Quasi-Experimental Evidence on the Neutrality of Money

Arindrajit Dube of the Institute for Research on Labor and Employment at UC Berkeley looks at how recession probabilities on Intrade changed immediately following the Fed's announcement of an emergency rate cut:

Market based evidence on the non-neutrality of monetary policy, by Arindrajit Dube: Ahem… we have for the first time used quasi-experimental evidence to estimate the impact of a large (and surprising) reduction in the federal funds rate on the probability of a recession.  Recent financial innovations allow us to use the market for “recession futures” to estimate impact of policy on implied probabilities in an event study framework.  Using hourly trading data from Intrade.com over a 48 hour period, we find that a 0.75 percentage point reduction in the federal funds rate on the morning of January 22 led almost instantaneously to a reduction in the implied probability of recession from 77% to 68%, and to around 70% after 24 hours of the announcement.

Please click here for larger version

We find the elasticity of recession odds to policy [d(probability of a recession)/ d(percentage point reduction in rate)]  to be between 9.3 and 12. This suggests that at best, a 400 basis point (4 percentage point) reduction in the federal funds rate can hope to reduce the odds of a looming recession by around 50%.  Our identification assumes that this was both unanticipated in terms of timing, and represents a net reduction in the medium term as compared to the forecasted path of the federal funds rate. If the surprise reduction in federal funds rate is partly substituting for a (now foregone) reduction in the future, the true effects of a medium-term reduction may be understated by our estimates. Overall, our results suggest that rational expectations are inconsistent with the theory of monetary policy neutrality.

Comments

[Bruce Wilder]

If traders hearing of the 75 basis point reduction thought it represented an acceleration of the 50-75 basis point reduction already expected for January 30, then I think we might calculate an elasticity of recession odds more nearly in the range of 2.5 to 3. No?

[Alex Tolley]

Even with Bruce's comments above, it assumes that the market is correctly determining the risks of recession. There are reasons to doubt this.

One indicator of recession used to be equity prices that would fall 6-12 months prior to a recession. That prices did not fall in 2007 suggests that the market was not factoring in a recession at any time in 2007 until the subprime debacle really started to unwind. This seems to suggest lack of the market's ability to forecast well.

Since recession futures should also be forecasts, why should we assume that they are in any way accurately determining the probability of a 2008 recession when the larger markets failed to do so?

[skeptonomist]

And what are the Las Vegas odds?

So now we are told that rational expectations is the sum of the opinions of gamblers. Why are they the only people who know what is going to happen? If such knowledge were general, there wouldn't be bubbles.

[dd]

Rational Expectations as the sum of expectations of a very special group of gamblers staked by IBs who reap an immediate benefit from lower short term interest rates. The Vegas scene is at least more democratic.

[Jay]

The volume at the announcement is $10,000 in total contract value. That is chump change, so I'm not sure this is the market I'd look to for risks of recession.

[arin dube]

> Bruce Wilder says...
> If traders hearing of the 75 basis point reduction
> thought it represented an acceleration of the 50-75
> basis point reduction already expected for January 30,
> then I think we might calculate an elasticity of
> recession odds more nearly in the range of 2.5 to 3. No?

Actually, it goes the other way. Imagine if the market expected the Fed to reduce the Fed Funds rate by 50 basis point on January 30. Instead what they got was a 75 basis point reduction on January 22. The timing difference should at most play a minor role in changing the odds of recession. So, in net, what they got was only a 25 basis point of UNANTICIPATED reduction in the FF rate in the medium term. Hence, the elasticity (using the 75 percent point reduction in the odds of recession) is 9/.25 = 36, instead of 12 (= 9/.75 as in my example). So the point is --- to the extent that part of rate cut was anticipated (or was simply shifted forward), the greater must be the the market's implicit belief of efficacy of the rate cut reduction in terms of reducing the odds of a recession.

To be clear, I am not suggesting you take the market's implicit belief as the word of God. Rather, I am simply pointing out that if you *do* believe in rational expecataions in the asset markets (as many economists do), it is inconsistent with the notion of monetary policy neutrality (as some economists do).

[nonotme]

How do you know that the relationship can be extrapolated to a a 400 basis point drop? (maybe that is explained in the original article?)

And how do you know that the market is not learning about the Fed's information from the change in rates, which reduces the effect of measured effects of the change in rates.

That is, the cut means that market has learned that the fed's probabilities are really 90% without the cut, so that the rate cut actually has a big effect... (convoluted, I know)

[Bruce Wilder]

Matthew Yglesias had a comment reflecting on the use of "markets" like intrade.com to predict the Presidential nominations and titled it, Predicting Conventional Wisdom.

