Recessions are predictions . . .
. . . of bad policy. That’s what Michael Woodford thinks, and I agree. Not all recessions. In 2001 economists didn’t even see a recession coming until about September, and the recession was over by November. I’m talking about severe recessions, those where there is the feeling of going over the crest in a roller coaster, then that “uh-oh moment,” followed by a sickening plunge. Like 1920-21, 1929-30, 1937-38, 1981-82, 2008-09. Can policy address the problem once it has occurred? Yes and no. Technically it can, but it is very unlikely to work in practice, precisely because it is very unlikely to be tried.
The markets are pretty smart, and when they perceive one of those uh-oh moments, they are usually correct. I mean all markets, BTW, not just the stock market, which often sees ghosts that aren’t there. If stocks, bonds, commercial RE and commodities all signal a big drop in NGDP, it’s almost certainly on the way. We’ve all seen the NGDP “downshift” graph that Beckworth, DeLong, Yglesias, Woolsey, Hendrickson, et al, keep showing. The one with no trend reversion. That’s what markets gradually realized in late 2008.
In a strange way this relates to Paul Krugman’s liquidity trap argument. His strongest argument is that conventional monetary stimulus is ineffective at the zero bound. Since conventional monetary stimulus is cutting nominal short term rates, no one can really contest this part of Krugman’s argument. His more controversial argument is that while there are some unconventional things that a central bank could do, they are not likely to do those unconventional things. In a sense that’s right, if they were likely to do those things the asset markets never would have crashed in the first place. Woodford showed that current AD is strongly influenced by expected future AD. But of course this is equally true of fiscal stimulus. The sickening plunge in asset prices and economic activity from August 2008 to March 2009 was an implied prediction that (take your pick) fiscal stimulus doesn’t work very well, or would not be done in large enough amounts (probably some of both.)
I doubt that a quasi-monetarist policy regime (say targeting NGDP expectations) would be effective in future severe recessions. Why not? Because if the regime was expected to be quasi-monetarist then the deep slump should never have happened. Alternatively, if it did happen then markets would presumably be forecasting that there would be no NGDP targeting. Only in the case where policy was quasi-monetarist, but not understood by markets to be quasi-monetarist, would this policy actually be adopted and work. By how likely is that? I keep coming back to 1933 because it’s a rare example of a huge stimulus surprise, although there are some smaller examples like QE2. In the real world policymakers are rarely able to surprise markets. Investors understand the policy regime, and draw the relevant conclusions. When they become pessimistic about policy, their fears are generally confirmed. We need to stop thinking about deep slumps as a sort of random “problem” that needs to be “fixed.” They need to be prevented; if they aren’t, they probably won’t be fixed.
I’m not sure whether Krugman really understands this point, despite all his excellent work on “expectations traps.” Consider the following:
Now bear in mind that in order to make a commitment to inflation work, central bankers not only have to stand up to the pressure of inflation hawks “” which is much harder when you’re having to testify to Congress than it is if you’re a Harvard professor “” but, even harder, they need to convince investors that they’ll stand up to that pressure, not just for a year or two, but for an extended period.
Now, the thing about fiscal expansion is that people don’t have to believe in it: if the government goes out and builds a lot of bridges, that puts people to work whether they trust the government’s commitment to continue the process or not. In fact, to the extent that there’s some Ricardian effect out there, fiscal policy works better, not worse, if people don’t believe it will continue.
I don’t know enough macro theory to say this is completely wrong, but at a minimum it is very misleading. He’s right that monetary stimulus is only effective if it is expected to be somewhat permanent. That’s even true (to slightly lesser extent) when we aren’t in a liquidity trap. It’s always (mostly) about the expected future path of monetary policy. The public might expect future monetary policymakers to sabotage current stimulus, and the mere expectation of that future sabotage would prevent prices from rising today. But that’s also true of fiscal stimulus. Both fiscal and monetary stimulus raise prices by shifting AD to the right. If the central bank is composed of inflation hawks that don’t want inflation, then fiscal stimulus won’t raise future inflation because the public will expect future central banks to sabotage fiscal stimulus. Of course one can argue about whether this offset or sabotage would be as complete for fiscal stimulus as for monetary stimulus. Krugman clearly thinks wouldn’t be; I think it would be even worse, as monetary policymakers worried about long run credibility would be unlikely to sabotage their own announced stimulus policy. But one certainly cannot pretend that it applies only to one type of stimulus. And it’s not just my model, in Krugman’s expectations trap model it is extremely important that future monetary policymakers are expected to behave as we’d like them to behave. Which once again suggests that we desperately need NGDP targeting, level targeting. That’s how you get the right expectations. If you do that, fiscal stimulus might work. But of course if you do that then fiscal stimulus wouldn’t be needed in the first place.
