Sunday, 11 May 2014

"Neo-Fisherites" and Fiscal Policy



I wanted to make an observation on what Noah Smith calls the "Neo-Fisherite" idea that raising interest rates raises inflation.  I don't intend to go over all aspects of it; suffice it to say, I think the idea is flawed.  I just wanted to look at one particular point which I think is interesting in that it says quite a lot about how economists of different schools frame questions differently.

The Fisher equation says that the expected real rate of interest is (approximately) equal to the nominal rate of interest less expected inflation.  The Neo-Fisherite idea is that if the central bank increases its nominal rate and holds it there, then the rate of inflation will adjust to restore the natural real rate of interest.  One way this might be achieved in theory would be through a sudden sharp fall in the price level as the immediate response to the change in interest.  From this point, the price level could then drift upwards at the increased inflation rate.  Needless to say, the sharp fall in prices is somewhat implausible.  What is rather more realistic in this scenario is that prices would edge down very slowly with depressed output in the mean time.

If this were the analysis, it would be fairly unremarkable.  What we would really be saying is that high interest rates lead to a prolonged period of deflation, which may eventually reach the point at which it can turn around.  For the Neo-Fisherite case to be interesting, it must involve no significant deflationary period.

There are all sorts of things to be said here, but one thing that intrigues me is what happens to government debt in this scenario.  If the central bank were to permanently raise its rate, then the price of long-dated bonds must fall.  So, without a sharp drop in the general price level, the immediate result of this policy would be a fall in the real value of government debt held by the private sector.  So we can't claim monetary neutrality - we have to accept there will be real effects.

You could deal with this by simply assuming all debt is short term or assuming that the revaluation has no effect on private sector behaviour.  However, to me this seems rather pointless, because you're then just assuming away most of the most important reasons why the Neo-Fisherite position shouldn't apply.

In Stephen Williamson's QE paper (which provides a detailed model in which the Neo-Fisherite case holds), this problem is avoided a different way.  With regards to fiscal policy, Wlliamson makes the assumption that taxes and transfers respond passively to central bank policy, so as to maintain a constant real value of government debt (unnumbered equation appearing between (25) and (26)).  So, because an increase in the central bank rate reduces the real value of debt, as described above, we are saying here that the government responds with expansionary fiscal policy.

So the argument appears to be that if the central bank raises rates and the government reacts by cutting taxes, then we will get inflation. 

Which is not really a very surprising result.  You could put that into an IS-LM / AS-AD model and get the result that raising interest rates would raise inflation.  You just need to assume that the fiscal policy is sufficiently expansionary to validate the central bank's inflation target, which is what Williamson's assumption achieves.

It may well be that this assumption is not essential for Williamson's result.  As I said, there are other ways of avoiding the problem, although to me all they do is make it more questionable whether that the result wouldn't apply in the real world.  It just seems really odd to me to make an assumption like this about fiscal policy and pay such little regard to it, when the same assumption would have massive consequences in a differently framed model.  

37 comments:

  1. What if the fed holds interest rates at 0 or goes negative due to low growth and prices don't deflate due to inflationary pressures from resources contraints, conflicts, environmental costs and social costs from underperforming economy? The fisher theorem cant hold as far as I can see.

    ReplyDelete
    Replies
    1. Yes, I think there are all sorts of problems with this idea.

      Delete
  2. Perhaps it is less controversial that falling interest rates (so the rate of change not the level) lead to deflation because they entice inward capital flows and so currency strength.
    Falling interest rates lead to increasing valuations for assets that give a cash flow. The Volcker/Thatcher rate hikes set things up for decades of falling interest rates with subsequent asset price inflation and inward capital flows.
    I guess it is true though that asset prices and gold prices would see a very sudden step change down if interest rates were hiked up a lot? That was the brief initial response to the Volcker/Thatcher rate hikes.

    ReplyDelete
    Replies
    1. I think it is fairly essential to the Neo-Fisherite idea that the central bank raises rates and keeps them there, and that everyone (eventually) expects rates to stay there. Raising rates with the expectation that they will then fall from that level is rather different.

