Yellin’ at Yellen II

First I would like to stress again that I have great respect of Janet Yellen. I try to only critize people whom I respect (my posts show just how respectful I am in general).

So this comment on her use of event studies is really a comment on event studies and not on Yellen in particular. I mainly objected to “Event studies can therefore be helpful.” That is not a strong claim and I don’t really disagree, but I would add “but not very helpful, because announcements can be declared non events on the grounds that they were anticipated, so conclusions based on event studies depend on judgment — something which they have in common with conclusions which aren’t based on data at all.”

More highly respectful verbal abuse* after the jump.

* that was meant as a joke.

I have the 7 year nominal treasury rate, the 7 year TIPS rate and the difference in this Fred graph which I don’t know how to embed. I see current rates are similar to those before the first announcement (August 2010). Looking at daily data, I think I see rates going up at the November announcement (that would correspond to investors thinking QE matters and there being less than expected). I don’t see much of any change pre-August to post November. The graph sure doesn’t look like a response to a larger open market operation than any in the 20th century. I think that money and 7 year notes are close substitutes now.

I had a methodological complaint about analysing four data points and ignoring two, because of an arguably valid claim that announcements were anticipated. Before moving on, I stress that, if QE has a fairly large fundamental effect, then the undetectably small response to the annoucement can be explained only if markets anticipated that QEII would be approximately half a trillion $ worth of purchases of long term treasuries. If the market had expected a trillion, and QE has a fairly large fundamental effect, then real interest rates should have jumped up. The actual market response to the announcement requires not only forward looking agents but also something fairly close to perfect forecasts of Fed policy.

I now ask myself if I can come up with another explanation of the data — one in which QEII would have had a small effect even if it hadn’t been anticipated ?

I have two explanations of the pattern. I present them partly to illustrate my objection to event studies.

The first is that the standard interpretation of event studies depends on the efficient markets hypothesis. The original idea is that the market knows the effect of policy, so we can learn what policy does by looking at the response of asset prices to the announcement. I don’t accept the premise, so I don’t accept the method.

It was clear that many people thought that the huge expansion of Fed liabilities would cause increased inflation. Some people predicted hyperinflation. Assume (as a modelling exercize) that about 10% of the risk bearing capacity of the market is controlled by people who believe in the quantity theory of high powered money — so a tripling of the supply of high powered money should cause almost a tripling of the price level some time soon. This implies a huge effect of QEI on expected inflation and on the spread between the return on regular nominal treasuries and CPI indexed TIPS.

Then inflation actually declined confirming Phillips curvaciouse predictions. No one admitted they had been wrong (people don’t do that) but people learned. The effect of QEI was not a fundamental effect. It wouldn’t have happened if people had rational expectations.

Then by last November people had learned and so QEII had a microscopic effect on asset prices.

It does not seem reasonable to assume that people know conditional expected values (I could stop there) conditional on events which have never occured in human history. That assumption is absolutely required in order to use two events to determine a structural fundamental effect of quantitative easing.

To try to summarize, the effect of QEI and QEI.5 could have been due to irrational belief in the quantity theory of money which no longer exists so they tell us nothing useful about the effect of future quantitative easing.

Was the effect in November microscopic because investors had correctly predicted something very close to half a trillion of purchases of 7 year notes or because they had decided that Fed purchases of 7 year notes matter very little ? I think there is no way to know.

A second problem with event studies is that one has to know which possible future events are like the past events in the data set. Yellen’s analysis and the analysis above depends on the assumption that quantitative easing is one thing. In particular it depends on the assumption that 7 year Treasury notes are just like Federal agency issued MBS which are just like commercial paper in 2008. QE involves the Fed issuing liabilities and purchasing assets, but the assets are different.

In Summer 2010, I saw no reason to think that it would matter much if the Fed bought hundreds of billions of 7 year Treasury notes. Nothing which has happened since should make me change my mind.

The effect of a given quantity of asset purchases on asset prices should depend on the elasticity of demand for those assets. The elasticity of demand for assets should depend on the covariance of their returns wwith that of the market portfolio. 7 year Treasury notes are perceived to bevery safe. MBS were not perceived to be safe at all when the Fed bought a $ trillion worth of them. Elementary theory suggests extremely different effects of QE I and QE II on asset prices. Yellen’s analysis is based on the assumption that the magnitude of the effects is known a priori to be similar. That makes no sense.

Also fall 2008 and winter 2008-9 were very different from summer 2010 and fall 2010. Many firms were desperate for liquid assets. They feared they might be subject to a run and didn’t wan’t to have to sell illiquid assets at fire sale prices. There was a huge spike in demand for high powered money. It is entirely possible that QE I was needed to keep the safe short term interest rate near zero. In other words, QE I might have been conventional policy which was extreme in scale, because market conditions were extreme.

In the end I don’t have a conclusion. I don’t claim that we can dismiss all evidence as irrelevant to the current policy debate because things were different then or because the policy is vaguely similar but not the same. But things were different then.

I believe that a sudden surprise announcement of another half trillion of purchases of 7 year Treasury notes would have a small effect on the real economy. That’s what I guessed last year and there certainly isn’t any new evidence that should cause me to change my view.