What sort of policy is best during a severe recession? The sort of policy that would have prevented it in the first place. In late 2008 we had a severe slump precisely because markets (correctly) predicted that the Fed would not take the steps necessary to prevent NGDP from falling, and staying on a lower trajectory. That’s why pessimists like Krugman are often right. A Fed that would provide effective stimulus in a deep slump is one that would have prevented the deep slump in the first place. We don’t have that sort of Fed; now it’s all about minimizing the damage, and educating policymakers so that we are better prepared for the next financial tsunami.
PS. This Matt Yglesias post makes some related arguments.
Tags:
21. March 2011 at 14:12
Scott
Your arguments seem very clear to me. A case in point is the 1987-07 period. In effect the “Great Moderation” only came about because the Fed was (unknowingly even to them) targeting NGDP along a level path! In 1998-04 there was first an overshoot followed by an undershoot. The Fed´s effort to bring NGDP back to its level path, which it succeeded in doing by 2005, was later (wrongly in my opinion) criticized even by some “New Monetarists” as being an important cause of the “Great Recession” through its effects on the house boom and bust. If it was explicitly targeting NGDP along a level path things would have been different (maybe the over and under shoots would have been smaller). And it is the MP by the Bernanke Fed, set according to an undisclosed target for an undisclosed index, which is behind the big “damage” to AD.
21. March 2011 at 14:46
“Woodford showed that current AD is strongly influenced by expected future AD.”
Didn’t he show (or rather assume) that current AD is strongly influenced by “past expectations of current and future real interest rates” (Interest & Prices p.325)?
I know you don’t like talking about interest rates, and I can see how a similar intuition might apply, but . . .
21. March 2011 at 14:51
Another thought-provoking post by Scott Sumner.
But I think it is true–global recessions mean market players believe that central bankers will not be fast enough on their feet, and indeed have botched it already.
Of course, it would be nice to dodge recessions entirely, and my layman’s guess is that NGDP targeting would do that.
BTW, an Andy Kessler has an op-ed in today’s WSJ to the effect that oil prices are up 30 percent above normal on speculation, thanks to a weak dollar and QE. We should raise interest rates, he says.
The right-wing, after disputing that oil markets could be gamed or respond to speculation, now says they can be speculated up, if QE is at fault.
If speculators can raise oil prices by 30 percent, who is to say they cannot boost oil prices by 50 percent on a “Peak Oil” scare (2008)?
And if that is true, why are we sending hundreds of billions of dollars outside the US economy, based on prices that speculators have boosted up?
21. March 2011 at 14:56
“Both fiscal and monetary stimulus raise prices by shifting AD to the right. If the central bank is composed of inflation hawks that don’t want inflation, then fiscal stimulus won’t raise future inflation because the public will expect future central banks to sabotage fiscal stimulus.”
The central bank is composed of people who want inflation to be near 2% but are willing to tolerate inflation rates significantly below 2% in order to avoid having to do too much unconventional policy. Therefore there is a range over which monetary policy is expected to accommodate fiscal policy. They won’t sabotage the fiscal stimulus, because they would have undertaken the comparable monetary policy themselves if they didn’t have such a distaste for unconventional policy.
If you look at what happened with monetary policy in 2008-2009, it wasn’t that monetary policymakers changed their preferences and suddenly decided they liked 1% inflation better than 2% inflation. It was that they encountered a situation where keeping up the inflation rate would require a lot of unconventional policy, and the market correctly perceived that they would be unwilling to undertake that policy to the extent necessary. Under those circumstances, fiscal policy would change expectations because it would reduce the amount of unconventional monetary policy necessary to achieve the Fed’s usual objective and therefore make the Fed more likely to take the policy actions necessary to achieve that objective.
21. March 2011 at 15:44
“Under those circumstances, fiscal policy would change expectations because it would reduce the amount of unconventional monetary policy necessary to achieve the Fed’s usual objective and therefore make the Fed more likely to take the policy actions necessary to achieve that objective.”
