As every economist knows, interest rates don't really exist. They are a mathematical construct derived from observed bond prices. Take a really simple example: suppose a bond (OK, a bill, if you want to be picky) promises to pay $100 one year from today. We observe that bond to be trading at a price of $95 today. Then with the aid of your calculator, you can derive the implied interest rate on the bond as i = ($100/$95)-1 = 0.053 = 5.3%. The calculation is a little more complicated for bonds that promise a stream of payments into the future, rather than just one lump sum, but it's just the same. We observe the bond price, we can calculate the implied interest rate, if we want to.
The same is true for all assets, not just bonds. The price of the asset is the observable reality, the interest rate on the asset is a secondary, derived reality. Except that in many assets that derivation is harder, and more ambiguous, than it is for bonds, or, at least, safe bonds. Take shares, for example. A share makes a more complicated promise about what it will pay, and when it will pay it. So we don't really know how big the flow of dividends will be, and when they will be paid. But if we make some assumptions about the expected values of the flow of dividends, and when they will be paid out, we can do the same maths and calculate the interest rate on the share.
We can also calculate the interest rate on a house, or a car, from their market prices, except it gets even more difficult and ambiguous. Houses and cars (if they are owner-occupied or owner-driven) pay dividends in kind (somewhere to live, something to drive) rather than in cash. We have to first convert them into a cash-equivalent, and then calculate the interest rate on the house or car.
Why bother. Asset prices are the observable reality. Let's just stick with what we can see, and forget about trying to calculate interest rates.
Andy Harless (H/T Mark Thoma) wrote a very good (and very brave) blog post the other day. The theoretical model on which he based his argument was, in one sense, absolutely standard. At root, it was exactly the same model we teach in intermediate macro. At root, it was exactly the same model that Paul Krugman says he uses (as a first approximation to reality). And Paul Krugman is certainly not alone in using this model.
But Andy made one slight twist, not in the formal structure of the model, but in how he presented it. He didn't talk about interest rates. He talked about asset prices instead.
Andy's "model" was really just a one-liner in his blog post. Paul Krugman's "model" is really just one curve. Very simple. And nothing wrong with that. The model is just the IS curve. Now when we normally draw an IS curve, and when Paul Krugman draws his curve, we draw a downward-sloping relation between aggregate demand and interest rates. Andy's twist was to "draw" an upward-sloping relation between aggregate demand and asset prices. (Only he didn't bother to draw it, he just used words).
By "drawing" his IS curve in a new way, Andy reached his conclusion much more quickly and clearly. In comments, I complimented him on his post, but made the mistake of suggesting it would be clearer still if he also mentioned interest rates. Then I realised how bad my suggestion was, and told him to forget it.
Why should it matter which way we draw the IS curve? (Leaving aside weirdo cases with multiple solutions to polynomial equations etc.) if you have asset prices you can mathematically derive the implied rates of interest, and if you have the interest rates you can mathematically derive the implied asset prices, and the two are negatively related, so it ought to be the same IS curve, just presented in two different ways.
Suppose we initially have aggregate demand at the right level. Then something causes the IS curve to shift left, and AD to fall. The job of the central bank is to lower interest rates to move us along the IS curve until AD gets back to the right level. Equivalently, the job of the central bank is to raise asset prices to move us along the IS curve until AD gets back to the right level.
At a fundamental level, these two statements of the role of monetary policy are logically equivalent. But God, the second way of saying it sounds weird. We are just not used to framing monetary policy that way.
But actually, the second way of framing the role of monetary policy is more immediate, in two ways.
First, because asset prices are the observed phenomenon, and interest rates a derived calculation.
Second, because the main proximate effect of monetary policy on AD is via Tobin's q -- when the price of existing assets rises relative to the marginal cost of producing new assets, firms will move along their MC curves and produce more new assets. Investment increases, in other words, and investment is a component of AD. And when the price of existing assets rises relative to the price of newly-produced consumption goods, both the income (wealth) and substitution effects lead households to increase their demand for newly produced consumption goods, and consumption is also a component of AD.
In fact, and contrary to what we are used to, it is much more natural to frame monetary policy in terms of asset prices than in terms of interest rates. Instead of thinking that monetary policy is interest rate policy, we should be thinking that monetary policy is asset price policy. It really, really, is.
Now, there are lots of different assets. Each has its own price; each has its own interest rate. By talking about monetary policy as changing interest rates, or asset prices, we are partaking in the usual macroeconomic vice of aggregation. No big deal. Or, rather, no bigger deal for asset prices than for interest rates.