The market efficiency hypothesis literature shows that markets incorporate new information as it becomes available, not that they have any sort of collective ESP or clairvoyance. The market for orange juice futures will react in a very nuanced way to news about the weather in Brazil or Florida or wherever oranges come from, including emerging weather forecasts. That doesn't mean that orange juice futures markets are some eerie counterpart to the Farmer's Almanac, able to predict weather conditions many months in advance.

Calculated Risk had a nice chart up showing how downturns in the S&P 500 lined up with recessions. Go look at it. You'll learn two important things.
1. Stock market declines coincide with, but do not predict recessions.
2. Martin Feldstein is lying scum, who deliberately falsified the dating of the end of the first Bush recession.

OK, the evidence of the chart is not conclusive on that second thesis.

Where recessions are concerned, I do think the meterological view of the economy is seriously misleading. We cannot do much about the weather -- it just happens. The economy, on the other hand, is something, which we, collectively, enact -- something we do, in other words.

There are several kinds of uncertainty involved in anticipating a recession, which are made worse by confusion over which are the source of confusion, at any one time.

One source of uncertainty is computational complexity. The economy is, itself, a computational engine, and there are peculiar difficulties with trying to simulate it. A model complex enough to be realistic in detail cannot be calculated faster than the real economy, and a model simple enough to be calculated faster than the real economy cannot be realistic enough in detail, to make the modeller confident with regard to timing.

But, worse than the problems of computational complexity are the problems of circular anticipation. Since a recession is something people collectively choose to do, forecast is a prediction of the choices people will make, which they have not yet made. And, the choices people make will, in turn, be outcomes of the predictions the choosers make about the outcomes of the choices.

Predicting that a recession will occur is not really so hard, if circumstances permit the anticipation that some quantitatively important trend will not continue for the simple reason that it cannot continue. If something cannot go on, it won't. No rocket science in that. But, when key decision-makers are ready to take on the pain of a reality-check can be hard to predict.

The recession currently underway was widely anticipated for that reason. Circumstances provide a surfeit of trends that could not continue, from the trade deficit to rapidly rising housing prices and a boom in home construction. Once the yield curve inverted, a response from the bankers was inevitable, even if the shape of the response and the timing and magnitude of the consequences of that response might have been beyond anyone's powers of calculation.

I think it interesting that so many were so reluctant to offer more than a fairly high probability of recession, even after the housing bubble finally began to deflate. It seems people have a basic confusion about the difference between the binary of whether a recession occurs and and the continuous variable of the severity of the recession.

I mentioned Martin Feldstein and his questionable intellectual integrity earlier. If the last recession lasted only 8 months, officially, and with the full benefit of hindsight, then maybe this recession won't even happen seems to be the reasoning. This is a source of "uncertainty" quite different from the ability to predict -- in fact, it derives from the ability to predict, the ability to predict the "perceptions" of a committed ideologue. It is the entirely predictable threat of a bully-in-power to shape conventional wisdom and consensus reality to suit some agenda, which gives rise to this sort of spurious "uncertainty".

But, of course, we all have agendas. The fact of an agenda, or that our agendas shape our perceptions of reality, and affect the names we apply to this or that, should not be an objection.

And, that brings me full circle to the "predictions" of markets. It is not the binary, that a market "predicts", it is the continuous variable of severity with which it is concerned, and as man is the measure of all things, so the market uses its own agenda to measure severity.

That's the fundamental problem with framing recession probability as a binary. If we were predicting, say, death or pregnancy, that would be different. You cannot be only a little bit pregnant, because it doesn't stop at a little bit, it continues right through wanting the car keys on Friday night and college tuition.

The binary probability of recession right now is 1, because it is an accomplished fact. We've (collectively) gone ahead and done it. It is as if we had accidently cut ourselves -- we fully expect to bleed. If the doctor comes along with bandage and pain killer, though, we may hope not to care too much.

Dr. Bernanke has come along with bandage and pain killer in big doses, and what Dube measures at Intrade (or what is seen at the NYSE) is the cheerfulness of the injured patient after treatment, not the probability of an injury, which has already taken place.

[Bruce Wilder]

Thanks to Arin Dube for the correction. As notsneaky can tell you, I often come up short on ratios.

[Bruce Wilder]

As an analgesic, money is never neutral.

[knzn]

As far as I can tell, all this shows is that a substantial fraction of prediction market participants think money is non-neutral. Considering that a large fraction of economists (including me) think money is non-neutral, it would be very surprising if a substantial fraction of prediction market participants did not also think so. My priors were overwhelmingly in favor of non-neutrality, but they have not become any more overwhelmingly in favor of it after reading this.