Fiscal stimulus would make the Fed more likely to engage in QE (since conventional poliies would not work)?
Or less likely (in the mistaken assumption that fiscal stimulus could handle the situation alone)?
21. March 2011 at 17:01
Marcus, That’s a good point.
Declan, I don’t have my Woodford in front of me, but that sounds right. Certainly his view is that current and future expected monetary policy strongly impacts AD. He defines monetary policy as the expected path of nominal interest rates relative to the Wicksellian equilibrium rate. I define it as the expected path on the base relative to the demand for base money (the Cambridge K.)
Benjamin, That 30% number sounds way too high for oil prices. They are mostly driven by industrial production in developing countries (plus supply shocks like Libya.)
Andy, Yes, but you’ve also presented a very strong argument for why the Fed would not sabotage itself, in which case Krugman is wrong. In this particular post I am not trying to argue fiscal policy is ineffective, I’m trying to argue that monetary policy is effective. I’m saying Krugman’s sabotage argument for monetary policy ineffectiveness is far-fetched and highly implausible. You are suggesting that it is also implausible when applied to fiscal stimulus. I’ve raised doubts about that view elsewhere, but here your argument against sabotage actually supports my view.
I think if fiscal policymakers had done nothing in early 2009 the Fed would have pulled out the “nuclear option” and not allowed a depression. I’ll never be able to prove that, but knowing Ben Bernanke I have always found it the most probable counterfactual. I’ve never heard a persuasive argument against it. That’s not to say fiscal policy wasn’t worth a shot as an insurance policy, that’s a difficult judgment call.
21. March 2011 at 18:19
Scott – yeah, but I don’t think you can just plug & play a different definition of monetary policy in W’s model like that – it is based on intertemporal maximisation given (real) interest rates. If you want to use NGDP (or whatever) instead wouldn’t you have to build a new model from the ground up?
Maybe a bit pedantic, but you have made that statement “Woodford shows . . .” quite often, and I think it is a pretty big leap.
21. March 2011 at 19:13
Declan, I need to stop making the claim if it is not true, but my understanding was that Woodford thought any policy that raised expected future AD also raised current AD. Isn’t that right? His co-author Eggertsson argued that fiscal stimulus expected to occur in the future will also tend to raise current AD. And Eggertsson basically uses Woodford’s model. But I don’t want to be sloppy, so if someone can show me that I am wrong, I’ll stop making that claim about Woodford, and assert it as my own brand new macro model.
Obviously corporate investment is heavily dependent on future expected AD.
21. March 2011 at 19:15
“[I]t is very unlikely to work in practice, precisely because it is very unlikely to be tried.” This is reminiscent of: “Have you stopped beating your wife?” So is: “I doubt that a quasi-monetarist policy regime (say targeting NGDP expectations) would be effective in future severe recessions. Why not? Because if the regime was expected to be quasi-monetarist then the deep slump should [i.e., would] never have happened.” These remarks are literally true, but they violate H. P. Grice’s “conversational implicature.”
By the way, your proposal is not so much that the Fed should target expected NGDP, but that the Fed should convince the markets that it is targeting expected NGDP. Of course, it would be practically impossible to do one of these without also doing the other; still, for you the latter has the priority.
22. March 2011 at 02:10
@Benjamin Cole
“Fiscal stimulus would make the Fed more likely to engage in QE (since conventional poliies would not work)?
Or less likely (in the mistaken assumption that fiscal stimulus could handle the situation alone)?”
Or possibly, modeling them using a more robust political economy, more likely… as the purpose, historically, of the central bank is to lower borrowing costs for the state, and when the state is running deficits of over 25%, regular open market purchases simply don’t cut it anymore.
22. March 2011 at 02:14
“Obviously corporate investment is heavily dependent on future expected AD.”
What percent of investment is expected AD and what percent is hedging against monetary expansion? It seems to me that the housing boom and the current metals boom is the latter.
22. March 2011 at 06:58
Philo, I sort of knew there was something illogical about the way I framed the argument. Here’s another way of putting it:
If a future Paul Krugman is pessimistic about the likelihood of Sumnernomics being tried in a deep slump, he’s probably right, as if central banks were likely to try it, they probably would have done so before the slump got deep.