      Delete
  3. Nick Edmonds,

    Have you read this?:
    http://informationtransfereconomics.blogspot.com/2014/05/a-neo-fisherite-rebellion-yes-please.html

    Or this?:
    http://informationtransfereconomics.blogspot.com/2014/05/blowing-anti-neo-fisherite-model-out-of.html

    Jason's information transfer model (ITM) has a unique view of expectations:
    http://informationtransfereconomics.blogspot.com/2014/05/the-effect-of-expectations-in-economics_5.html

    And the ITM finds that across various nations the ratio of the currency in circulation to NGDP seems important (in terms of what increases to MB accomplish). Take a look at the final chart here in particular:
    http://informationtransfereconomics.blogspot.com/2014/01/it-really-does-seem-to-be-about-size-of.html

    ReplyDelete
    Replies
    1. Tom Brown, is that last link a bit like Gibson's Paradox that held during the 1821-1913 "true gold standard" period? Then, nominal interest rates were very tightly correlated to the price level but entirely uncorrelated to inflation/deflation. The Bank of England could not provide enough bank reserves when the price level (aka NGDP) was high because they were constrained by the limited stock of monetary gold so interest rates became high whenever NGDP was high enough to provide a strain on settlement of payments using the limited stock of bank reserves. Likewise today, monetary policy has traction only when the stock of bank reserves is a limiting factor.

      Delete
    2. Unfortunately, this stuff is just too unconnected with stuff I'm familiar with. I think it would take me some time to work out what I thought of it.

      Delete
    3. @stone, It's 3 AM where I am (California)... I read you comment a couple of times but I can't say for sure there's a connection to what you describe (but it could just be my tired brain!). You might ask Jason... he's happy to answer questions. I'm *slowly* learning more about his model and his approach, but he's the expert. I can tell you that on the last link, on the first chart, Japan and the US today, and the US back during 1929 to 1944, were past the knee of the curve (right hand side of plot), and thus increasing MB does not seem to affect the price level (like it does still in Canada or the US during the 1970s). The reason for this in terms of the ITM has to do with the information content of those additional reserves (according the theory). I'd better stop now before I butcher Jason's theory. As far as I know he's the only one out there using this kind of information transfer model for econ and macro. His blog is pretty fascinating. I can follow some of the posts better than others. You'll see I ask him a lot of questions in the comments section.

      For example, this one I can follow pretty well:
      http://informationtransfereconomics.blogspot.com/2014/05/expectations-destroy-information.html

      He applies the results of that to several subsequent posts in which he uses data rather than just a simplified model.

      His take on the US (and Japan) now is that we're stuck in this low inflation situation. We could escape it with a bit of hyperinflation or *I think* with the appropriate fiscal policy (again according to his model).

      @Nick E: yes it is quite different. Since I'm very low on the totem pole in terms of understanding anybody's models at this point, I'm equally lost looking at his stuff as I am at yours or at the model that Simon Wren-Lewis teaches.

      Jason does attempt to show how the ITM can reconstruct more traditional econ results though, in certain situations: for example the QTM. He also looks in detail at some Nick Rowe posts. Here's a recent example:
      http://informationtransfereconomics.blogspot.com/2014/05/equilibrium-in-two-good-market.html

      Honestly, I didn't delve into that last one, since Rowe's original was a bit beyond me... I might be able to glean a bit more from it if I go back and study it a bit (Rowe's that is), but perhaps someone who understands Rowe's original post might be able to get a bit more out of Jason's. What attracted me to Jason's blog was a previous look he had at a Rowe post... and then I saw his take on expectations (the most mysterious part of econ to me so far) and I was intrigued by the way he approached it, and his criticisms of the way they are treated, especially by some of the MMists.

      So ultimately, I don't know if it's worth learning the ITM. But I like that it seems to provide more concrete answers where NKers or MMists might invoke the mysterious "expectations" answer. How well it stands up, I guess we'll find out.

      Thanks for taking a glance though! I value your opinion.

