Sunday, September 20, 2015

The ZIRP Blues

Here's the time series of the fed funds rate and inflation rate in the United States, from the time Paul Volcker became Fed Chair:
Suppose an alien with a high IQ lands in my back yard. I show her this picture, and explain that the central bank moves the fed funds rate up and down so as to control inflation. Ms. Alien points out that the fed funds rate and inflation were in the neighborhood of 10% in August 1979. Now, 36 years later, the fed funds rate and the inflation rate are close to zero. So, says Ms. Alien, it looks like the central bank spent 36 years fighting the inflation rate down to zero.

Ms. Alien would be surprised to learn that most people are not happy with the current state of affairs. There are always exceptions, of course - in this case, John Cochrane. But popular views on current U.S. monetary policy fall basically into two camps:

1. Phillips curve A: These people think inflation is too low. But eventually the Phillips curve will re-assert itself, and inflation will rise of its own accord. When that happens, we can worry about liftoff - an increase in interest rates to hold inflation down.
2. Phillips curve B: These people also think inflation is too low, that, eventually, the Phillips curve will re-assert itself, and that inflation will rise of its own accord. But a Phillips curve B type thinks that we need to get ahead of the game. Milton Friedman told us that there are "long and variable lags" associated with monetary policy. If we wait too long, then monetary policy will be scrambling to keep up with higher inflation, and interest rates will need to climb at a high rate, at the expense of real economic activity.

The Phillips curve A group includes Summers, Stiglitz, and Krugman, who states that we should "wait until you see the whites of inflation’s eyes." Members of the Phillips curve A and B camps have to somehow come to grips with the Phillips curve we see in the recent data, which looks like this:
The line joins the points in the scatter plot in temporal sequence, roughly from right to left. Krugman's point in his piece is that the natural rate of unemployment (NAIRU) has been receding as we get closer to it. In this view, we're supposed to have faith that the Phillips curve looks like this:

An alternative to Phillips A/B is the neo-Fisherian view. As John Cochrane says:
But if a 0% interest rate peg is stable, then so is a 1% interest rate peg. It follows that raising rates 1% will eventually raise inflation 1%. New Keynesian models echo this consequence of experience. And then the Fed will congratulate itself for foreseeing the inflation that, in fact, it caused.
Cochrane's saying that central bankers have to come to terms with the Fisher effect. If the short-term nominal interest rate is low for a long time, we should not be surprised that the inflation rate is low. And John is quite happy with low inflation. While the Phillips curve A and B camps fight it out over how to get inflation up, and sing the ZIRP (zero interest rate policy) blues, he's hoping they never figure it out.

There's a more subtle idea in the quote from Cochrane above, which is that a neo-Fisherian could find common cause with the Phillips curve B camp. They could all agree to liftoff, the inflation rate could rise due to the Fisher effect, and the central bank "will congratulate itself for foreseeing the inflation that, in fact, it caused."

If you're wondering what central bankers are thinking, a nice summary of conventional views is in a speech by Andy Haldane, Chief Economist at the Bank of England. It's a long speech, by U.S. central banker standards, but certainly thorough. Much of the speech focuses on the "problem" of the zero lower bound (ZLB). In most of the monetary models we write down, and in the traditional thinking of central bankers, zero is a lower bound on central bank's policy interest rate. The ZLB is thought to be a problem as, once the central bank reaches it, its policy options are limited. If one takes this seriously, there are two responses: (i) stay away from the ZLB; (ii) get more creative about policy options at the ZLB.

How do we stay away from the ZLB? Haldane tells us why we're now seeing ZLB policies:
... by lowering steady-state levels of nominal interest rates, lower inflation targets ... increased the probability of the ZLB constraint binding.
He's saying that low inflation targets, i.e. average rates of inflation that are low, imply lower nominal interest rates. So,
... one option for loosening [the ZLB] constraint would simply be to revise upwards inflation targets. For example, raising inflation targets to 4% from 2% would provide 2 extra percentage points of interest rate wiggle room.
So this is entirely consistent with John Cochrane and the neo-Fisherians. If the central bank's inflation target is higher by two percentage points, then the nominal interest rate must on average be higher by two percentage points, and the chances that monetary policy will take us to the ZLB should be much smaller.

But, Haldane is certainly not a neo-Fisherian. He's more in the Phillips curve A camp, as this is his policy recommendation:
In my view, the balance of risks to UK growth, and to UK inflation at the two-year horizon, is skewed squarely and significantly to the downside.

