Storytelling in Kennywood Park

Monetary policy works in practice, not in theory.

Trevor Chow

Can monetary policy control the path of nominal GDP? I have no idea anymore and am wildly confused, so this is an attempt to tell a few just-so stories and see how they land.

Disclaimer: This is wildly wildly wildly speculative, especially story E.

Story A

  1. The Fed is the monopoly supplier of base money
  2. The Fed can set the ratio of supplies of base money to T-bills by open market operations
  3. The value of base money is defined as its the purchasing power/exchange rate
  4. The Fed can set the value of base money in terms of T-bills i.e. 1/price of T-bills
  5. The Fed can set the price of T-bills
  6. The Fed can set the nominal interest rate of T-bills
  7. NGDP depends on the nominal interest rate of T-bills
  8. The Fed can set nGDP

This is an Old Keynesian story.

But Trevor, the Old Keynesians complained about monetary policy precisely because of this story. It falls apart in a liquidity trap. When the spread between base money and risk-free debt goes to zero, money strictly dominates due to the convenience yield. And it’s kind of bonkers to assume that nGDP depends meaningfully on the nominal interest rate of T-bills. This seems more like a relic of having one interest rate in IS-LM models than a realistic depiction of the world.

Story B

  1. The Fed is the monopoly supplier of base money
  2. The value of base money is defined as its the purchasing power/exchange rate
  3. The Fed can set the value of base money in terms of nominal output i.e. 1/nGDP
  4. The Fed can set nGDP

This is an Old Monetarist story and it avoids the Old Keynesian trap.

But Trevor, this is basically definitional. You’ve defined some ratio and claimed that because you can control one side of the ratio, you can control the ratio. But the other side could just adjust accordingly, and you can’t tell unless you give some causal mechanism. Notice which line you skipped from story A: the line about how open market operations work! You can’t just buy chunks of nGDP.

Story C

  1. The Fed is the monopoly supplier of base money
  2. The value of base money is defined as its the purchasing power/exchange rate
  3. The Fed can set the value of base money in terms of assets i.e. 1/price of assets
  4. The Fed can set the price of assets
  5. NGDP depends on the price of assets
  6. The Fed can set nGDP

This is a Market Monetarist story and it adds some details to the Old Monetarist hot potato story. Empirically, the entire literature on the Quantity Theory (think McCandless and Weber etc.) suggests that central banks can control nominal quantities.

But Trevor, central banks don’t actually do this. I have no doubt that if a central bank went about buying everything it could get its hands on by printing money, their prices would get bid up. But the fact that a massive increase in base money would raise nominal quantities does not mean central banks can control nGDP on a meaningful level under its ordinary practices. The ability to wield a machete is no evidence that one can utilise a scalpel.

Story D

  1. The Fed can set the discount rate, the IOR rate and the ONRRP rate
  2. The effective federal funds rate will be between the IOR and ONRRP rate by arbitrage
  3. The Fed can set the effective federal funds rate
  4. Commercial bank lending depends on the cost of financing
  5. One cost of financing is the effective federal funds rate
  6. Broad money creation depends on commercial bank lending
  7. NGDP depends on the amount of broad money in the economy
  8. The Fed can set nGDP

This is a New Keynesian story with slightly more institutional structure than usual. It’s calibrated to monetary policy as it currently stands in the US (hence the leaky floor). For empirics, refer to the appendix.

But Trevor, once you include the institutional details you can see why the causal mechanism is much weaker. Arbitrage of the federal funds rate is imperfect. Commercial banks don’t just rely on that for financing, and their lending decisions have a lot to do with profitable opportunities too. Bank lending doesn’t necessarily create broad money - if people decide to pay off debts, that money vanishes as quickly as it is created. Hence the amount of broad money in the economy is deeply endogenous at any particular point. It may even be entirely endogenous!

Story E

  1. Money is defined as a store of value, a unit of account and a medium of exchange
  2. None of these imply that control of base money allows the control of nGDP
  3. Goodhart’s Law: Any observed statistical regularity will collapse if used for control purposes
  4. This is just vanilla Lucas critique commentary
  5. Try to name a Fed-controlled monetary aggregate which correlates strongly with nGDP
  6. It’s hard because the Fed controls base money, nGDP is affected by broad money and there isn’t a stable exploitable relationship between the two

Why might this be the case?

  1. There are many stores of value available to the public nowadays
  2. Total nominal spending depends on total wealth across all stores of value
  3. The Fed’s control over one such assets is basically irrelevant
  4. The causal channel cannot be money as a store of value
  5. The role of the dollar as a unit of account is social convention
  6. The Fed’s OMOs do not affect the unit of account
  7. At best they affect the media of exchange’s relation to the unit of account

Why does the medium of exchange defy control?

  1. There are many media of exchange available to the public nowadays
  2. Suppose there is a stable relationship between nGDP and the quantity of 10 dollar bills
  3. If the Fed tried OMOs of 5 dollar bills in exchange for 10 dollar bills
  4. It seems unlikely nGDP rises just because the quantity of 10 dollar bills rises
  5. Instead, the velocity of 10 dollar bills would just fall and people would use other forms of payment
  6. Asset swaps of nigh-identical assets don’t do anything, including for T-bills and base money (especially given IOR)

Why might portfolio rebalancing not work?