Yes, my policy is actually about expected NGDP targeting, and I feel strongly that the Fed can only convince markets it is doing that, IF IT IS ACTUALLY DOING THAT.
Doc Merlin, Housing responds to both inflation and real income growth. Recall that inflation also raises nominal rates, so it’s not an unmixed blessing for housing. But I was primarily thinking of investment in factories and offices.
22. March 2011 at 07:24
You can target N(gDp) if you can accept 1987 (Black Monday) type disruptions. In the 2 months prior to the downdraft, the proxy for real-output, fell at its sharpest rate of decline since the Great Depression.
N(gDp) in contrast, demonstrated no comparable contraction.
22. March 2011 at 09:31
flow5,
Couldn’t one argue that, had central banks not reacted quite so defensively to Black Monday then the late 1980s overheating wouldn’t have been quite so severe?
In the UK, at least, NGDP in 1988 went quite out of control and the entire quinquiennium was depressingly reminiscent of the Barber Boom in the early 1970s, which was also driven by excessively lax monetary policy.
I think that using monetary policy targeting ANY real variable is a bad idea, whether that’s employment, real output or exports.
22. March 2011 at 12:33
W. Peden:
I don’t know anything about the UK. Reading about money & central banking was about making money. But are you talking about 1990’s? In the 1990’s reservable liabilities were reduced, reserve ratios were reduced, & increasing levels of vault cash (including ATM networks) plus retail deposit sweep programs, eliminated all reserve requirements for the majority of commercial banks. The boom-bust in housing was then the inevitable consequence.
Monetary policy objectives should not be in terms of any particular rate or range of growth of any monetary aggregate. Rather, policy should be formulated in terms of desired roc’s in monetary flows (MVt) relative to roc’s in real GDP.
Nominal GDP, can be used as a proxy figure for roc’s in all transactions. Roc’s in real GDP have to be used, of course, as a policy standard.
Because of monopoly elements, and other structural defects, which raise costs, and prices, unnecessarily, and inhibit downward price flexibility in our markets, it is necessary to follow a monetary policy which will permit the roc in monetary flows (MVt), to exceed the roc in real GDP by c. 2 – 3 percentage points.
22. March 2011 at 13:47
@Scott:
“Doc Merlin, Housing responds to both inflation and real income growth. Recall that inflation also raises nominal rates, so it’s not an unmixed blessing for housing. But I was primarily thinking of investment in factories and offices.”
Price inflation raises nominal rates, but monetary expansion drops them in the short term, because it increases the supply curve of loanable funds.
22. March 2011 at 14:08
flow5,
The UK has had three big housing boom-and-busts in the past 40 years:
1971-1975: The Barber Boom and bust. Broad money grew by nearly 30% per year, quantitative controls were abandoned and skewflation boosts housing prices. The crisis point is the departure from the European Currency Snake in 1972, which removes the last possible control on a loose monetary policy for the UK.
1985-1991: The Lawsom Boom and bust. Broad money explodes in 1985 and £M3 targets are phased out. The crisis point was Black Monday in 1987, to which Lawson responded with a big monetary expansion. Once output fell in 1989, the UK inflation rate shot up towards 10% again.
2002-2008: The Brown boom and bust. After years of strong growth that was contained by a period of unusual monetary stability by UK standards, RPI targeting is replaced by the narrower CPI targeting and this permits another bout of skewflation.
I find the Lawson and Brown booms interesting, because they show how one can have speculative bubbles in conditions of relatively low inflation.
I agree that NGDP is superior to any monetary aggregate growth target as the prime target of policy. I do believe, however, that a NGDP target should be supplemented with monitoring of broad money in order to prevent large but temporary adjustments in velocity from sending distorting signals. However, that’s an idea that I’m still thinking about and given how unfashionable monetary aggregates are outside of the Bundesbank I doubt it will ever even be seriously considered by someone intelligent enough to analyse such an approach.
Anyway, my point was that a NGDP target would not have suggested excessive looseness in 1987-1988, because NGDP targeting central banks wouldn’t overreact to temporary changes in real factors.
22. March 2011 at 18:22
Oh, good lord Scott… I mean c’mon!
This is the perfect post to go back past 2008 into 2004-2006, when the real expectations of Fed bias were confirmed and rewarded.
I’m telling ya, your work will be exponentially more compelling if you can show what it looks like in recent history when the Fed should have TIGHTENED.
