative to the NGDP target"
    Sumner and Rowe seem to think the approach of specifying rates is wrongheaded.
    http://worthwhile.typepad.com/worthwhile_canadian_initi/2009/04/mechanical-metaphors-for-monetary-policy.html
    To them the zero-bound "trap" is just an illusion foisted on us by leftover Keynesianism.

    ReplyDelete
  2. The key, I think, is that those who call for NGDP targeting systematically employ (1) forward-looking variables and (2) level path targeting.

    The first feature is standard and I don't want to spend much time on it.

    The second feature states that the Fed wouldn't target 5% NGDP growth per period, but an average of 5% NGDP growth over time. This gives the policy rule "memory": if the Fed undershot its target in a prior period, it would be obliged to overshoot its target this period or in future periods so as to keep the time average of NGDP growth at target.

    This has effects on private-sector expectations. If current NGDP growth is below target, E_t(NGDP growth_{t+1}) will rise in anticipation of expected monetary easing. Under normal inflation targeting, even if inflation is below target today E_t(pi_{t+1}) remains at target. Given that a major component of aggregate demand today is expected aggregate demand in the future, level targeting provides for more effective stabilization policy than rate targeting.

    There is a difference at the zero lower bound when you have memory. Consider a first central bank that targets NGDP growth as a rate. That central bank will begin to increase interest rates as soon as NGDP growth rises above its target rate. However, a second central bank which targets not the flow NGDP growth rate but the average NGDP growth rate (or, equivalently, the expected future average NGDP growth rate) will stay at the zero bound even after NGDP growth rises to the target level. In essence it is a credible commitment to keep the interest rate lower, for a longer period of time.

    ReplyDelete
  3. "Taylor Rule is backward looking, NGDP target is forward looking."

    What's that mean? I have a target, and in the present I have to adjust some policy tool(s) in response to the state of the world. How is one rule backward-looking and the other forward-looking?

    "Sumner and Rowe seem to think the approach of specifying rates is wrongheaded."

    So they want to do what instead? What's the operating rule you issue to the Open Market Desk?

    ReplyDelete
  4. Steve,

    As a faithful follower of your blog, I am grateful for the mention. Here are some quick thoughts.

    NGDP level targeting is foremost about properly setting expectations. An explicit, publicly announced and understood NGDP level target would clearly signal that the Fed is committed to buying up as many assets as needed until some trend path on nominal GDP is achieved. Knowing the Fed's commitment to this regime would cause investors to begin rebalancing their portofolios automatically. This rebalancing would be far more pronounced than anything done under QE2 or Operation Twist, since the rebalancing is tied to an explicit objective. Investors would have an incentive to rebalance until the Fed's target NGDP target path was hit. FDR did something similar to this with a price level target between 1933 and 1936. FDR explicitly announced it, backed it with policy changes, and in turn it worked incredibly well in far worse circumstances. See Eggertson (2008). I also believe something similar was done in Sweden over the last few years and it was capable of restoring robust nominal spending. Both examples suggest a NGDP level target would gain real traction here too.

    Taylor Rules aren't as clear in communicating a target path, though the output gap does imply some kind of catch up growth. It is easier to signal something as transparent as a current dollar spending path then trying to signal a closing of the output gap. Taylor Rules are also limited because there are competing measures of the output gap, inflation, and the neutral federal funds rate. A nominal GDP level target ignores those issues by focusing on nominal spending alone and letting the division between output and the price level fall wherever it does (although you do need to specify your trend growth rate based on some criteria). Thus, when for example a negative supply shock hits the Fed would not need to get all worked up about inflation, but rather focus on stabilizing nominal expenditures.

    Finally, in an ideal world NGDP level targeting would be done with NGDP futures contract. The Fed would target the market's forecast of NGDP through such a contract. Unfortunately, no such contract yet exists. Hopefully, one day it will.

    ReplyDelete
  5. Integral,

    1. The rule always has to be specified in terms of the observed state. You can write down a rule that includes a "forward-looking" variable such as a forecast of something, but that forecast has to come from the history that I see.
    2. I can write down a Taylor rule in terms of price levels relative to a target rather than inflation rates relative to inflation rate targets. NGDP has to be a special case. What makes that special case optimal?

    David,

    Why not just specify a price level target path? What makes you want real GDP in there, and included in this particular way? Do you have a particular view on the short-run nonneutrality of money, where it is coming from, and how you want to intervene to exploit it?