  1. Traditionally, OMOs disturb the asset portfolios of investors and they want to correct for this
  2. Under the portfolio balancing effect, they buy up bonds till the higher price and lower yield re-equilibrates
  3. But when the Fed buys bonds via OMOs, these bonds don’t disappear and instead land on the Fed’s balance sheet, wich returns revenues to the Treasury
  4. The bond risk simply becomes a tax risk due to a change in the government’s balance sheet
  5. If private investors see through this, they will neutralise any changes in government debt holdings

How do money and credit adjust?

  1. Hence the Fed cannot control broad money and nGDP due to this Wallace neutrality
  2. This is a specific example of a deeper claim that moneyness is a spectrum
  3. Money is a means of final settlement, while credit is a promise to pay money i.e. to delay settlement
  4. But what counts as “final settlement” depends on where you stand, and so does what is money
  5. While money at any level might be scarce, credit is elastic i.e. monetary policy’s power is by affecting credit creation

This is a farrago of orthodox and heterodox ideas which I am at least sympathetic to, even if I remain skeptical.

But Trevor, this is wildly counterfactual. Money is by all empirical measures non-neutral in the short run and neutral in the long run - and monetary policy can affect credit creation. In practice, Wallace neutrality does not seem true. For one, people do not “see through” government balance sheets a la Modigliani-Miller. Base money isn’t just held for reasons of pecuniary returns, which is one of the conditions for Wallace neutrality. And if OMOs or QE allow some sort of liquidity or risk transformation for heterogenous consumers, it may change the nominal and real wealth and debt.

Let me sort out what I believe out of this. Firstly, money can be non-neutral even when all of this is true, if you believe that its only use is to coordinate people’s expectations. In other words, it picks the good sunspot for us and works because everyone thinks everyone else acts as though it works. So the non-neutrality of money is observationally equivalent for this story and the conventional one.

Secondly, Wallace neutrality is indeed probably nonsense in real life. But what do asset swaps do? Insofar as they occur at market price, I don’t thik they really change bank capital. I guess ordinary OMOs make lending a bit cheaper or bind them to the IOR/ONRRP rate, though the idea that this one federal funds rate is mechanically significant for credit creation in the broader economy is somewhat suspicious. In part, this is because banks finance much of their lending nowadays via borrowing from the public.

I think a more plausible answer has to do with risk premiums and transformation (think Araujo, Schommer and Woodford 2015). One way to break Wallace neutrality is if investors who would ordinarily neutralise central bank asset swaps are financial constrained and cannot do so. But that means you’re essentially forcing people to bear or offload risks they otherwise wouldn’t want to, and as a corollary, you’re shifting the risk which the Fed takes. This may be a good thing - in a closed economy, net inside debt is always zero and it’s always the distribution that matters anyways. But the point is that this is only a possible and good thing conditionally, depending on how this risk is being shifted around and across whom.

One exception to all of this: helicopter drops. The reason helicopter drops work is because they are designed to commit governments to not raising taxes to cover the money, so the Ricardian motive implicit in the Wallace neutrality story disappears. In that way, credit discreation might not aim counteract money creation. This is a very Fiscal Theory story, because it alludes to the fact that the ultimate effects depend not on the central bank liabilities issued but on the government assets backing those liabilities. Incidentally, this is why such a model doesn’t rule out hyperinflation, which is a fiscal phenomenon. In that sense, inflation is a fugazi. It’s fairy dust. It prevails at some low and stable level because everyone thinks it will, unless it’s kicked out of gear by some fiscal regime shift.

So yes, I did say story E was going to be hetero-orthodox!


Empirically, the debate around story D is the bulk of the canonical literature. To begin with, we have the traditional identification of federal funds rate shocks: vector autoregressions (Christiano Eichenbaum and Evans 1999; Bernanke, Boivin and Eliasz 2005; Boivin, Kiley and Mishkin 2010; Amir and Uhlig 2015), high frequency identification (Faust, Swanson and Wright 2004; Gurkaynak, Sack and Swanson 2005; Barakchian and Crow 2013; Gertler and Karadi 2015; Swanson 2017) as well as the classic Romer-Romer narrative method (1989, 1994, 2004 and 2013).

And we have plenty of mechanisms for this: the traditional interest rate channel on consumption (Wong 2018), on investment (Guiso et al. 2002; Cloyne et al. 2018), the lending channel (Bernanke and Blinder 1992; Kashyap, Stein and Wilcox 1993; Bernanke, Gertler and Gilchrist 1999; Kashyap and Stein 2000), the exchange rate channel (Svensson 2003) and plenty else from the more modern HANK literature (Moll 2019). All of these suggest that expansionary federal funds rate changes are expansionary.

On the other hand, I could just as easily give papers contradicting what I’ve said so far: that a federal funds shock goes the opposite direction (Nakamura and Steinsson 2018), that consumption isn’t terribly interest sensitive (Elmendorf 1996; modern estimates of credit-constrained or habit-following Euler equations), that investment isn’t either (Caballero 1999; Sharpe and Suarez 2014; Lane and Rosewall 2015) and that the bank lending channel is dubious too (Oliner and Rudebusch 1996). And while I am inclined to believe the overall literature does side with the usual view, the fact remains that macroeconomic identification is difficult and confusing (see: the price puzzle or the well-known issues with Romer-Romer) so our priors should be pretty weak.