      Delete
    4. Tom Brown, I totally agree with you about expectations being a somewhat iffy mechanism to evoke. I'm just thinking that there is a mundane explanation for the phenomenon you describe in that, if there are enough bank reserves such that they don't have any scarcity, then adding more won't make any difference. If there is a tight shortage of bank reserves then relieving that shortage will have a great effect -allowing interest rates to fall and credit growth to resume. It is not to do with information or expectations or anything fancy. It is merely to do with banks needing to obtain enough reserves from each other on a day to day basis to operate the settlement process.

      Delete
    5. stone I'm going to avoid answering your question until I understand a bit more about ITM, but in the mean time here's another one that might interest you: a bit of circular reasoning on Sumner's part?
      http://informationtransfereconomics.blogspot.com/2014/05/adventures-in-circular-reasoning.html

      Also, bad ad hoc vs good ad hoc:
      http://informationtransfereconomics.blogspot.com/2014/05/good-ad-hoc-vs-bad-ad-hoc.html

      Jason proposed that the authors of a paper Sadowski gave me a link to (on expectations) were guilty of "ad hoc in the worst way":
      http://informationtransfereconomics.blogspot.com/2014/05/good-ad-hoc-vs-bad-ad-hoc.html

      Delete
    6. stone: I was unaware of Gibson's paradox but if the "information transfer" picture is correct, then Gibson's paradox would be consistent with a high currency to NGDP ratio making the situation then much like today. In fact the information transfer picture "resolves" Gibson's paradox by saying the quantity theory view holds when the ratio of currency to NGDP is low and the paradoxical result holds when ratio is high.

      http://informationtransfereconomics.blogspot.com/2014/03/the-effects-that-move-interest-rates.html

      Delete
    7. I suppose the crucial point is that the relationship between the amount of monetary base and its effect of interest rates is so non-linear. If there is plenty then interest rates will be stuck on the zero bound (like in Japan) unless interest is paid on excess reserves by the central bank as a floor system (as in UK and USA). If there are insufficient bank reserves, then interest rates will spike up and the effect of further decreases will be an extremely steep and ever steeper increase (as during Volcker tightening in 1980).
      The relative variables are the size of the stock of bank reserves, the amount of payments and the time span between settlements. Extending settlement time will allow a given amount of bank reserves to suffice. Having less payments made (lower NGDP) will mean that the discrepancy between payments in and out for each bank over a given span of time between settlements will be less so less bank reserves will be needed.
      I can't see that anything more theoretical is needed over and above that extremely mundane operational view. None of what we see seems paradoxical to me from the viewpoint of that mundane operational view.

      Delete
    8. Of course there is the added dimension of why the choice is made to sometimes limit the amount of bank reserves made available. That is very much political economy. I think an un-muddled view requires requires an acknowledgement of just how political that decision process is. For me, the clearest view of all of that I've seen was from this 1943 essay by Michal Kalecki
      http://mrzine.monthlyreview.org/2010/kalecki220510.html

      It also entails considerations of currency exchange rates, so conflicts between various interest groups within a country and also between countries. To me the whole Volcker tightening and Great Moderation was about ensuring that the developed world got all of the global output whilst the 1970s stagnation was all about the rest of the world starting to get a share of what the developed world previously had had to itself.
      http://directeconomicdemocracy.wordpress.com/2013/05/09/isnt-a-financialized-economy-the-goose-that-lays-our-golden-eggs/

      Delete
    9. I should have written "1970s stagflation" not "1970s stagnation". -sorry

      Delete
    10. Just to clarify, I'm not saying that understanding of this requires any particular personal political outlook. I'm just saying that it only makes sense when it is appreciated that the mechanism is a politically motivated choice as to when and whether to impose a liquidity constraint.

      It is a bit like how biology is unfathomable if you don't view it in the context of evolution. You don't need to have an opinion as to whether viruses are nice or not, you just have to appreciate that things are as they are in biology due to evolution. Likewise things are like they are in economics due to politics.

