Thursday, September 22, 2011

Post-FOMC: What Does the Fed Think It Is Up To?

I'll try to dissect the FOMC statement from yesterday.

The primary policy change is:
The Committee intends to purchase, by the end of June 2012, $400 billion of Treasury securities with remaining maturities of 6 years to 30 years and to sell an equal amount of Treasury securities with remaining maturities of 3 years or less.
While some people want to interpret this as different from QE2, which was a swap of $600 billion in reserves for long-maturity Treasury bonds over a period of about 8 months, a swap of short-maturity Treasury bonds for long-maturity Treasury bonds amounts to essentially the same thing under current conditions. The only differences are that the purchase is 2/3 of QE2, and takes place over a longer period of time.

While this asset swap was widely anticipated, the other policy change was not:
To help support conditions in mortgage markets, the Committee will now reinvest principal payments from its holdings of agency debt and agency mortgage-backed securities in agency mortgage-backed securities.
Since mid-2010, Fed policy had been to reinvest these principal payments in long Treasury bonds.

My contention is that both of these interventions are irrelevant, and will have no effect on current or future prices and real activity. First, as I argue here, the asset swaps cannot matter. Second, while the QE1 purchases of mortgage-backed securities (MBS) may have mattered (possibly in some bad ways), under current conditions MBS purchases by the Fed cannot make any difference unless the Fed purchases dominate the market, which they will not.

So what does the FOMC think it is doing? First, it justifies the interventions in terms of its dual mandate:
Consistent with its statutory mandate, the Committee seeks to foster maximum employment and price stability. The Committee continues to expect some pickup in the pace of recovery over coming quarters but anticipates that the unemployment rate will decline only gradually toward levels that the Committee judges to be consistent with its dual mandate. Moreover, there are significant downside risks to the economic outlook, including strains in global financial markets. The Committee also anticipates that inflation will settle, over coming quarters, at levels at or below those consistent with the Committee's dual mandate as the effects of past energy and other commodity price increases dissipate further.
As we know, FOMC statements post-financial crisis now are much more explicit about the dual mandate, in including the "maximum employment" language. Further, note the emphasis on future inflation here. The FOMC is telling us that what matters is the Fed's forecast of future inflation (which is low) not current inflation (which is high). This is quite different from what we saw in Bernanke's justification for QE2:
Today, most measures of underlying inflation are running somewhat below 2 percent, or a bit lower than the rate most Fed policymakers see as being most consistent with healthy economic growth in the long run.
Thus, in November 2010, Bernanke wanted to convince us that QE2 was reasonable by appealing to current inflation observations; now he wants to convince us based on the inflation rate that we have not yet seen. My interpretation of this is that he does not actually care that much about inflation, but is focused on the second part of the mandate (real activity), as Charles Evans (one of the members who voted for the policy) certainly is.

Now, an interesting part of Bernanke's Washington Post piece that justified QE2 was this:
This approach eased financial conditions in the past and, so far, looks to be effective again. Stock prices rose and long-term interest rates fell when investors began to anticipate the most recent action. Easier financial conditions will promote economic growth. For example, lower mortgage rates will make housing more affordable and allow more homeowners to refinance. Lower corporate bond rates will encourage investment. And higher stock prices will boost consumer wealth and help increase confidence, which can also spur spending. Increased spending will lead to higher incomes and profits that, in a virtuous circle, will further support economic expansion.
Maybe Bernanke can explain to us why the Fed's announcement this time coincided with a large drop in stock prices and a narrowing of the spread between nominal Treasury bond yields and TIPS yields (the break-even inflation rate).

I am collecting a set of rules for central bankers. Here are some of them:

1. Don't claim credit for things you cannot control.
2. Don't claim property rights over things you cannot control.
3. Don't engage in interventions when you have no idea what the consequences are.

Bernanke and company have broken all three of those rules.

1. It is dangerous to claim credit for stock price appreciation.
2. The FOMC claims it is intervening to lower the unemployment rate. Under current conditions, it cannot do that.
3. The FOMC does not understand the effects of its policies.

