25 January 2016

Objectives vs. Tools of Monetary Policy

In the comments of one of Nick Rowe's recent posts, Scott Sumner has accused me of confusing objectives and tools of monetary policy:
You are looking at the causal effects of QE, whereas it makes more sense to view QE as the effect of a tight monetary policy that drives rates to zero. If you do a more expansionary monetary policy, such as currency depreciation, then you do not need as much QE. QE is a defensive mechanism, monetary policy needs to be viewed in terms of the policy goals of the central bank, and in terms of whether it will do whatever it takes to reach those goals.
Basically, Scott is suggesting that quantitative easing isn't actually a monetary policy, and is instead the natural conclusion to what he does view as monetary policy -- currency depreciation. Here, Sumner provides an interesting set of definition for what constitutes monetary policy and, more generally, what can reasonably be considered exogenous to a central bank.

In his mind, exchange rates are basically exogenous to the extent that central banks try to influence them. This is evident from his implicit assertion that, if central banks are "doing whatever it takes to reach [their] goals," they will invariably reach those goals. Of course, this isn't necessarily news, everyone has know Sumner's opinion that central banks are nearly omnipotent for quite some time, but this time he has laid it out more directly.

According to Sumner, the evolution of any nominal variable over time can be completely controlled by a central bank and, as such, can be used as a point of criticism for that central bank: "monetary policy needs to be viewed in terms of the policy goals of the central bank." As such, the actual polices that central banks follow are completely irrelevant; it doesn't matter what the path of interest rates is, the correct judge of current Federal Reserve policy (for example) is whether or not inflation is on target.

Of course, I, along with I hope the majority of people, don't see monetary policy in this light. Sumner seems to have made a point of confusing monetary policy -- e.g., QE, interest rate setting, open market operations -- with whatever nominal variable he happens to care about at the moment -- in this case exchange rates. This separation is important; it allows us to understand more directly a central bank's goals and how it intends to achieve those goals.

Evidently, Scott could care less about the how and only wants us to focus on the goals. He basically has reduced his thinking about monetary policy to the point that he views NGDP as an instrument of the central bank -- effectively an exogenous variable -- rather than a variable that a central bank may act to control. This level of abstraction from the operation of monetary policy, in my opinion even more grievous than the New Keynesian obsession with the nominal interest rate, is what allows Market Monetarists to callously ignore every model that doesn't allow exogenous NGDP that says the zero lower bound actually represents a constraint on monetary policy.

If central banks could make NGDP exogenous, would they be able to make NGDP exogenous? Naturally, but no one should care about the answer to such a redundant question, yet this is effectively the answer that you get from Sumner; he'll simply assert that "the BOC can always depreciate the Canadian dollar. The zero bound is not an issue in Canada" (from an earlier comment on the same post). Naturally, we should all trust Sumner's clairvoyance on this issue, clearly no argument about monetary policy effectiveness is necessary (see my first comment on Nick Rowe's post, if you want one anyway) and we can rest assured that fiscal policy is never necessary.

Ideally, considering the ability of monetary policy to effectively deal with challenges should be at least of some consideration and, since monetary policy has proved theoretically capable of offsetting the demand-side effects of fiscal stimulus among other shocks, the only point at which this can be of much concern is the zero lower bound. Both Sumner's and Rowe's refusal to give theoretical arguments against me in this area is rather troubling, evidently just assuming monetary policy is effective in every circumstance is completely acceptable.


  1. John: I have a limited stock of apples in my basement, and an unlimited stock of bits of paper with "IOU NR" written on them. For some reason people like owning my IOUs. If I swap my IOUs 1 for 1 with apples, in either direction, that is what they are worth (unless I run out of apples, in which case my IOUs depreciate). I cannot run out of bits of paper with "IOU NR" written on them. In what world would the demand for my bits of paper, at any given exchange rate I announce, be infinite?