    ReplyDelete
  6. I prefer a real GDP target.

    ReplyDelete
  7. Its fascinating how we go through these monetary policy fads - largely due to great marketing skills. Woodford, Taylor, Beckford and Sumner should really have positions in the marketing department. Through all the fashions the real economy keeps on growing at 2% trend, 130 years and counting, with a few productivity blips here and there.
    Really, Prescott is the only intellectually honest macro guy out there. Money is second order in the scheme of things.

    ReplyDelete
  8. I think what is true is that, if the central bank is behaving appropriately, we should hardly notice it is there. I think it is also true that the ability of a central bank to bringing about improvements in economic welfare, working through real activity, is extremely limited. However, a central bank is also capable of inflicting significant damage if it behaves badly.

    ReplyDelete
  9. There is a subtle difference in the rules. The standard Taylor rule has the Fed reacting to deviations of inflation from target inflation. The NGDP targeting rule has the Fed reacting to deviations of the price *level* from target (the p(t-1) terms in the second bracket cancel out).

    This may seem like a minor difference, but it's not from the perspective of determinacy. In particular, any positive coefficient on the price level targeting rule leads to determinacy; for the inflation-based rule some restrictions have to be satisfied. This means you could get sunspots and non-optimal, non-fundamental fluctuations.

    Another advantage of the price-level rule is that it, in a sense, forces you to commit to fixing past "mistakes". The same is not true with the inflation rate rule.

    Of course, as you point out, neither of these rules are "optimal" in the sense of maximizing welfare. But they are simple and implementable. That's why they've become so popular.

    ReplyDelete
  10. "those who call for NGDP targeting systematically employ...forward-looking variables"

    as do those who favor inflation targeting.

    ReplyDelete
  11. Nominal spending is equal to the money supply multiplied by velocity.

    For the zero nominal bound to prevent the Fed reaching is NGDP target, it would have to be true that any increase in the money supply would be exactly offset by a decrease in velocity. While this could happen when the asset the Fed purchases is a perfect substitute for money, because those who previously held the asset now just sit on the money instead, it can be avoided simply by purchasing other types of assets from people who want money to spend rather than to hold.

    Alternatively, my idea would be to have the Fed set interest on reserve balances, for example, 2 percent below the interest rate on short-term government bonds. This means that when the interest rate on such bonds approaches zero, interest on reserves will go negative. This institutional change should effectively eliminate the zero nominal bound, because base money and T-bills should never become near perfect substitutes.

    ReplyDelete
  12. Steve,

    You wrote, "Do you have a particular view on the short-run nonneutrality of money, where it is coming from, and how you want to intervene to exploit it?"

    No. For those that believe money is non-neutral, this seems to be a benefit of this approach relative to the Taylor rule. For example, most monetary economists believe that money isn't neutral in the short run. However, they greatly differ with regards to the source of non-neutral (limited participation, sticky prices, sticky wages, etc.). A Taylor rule forces you to have a precise framework that explains the division between output and prices. A nominal GDP target doesn't require a specification of the division between prices and output. This was one of McCallum's main arguments for supporting a nominal GDP target.

    Also, you are correct that in the long run a nominal GDP target is equivalent to a price level target. However, what happens when the price level rises due to factors unrelated to money growth? For example, if the price of oil is rising because of increasing demand in places like China, India, and elsewhere this (according to the literature on oil prices and business cycles) creates higher costs for firms, which are born out in slower production and higher output prices. Under a price level target, the central bank would have to tighten policy. Under a nominal income target, the policy response would depend on the degree to which the reduction in output would offset the increase the in price level.

    ReplyDelete
  13. "Through all the fashions the real economy keeps on growing at 2% trend, 130 years and counting, with a few productivity blips here and there."

    What monetary economists believe that money influences the trend of real GDP? None that I know of.

    ReplyDelete
  14. "Another advantage of the price-level rule is that it, in a sense, forces you to commit to fixing past "mistakes". The same is not true with the inflation rate rule."

    Yes, that's true, but what is output doing in the rule?

    "Of course, as you point out, neither of these rules are "optimal" in the sense of maximizing welfare. But they are simple and implementable."

    1. So how much welfare are you willing to sacrifice for simplicity? How much do you think you are sacrificing in this case?
    2. Part of my argument is that there are circumstances where this is not implementable. Sometimes you just can't hit the target.

    ReplyDelete
  15. "Part of my argument is that there are circumstances where this is not implementable. Sometimes you just can't hit the target."