      Delete
    11. @stone This "mundane operational view" seems like a just-so story. It doesn't seem to explain anything; it just tells us that's the way things are ... how was this mundane operational view constructed? Where does it come from? It doesn't even tell us how much interest rates would change given a change in NGDP or the base. I have some comments on the specific pieces below, a little out of order. The answers come from the new theory.

      The relative variables are the size of the stock of bank reserves, the amount of payments and the time span between settlements. Extending settlement time will allow a given amount of bank reserves to suffice. Having less payments made (lower NGDP) will mean that the discrepancy between payments in and out for each bank over a given span of time between settlements will be less so less bank reserves will be needed.

      Why are those the relative variables? Assumption? [Answer: the variables are NGDP and the monetary base because NGDP represents aggregate demand and the base represents the aggregate supply and it is at its heart a supply and demand argument.]

      I suppose the crucial point is that the relationship between the amount of monetary base and its effect of interest rates is so non-linear.

      Ok it's nonlinear -- what is the non-linearity? Is it r ~ MB^2 or r ~ log MB? r ~ 1/MB ... there are an infinite number of nonlinear functions. [Answer: log r ~ a log (NGDP/MB) + b]

      If there is plenty then interest rates will be stuck on the zero bound (like in Japan) unless interest is paid on excess reserves by the central bank as a floor system (as in UK and USA). If there are insufficient bank reserves, then interest rates will spike up and the effect of further decreases will be an extremely steep and ever steeper increase (as during Volcker tightening in 1980).

      I assume you mean plenty of base money or insufficient base money relative to the amount/time span between payments as you stated above. How do you model this? Intertemporal optimization? There is also a missing time variable. The words plenty/insufficient should be compared with a dimensionless variable -- MB/NGDP works. But the time between payments -- what other time scale do you use to make that dimensionless? dMB/dt? The time for IOR to pay out? [Answer: MB/NGDP and log MB/log NGDP are all you need.]

      IOR is an ad hoc mechanism and appears to not have a large effect on interest rates (the BoJ did not pay IOR from 2000-2008, but did before and does now ... with no change in the behavior of interest rates). Why does IOR matter? [Answer: IOR does not matter to first order.]

      All of these relationships can be reduced to three functions with the new theory:

      rl = R(NGDP, M0)
      rs = R(NGDP, MB)
      P = P(NGDP, M0)

      And you even get an added bonus of being able to explain the price level and short and long term interest rates! How does this "mundane operational view" explain the price level or the difference between short and long term interest rates?

      Delete
    12. I'm going to add that by "M0" I think Jason means currency in circulation, which is paper notes and coins either in bank vaults or held by the non-bank public: basically MB minus the Fed deposits.

      Jason, regarding IOR's effect on interest rates, when excess reserves (ER) are plentiful, then it does seem to set the overnight rate I think, i.e. FFR ~= IOR rate. I understand the real overnight rate is a little bit less than the IOR rate, and I think this has something to do with non-bank Fed deposit holders, which I don't think get paid IOR: I'm not very clear on this, but *I think* these deposit holders loan out their deposits for a bit less than the IOR rate... if anybody knows the real story there I'd be glad to hear it). So when you talk about "interest rates" I assume you're not talking about the FFR, correct? What do you mean? Like a 10 year T-bond or something? The rest of the yield curve in general?

      Delete
    13. Jason, just now noticed you have both rl and rs? Long and short term interest rates resp?

      Delete
    14. @Jason, thanks for so robustly challenging what seemed to me to be self evident. I'm a bit tied up now (time difference to here in UK) but I am really interested in getting my head around all of this.

      I saw the "non-linear" relationship between monetary base and interest rates as being a bit like how as a tunnel gets narrower, traffic slows -very messy. It is not like water in a tunnel, it is like trucks in a tunnel. Obviously if there was only one bank -no bank reserves would ever be needed for settlement. Obviously if frequency of settlement was infinitely long -no bank reserves would ever be needed.

      I just saw long term interest rates as being like a moving average of short term interest rates (with a time window of a few years).

      Delete
    15. Jason, in terms of IOR, have you examined the case of Sweden's recent (within the past five years) experiment with negative IOR rates? The MMists tout that as a success. I'm pretty sure that program is no longer in effect. I think it was pushed by the now former Riksbank deputy governor Lars Svensson (who's a big critic of the current Riksbank leadership).