    Now suppose you are just like me, except you swap your IOUs for bananas.

    If we both try to depreciate our exchange rates against each other's IOU, at least one of us will fail. But nothing can prevent us depreciating our exchange rates against apples and bananas.

    1. Nick, I'm not sure I fully understand your sketch. I get that the IOUs are money, but I don't know what the apples and bananas map to in the real world.

      Otherwise, I'm pretty sure that the statement of most importance you make is "In what world would the demand for my bits of paper, at any given exchange rate I announce, be infinite?"

      Basically, in what situation does money demand ever become infinite?

      Infinite money demand immediately makes me think of MIUF (which I think also serves as a reduced form of shopping time and/or transaction cost models) in which money demand becomes infinite as the opportunity cost of holding money approaches zero. Maybe people really enjoy holding your IOUs (or they make shopping a lot easier or less time consuming), so they'll demand infinite amounts if there are no other assets that pay more interest.

      Otherwise, maybe your IOUs are simply required to facilitate transactions and everyone chooses to hold only enough to pay for their planned consumption unless they aren't forsaking interest from some other asset by holding the IOUs. In this case, they will simply take whatever amount of money you supply them, but they won't necessarily use it to spend; money demand is indeterminate and so is the price level.

  2. "... he views NGDP as an instrument of the central bank"

    Thus can I assume by fiat that the unemployment rate, the median wage rate, or RGDP is an instrument of the central bank and then criticize it mercilessly when if fails (refuses) to make us all wealthy and happily employed beyond our wildest dreams? ;D

    1. But more seriously, the MMists sometimes fall back on "The Fed could buy the whole world. Are you telling me inflation won't rise then?" (paraphrasing). It is hard to imagine inflation refusing to budge as more and more of the world's assets are gobbled up by the Fed. And with it goes NGDP I'd suppose. But then in that circumstance it's behaving in a manner totally unlike its normal mode of operation. There's a transition region between the two modes I'd guess.

    2. I wonder where the models break down with bad CB behavior though. According to models I cannot stop the flow of electric current through a coil instantaneously, so what happens when I physically break the connection while current is flowing through one? Do I get a spike to infinite voltage, and the resultant destruction of the universe? Or do we get a few 10s of 1000s of volts and a 1/4" long spark?

    3. Sadly, I'm not sure Sumner would find much wrong with your first comment...

      I keep meaning to write down a model where there are more assets than just money and government bonds. Needless to say, my intuition is that, at least in cash in advance models, the composition of the central bank balance sheet is irrelevant -- the central bank can buy the world, and nothing will happen to inflation at the zero lower bound.

      Regardless, you're right: this is a complete regime change for the central bank in question and it basically requires that is abandons it long term goals -- e.g. an inflation target -- in favor of keeping current output on target. Basically this is Krugman's result in his 1998 paper; the central bank won't allow future inflation to rise, so any monetary expansions are temporary and temporary expansions that are too large result in the zero lower bound and can do nothing to increase the price level at this point.

    4. Thanks John. What are your thoughts about Roger Farmer's "belief function?" Also, what do you make of Beckworth's NY Times Op Ed today?

    5. Beckworth seems grounded in the notion that, if the nominal interest rate was above the natural rate in the previous period, then this will have caused a decline in the current natural rate. Aside from simply being convenient, I don't know why MM's would assume this. It appears as if they are suggesting that either negative monetary policy shocks increase potential output (nonsense) or that people are really backward looking (more reasonable).

      I think the idea that people are backward looking is plausible, but it certainly doesn't square with the other tenets of MMism, so I don't see why they'd be espousing a theory that requires it.

      Of course, another possible mechanism for previously tight monetary policy causing declines in the natural rate is that it happens by magic and that they have come to know this through divine revelation, but then MMism sounds more like a religion than an economic school of thought.