    Doesn't this apply to price level targets as well.

    ReplyDelete
  16. Lee,

    You can't peg the interest on reserves to the Treasury. Bond yields are effected by monetary policy, which would in turn be effected by bond yields, which in turn would be effected by monetary policy, which...

    ReplyDelete
  17. Both the Talor Rule & NgDp targeting are backward looking, i.e., rely on variables reported with a lag - which can't be projected forward with any certainty.

    But focusing on NgDp will prevent the economy from falling into a black hole.

    ReplyDelete
  18. Bill Woolsey responds:
    http://monetaryfreedom-billwoolsey.blogspot.com/2011/10/williamson-on-ngdp-targeting.html

    ReplyDelete
  19. Benjamin may be a layman who hasn't studied as much economics as folks like our host, but he's not gibbering or foaming at the mouth or citing completely marginal thinkers. And he doesn't resort to insulting people who disagree with him. If you think his argument is wrong, just explain why. Say you disagree that his hypothetical bond-buying program will create inflation, or that the Fed is legally restricted from some action he proposes, or say that more inflation will not help the "real" economy and that is the source of our unemployment or output-gap problem. I know it requires a bit more effort than snark, but this is an econ blog not Gawker or whatever else the kids these days read.

    ReplyDelete
  20. "Doesn't this apply to price level targets as well."

    Yes.

    ReplyDelete
  21. Wonks,

    What is Woolsey's model? It was hard to understand what he was objecting to.

    ReplyDelete
  22. Evidently, censorship has become a method of "winning" arguments.

    ReplyDelete
  23. Stephen,
    I don't quite understand your point. Are you really saying that you don't believe the central bank can create any arbitrary level of inflation? I must be misunderstanding you.

    ReplyDelete
  24. Nick Rowe outlines some concrete steps for E(NGDP) targeting:
    http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/10/engdp-level-path-targeting-for-the-people-of-the-concrete-steppes-.html

    I'll tell Woolsey you found his model unclear. He thinks in your implicit model the Fed would wind up purchasing all the assets in the economy (though maybe his model of your model differs too much from your actual model). Nick Rowe argues against that outcome in the above link.

    ReplyDelete
  25. Nominal GDP targeting will never work outside of economic models.

    In real-time the estimates of GDP are often way off, and the Fed can't trust these numbers for policy purposes. Remember this year's annual BEA revision of the GDP data?

    The entire argument rests on the Phillips being true.

    Perhaps worse, inflation expectations could become unanchored by the policy induced swings in inflation.

    KP

    ReplyDelete
  26. do you really think there is nothing the Fed could do if it wanted to achieve a 5% NDGP growth target? That's silly. If the Fed announced a massive permanent increase in the money supply, bought treasuries, mortgages, heck even notebooks and pencils, until the 5% target was achieved, I guarantee you we would get to the 5% target.

    ReplyDelete
  27. "I don't quite understand your point. Are you really saying that you don't believe the central bank can create any arbitrary level of inflation? I must be misunderstanding you."

    No, that's correct. Most of the argument is in here:

    http://newmonetarism.blogspot.com/2011/08/liquidity-traps-money-inflation-and.html

    ReplyDelete
  28. The entire argument rests on the Phillips being true.

    Perhaps worse, inflation expectations could become unanchored by the policy induced swings in inflation.


    scary sounding statements based on no logic or evidence whatsoever. that's different than the current defacto core inflation targeting how exactly? CPI says rents are up 7% while housing prices are off 30+% and mortage rates are lowest in 30 years... but the housing component of the CPI is going up? GDP gets revised but CPI is perfect?

    If the NGDP target is 5% growth, inflation is not much higher than 5%, and likely significantly lower, like 2%. how is that unanchored? If inflation + RDGP are too high the fed tightens, same as they do now. If RDGP is above "full employment" and inflation pressures are builidng, the fed tightens, same as now.

    ReplyDelete
  29. dwb,

    No way. Again, see:

    http://newmonetarism.blogspot.com/2011/08/liquidity-traps-money-inflation-and.html

    ReplyDelete
  30. "Nominal GDP targeting will never work outside of economic models."

    Actually, I have not yet seen an argument for NGDP targeting tied to a serious model.