Fisher, Plosser, and Kocherlakota are on the right side of the fence in dissenting on this decision, and I think we should support them.

42 comments:

  1. On rule 3.: no role for experimentation?

    ReplyDelete
  2. ^ by definition, no.

    ReplyDelete
  3. I think experimentation with your citizens is a very bad thing.. this is why biological scientists use mice instead of people..

    Steve, there is something else regarding this move by the FED that I think it's quite risky: concentrate the debt in the short term.. as we have seen in the crisis in emerging market, once people do not trust in the ability of the government to pay the debt (or to maintain a low inflation rate) people are going to require huge returns to renew that debt... Anyway.. I can't believe the US is conducting these crazy policies..

    ReplyDelete
  4. If the Fed is going to do something new, they at least should have good theory backing it up. For example, some countries (e.g. New Zealand, Canada) in the past moved to systems with no reserve requirements and interest on reserves. Those were experiments, but there was sound theory to tell us what would happen, and then the central banks worked the bugs out. Quantitative easing does not have sound theory behind it.

    ReplyDelete
  5. But medical scientists use people when they do clinical trials. As do development economists.

    I am serious: There is an ethical case for experimenting with human subjects if the gains are potentially large enough. Bernanke et al. seem to think that the gains of QE in whatever form are large since they would have the power of reducing unemployment. Isn't it worth a try? If not why not? What are the parameters? Where do we set the limits?

    It seems like we could benefit of having an adult conversation about the topic instead of just putting a blanket prohibition and sticking to the status quo even if it is a pretty dismal one.

    ReplyDelete
  6. last anonymous,

    The Fed's actions make the outstanding debt of the Fed and the Treasury of shorter maturity. An interesting point (Jim Hamilton has discussed this) is that the Treasury has been moving in the other direction, by issuing more long-term debt. The Fed thinks the public should hold more short-maturity debt; the Treasury thinks the public should hold more long-maturity debt. Obviously someone is wrong. As you point out, if the outstanding debt is longer maturity, that can insulate the government against short-term debt crises where debt-holders start demanding large risk premia. Whether that matters for the US is not clear. You never know of course.

    ReplyDelete
  7. Another point:

    Note that the Fed is not saying: "This is an experiment, let's see if it works." They are essentially saying they know it works.

    ReplyDelete
  8. How about these two rules?

    If:
    1) Fed policy does not affect demand for bank reserves, and/or
    2) Fed policy does not significantly reduce market liquidity risk premia,

    Then:
    Fed policy is either likely to be ineffective, or likely to be fiscal policy, or both.

    ReplyDelete
  9. Sorry, that qualifier should not be an "and/or", but an "and" instead.

    ReplyDelete
  10. Steve, you wrote, "...these interventions...will have no effect on current or future prices". But the move (and the nature of the move) in Treasurys (specifically the 30y) following the announcement suggests the intervention (or rather the announcement thereof) did affect prices.

    ReplyDelete
  11. So maybe you want to attribute all the asset price movements in the last couple of days to the Fed announcement? The S&P 500 moved down 6%. Seems that has the wrong sign.

    ReplyDelete
  12. Steve: asset price movement = f(actual announcement - expected announcement). If the market was hoping for something stronger, the fall in asset prices is what you would expect.

    ReplyDelete
  13. To be more specific, suppose expected future Nominal GDP fell (partly real Y, partly P) because the announcement was weaker than expected, then you would expect 30 year government bond prices to rise and the S&P500 to fall.

    ReplyDelete
  14. Nick,

    I'm curious why I should take your position seriously since it doesn't seem to fit the fundamental tenant of science - falsifiability. As far as I can tell, you can rationalize anything. If asset prices go up, then your position that easing helps is proven correct. If asset prices go down, then you say the annoucement was smaller than expected, again proving your theory correct. Since any movement in asset prices can be vaguely rationalized in your framework, this looks more like faith than science. Explain what reaction in asset prices (or some other relevant real variable) would falsify your theory? What possible data reaction would suggest your theory is wrong? If you can't point to that, then what you are doing is religious proselytizing, not empirical science.