Against that backdrop, the case for raising UK interest rates in the current environment is, for me, some way from being made. One reason not to do so is that, were the downside risks I have discussed to materialise, there could be a need to loosen rather than tighten the monetary reins as a next step to support UK growth and return inflation to target.
Haldane makes it clear that he thinks the way to "return inflation to target," i.e. 2%, is not to let the central bank's interest rate target go up. And, as I wrote here, it's not as if the UK data will make you a believer in the Phillips curve. Here's the policy problem the Bank of England faces:
The policy interest rate target is currently at 0.5% in the UK but, as in the U.S., the inflation target is at 2% and actual inflation is hovering around 0%.

Haldane discusses ways in which central banks can get creative when confronted with the ZLB. The options that have been discussed (and in some cases implemented by some central banks) are:

1. Quantitative Easing: The idea here is that, at the ZLB, purchases by the central bank of short-term government debt are essentially irrelevant, as there is no fundamental difference between short-term government debt and reserves at the ZLB. But, the central bank could purchase long-maturity government debt or other assets at the ZLB. Perhaps that does something? Post-Great Recession, the U.S. of course acquired a large portfolio of long-maturity Treasury securities and mortgage-backed securities, and maintains the nominal value of that portfolio of assets through a reinvestment policy that is still in place. Whatever the effects of U.S. QE programs, it's an inescapable reality that inflation is close to zero. But, even larger asset purchases were carried out by the Swiss National Bank, and the Bank of Japan. Here's what's happened in Switzerland:
In this case, both the policy rate and the inflation rate are well below zero. The Swiss National Bank has a goal of price stability, which it defines as less than 2% inflation. I'm not sure if they are OK with an inflation rate less than -1%.

The Bank of Japan began a program of "qualitative and quantitative monetary easing" in April of 2013. Here's the overnight interest rate and inflation rate time series for Japan:
I've included the whole 20-year period over which Japan's overnight interest rate was below 1%. Japan is, as you know, our stock example of what ZIRP produces. But what of the effects of the Bank of Japan's recent QE experiment? Don't be deceived by that burst of inflation in 2014. In April 2014, the consumption tax in Japan went up from 5% to 8%, and that feeds directly into the CPI - the prices in the index are measured after-tax. If we look at the CPI levels since the beginning of the QE program in April 2013, you can see that more clearly:
So, from April 2013 to July 2015, the CPI increased about 4%. If 3 percentage points of that is simply due to the consumption tax increase, then we're left with less than 1/2% per year in inflation since the QE program began. The Bank of Japan's inflation target is 2%, which it is missing by a wide margin on the low side, in spite of an increase in the monetary base in Japan that looks like this:
You can't blame John Cochrane for stating the following, with respect to the U.S.:
Even the strongest empirical research argues that QE bond buying announcements lowered rates on specific issues a few tenths of a percentage point for a few months. But that's not much effect for your $3 trillion. And it does not verify the much larger reach-for-yield, bubble-inducing, or other effects.

An acid test: If QE is indeed so powerful, why did the Fed not just announce, say, a 1% 10 year rate, and buy whatever it takes to get that price? A likely answer: they feared that they would have been steamrolled with demand. And then, the markets would have found out that the Fed can’t really control 10 year rates. Successful soothsayers stay in the shadows of doubt.

I've written down a model of QE, in which swaps of short-maturity assets for long-maturity assets by the central bank can have real effects. Basically, this increases the stock of effective collateral in the economy, relaxes collateral constraints, and increases the real interest rate. It's a good thing. But, if the nominal interest rate is pegged at zero, this will lower the inflation rate.

2. Lower the lower bound: If the ZLB is a problem, possibly we can make the problem go away by relaxing the bound. In models we write down, the zero lower bound arises because it is costless to hold currency which, given current technological constraints, cannot bear interest. When the central bank has excess reserves outstanding in the financial system, if an attempt were made to charge financial institutions for the privilege of holding reserves with the central bank, these institutions would opt to hold currency instead - in some of our models. But, in the real world it is not costless to hold currency. Making interbank transactions using currency is impractical, as millions of dollars in currency takes up a lot of space, and because real resources would have to be expended in preventing theft. This implies that market nominal interest rates can be negative and, indeed, some jurisdictions have opted for negative interest rates on reserve balances held at the central bank. One of those, as you can see in the chart above, is Switzerland, where the inflation rate is now below 1%. Another is the Euro area:
European overnight interest rates have not gone as low as in Switzerland, nor is the inflation rate as low, but it's a similar picture - not much inflation.

