Monetary Policy Targets
Pondering my last post, I got to thinking: Why stop with LIBOR? Why not target commercial paper yields? Or even maturing junk bond yields? In practice, there are a couple of reasons I wouldn’t suggest such things. First, I don’t trust the rating agencies enough to make them partners in monetary policy. (The definition of what is being targeted would have to be standardized on the basis of ratings, which means that any upgrade or downgrade would change the de facto target.) Second, market-specific problems could distort the economic meaning of the policy. (The latter is already an issue, though, because problems specific to the banking system can distort the economic impact of a federal funds rate target.) But I still think that, in theory, targeting risky interest rates is a good idea.
Here’s my argument: The Fed sets its policy based on an economic forecast, which contains some implicit or explicit assumption about the price of risk. If the price of risk changes, there ought to be some automatic mechanism for altering policy so as to offset the change in the forecast. Targeting risky interest rates would be such a mechanism. (This is essentially the argument I made last month about why a cut in September would make sense.) In fact, the Fed already does this to some extent by targeting the federal funds rate: if the price of risk rises, so does the spread of the funds rate over the risk-free T-bill rate, and the Fed acts to push down the risk-free rate so as to keep the funds rate constant.
But if automatic policy adjustment is a good idea, why stop with risky interest rates? Why not, instead of targeting a single interest rate or reserve aggregate, target some linear (or nonlinear) combination of economic and financial data. (I’ve always thought, for example, that while targeting a single monetary aggregate makes little sense, the aggregates do contain information about how Fed policy is affecting the economy. I’d like to see them be part of such a composite target.) Seems to me that, with today’s technology, it is quite feasible to put monetary policy on autopilot between meetings instead of fixing some particular interest rate (which is kind of like just a primitive version of autopilot). And the formula could be announced and followed by the markets, as any competent quant could program her spreadsheet to mimic the Fed’s spreadsheet. And the decision at a meeting would be, not whether to change an interest rate target, but whether to adjust the autopilot parameters. OK, I’m dreaming, but…
Here’s my argument: The Fed sets its policy based on an economic forecast, which contains some implicit or explicit assumption about the price of risk. If the price of risk changes, there ought to be some automatic mechanism for altering policy so as to offset the change in the forecast. Targeting risky interest rates would be such a mechanism. (This is essentially the argument I made last month about why a cut in September would make sense.) In fact, the Fed already does this to some extent by targeting the federal funds rate: if the price of risk rises, so does the spread of the funds rate over the risk-free T-bill rate, and the Fed acts to push down the risk-free rate so as to keep the funds rate constant.
But if automatic policy adjustment is a good idea, why stop with risky interest rates? Why not, instead of targeting a single interest rate or reserve aggregate, target some linear (or nonlinear) combination of economic and financial data. (I’ve always thought, for example, that while targeting a single monetary aggregate makes little sense, the aggregates do contain information about how Fed policy is affecting the economy. I’d like to see them be part of such a composite target.) Seems to me that, with today’s technology, it is quite feasible to put monetary policy on autopilot between meetings instead of fixing some particular interest rate (which is kind of like just a primitive version of autopilot). And the formula could be announced and followed by the markets, as any competent quant could program her spreadsheet to mimic the Fed’s spreadsheet. And the decision at a meeting would be, not whether to change an interest rate target, but whether to adjust the autopilot parameters. OK, I’m dreaming, but…
Labels: economics, finance, interest rates, macroeconomics, monetary policy
posted by knzn at 9/08/2007 05:26:00 PM
8 Comments:
Second, market-specific problems could distort the economic meaning of the policy. (The latter is already an issue, though, because problems specific to the banking system can distort the economic impact of a federal funds rate target.)
There you go bringing up the "zero interest rate floor" again from the last post. This is not much of an "issue" at the time horizons for the effectiveness of monetary policy. We had a few days where a few trades were conducted at low prices because a constraint wasn't binding. It was, in fact, proof that the liquidity injection was sufficient to get through the few days of greatest strain on the system. This is a bad thing?
If the price of risk changes, there ought to be some automatic mechanism for altering policy so as to offset the change in the forecast. Targeting risky interest rates would be such a mechanism.
But the Lucas Critique applies here. If you base monetary policy on the price of risk, you will alter the behavior of market participants with regard to risk. That's a big concern.
Why not, instead of targeting a single interest rate or reserve aggregate, target some linear (or nonlinear) combination of economic and financial data.
Because you cannot consistently control too many things at once. The simple, canonical example is that you can't fix the inflation rate and the exchange rate simultaneously without capital controls. Likewise, you can't fix the T-bill rate (or the spread) and keep control over reserves. The markets are interconnected, and when you tug on one string, you unleash forces in places you hadn't expected. Tug on multiple strings and some cancel out while others unravel.
I suppose if an omniscient, benevolent social planner knew how all the threads were interconnected, then an internally consistent spreadsheet could be constructed.
Yeah, I think that meets the definition of dreaming.
I was checking out the time path of the 3M t-bill and the funds target.
It looks to me like the t-bill leads the funds target downward into recessions and though it is harder to tell it seems to lead the target upwards during inflationary periods.
Doesn' this mean targeting the t-bill will make credit markets less responsive to changes in the economy, since expectations of future policy could not be priced into the t-bill.
Doesn't this imply that targeting
I agree with a September cut.
There you go bringing up the "zero interest rate floor" again from the last post.
That isn't really what I meant. I was more referring to a hypothetical situation where (for example) funds become tight in the banking system but alternative sources of credit are still easy. I don't know if this has happened, but with all the securitization and "non-bank banks" and such, it certainly seems possible.