    ReplyDelete
  31. Stephen,
    I read your article that you say explains why the fed could not create any arbitrary amount of inflation. It does not. You preclude all sorts of things the fed could do...say buy up all commercial paper that anyone wants to offer. Your argument seems to be the fed can't do anything if they restrict themselves to the stuff they have already done. Well no one can disagree. But why is this an interesting comment?

    ReplyDelete
  32. "You preclude all sorts of things the fed could do...say buy up all commercial paper that anyone wants to offer."

    At best that does nothing. At worst, it reallocates credit, and does nothing useful for us on net.

    ReplyDelete
  33. Those pushing NGDP targeting seem to be rather sanguine about the decomposition of NGDP growth into RGDP growth and inflation. But presumably there are welfare consequences between ending up with 3% RGDP growth and 2% inflation versus -1% RGDP growth and 6% inflation.

    ReplyDelete
  34. Stephen,
    You are changing the subject. I never said massive fed action to cause inflation does anything "useful on net," I just said it would increase inflation. Where Zimbabwe has gone we could surely follow.

    ReplyDelete
  35. Steve,

    "Actually, I have not yet seen an argument for NGDP targeting tied to a serious model."

    McCallum and Nelson (JME, 1999): http://ideas.repec.org/p/hhs/iiessp/0644.html

    ReplyDelete
  36. Robb,

    One way to put it is that, given the large stock of excess reserves in the system, inflation has come unhinged from policy actions. I think there are contingent paths for policy that would allow the Fed to control inflation from here on out, but NGDP targeting won't do it, and neither will buying commercial paper.

    ReplyDelete
  37. "McCallum and Nelson (JME, 1999): http://ideas.repec.org/p/hhs/iiessp/0644.html"

    I think you and I have different notions of "serious."

    ReplyDelete
  38. Apropos of the point I raised above, McCallum and Nelson (JME, 1999) write, "real output and employment fluctuations might be smaller on average than with pure inflation targeting ... [this] result can not be assured, because of the profession’s ignorance concerning the mechanism by which nominal income growth is split between inflation and real output growth components." The "ignorance" M&N refer to appears to be rarely acknowledged in the claims of Sumner & Co.

    ReplyDelete
  39. Wouldn't negative interest (or a tax) on excess reserves push a lot of that out and cause inflation? I know you've said the Fed isn't authorized to do that, but they're already breaking the law.

    ReplyDelete
  40. Phil:

    Everything rests on the Phillips curve being correct: economic activity and inflation co-move positively. There can never be a contradiction between them.

    Thus, the NGDP targeting people don't believe that they have to worry about your decomposition.

    KP

    ReplyDelete
  41. "Wouldn't negative interest (or a tax) on excess reserves push a lot of that out and cause inflation? I know you've said the Fed isn't authorized to do that, but they're already breaking the law."

    Yes, if you allow the Fed to tax reserves, there's no liquidity trap. Problem solved. The law that permits interest on reserves seems deeply flawed, but I doubt that the Fed would want to violate it in such an obvious way as to charge fees for holding reserve balances.

    On another note, don't you think that a problem with NGDP targeting is the problem with other simple rules. Here I'm thinking of Friedman-style money growth targeting. Friedman wanted to think that there was a simple money demand function that was structurally invariant to the policy intervention (changing to a policy rule with money growth targeting), and also to technological and regulatory change. Not the case. There's the same problem with NGDP targeting. This seems to require that you think of the long run real GDP trend as fixed. Also I think there has to be a fixed relationship between the fluctuations in economic welfare and fluctuations in nominal GDP, about trend. That seems far-fetched.

    ReplyDelete
  42. Steve,

    Why don't you consider McCallum and Nelson to be serious?

    I can understand if you have quibbles regarding the role of money and sticky prices, etc. in these types of models -- I share that view. Is that what you are talking about when you say the model isn't serious? If so, then I do think we have different meanings of "serious." For example, I think Woodford does serious analysis. I think McCallum and Nelson do serious analysis. I might differ with them in regards to the usefulness of such analysis, but that reflects my preferences, not a level of seriousness. They would obviously disagree.

    ReplyDelete
  43. I was half-joking. The idea in McCallum and Nelson's work, some of McCallum's work, and the Taylor-influenced literature is one of robustness. We're uncertain about what a good macro model is. If I can demonstrate that a particular policy rule does well in a wide class of models, under some well-defined criterion, then I'll think of that rule as robust, and argue that policymakers should use it. McCallum and Nelson set up a particular sticky-price model, and simulate it under a NGDP target and compare that to some alternatives. The criterion seems to be how well it does in hitting an inflation target while also keeping the variance of the output gap low.