    ReplyDelete
  15. Correct me if I am wrong, but isn't the Fed also assuming relatively strong segmentation between different maturity "tranches"? If segmentation is weak, selling shorter term securities and buying longer term securities may have little overall impact on interest rates, right?

    ReplyDelete
  16. Steve,

    i. purchases in the 30y were much larger than the market expected
    ii. 30s outperformed on the curve, beginning immediately after the Fed announcement. Volumes were high

    I happen to think i and ii are related. I am not trying to "attribute all the asset price movements in the last couple of days to the Fed announcement", just one.

    ReplyDelete
  17. I'm with Ted on this one. To sort out the effects empirically, we need a structural model. No one has provided one. All of the "evidence" is from event studies, which are essentially worthless.

    ReplyDelete
  18. Justin,

    Yes, exactly. The Fed's arguments rely on market segmentation. Put another way, they are saying that the Fed can intermediate across maturities of Treasuries, and that will matter. The problem is that the private sector can do that intermediation just as well. The Fed does more of this intermediation, and the private sector undoes that by doing less.

    ReplyDelete
  19. Ted: fair point.

    I think the answer is to survey people before the announcement, and ask them what they expect the announcement to be. Then compare the actual announcement to what people said in the survey. This isn't perfect of course, because the survey will give a range of answers, and it's not obvious what statistic (mean? median?) to use as a proxy for the expected announcement. (Theory can answer that question in principle, but not in practice).

    Alternatively, we look at announcements that seem to be totally unexpected, where the most plausible counterfactual was that people expected that the world would continue just as it had been.

    People who do "event studies" on financial markets (not me) are the ones who could best answer your question.

    But this is really a part of a more general question: how can we ever really tell if any policy worked? We don't have a control group and a randomised experiment, so can't really tell what would have happened otherwise.

    Sometimes we look for "natural experiments". Sometimes we get the theory to tell us what co-movements to expect. My theory says that a successful monetary policy announcement (one that was better than expected) should cause: expected future price level to rise; expected future real income to rise; stock prices to rise; real commodity prices (like oil) to rise; and the yield spreads between risky and safe bonds should come down. That's falsifiable. (But in this game, since we are so bad at it, I would count getting 4 out of 5 right as doing OK!)

    ReplyDelete
  20. "how can we ever really tell if any policy worked?"

    You need a theory that you somehow trust, immune to the Lucas critique for the policy intervention in question.

    ReplyDelete
  21. Steve: agreed. Trouble is, all of us have peeked at the data before we write our theories. So we know our theory will match (roughly) the data we peeked at before writing the theory. Plus, all theories have unobservable exogenous variables.

    To my mind, the one stylised "fact" about the business cycle that too many theories ignore is this: in recessions, it is harder than normal for an extra seller of illiquid goods to find an extra buyer, and easier than normal for an extra buyer of illiquid goods to find an extra seller.

    In some future day, you, or someone like you who works on liquidity and search, will come up with a model that captures that stylised fact (and is otherwise vaguely reasonable as a model).

    When that day comes, I will not say "You were right and I was wrong". Instead I will say "That's what I've been trying to say all along!".

    But until that day comes, the best I can do is: monetary exchange; sticky prices; Q=min{Qd,Qs}.

    ReplyDelete
  22. "Second, while the QE1 purchases of mortgage-backed securities (MBS) may have mattered (possibly in some bad ways), under current conditions MBS purchases by the Fed cannot make any difference unless the Fed purchases dominate the market, which they will not."

    Steve, what do you mean by "dominate"? At what point does one dominate or not dominate a market? If the Fed does or doesn't dominate the market, how does this make a difference vis a vis MBS purchases?

    ReplyDelete
  23. Actually, I can do a *little* bit better than "monetary exchange; sticky prices; Q=min{Qd,Qs}".

    http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/04/a-monetarist-search-model-of-keynesian-unemployment.html

    It does capture that stylised fact (if not much else). You have the ability to do a much better job of it.