I suppose it shows the limits of the rational expectations concept. A lot of people think money is neutral, and a lot of people think it's non-neutral; one of these groups must be wrong, and the other right, which means the average forecast must be biased, since that average will include a lot of people who were wrong in a specific way and nobody who was wrong in the opposite way (since everyone who believed the opposite was right). All you really need is a thought experiment to show that rational expectations is wrong.

[gordon]

I have to point out that nobody, absolutely nobody, not even Bruce Wilder (with whose comments I am usually in agreement) is entitled to say "We cannot do much about the weather -- it just happens" any more.

[Bruce Wilder]

gordon: "nobody . . . is entitled to say 'We cannot do much about the weather -- it just happens' any more."

We definitely need to be working on control of the climate, but I expect the weather will remain beyond our immediate control.

[Bruce Wilder]

knzn: "A lot of people think money is neutral, and a lot of people think it's non-neutral; one of these groups must be wrong, and the other right, which means the average forecast must be biased, since that average will include a lot of people who were wrong in a specific way and nobody who was wrong in the opposite way (since everyone who believed the opposite was right)."

If there is money to be made in being right, and only one opinion is always right, the wrong will go broke and leave the market. Price will form with people of only one opinion type participating; people of the other opinion type will get jobs as CNBC pundits and University of Chicago professors and kibbitz shamelessly.

If both opinions are circumstantially true some of the time, and the market allows both short and long bets, then an unbiased equilibrium implied forecast by the market might be possible. The forecast may never actually prove to be "correct" in individual cases, if an individual case is always one or the other, but the market's estimate could nevertheless be unbiased in relation to the whole class. And, both opinions would sometimes make money.

[Lafayette]

Article: Using hourly trading data from Intrade.com over a 48 hour period, we find that a 0.75 percentage point reduction in the federal funds rate on the morning of January 22 led almost instantaneously to a reduction in the implied probability of recession from 77% to 68%, and to around 70% after 24 hours of the announcement.

Ohhh, Wow!

A three-quarter basis point reduction in the Fed money rate reduced the recession probability by 7%!

Will wonders never cease ....

This suggests that at best, a 400 basis point (4 percentage point) reduction in the federal funds rate can hope to reduce the odds of a looming recession by around 50%.

Uh, oh. He's proposing a negative interest rate?

Think of what a negative Fed rate would do for the probability? Then remember Japan's decade of economic stagnation when it was tried there ...

Back to the drawing boards.

[Alex Tolley]

BW: "1. Stock market declines coincide with, but do not predict recessions."

No, the chart clearly shows that the markets declined steeply from their peaks before the recessions. They often continue to decline during the recession, but that is not relevant to the historical ability of the market to forecast a recession prior to it happening.

BW: "One source of uncertainty is computational complexity. The economy is, itself, a computational engine, and there are peculiar difficulties with trying to simulate it."

Lots of natural phenomena exhibit computational complexity, nevertheless we can make predictions of: the weather, human health, human psychology.

BW: "The recession currently underway was widely anticipated".

I don't think so. You can check the Google trends for the word "recession" here:

http://www.google.com/trends?q=recession&geo=usa&sa=N

which clearly shows only low levels of interest even in late 2007. Anecdotally, looking back over the last 8 months of old BusinessWeek magazines, there is next to no mention of recession. My original post on the market behavior not indicating a recession stands.

I find your argument about "framing recession probability as a binary" confusing. Recessions are a binary event - either the economy meets the criteria for declaring a rececession or it doesn't. The probability of that event is not binary, by definition. One can argue about whether it makes any sense to talk about recession probabilities, but that is a different topic.

[knzn]

Bruce, I disagree. First of all, the proposition "Money is neutral" has to be either true or false. (It could be false but there are many situations where monetary impulses happen to have zero effect.) If there is enough noise and enough other factors to sort out, the people with the wrong opinion will not go broke in any reasonable time frame, but over any sufficiently long time frame, they will have lower risk-adjusted returns (unless the wrong belief is correlated with a right belief about something else).

[gordon]

Bruce Wilder, please read the IPCC reports urgently!

[skeptonomist]

Bookies don't actually bet on events - they adjust the odds so they are covered, and take a percentage of the action. Are major players doing this in the futures markets?

Professional gamblers who are not bookies make their living from suckers - people who don't have any idea what the odds are (or don't care) and are effectively just throwing their money away. You can see lots of suckers in any casino. Who are the suckers in the futures markets, if smart people are making money?