    1. The criterion for evaluating good performance has to be model-specific. I'm not sure how you make these cross-model comparisons in a reasonable way.
    2. Even in McCallum and Nelson's framework, they're not telling us what policy is optimal, or how the NGDP target does relative to an optimal policy.
    3. I find it hard to believe that any of the models we play with have the property that the welfare losses from suboptimal monetary policy depend only on the gap between nominal GDP and a constant nominal GDP growth path.

    ReplyDelete
  44. If we look in the long term, the GDP trend does seem to be fairly constant. The Great Depressin seems like a blip, which may be why Friedman came up with a "plucking model".

    NGDP targeting may be vulnerable to the Lucas Critique. I think the market monetarists would say the various measures of money were mistaken targets because we don't care about money in itself, it's just a means to an end. NGDP (or it's relation to the long-run trend) is their representation of aggregate demand and so if they want to tinker with aggregate demand they tinker with NGDP. It could all be pointless if, like in Hume's thought experiment, it just resulted in everyone adding a zero on the end of all prices (if a ten fold increase is ridiculous, then these are agents who price in the more sensible binary system) without having any real effect. Their hypothesis is that we aren't in the typical market clearing situation precisely because prices aren't that flexible.

    ReplyDelete
  45. "If we look in the long term, the GDP trend does seem to be fairly constant."

    Friedman looked at the historical data and said: "money demand seems pretty stable."

    ReplyDelete
  46. there are too many things wrong with your assertions.


    "The Fed has no hope of moving asset prices." The spread between mortgages and treasuries is higher now than in the spring. I guarantee if the fed bought 1 Tn in MBS, MBS prices would move (the spread would compress). If the Fed bought asset backed securities or Greek or Italian debt, those prices would move too.

    your premise that the system is "awash with reserves to the point where the marginal value of reserves in financial transactions is essentially zero" is contradicted by your own evidence. the fact that reserves are marginally more expensive (i.e. scarce) than t-bills, even if only a few bps, tells you money is still tight. The private sector cannot "create reserves" as you claim. I agree its confounding to have IOR at 25bps (btw this is a policy not a law), but nevertheless reserves are still relatively scarce. the fact that we have a lot does not prove we have enough.

    the "other liquid assets" which have "different liquidity properties" but are "essentially identical" for exchange is not accurate and is a gross oversimplification. MBS, ABS, most of the things you mention all get collateral haircuts. the yield spread on these securities are all highly correlated with other measures of financial stress (the TED spread, the OIS spread, the VIX, AAA-BBB credit spread etc). Those are assets whose price will fall in times of stress, and are NOT useful for "exchange."

    "At the zero lower bound the fed cannot achieve a higher price level except for talk about the future."

    That is exactly the point of NGDP targeting. The fed commits to doing anything and everything. there is nothing preventing the Fed from buying corporate debt, muni debt, or many other things. Maybe in a grossly oversimplified and inaccurate world of reserves and dairy farms, sure. But: there is a huge logical and evidentiary gap between an agnostic and an atheist.

    and, if you are an atheist on QE (no effect), hen its simultaneously pretty irrational to be against something like more QE that you claim will have no effect.

    ReplyDelete
  47. "I guarantee if the fed bought 1 Tn in MBS, MBS prices would move (the spread would compress). If the Fed bought asset backed securities or Greek or Italian debt, those prices would move too."

    You guarantee it, so I guess it must be true.

    "your premise that the system is "awash with reserves to the point where the marginal value of reserves in financial transactions is essentially zero" is contradicted by your own evidence. the fact that reserves are marginally more expensive (i.e. scarce) than t-bills"

    No, it's the other way around. The interest rate on reserves is higher than the interest rate on T-bills, so the price of reserves is lower than the price of T-bills, i.e. T-bills carry a higher liquidity premium.

    "That is exactly the point of NGDP targeting. The fed commits to doing anything and everything."

    You're not serious, are you? The Fed should commit to intermediating everything? Do you understand why it is a bad idea for the Fed to be buying private assets?

    ReplyDelete
  48. Friedman looked at the historical data and said: "money demand seems pretty stable."