      Delete
    16. Tom -- if you look at the difference between log EFFR and log IOR, it is more dramatic. EFFR is the effective FFR. And for the long rates I used rl = 10 year treasury and short rates either the EFFR or the 3-month secondary market rate.

      I will see if I can find a case where IOR seems to have an impact.

      stone -- the truck model sounds like queuing theory ... Related to communications theory and information theory :)

      Delete
    17. Jason, Thanks. Have you ever incorporated an overlapping generations model into your ITM framework?

      Delete
    18. Overlapping generations is an agent-based model that makes specific assumptions about how humans behave. One use of it is to come up with a reason why money works as a medium of exchange. Apparently, in the model, money gets its value because the future generations expect it to have value in the future. In OG models without expectations, the value of money has a tendency to collapse to zero. But expectations allow you to avoid that solution (by assumption).

      Delete
    19. Tom and Stone, here is a graph of the effective fed funds rate alongside the 3-month treasury rate, IOR and the information transfer model fit:

      http://research.stlouisfed.org/fred2/graph/?g=ACo

      Delete
    20. Jason, I am keen to read through your blog more and to try and get my head around your idea. On first impressions though it would gain a lot if the descriptions were more grounded in what the fed and the banks actually do.
      eg one thing that puzzled me is that you thought, "further quantitative easing (QE) would likely not stimulate the economy. However, printing currency could lower long term rates and increase inflation."
      To me that risks getting the causality back to front. The fed supplies currency at demand and takes back excess unwanted currency (there is far more currency during the holiday period and it is taken back afterwards). So more currency is symptomatic of more spending but that is quite different from implying that somehow foisting more currency on the population (how?) would induce more spending.

      eg check out:
      http://www.newyorkfed.org/aboutthefed/fedpoint/fed01.html
      "When the public's demand for cash declines—after the holiday season, for example—banks find they have more cash than they need and they deposit the excess at the Fed. Because banks pay the Fed for cash by having their reserve accounts debited, the level of reserves in the nation's banking system drops when the public's demand for cash rises; similarly, the level rises again when the public's demand for cash subsides and banks ship cash back to the Fed. The Fed offsets variations in the public's demand for cash that could introduce volatility into credit markets by implementing open market operations."

      Delete
    21. Jason - I think the point of the OLG models of money is to show why people might use it as a store of value, rather than as a medium of exchange. I don't think all of them achieve this very well. However, once you have established the benefits of monetary assets as a store of value, you can then ask whether they would form a better exchange medium than anything else available. Why is it more beneficial to use bank acounts for payments than exchange of apples, say?

      Delete
    22. Jason, nice graph. So the the effective FFR is an attempt to determine what's actually being paid for those funds or "cash" as the notes page on the FRED graph says (I'm almost positive they don't actually mean paper notes and coin by "cash"... which drives me nuts... because you can find other places where this is spelled out more clearly on Fed websites).

      Delete
    23. stone, I'm going to go out on a limb and suggest that Jason is probably aware of the causality problem you bring up. I think using currency in circulation rather than MB in some of his formulations was a case where he found a better empirical fit with currency in circulation. So when I read a sentence like that from Jason I don't necessarily think to myself that he's recommending that the Fed somehow force more curreny into circulation, but rather that if more currency were to be in circulation that would be correlated with the effects he describes.

      Actually, he has an even more radical idea for how the US and Japan might "escape" their current predicament:

      http://informationtransfereconomics.blogspot.com/2013/09/exit-through-hyperinflation.html

      But take what I say with a grain of salt, because I'm only a step or two ahead of you in familiarizing myself with Jason's ideas, and I may not be representing him well here!

      Delete
    24. @Nick -- yes, I sort of skipped over that piece of the argument. Expectations allow money to function as a store of value (what I meant by the colloquial "to have value") and that value allows it to function as a medium of exchange. I wasn't making a judgment on the validity of the argument, just that it exists and what it sort of says.