      Regarding Farmer, I think the belief function approach for expectations is interesting, but (since I am definitely part of the mainstream) I still think that rational expectations are more structural (to the model, of course); 'belief functions' are a little too ad hoc for my taste.

    6. "It appears as if they are suggesting that either negative monetary policy shocks increase potential output (nonsense)"

      Can you elaborate a bit on that?

    7. So, there are two ways to look at this:

      1) The New Keynesian Way:

      Start with the basic consumption euler equation:

      (1) 1/c_t = B 1/c_t+1 (1 + r_t)

      where c is consumption, B is the discount rate, and r is the real interest rate.

      A similar relation can be written for 'potential consumption' - the level of consumption that prevails when prices aren't sticky:

      (2) 1/c*_t = B 1/c*_t+1 (1 + rn_t)

      if the interest rate is at the natural, then c_t+1/c_t = c*_t+1/c*_t. We can linearize (1) and (2) to write the real interest rate and natural interest rate in terms of expected growth of consumption and 'natural' consumption:

      (3) r_t = r* + g_t+1

      where g_t+1 is the growth of consumption over the next period and r* = 1/B - 1 is the discount rate.


      (4) rn_t = r* + g*_t+1

      Basically, the natural rate of interest is increasing in the expected growth rate of potential consumption; if the natural rate falls, then that means that current natural consumption is high relative to future natural consumption.

      The Market Monetarist argument is that tight monetary policy in the previous period causes the current natural rate to fall. This implies one of two things:

      1) tight monetary policy last period caused current potential consumption to rise relative to trend.

      2) tight monetary policy last period caused future potential consumption to fall relative to trend.

      In other words, either past tight money raises potential GDP now or lowers it two period in the future. Both of these are equally implausible in my view.

      2) The Neoclassical way:

      The natural real rate of interest is the marginal product of the natural level of capital.

      Start with the definition for natural output: Y* = F(K*,L*) = K*^a L*^1-a

      where Y* is potential output, K* is the capital stock consistent with potential output, L* is the level of employment consistent with potential output, and 0 < a < 1 is capital's share in production.

      The representative firm maximizes profits (Y* - w*L* - rn K*) subject to the production function which yields the following FOC for the natural real interest rate:

      rn = a(K*/L*)^a-1

      In words, the natural real interest rate is a decreasing function of the natural capital stock and an increasing function of the natural labor supply.

      Market Monetarists suggest that this rate falls when money was tight in the last period, which implies that previously tight money requires the natural capital to labor ratio to increase. Alternatively, since Y* = (K*/L*)^a(L*), this means that potential output must increase if money was tight in the last period (unless the increase in K*/L* was entirely caused by a falling natural labor supply, which would be really strange if caused by monetary policy). Apparently tight money either causes a whole lot more demand for capital all of a sudden or causes the natural level of employment to fall.

    8. Wow, great. Thanks John. It's easier for me to grasp the neoclassical way, but that explains it.

    9. John, Jason Smith thinks MM isn't falsifiable. Do you share that view?

    10. I basically agree with Jason; MMists will just claim that the central bank didn't want it enough when inflation is below target, despite any concrete steppes that the central bank actually took.

      I do think that Nick Rowe is consistently more reasonable than, e.g., Sumner or Beckworth, though. At least Nick bothered to respond here with a theoretical argument; Sumner would just tell me to look at the announcement effects of QE and pretend that there's no theoretical case for a liquidity trap.

    11. Tom,

      See my comment on Sumner's recent post for a little more explanation for why exactly I think MMism is not falsifiable and/or completely wrong.

    12. Thanks John! Your concluding paragraph made me laugh:

      "please, for the love of sanity..."

      I saw one sentence fragment in there that reminded me of the wrath of Rowe. It was actually the first post of Rowe's I ever read, and it's stuck with me all this time. But first your sentence fragment:

      "(leaving MB demand determined)"

      I know what you mean, but here's the Rowe post that made me laugh so long ago (being one of the 1st macro posts I ever read, and not being the sharpest tool in the shed, I probably read it three times before I had an inkling of what he was talking about).