    Yes, as did Meltzer and Laidler and Lucas. It is also not entirely clear that these guys were incorrect. Most studies that say that money demand isn't stable use cointegration techniques. Using these techniques, they fail to find stable cointegrating relationships and conclude that money demand is unstable because of financial innovations. As McCallum has detailed, however, these statements are contradictory. If the demand for money is influenced by financial innovations, which I think that most of us would agree with a reasonable proposition, this would necessarily imply that one would not find evidence of cointegration. Thus, saying that money demand is unstable based on an absence of cointegration is not correct.

    More importantly, Steve, have your TA get the data from some of these "unstable money demand" papers and try to replicate their results -- they are not robust.

    ReplyDelete
  49. if you want to have debate about whether its appropriate for the fed to buy other assets like MBS, thats a useful debate. If you want to debate elasticity- whether it would take 1 Tn or 10Tn in more QE, that's also interesting. But you are not making that argument, you are making the argument that they can't, period. "The Fed has no hope of moving asset prices." Such a statement is patently, absurdly, and easily proven false. The Fed is being held back, not because it can't target long term nominal rates, or move asset prices, but because of the potential distortions it might causes (a legitimate and serious concern). The Fed could target NGDP, its a question of whether the benefits are worth the costs, depending on what it would have to do to get there, for how long, and how hard it will be to exit.

    Yes, it may have to raise it's balance sheet to 5Tn and accomodate some likely commodities price rises. But at full employment, commodities prices will be higher, there is no getting around that.

    but saying that it can't be done... well like i said, relish the opportunity to be proven right if they embark on another round of QE.

    ReplyDelete
  50. anonymous 8:27,

    My point was that Friedman first asserted this, then central banks tried targeting monetary aggregates and it didn't work. Why? Because money demand is not stable. There is good theory and empirical evidence for why that is the case. You're pretty hard core if you want to argue otherwise.

    dwb,

    You are just making assertions. I don't see an economic argument in there.

    ReplyDelete
  51. Why is anyone worried about "controlling inflation" when core is at 2 percent for last year, and many conservative economists have said that the CPI probably overstates true inflation (due to rapidly evolving goods and services). When GDP is 15 percent below trend and unemployment is at 8 percent.

    Additionally, far smaller fractions of US labor force are unionized than previous eras.

    Additionally, goods, services, capital and labor easily cross into the USA (less os lately for labor, perhaps)--and price surges are met with fresh supply, much more so than previous decades.

    A peevish fixation on inflation, rather than economic growth, innovation, and commercial freedom is becoming the intellectual death knell of right-wing economists.

    The three-year period ended August 2011 was the lowest CPI for a three-year period in the postwar era. If you are doing the "Chicken Little" dance now on inflation, when would you be happy? In Japan after 20 years of deflation?

    Several years of moderate inflation and strong economic would be a great tonic for the USA.

    ReplyDelete
  52. Again, you said "the fed has no hope of targeting asset prices." The money supply is potentially infinite. Infinite demand chasing finite supply must necessarily move prices. Any model may be inelastic locally, but no one would actually extrapolate that to the case, where say the Fed beefs up its balanace sheet to, say, 50 Tn.
    If the Fed expanded its balance sheet to 7 Tn, buying MBS, the price of MBS would go up (yields decline) because prices are not inelastic. The total stock of GSE MBS is in that neighborhood, so the Fed now owns all GSE MBS. At the margin, the price of that last MBS tranche is virtually zero. Don't take my word for it, ask your favorite bond trader or FOMC member what would happen if the Fed announced 7Tn in MBS purchases. Then another 7 Tn in corporate debt purchases, then 20Tn in stock market purchases. maybe a cartoon model of says asset prices don't move, but taken to its logical conclusion, its empirically false. There's been plenty of rebuttal from other economists.

    ReplyDelete
  53. My point was that Friedman first asserted this, then central banks tried targeting monetary aggregates and it didn't work. Why? Because money demand is not stable. There is good theory and empirical evidence for why that is the case. You're pretty hard core if you want to argue otherwise.

    I am very familiar with this literature and the problem is that most people are talking past one another. Read Lucas, Laidler, Meltzer, and Friedman. They are arguing that money demand is stable over the long-run. Then, look at the empirical evidence of those who say that they are unstable -- they use much higher frequency data.

    Laidler discusses this here:

    http://research.stlouisfed.org/publications/review/90/03/Legacy_Mar_Apr1990.pdf

    Also, Lucas was clearly aware of this issue as well:

    http://ideas.repec.org/a/eee/crcspp/v29y1988ip137-167.html

    Finally, I think that we would agree that there is some good theory to suggest money demand is unstable. I would disagree that there is good empirical evidence. Again, many of these studies are not robust.