Relaxing the lower bound meets with a difficulty similar to that for QE - in the long run, this just serves to make inflation lower. To see this, consider a very crude monetary model - cash-in-advance. There's a representative consumer who gets utility u(c) from consumption goods c, and suffers disutility v(n) from supplying n units of labor, which produces n units of consumption goods. Consumption goods must be purchased with cash. There are also one period bonds, which sell at a price q at the beginning of the period, and pay off one unit of cash next period. Cash and bonds are held across periods, and fraction t of cash holdings held between periods is stolen. Suppose for simplicity that thieves steal money and burn it. To make things easy, look at an equilibrium in which the money growth rate is a constant, i. Letting B denote the discount factor, in equilibrium the price of the bond is given by

(1) q = B/(1+i)

That's just the Fisher relation. There are no liquidity effects in this model, and in equilibrium the nominal interest rate is (roughly) given by

(2) R = p + i,

where p = 1/B -1 is the real interest rate. In equilibrium c = n, i.e. all output is consumed, and c is determined by

(3) v'(c) = [B(1-t)u'(c)]/(1+i)

What's the lower bound on the nominal interest rate. It's R* = - t, that is, it's determined by the cost of holding cash. And, if the nominal interest rate is at its lower bound, R*, then the inflation rate is

(4) i* = - p - t,

so lowering the lower bound only serves to decrease the inflation rate. You can add bells and whistles - reasons for the real interest rate to be low, endogenous theft of currency, short run non-neutralities of money, or whatever, and I think the basic idea will go through.

Another suggested approach to increasing the inflation rate, given ZIRP, is:

3. Helicopter Drops: The "helicopter drop" was a thought experiment in Milton Friedman's "Optimum Quantity of Money" essay. In the thought experiment, Friedman asks you to consider what would happen if the government sent out helicopters to spew money across the countryside. People would pick up the money, spend it, and prices would go up, etc. Surely, if inflation is perceived to be too low, and we're at a loss as to how to increase it, we should be thinking about this, the argument goes. Can't the government just send people checks and make inflation go up?

Paul Krugman has a suggestion along these lines, for Japan, though what he's suggesting is not Friedman's helicopter transfers (which increase the government budget deficit), but increases in spending on goods and services, financed by printing money:

What’s remarkable about this record of dubious achievement is that there actually is a surefire way to fight deflation: When you print money, don’t use it to buy assets; use it to buy stuff. That is, run budget deficits paid for with the printing press.
Actually, that's exactly what has been going on in Japan. The Japanese government has been running a deficit, the quantity of government debt outstanding is very large (in excess of 200% of GDP) and, as we can see in the chart above, the monetary base is growing at a very high rate. That's what printing money amounts to. But, the central bank can only control the total quantity of outside money in existence, not its composition. How outside money is split between currency and reserves is determined by the banks who hold the reserves and the private firms and consumers who hold the currency. The central bank can do all the money printing it wants, but if the new money sits as reserves, as appears to be happening, it's not going to have the effect that Krugman wants.

Increasing interest rates is hard for central bankers. A decrease in rates rarely produces any flack, but central banks have few supporters when they talk about rate increases. Media pieces like this one in the NYT and this one in the Economist propagate the idea that interest rate increases are fraught with peril. One example people like to use is tightening by the Swedish Riksbank in 2010-2011. Here's the relevant chart:
The tightening that occurred was an increase of 1.75 percentage points in the Riksbank's target interest rate, in quarter-point steps, from July 2010 to August 2011. In the realm of central bank tightening phases, this isn't a big deal. Compare it to the previous tightening phase in Sweden, or the 4.25 percentage point increase that occurred in the U.S. over the 2004-2006 period. But, the Riksbank caught hell from Lars Svennson as a result. The Riksbank seems to have more or less followed Lars's advice since, but as you can see it is now keeping company with other central banks, with a negative policy rate, and inflation close to zero - two percentage points south of its target.

What are we to conclude? Central banks are not forced to adopt ZIRP, or NIRP (negative interest rate policy). ZIRP and NIRP are choices. And, after 20 years of Japanese experience with ZIRP, and/or familiarity with standard monetary models, we should not be surprised when ZIRP produces low inflation. We should also not be surprised that NIRP produces even lower inflation. Further, experience with QE should make us question whether large scale asset purchases, given ZIRP or NIRP, will produce higher inflation. The world's central bankers may eventually try all other possible options and be left with only two: (i) Embrace ZIRP, but recognize that this means a decrease in the inflation target - zero might be about right; (ii) Come to terms with the possibility that the Phillips curve will never re-assert itself, and there is no way to achieve a 2% inflation target other than having a nominal interest rate target well above zero, on average. To get there from here may require "tightening" in the face of low inflation.


  1. Everything is clear to me but I have a problem with your assertion about helicopter drops. I have worked with some models with explicit monetary and fiscal policy rules, and as far as I know, interest rate pegs are stable if fiscal policy is passive, in the terminology of Leeper, i.e. the government sets the budget deficits and does not respond to debt increases by increasing taxes. This is also the fiscal theory of the price level. In this setting if you increase consolidated government debt and spend more (increasing the present value of budget deficits) you increase inflation, because the present value of debt falls. So under a stable interest rate peg, helicopter drops should increase inflation.

