Section II uses autoregressions with error correction terms to document that the day-to-day
variation in rates for maturities of a month or more has little or nothing to do with the Fed’s target rate.
This is consistent with a Fed that has little control of rates, but we shall see that it is also consistent with a
powerful Fed whose predictable actions with respect to TF are built in advance into interest rates.
Thanks to John Cochrane for the pointer. Note that TF is the Fed funds rate. My remarks:
1. I do not trust any time series regressions.
2. I think it is interesting that a framework in which the Fed has no influence is observationally close to a framework in which the Fed controls interest rates but market participants anticipate the Fed’s actions very well.
3. We may be in an environment today in which long-term rates over-react to short-term changes in the Fed funds rate.
4. Still, I take the view that the Fed is not big enough in the financial markets to have much durable influence on market interest rates.
“I do not trust any time series regressions.”
Why not?
At high frequencies (quarterly data, monthly data, daily data), you get noise. At low frequencies (annual data or longer) you get spurious correlation
I humbly request that our host expound on this claim. It’s not clear to me what is meant.
And if the Fed is really impotent and there should be no indicators of ‘reaction’ at all to anything they say or do, then what is an ‘over-reaction’?
I think that there is a tendency for market reactions to events to be many times their possible fundamental impact. Shiller takes that view. It could be wrong, but it seems right. What that means is that short-term market reactions are likely to be largely reversed over time (although this holds only if other things are equal, which they are not. If you try to bet on it, the next shock could go against you).
Anything that investors think matters will matter in the very short run, and that can include the Fed, even if what the Fed does has no long-run impact.
Maybe we are seeing both.
Previously banks had little incentive to hold excess reserves and as a consequence the fed funds market was an extremely thin market where it was easy for the fed to manipulate the fed funds rate.
But now banks hold massive excess reserves and earn interest on them from the Fed. So it may not be as easy for the fed to raise or lower the fed funds rate as it once was. At a minimum they will probably have to change the rate they pay on bank reserves. But when the fed does actually decide to raise rates it may turn out to be an interesting experiment. Maybe it will be similar to the early 1980s when they discovered it was much harder to stop home building without Reg Q than they had previously believed.
Of your points, I’m on board wholeheartedly for (1), somewhat for (2), and maybe for (3).
But I have to take extreme, extreme exception to (4): “the Fed is not big enough in the financial markets to have much durable influence on market interest rates”. I just don’t understand what underlying model of the world could support this claim.
As I see it, there are three possible arguments you could make:
(A) The Fed has limited power to control nominal interest rates even in short-term money markets, because other financial institutions will offset its actions.
(B) Financial market segmentation is so extreme that even if the Fed successfully controls rates in its corner of the money markets, this has little impact on other rates.
(C) Prices and wages are very flexible, so any attempt by the Fed to move real interest rates away from their natural rates will provoke a strong inflationary or deflationary reaction.
When dismissing the effectiveness of monetary policy, I think it’s pretty clear that one has to argue A, B, or C. In fact, if you do allow the Fed to control the money market interest rate (contra A) and you have frictionless financial markets (contra B) together with nominal rigidities (contra C), then it’s easy to prove that it can have a large impact. See any New Keynesian model.
Generally I understand people like Black and Fama as arguing some mix of A and B, mainly A. But the weird part is that A is the most easily rejected: for one thing, you can just look at a graph of the effective federal funds rate and see that (especially in recent years) it’s moved closely — in artificial-looking steps that are surely not the recent of some fundamental market rate process — with the FOMC’s target rate. You can then see that this rate and its near-term trajectory is transmitted almost one-for-one to short-term Treasuries, CD rates, etc., and that there is a heavy influence on mortgage rates, corporate bond rates, and so on as well. The extreme market segmentation needed for B clearly doesn’t seem to be there.
At some level the wrongness of A is even more basic. The Fed operates a ledger for reserves; nominal money market rates give the rate of exchange between an entry on this ledger today and an entry on this ledger tomorrow. Why *shouldn’t* the ledger operator be able to control this rate of exchange? There are a million ways that it can do so – it has a lot of power!
The traditional way is to pay zero on reserves and then to manipulate the scarcity of reserves, which are desired for various purposes (including to fulfill legal requirements), in order to achieve the desired interest rate as a scarcity premium. But you can also just saturate the market with reserves and then pay interest on them, which is the Fed’s current plan going forward. And so on… the bottom line is that if it wants to do so, a ledger operator can easily control the intertemporal rate of exchange on its ledger. I just don’t see how this is supposed to be controversial.
(A comparison with conventional banks: a private bank too operates a ledger of sorts, and can control the interest rates for positive and negative balances on that ledger. But if the bank has deposits redeemable in terms of dollars, this power is very limited, because if the bank steps away very far from market rates it will either provoke mass deposits or mass redemptions. By contrast, the Fed’s deposits — base money — aren’t redeemable in terms of anything, so the Fed has no such constraint.)
The only remaining option is C, to argue against significant nominal rigidity. This is not quite as easy to argue using the data, but there are still a lot of clear points in favor of nominal rigidity. For instance, McKay, Nakamura, and Steinsson find that shocks to nominal bond rates on Fed announcement days translate almost one-for-one into real bond rates, with nominal rigidity apparently preventing much response in expected inflation. Mussa famously found (as a redo of his analysis in Europe today would confirm even more strongly) that real exchange rates are vastly more volatile under fixed nominal exchange rates than under floating rates. And so on… the evidence for stickiness is really, really strong.
(And even if C is a possibility, as long as inflation stays within reasonable bounds, we might as well act as if it’s not true: if prices are so flexible that monetary policy doesn’t matter much for real outcomes, then there’s no harm in getting the policy wrong either way, while if prices are quite sticky, then there’s massive harm.)
To be honest, I find statements about the Fed’s ineffectiveness quite painful – to me, the case seems so watertight on both theoretical and empirical grounds that I’m really, really mystified as to what the holdup is supposed to be.
Matt I couldn’t agree more.
Your point about full control of the ledger for reserves is what I always turn to when I read any variation of “the Fed is impotent hypotheisis.” Does anyone truely believe that if the Fed set the interest rate on reserves to 5% tomorrow that financial markets whould shrug and the economy would just continue to plug along at a 3% NGDP growth rate?