    ReplyDelete
  24. I am sure someone has said this before, but event studies are like the proverbial drunk looking for his keys under the lamppost. It is impossible to know what markets expected beforehand, and what other influences are brought to bear. Further, they ignore levels: if the S&P500 falls 10% in two weeks in expectation of an announcement and rises 2% the day of, this will be flagged as a "positive response".

    The only question is, if they are so obviously flawed, why do journals publish them? It seems once they do, other economists (like Bernanke) are free to cite them as probative.

    ReplyDelete
  25. Nick : "I think the answer is to survey people before the announcement, and ask them what they expect the announcement to be.

    John Cochrane: "Rats in mazes do a good job of obeying the laws of economics, but they’re not so good at responding to surveys."
    http://www.bloomberg.com/news/2011-09-22/why-identifying-a-bubble-is-so-much-trouble-john-h-cochrane.html

    ReplyDelete
  26. All economic data, ultimately, is survey data. We ask people questions, and record the answers they give us. And call it "GDP", or "employment", or whatever.

    ReplyDelete
  27. I have another rule:
    4. Understand that it is very hard to forecast inflation.

    ReplyDelete
  28. Nick,

    "Trouble is, all of us have peeked at the data before we write our theories."

    Of course you did. Why is that "trouble?"

    "...it is harder than normal for an extra seller of illiquid goods to find an extra buyer, and easier than normal for an extra buyer of illiquid goods to find an extra seller."

    I have a vague memory of someone using a search model to think about this regularity in the housing market, and I think they had some success. Not sure why this observation wants you to think about sticky prices and wages.

    "All economic data, ultimately, is survey data. We ask people questions, and record the answers they give us. And call it "GDP", or "employment", or whatever."

    That does not have anything to do with the issue that Cochrane is getting at. Actually, not all data is survey data. Sometimes we are literally counting, for example housing starts. Sometimes the survey has to do with simple questions. For example, the unemployment survey has clear-cut questions. The probability is low that someone will misunderstand the questions, and there are yes/no answers. Some surveys seem pretty useless, which is what Cochrane is getting at - inflation expectations surveys for example. Who cares? Truman Bewley seemed to think that if he asked business people questions about how they set prices and wages, that would be informative, and he wrote a whole book about it. What a waste of time.

    ReplyDelete
  29. JP,

    I'm thinking that when QE1 happened, most mortgages in the US ended up either being held by the GSEs, or were sold to the Fed as MBS.

    ReplyDelete
  30. "Why is that "trouble?""

    see the Econometrica paper here: http://www.ssc.wisc.edu/~bhansen/718/White2000.pdf

    it has enough math even for you, professor williamson!

    ReplyDelete
  31. This seems related to things that Leamer worked on. I can go on a fishing expedition where I run a set of regressions looking for big t-statistics. I stop when I get the results I want, then I report only the last regression I ran. What I have in mind is something that I think is different. I am trying to construct a theory to explain what is going on in the world. When I'm working on the theory, I have some idea of what is in the data. Then I try to match up my theory with the actual data, and maybe I find some problems, in which case I refine the theory. There is an interplay there between the theory and the data that ultimately gives you a good theory. I don't create the theory in a vacuum.

    ReplyDelete
  32. Ok, but why would domination be important? If Fed purchases of MBS were to dominate that market, how might this change things from a situation in which their purchases didn't dominate the market?

    ReplyDelete
  33. Steve: "Of course you did. Why is that "trouble?" "

    Just that we can have more than one theory, and they all fit the data. Observational equivalence. That's the trouble I had in mind.

    "Not sure why this observation wants you to think about sticky prices and wages."

    Start in equilibrium. Then something changes that causes the market-clearing equilibrium price to fall, but the actual price falls too slowly. So there's excess supply. Which means it is very hard (impossible) for an extra seller to find an extra buyer. But the extra buyer can find more extra sellers than he needs. And there's a recession, because Q=min{Qd,Qs} and Qd is below the market-clearing level.

    So the really crude sticky price model "explains" the stylised fact that it gets harder to sell, and easier to buy, goods in a recession. Except for money, which is the most liquid good, where it's the exact opposite. Because selling goods is buying money.