    1. Well, your model sounds interesting, but in Japan they are doing the experiment and it's not producing inflation. In terms of what central banks actually do, a helicopter drop would be: (1) fiscal authority issues debt; (2) central bank issues reserves to buy the debt. It should not make any difference whether the central bank actually buys the debt with currency - people will just deposit the currency in the bank and it will ultimately be held as reserves.

  2. Stephen, help me better understand the implications of neo-fisherism (NF) for the current situation facing the Fed. On one hand, the NF argument is that by raising interest rates the Fed will also raise inflation. On the other hand, it seems the underpinnings of the NF model is some kind of fiscal theory of the price level which implies (and you have recently endorsed in response to a Krugman post on the shortage of safe assets) permanently increasing the amount of government debt will raise inflation. Do these choices correctly characterize the implications of NF for the present situation?

    1. This should help:

      There's no special "NF model." The reasoning comes from most every conventional (and serious) monetary model. Nothing relies on a fiscal theory of the price level argument.

  3. Presumably, serious monetary theorists like Woodford have issues with NF. How would you respond to their views?

    1. Mike understands that his models have NF properties. His response is to make up another model:

  4. Regarding Japan; what's going on is still just QE, albeit on an abit larger scale. The narrative is still that the public debt will eventually be repaid and QE thus reversed; once they hit their inflation target they would stop QE.

    Helicopter money would be a transfer of money to the treasury by the central bank, and not an open market operation; or equivalently cancellation of the treasury's debt that was bought under QE. This would create "true" outside money that is a net asset for all agents in the economy, that is both the private and public sector.
    It is hard not to believe that this eventually produces inflation, since otherwise Japan could just set all taxes to 0% and print the whole of the yearly budget. I realize helicopter money in this way is illegal in most developed countries and central banks cannot perform it.

    1. "... the public debt will eventually be repaid and QE thus reversed; once they hit their inflation target they would stop QE."

      I think we have no idea what they intend to do. Whether this is permanent or temporary remains an open question. And, if they maintain ZIRP, it's permanent, as they'll never hit the 2% inflation target and sustain it.

      "...a transfer of money to the treasury by the central bank..."

      Same thing. Whether the treasury issues the debt to finance spending and the central bank buys it, or the central bank gives the treasury money (currency) to spend makes no difference. People will just deposit the currency in the bank, and it will end up in reserve accounts.

    2. It is not really the same thing. If the transfer occurs in such a way that it's not recorded as a loss on the CB balance sheet, so as not to imply an equivalent loss of remittances to the treasury in the future, this would be helicopter money or monetary finance. It creates net assets for the economic agents in the economy and it also cannot be reversed.

      QE is reversed as soon as the bond is repaid, is not permanent and doesn't create net wealth.

      Kuroda said they don't plan to do monetary finance and that the government plans to eventually balance the budget:

    3. " If the transfer occurs in such a way that it's not recorded as a loss on the CB balance sheet, so as not to imply an equivalent loss of remittances to the treasury in the future, this would be helicopter money or monetary finance. It creates net assets for the economic agents in the economy and it also cannot be reversed."

      How the treasury and the central bank do the accounting makes no difference. We could say that the central bank buys a perpetual government bond (a consol, say) from the Treasury with currency. So, the bond is never retired - the money injection is permanent. Same difference. If the ZLB is permanent, the currency will be deposited in banks and held as reserves.

  5. Had a related question along this line and so will tag along on this thread...

    So the Fed buys every dollar of government debt and declares the government absolved of repayment. No inflation? The Fed then moves on to corporate / personal debts and similarly absolves the debtor of repayment. We'll be a debt-free nation with many trillions of new currency parked in reserves and no hyperinflation. Why not do that?

    1. No one is being absolved. This is just the central bank converting government debt into reserve balances.

  6. It depends - should the central bank commit to permanent monetary injections, the debtor is effectively absolved (and consequently the required level of monetary injection to hit inflation targets would be much smaller). Such permanence has been noted as a requirement to maintain the effectiveness of monetary policy while at the ZLB.

    Regardless of the likelihood of such a policy, I suspect that it could result in 0% nominal short term rates and unbounded inflation. How does that square with conventional and serious monetary models?

    1. "...the debtor is effectively absolved..."

      What debtor?

      "Such permanence has been noted as a requirement to maintain the effectiveness of monetary policy while at the ZLB."

      Noted by who? The idea is that monetary policy is not "effective" at the ZLB, if effective means making inflation go up.

      "How does that square with conventional and serious monetary models?"

      I guess we would say it does not square.