But nominal interest rates have a zero lower bound. Equivalently, asset prices have an upper bound (though that can be infinite, in the case of infinitely-lived assets). Beyond those bounds, cash will dominate. Short-term US dollar Treasury Bills are at that upper bound. Big deal. Lots of assets are nowhere near their theoretical upper bounds. They are lower than they used to be. And the current price of an asset depends on the expected future price of an asset, so expected future monetary policy can have a direct effect.
We have fallen into the habit of thinking of monetary policy as a (contingent) time-path of current and future interest rates. There is absolutely no reason why we cannot change our habits, and switch to thinking of monetary policy as a (contingent) time-path of current and future asset prices. By "we" here I mean not just "we economists", and "we central bankers", but also "we investors and consumers".
I too am beginning to despair of (especially) US and ECB monetary policy. What makes my despair worse is my belief that it is not even ignorance that is preventing us escaping a trap of our own making. It is our habitual way of thinking about monetary policy. It limits what we can say about policy, and so limits how we can communicate policy, and so limits what commitments we can make about future policy, and limits what we think of as "conventional" vs. "unconventional" policy on which particular assets to buy and sell.
The Fed is playing a losing hand in a card game, and won't change to a different game, where holding and playing the exact same cards could mean winning rather than losing.
(Sorry, I searched but can't find the name of the commenter on Scott Sumner's blog who gave me that great card game metaphor).
From what I hear, Sweden's central bank has instituted a negative interest rate. Which is equivalent to negative asset values. Comments?
Thanks. :)
Posted by: Min | August 25, 2010 at 12:07 PM
It implies that the asset is really a liability and the buyer truly is purchasing a bill of goods.
Posted by: Determinant | August 25, 2010 at 12:16 PM
The Fed needs to play Calvin Ball!
BTW, this is all getting to be very depressing. I have zero formal econ training and Kocherlakota's speech was enough to make me want to open a vein.
Posted by: Patrick | August 25, 2010 at 12:22 PM
Min, Determinant: It depends, roughly speaking, on the "duration" of the asset.
For a 1 year Tbill, that pays $100 one year from now, a -1% interest rate would mean the market price is (approx) $101.
At the other extreme take an asset that pays $1 per year forever. At 10% the price is $10; at 5% $20, at 1% $100, then in the limit, as the interest rate goes to 0%, the asset price goes to infinity. At negative interest rates, the math is undefined (You try to add up an infinite series of numbers that never converges, even though a negative asset price seems to work, because the normal formula, when it does converge, is P=$1/r).
Patrick: his speech made me want to open 2 veins: one for the economy; and one for formal economic training. See my latest post.
Posted by: Nick Rowe | August 25, 2010 at 12:33 PM
Nick,
You are talking about yields here, not interest rates. Please at least use the right terminology.
You calculated a yield on the bond, not an interest rate.
This is partly about accounting as well as economics.
The interest rate is a stated function of the book value.
The yield is a derived function of the market value.
Dividends are not interest.
Beyond the accounting aspect, you still can’t generalize the way you do, because economics apart from the accounting impinges as well. Assets exist that have a “floating” interest rate, tied to something like the Fed funds rate, or Libor, or the prime rate. You can’t derive either the interest rate or the yield merely by observing the asset price. The interest rate in those cases is designed specifically to hedge the asset price from interest rate risk. And assets that are priced this way exist in the $ trillions. In Canada, there are $ hundreds of billions in variable rate mortgages where a relatively constant asset/liability price can’t be taken at face value by the Bank of Canada in considering monetary policy. They must consider the effect of monetary policy on interest payments made on liabilities whose present value is relatively immune from monetary policy. For this reason at least, it would be utterly foolish to base monetary policy entirely on asset prices. At the other end of the spectrum, you have zero coupon instruments that pay no interest and whose yield is entirely a function of an interest-free asset price.
Posted by: JKH | August 25, 2010 at 01:20 PM
JKH: Welcome back here!
Given what you say about terminology, economists are (at least nearly) *always* talking about "yields" when they say "interest rates". When we say "The BoE is using QE to try to push down interest rates on long term bonds" we should presumably be saying "yields" instead. But this is standard terminology in macro. It maybe works OK for us because we almost never need to talk about what you call the interest rate on a bond, the "official" rate, as a percentage of the face value.
Yes, perhaps I did overstate my case, by ignoring the floaters. But AD depends more on longer term rates than on zero-horizon interest rates. The biggie, empirically, is supposed to be the price of existing investment goods (houses, machinery, etc.) relative to the MC of producing new ones.
Posted by: Nick Rowe | August 25, 2010 at 02:15 PM
Nick Rowe: "At negative interest rates, the math is undefined (You try to add up an infinite series of numbers that never converges, even though a negative asset price seems to work, because the normal formula, when it does converge, is P=$1/r)."