    Stable money demand doesn't imply that we should target monetary aggregates. I think there are many reasons why we shouldn't that are independent of whether money demand is stable.

    ReplyDelete
  54. "Read Lucas, Laidler, Meltzer, and Friedman."

    Exactly. The hard core quantity theorists.

    ReplyDelete
  55. "Read Lucas, Laidler, Meltzer, and Friedman."

    Exactly. The hard core quantity theorists.


    Are they quantity theorists because they looked at the data or did they interpret the data the way that they did because they are quantity theorists? Your response seems to suggest the latter. I'm willing to give people the benefit of the doubt on both sides of the debate and assume that their conclusions aren't driven by their biases.

    The point that I am making is as follows.

    1. It is not clear if money demand is unstable. Evidence using annual data suggests that it is stable. Evidence that uses monthly and quarterly data is mixed, but it leans toward instability (although some of these findings are not robust).

    2. Even if money demand is stable in the long run, it does not imply that the central bank should target monetary aggregates.

    At least on point 2 I think that we would agree.

    ReplyDelete
  56. "Hard core" meaning the primary advocates. The rest of the profession has moved on.

    1. Adding up some asset quantities and calling that stuff money is not a useful exercise. Many assets not in M1 or M2 nevertheless play important roles in various kinds of exchange, and every asset's role in exchange is important in determining its price.
    2. The "demand for money" is not a useful concept, in part because of (1), and also because that relationship is not structural. We can't think about the demand for an asset in the same way as we think about the demand for broccoli.

    ReplyDelete
  57. 1. Adding up some asset quantities and calling that stuff money is not a useful exercise. Many assets not in M1 or M2 nevertheless play important roles in various kinds of exchange, and every asset's role in exchange is important in determining its price.
    2. The "demand for money" is not a useful concept, in part because of (1), and also because that relationship is not structural. We can't think about the demand for an asset in the same way as we think about the demand for broccoli.


    Now, we are getting somewhere. I agree on both points. Although I will note that Bill Barnett and his cohorts have been working on ways to better measure the money supply -- both from a mathematical and from an economic perspective.

    ReplyDelete
  58. I tend to think of Barnett's approach as applying directly the aggregation theory that was developed to think about broccoli and carrots to think about assets. Not a good idea. Further, what's the goal anyway? What are you going to do with the monetary aggregate once you construct it?

    ReplyDelete
  59. Steve,

    I think that you are partially correct (actually, I should say I partially agree with you, whether you are correct is independent of what I think). Empirically, I see the usefulness of money as an indicator variable. Nonetheless, for it to be meaningful in this sense, we need to have some idea about what we are measuring and how and why we are measuring it. Barnett's approach doesn't completely answer those objectives yet as a lot of what has been done is to develop acceptable index number counterparts to existing monetary aggregates.

    The question is then whether we can get some measure of money that isn't simply based on intuitive ideas about liquidity or other asset characteristics. You would likely argue that this is impossible, irrelevant or both and you might be correct. I am slightly more optimistic.

    In the past, you have mentioned using currency. I think that this is intriguing and I think that within the context of many monetary models that is actually how money is defined (whether or not this is carried over into the corresponding empirical analysis, which it is often not...ugh). My only problem with currency as an indicator variable is that I am not sure how well it performs empirically as a sign of future inflation. For example, if currency in circulation rises, how do I know that it is the result of monetary policy or a change in the demand for currency relative to other assets? [If you look at the episodes in Friedman and Schwartz's Monetary History, for example, a great deal of contractions coincided with increases in the demand for currency relative to other assets, such as demand deposits.]

    Regardless, there is much work to be done and I think there is a lot to glean from the search paradigm. On that, I know that we agree.

    ReplyDelete
  60. Yes, I agree for the most part.

    "...I think that within the context of many monetary models that is actually how money is defined (whether or not this is carried over into the corresponding empirical analysis, which it is often not...ugh)."

    Yes, this is one of my pet peeves. In the model, the stuff used in exchange looks like currency, and there are no financial intermediaries anywhere in sight. Then, the empirical work uses M1. Ugh for sure.

    "...if currency in circulation rises, how do I know that it is the result of monetary policy or a change in the demand for currency relative to other assets?"

    Yes, exactly. For the US, we don't even know how much of the outstanding stock of currency resides domestically.

    ReplyDelete