      I still think of that progress in other contexts: not even not even wrong => not even wrong => wrong => not wrong

      But anyway, back to your comment and your response to E. Harding. I think it's very clear what you're saying there about exogenous variables. You've no doubt heard the MMs discuss Chuck Norris though, right? Here's some classic Nick Rowe:

      "MM is more about interpreting existing models differently and interpreting the existing data differently. Monetary policy is 99% expectations, so how monetary policy is communicated is 99% of how it works. And market prices (especially asset prices) are the main source of information we have on how that communication is being received."

      When I pointed that out to Jason, he responded:

      "If economics is 99% expectations then economics is 99% question begging."

      You seem to be saying that there's a gap in the MM description of how communication through the channel Nick describes is accomplished. MMs might say that Chuck Norris doesn't have to worry about such details: don't worry about how, he gets the communicatin' done all right! Or like I read today somewhere "Trump doesn't follow rules. Rules follow Trump."

    13. Sumner brings up Lars Svensson in his response. He wants to know if you think he's a crackpot. What is your view of Lars (if you have one)?

    14. My disagreement with MM's stems basically from the difference between our attitudes about what needs to be modeled: I think the effectiveness of monetary policy needs to be derived, they are happy to simply assume it. Normally, they are probably justified in this assumption, but the failure of QE to generate large amounts of inflation, for instance, explains why we need to have models that explain what happens with monetary policy at the zero lower bound and, provided those models qualitatively match up with reality (e.g., they predict that large increases in MB at the zero lower bound are useless), they should be used to determine the correct policy response.

      Regarding Svensson, I personally don't like his propensity to skip directly to the 3 equation linearized NK model and to suggest that price level targeting guarantees escape from a liquidity trap based on a model where liquidity traps don't actually exist (baseline NK models without money). Otherwise, I don't have a problem with him; at least when he suggests targeting the forecast, he doesn't implicitly mean pegging the forecast, like Sumner does but won't admit.

      Basically, I think that MM's make unfair assumptions: there reasoning often amounts to "assuming monetary policy is always effective, monetary policy is always effective." Naturally, everyone knows this, but no one should care. It is necessary to actually derive monetary policy effectiveness, and I've never seen Sumner or Rowe (don't get me started on Beckworth and Lars Christensen) even bother to do this. In fact, Nick frequently assumes that inflation targeting is equivalent to inflation pegging in his posts, but at least he's explicit and his results are usually robust to the assumption -- it serves to simplify instead of dodge.

    15. Regarding "leaving MB demand determined," I think this kind of reasoning is only acceptable when there is a direct link between the price that the supplier (in this case the central bank) and the quantity supplied/demanded of the good in question (in this case money). MM's like to extend this to any nominal variable: in their view the central bank can do this to any nominal variable (which is why Sumner has no problem saying money is tight when GDP growth is low) even though it is only reasonable when there is a direct link between the pegged variable and what the central bank can actually control: the monetary base. You could almost think of money demand as a completely normal market -- the nominal interest rate is the price of money, the central bank is the monopoly supplier, and agents demand for money is decreasing in the price of money. Since the central bank is a monopolist, it can set the price in the market to whatever it wants, thus allowing agents to hold whatever amount of money they want to given this price.

    16. Thanks for your views John. What about your thoughts on Glasner and Nunes?

    17. I don't read much from either Glasner or Nunes, so I can't really go into much depth on either, but I will say that the little I've read from Glasner annoys me less than Sumner (but interests me less than Rowe) and that Nunes (who I've read even less of) appears to be guilty of everything Beckworth and Sumner are.

    18. BTW, Glasner critiques the NY Times Beckworth & Ponnuru piece in his latest.

  3. Replies
    1. Unfortunately, Williamson has his own problems. Take, for instance, Neo-Fisherism.