    What I want is some sort of search model, with monetary exchange built in, which can fit the same stylised fact without ad hoc price stickiness.

    ReplyDelete
  34. JP,

    The MBS the Fed will be buying are close substitutes for other assets, particularly Treasury bonds. If they Fed does not buy all of them, then the price of the MBS is determined by the price of the close substitute, and I'm arguing you have irrelevance. If the Fed buys everything, then they control the price. Of course then there is the issue of existing vs. newly-issued MBS, and maybe the Fed has to buy absolutely everything to actually have an effect.

    ReplyDelete
  35. Nick,

    1. On trouble: That's a problem whether you're looking at the data in advance or not. Observational equivalence is observational equivalence. Then you have to look for more data.

    2. "...but the actual price falls too slowly."

    So you have the wrong model.

    ReplyDelete
  36. "My contention is that both of these interventions are irrelevant, and will have no effect on current or future prices and real activity."

    Are you saying that the twist operation will have no effect on the general price level, or no effect on relative prices?

    Because 30 year bond futures exploded upwards in the seconds following the 2:15 announcement while 5 year bond futures got crushed. This never happens. So relative prices have definitely been affected by the announcement.

    ReplyDelete
  37. "Are you saying that the twist operation will have no effect on the general price level, or no effect on relative prices?"

    No effect on anything.

    "So relative prices have definitely been affected by the announcement."

    That's what we have been discussing. You don't know what moved the prices.

    ReplyDelete
  38. "Trouble is, all of us have peeked at the data before we write our theories."

    This is no different than physical science. The physics, chemistry, medicine, etc. theories were devised by looking at what happens empirically and devising explanations why that fit the phenomena -- and other evidence from reality -- with hopefully mild assumptions, at least what we see actually exists, etc.

    This is fine, if it really fits the empirical data -- AS WELL AS the tons of other data and logic we've accumulated about the world and human beings, and then you also just keep watching it to see of it continues to do well.

    ReplyDelete
  39. Ok, well I'll have to respectfully take a pass on your theory then. It's reasonable to me that what transpired in the markets in the 180 seconds following the announcement was due to the effect of that announcement. Given the fact that US 10 and 30 year bond futures were up significantly in that interval, as were German long bonds and German short bonds, yet US 5 and 2 year bond futures prices were down dramatically, tells me that Operation Twist had an effect on relative bond prices.

    ReplyDelete
  40. JP Koning,

    Virtually all of the drop in l.t. Treasury yields after the "Twist" announcement came from lower inflation expectations. It seems it was not "Twist" that moved l.t. yields, but the signal the announcement sent about the path of future policy. Specifically, the bond market may have reacted to the Fed's failure to deliver on an expected IOR cut to zero. This expected cut would also have depressed s.t. yields, such that its absence would cause those yields to rise.

    Saying "Twist" did not affect relative prices is not the same thing as saying the announcement, as a whole, had no effect.

    ReplyDelete
  41. anon1,

    Yes, agreed. If you look at how all the asset prices moved, it does not make sense in terms of any theory that the Fed has in mind (and I'm using the word theory very loosely here - more about that later) about how this intervention is supposed to work. You can argue the same for QE2 for example, that any of these announcement effects that people are picking up come from a change in views about the path for future policy.

    ReplyDelete
  42. "Specifically, the bond market may have reacted to the Fed's failure to deliver on an expected IOR cut to zero."

    Fair enough. But I don't think I heard anyone of significance who was expecting a cut to the IOR.

    Most of the big firms were expecting some sort of twist operation; the size was the uncertainty.

    A minority expected QE3.

    Here's my "model" for what happened. The moment the announcement came out, speculators front-run the Fed by buying up 10 and 30 year bonds and shorting 2 and 5 year bonds. It makes sense to get in front of big traders (like the Fed) and profit off them, especially if they make it easy by telegraphing their moves. So in the end, it was the signal that the announcement sent about the path of future purchases/sales that twisted yields.

    ReplyDelete