23. March 2011 at 14:12
flow5, I welcome years like 1987, which was a very good year for the macro economy, with stable employment and stable inflation.
Doc Merlin, Monetary stimulus only lowers rates for a very short period.
Morgan, David Beckworth is “exponentially more compelling.”
23. March 2011 at 19:03
Scott,
“my understanding was that Woodford thought any policy that raised expected future AD also raised current AD.”
I don’t think this is wrong, but it leaves out the crucial part in Woodford’s logic (interest rates), and this is precisely the part that you want to discard!
I agree that, according to the standard Euler condition for expenditure growth, anything that raises expected future income while leaving the interest rate (and everything else) constant must raise current expenditure. But if that’s all you are saying, why refer specifically to Woodford? I thought it was basic to all intertemporal maximising models. (More like “modern macro assumes . . .” than “Woodford showed . . .”)
More importantly, monetary policy can only raise expected future AD in Woodford’s model by changing the path of expected future interest rates, but you want to tell a story where monetary policy boosts expected future AD directly without worrying about interest rates. Maybe that is a better way to tell it, I don’t know. But if you don’t have interest rates, then you don’t have the Euler condition, so you don’t get to tell the above story! It is like you have taken out the clockwork but you expect the clock to keep working. (I seem to remember a post where you wanted to drop interest rates from macro but keep them for modelling intertemporal decisions. But if you want expected future policy to affect current outcomes, then you can’t do that, because you can’t do the macro without modelling intertemporal decisions.)
24. March 2011 at 09:15
W. Peden:
You use bank debits (i.e., money X velocity) to target nominal gDp.
“supplemented with monitoring of broad money in order to prevent large but temporary adjustments in velocity from sending distorting signals”
24. March 2011 at 10:36
Declan, OK, Maybe I shouldn’t cite Woodford, but rather cite macro theory. But then why do so few economists understand this basic macro theory? Didn’t Krugman argue in March 2009 that one could hardly expect fiscal stimulus to have worked yet, after all, the money hadn’t been spent yet? Why didn’t Krugman understand basic macro theory?
It’s easy to have monetary policy raise future NGDP without interest rates being involved, as the excess cash balance mechanism alone will get you that result. But in any sticky price model interest rates will change until prices adjust.
My view is that the key insight of Woodford (which Milton Friedman often overlooked) is that what matters is not the current stance of monetary policy, but rather changes in the expected future path of policy. To me that’s the key insight. Whether you descibe that as a path of short term interest rates, or a path of the monetary base, seems less important to me. But maybe others don’t see it that way.
24. March 2011 at 15:56
No argument with the insight! I just think Woodford gets there a very different way,* and you cite him like he is saying something much closer to you than he actually is. (Even if you increase future NGDP by the cash balance mechanism, I think you have to think worry about interest rates if you want future NGDP to affect current NGDP. A short run boost to demand won’t affect lifetime income all that much, so that channel won’t raise current expenditure much.)
Krugman: “I don’t actually believe in all the intertemporal maximization that goes on in these models”.
http://krugman.blogs.nytimes.com/2011/03/21/nobodies-of-macroeconomics-very-wonkish/
*By assuming that interest-sensitive expenditure has to be determined d periods in advance, and that people make these decisions in a forward-looking manner, so that it is expectations of interest rates as of time t-d that affect expenditure at time t: “an unexpected change in monetary policy in period t can have no effect on aggregate real expenditure before period t+d.” p325
26. March 2011 at 05:28
Declan, If future NGDP rises, and nominal interest rates are unchanged, then the nominal interest rate minus the NGDP growth rate has fallen. That’s my preferred definition of the “real interest rate.” So that provides a transmission mechanism. There’s also intertemporal arbitrage. If higher future NGDP mean higher future prices, that makes current prices rise. If it means higher future output, that makes current investment rise.
11. September 2011 at 08:43
[…] However, serious economic downturns are almost invariably the result of bad policy, and economists and economics will be berated for failing to give successful advice. (Indeed, one view is that serious downturns can be manifestations of market predictions of bad policy.) […]
1. October 2011 at 17:59
[…] a recent comment Lorenzo reminded me of this Scott Sumner post from a few months […]
26. March 2015 at 11:21
[…] all the “dreaming” going on, I was reminded of this passage in an old Scott Sumner […]