But the Lucas Critique applies here. If you base monetary policy on the price of risk, you will alter the behavior of market participants with regard to risk. That's a big concern.
I don't see it as a problem. If the goods and labor markets had perfect price adjustment and the Fed could perfectly control some ideal price index, then the incipient deflationary impact from an increase in the price of risk would immediately result in a reduction of the risk-free rate. I want to make the real world more like that perfect one. True, if the Fed targeted the CP rate, CP holders would have less interest rate risk, but they would have the same default risk (except to the extent that the policy, by stabilizing business cycle conditions, makes default less likely -- but that's a good thing).
Because you cannot consistently control too many things at once.
You can control (to some degree) a mathematical function of many things. Your spreadsheet updates the value of the function continuously as the data inputs change, and you add reserves when the value of the function goes above the target and withdraw reserves when the value goes below the target.
That's roughly what the Fed did (in reverse, of course) when it was targeting M1, although the function in that case was just a simple sum of the different sub-aggregates.
Now that I think about it, that's basically what a Taylor rule is, too, except that the Taylor rule goes further by specifying exactly how much to react, and it assumes perfect control over some interest rate (which gets us back to the original problem).
But Karl has a point, I think: By targeting an overnight rate, the Fed is able to outsource term interest rate policy to people who, collectively, have a lot more resources than it does. Which means the Fed doesn't have to be constantly preoccupied with possibly needing to adjust its target, because the market basically does the job between meetings. I'll have to think about that.
Karl's point, which you acknowledge, is actually very similar to mine. Specifically, it is also a version of the Lucas Critique. Karl says:
Doesn' this mean targeting the t-bill will make credit markets less responsive to changes in the economy, since expectations of future policy could not be priced into the t-bill.
Yes and no. They couldn't be priced into the T-bill, of course, because that's what you'd be fixing. But expectations related to other asset prices would adjust. And I don't think it would be a simple matter to trace out those effects. This is what I mean by pulling on one thread causing another to unravel. The bottom line is that the actual effect of targeting some arbitrary asset price with monetary policy will lead to changes in expectations and behavior in other markets to which that asset is linked. The result would be extremely hard to predict...much harder than you are acknowledging.
Now this time you say,
You can control (to some degree) a mathematical function of many things.
But that's not how I read your original...
Why not, instead of targeting a single interest rate or reserve aggregate, target some linear (or nonlinear) combination of economic and financial data. (I’ve always thought, for example, that while targeting a single monetary aggregate makes little sense, the aggregates do contain information about how Fed policy is affecting the economy. I’d like to see them be part of such a composite target.)
Which is it? Many variables as arguments into a function that yields a target for a single policy variable (interest rate, aggregate, or whatever), or targeting a multiple things at once? Your original post clearly says the latter. When I balk, you say that you're talking about the former, like a Taylor rule which is clearly something different (and not at all objectionable).
Do you acknowledge that targeting multiple variables would be problematic (c.f. Poole 1970)?
Of course I have no problem with a Taylor rule.
When I said, "some linear (or nonlinear) combination of economic and financial data," I meant a mathematical function, such as SUM(w[i]x[i]), where the x's are the individual variables you are combining, and the w's are chosen as a set of parameters embodying the current policy. My idea was just that you push reserves when the function goes above its target and pull reserves when the function goes below. (Just as a formal matter, the x's need to be defined such that they tend to fall when you push reserves, so for example, you would choose "yen per dollar" rather than "dollars per yen" and so on.)
One step further is to make the amount of reserves you push/pull a function of the amount by which the objective deviates from its target. (And of course the instrument doesn't need to be reserves; it could be the T-bill yield. I guess Taylor treats the federal funds rate as the instrument, but that's abstracting a bit, since the Fed doesn't really have its hand directly on that instrument. So a Taylor rule in practice involves two levels of instrumentation.)
I have issues with the Taylor rule in particular (which I'll probably discuss in a later post), but what I'm suggesting is essentially a between-meeting operating procedure that would have a form analogous to the Taylor rule.
Ok, our misunderstanding was over the word "target". When I see "target" as in "target a single interest rate" or "target some linear (or nonlinear) combination of economic and financial data", I read that as a statement about the desired outcome. As in, the target for the fed funds rate is 5.25%. That's the desired outcome. For an inflation targeting scheme, you might have a target of 2% inflation. Again, the target is a (single) desired outcome. I balked at the idea of having multiple targeted outcomes. You can have multiple inputs, but in general you can't induce the desired change in all the inputs simultaneously. That's how I read your original post. Sorry for the misunderstanding.
So yes, you can control a function of many things. I'm not convinced that it would be desirable to have a lot of things in that function though. As you start to construct the mapping and add variables I fear that it would start to look like a Ptolemaic system of epicycles upon epicycles. The model would be a complicated and in need of frequent revision.
Whereas something like a simple inflation target would be simpler and would be almost as good (and perhaps better) at achieving the goal and with less need for revision.
I don't think we understand the economy well enough to construct a one-to-one mapping of, say, more than three variables to a policy instrument. Do you? And what happens the first time that the model kicks out a result that doesn't square with the "boots on the ground" instinct of the central bankers? The more variables in the model, I think the more likely that scenario becomes.
網頁設計，FreeForm，徵信,翻譯,外遇、電扇、電風扇、涼風扇,酒瓶雕刻、禮贈品、禮品、贈品
Post a Comment
<< Home