      @stone -- Tom is correct about the empirical motivation. The causality could potentially go from the central bank setting interest rates, allowing the base (and NGDP) to adjust endogenously (I think this is Nick Rowe's version that I look at here). The equation I used above doesn't really say much about causality. The Fed doesn't usually conduct monetary policy via the currency component of the base, so I'd imagine they'd have to do something different -- like convert reserve balances into physical currency, buying assets with currency or printing money and mailing it to people. However it is not entirely out of the realm of possibility that the Treasury could simply print the currency, store it somewhere and it could cause the price level to rise and long term interest rates to fall. As a side note -- the seasonal fluctuations in the price level are commensurate with the seasonal fluctuations of the currency in circulation in the model (well, potenially).

      Delete
    25. @Jason, I think the currency thing is just a small picky point and the rest of what you are saying is interesting and I want to check it out more. But I do still struggle with what you are saying about the currency. If the causality direction is that currency in circulation is merely symptomatic of NGDP, then any attempt to boost NGDP by exchanging currency for other money would merely cause the break down of the correlation you observed up until now. Your equation would need to use something else other than currency as a proxy for consumer demand or whatever.

      Delete
    26. Please excuse my goofy analogy but to me it is a bit like how although sick dogs have dry noses and healthy dogs have wet noses, wetting a sick dog's nose with some water won't make it any less sick.

      Delete
    27. @stone -- the potential for the relationship to break down is the basis of the Lucas critique. The information theory behind the model provides the 'microfoundations' that tell us the relationship won't break down** and that causality goes in either direction. The relevant analogy is an entropic force, not the dog's nose. Imagine a semi-permeable membrane with salt water on either side in equilibrium. Adding water to one side will cause water to move to the other to bring the concentrations back to equilibrium. However, adding salt to one side will cause water to move from the other side to that side. The entropic force works in both directions. Effectively you changed the number of possible states on each side of the membrane, making the final result the more likely state.

      In the case of money, simply printing the money (when the base is small compared to NGDP) will allow new possible of states with a higher NGDP and the economy will randomly walk there. In this model, the equilibrium is a dynamic equilibrium with money moving everywhere in the economy and finding every possible state -- like a smell of bacon cooking filling a house. No mechanism is causing the smell of bacon to fill the house. It fills the house because there are more states with the scent of bacon everywhere in the house than there are states with it near the source. The individual molecules are just doing their own Newtonian thing.

      http://en.wikipedia.org/wiki/Entropic_force

      ** If the relationship does break down, that tells us the theory is wrong.

      Delete
    28. Jason, I totally agree that "spending capacity in the hands of people wishing to spend it on consumption" could be much like the water in your semi permeable membrane analogy and it could feed through to adjust the price level. The real life example that springs to mind is how Deutsche Marks were introduced in 1948 with everyone just being given DM60 to start things off. That was a flat outright gift (theft ?!?), not an exchange for any other form of spending capacity. It was not monetary policy. It was redistributionist fiscal policy. That got aggregate demand back on track and started off the post war economic miracle.

      What I fear could break down in a Lucas critique type of way is the way that currency in circulation stands in as a proxy for "spending capacity in the hands of people wishing to spend it on consumption". You empirically found that, over the historical period you looked at, currency in circulation has been a good proxy but there seems no reason to be confident that it would continue to do so if efforts were made to induce exchange of currency for other forms of money as a way to boost NGDP. That could be like wetting the sick dog's nose.

      Delete
    29. It could be like the dog's nose. In that case the underlying 'microfoundations' are wrong and the theory is wrong. If the theory is correct, then it doesn't matter how the money filters out (the theory is agnostic about the mechanism).

      Delete
    30. I guess the issue is what causes money to filter out. Doing an asset swap, exchanging currency for some other store of value might just result in the currency being held as a store of value and so not filtering out at all.

      Delete
  4. Nice post on a puzzling idea.

    If it is possible for a subject matter to be firmly framed in convolution, this one is a good candidate.

    ReplyDelete
    Replies
    1. Yes, I think that sums it up pretty well.

      Delete