Thanks, Nick. :)
So what is happening with Sweden?
Posted by: Min | August 25, 2010 at 02:58 PM
It's only undefined if the interest will, with positive probability, stay negative forever. If this is a probability zero event then the series can still converge.
Posted by: Adam P | August 25, 2010 at 03:05 PM
Nick,
No problem with the terminology of pushing down interest rates. That can apply directly to the interest rate on a new bond, or indirectly via the yield on an existing bond (by changing its price rather than its interest rate). It’s just that you were talking in the beginning about observed bond prices.
I understand your answer above, but now I’m a bit confused about the post in general.
How does your post relate to the debate about whether central banks should be worried about asset price inflation in addition to CPI type inflation?
Posted by: JKH | August 25, 2010 at 03:09 PM
Min: what Adam said. In Sweden, asset prices are (presumably) still finite, because only short term interest rates are negative, long term ones are positive, and future short term rates are expected to be positive. But a promise to pay $100 3 months from now may be worth more than $100.
JKH: My post is not directly related to the debate over whether central banks should be targeting 2% (say) CPI inflation or 2% CPI+asset price inflation. (It might have implications for that debate, but I'm not mentally ready to get my head around them yet). It just recognises that if demand falls, then the job of the central bank can be described as cutting (real) interest rates to get demand back to normal, or, equivalently, by raising (real) asset prices ("real" means "relative to the CPI") to get demand back to normal. Two different ways of saying the same thing. Andy's post is a nice clear example of this way of thinking.
Posted by: Nick Rowe | August 25, 2010 at 03:52 PM
Hey Nick,
What if you push AD to the right continuously?
Posted by: Rahul | August 25, 2010 at 09:44 PM
Rahul" It depends what you mean by that:
1. If some exogenous force pushes the IS curve to the right continuously, then the central bank will need to ensure that real asset price fall continuously (or real interest rates rise continuously) to offset it.
2. If the central bank decides to move along the IS curve to the right continuously, by continuously raising real asset prices (or continuously lowering real interest rates), this will cause ever-increasing excess demand, and ever-increasing inflation, until the monetary system explodes.
Posted by: Nick Rowe | August 25, 2010 at 09:56 PM
Given that most of the conventional discount curve is built out of interest rate instruments (swaps), I think the original statement is a bit strong.
Posted by: Matt | August 25, 2010 at 11:28 PM
"we should be thinking that monetary policy is asset price policy."
Right.
As Hayek and Mises explained 85 years ago.
Altering the flow and size of money streams changes the relative valuations of different asset classes, throwing the economy into discoordination, creating recalculation problems.
Think of housing assets and housing related financial assets.
Posted by: Greg Ransom | August 26, 2010 at 02:15 AM
Would it be more accurate to say that "the job of the central bank is to raise [current] asset prices [relative to future asset prices] to move us along the IS curve until AD gets back to the right level"?
It is not only the currently observable asset price that matters, but also (unobservable) expectations about future asset price?
Sorry if I am being confusing. Struggling to understand the post.
Kien
Posted by: Kien | August 26, 2010 at 03:23 AM
Nick,
I liked your explanation, though I'm still struggling to understand the "right" level of AD - " The job of the central bank is to lower interest rates to move us along the IS curve until AD gets back to the right level."
This seems to me to be metaphysical terminology. Can you explain why, if the society so chooses collectively, that AD can not be permanently lower than potential GDP? Is it because this is irrational? Or is it because then some members of society will be unemployed, and that is always bad?
I think I do know the answer, sort of, since it is irrational, and represents a waste of resources - and therefore it represents not a shift in consumption from the present to the future, as individuals hoarding money think they are doing, but rather a waste of their own money - but I am interested as to the "correct" or professional answer. I am a mere BA in Economics.
Thanks,
Chaitanya
Posted by: Chaitanya | August 26, 2010 at 06:02 AM
I do not agree with JKH. In finance, we always talk about rates. I have never heard anybody at a broker-dealer use the term "yield" for his own purposes. The term is ill-defined, but usually means the internal rate of return - a meaningless number you quote to the buy-side.
You are obviously correct about the primacy of bond prices over rates. A rate is mostly a construct for comparing bond prices, useless unless you know the conventions associated with it. You cannot buy and hold a "rate" - it is not a traded asset. But you have to be careful here; what we really mean is that rates are not *cash and carry* assets. Spot rates are directly observable even though they cannot be bought or sold and forward rates are the most traded asset in the world. The modern course of IR modeling has been to model these observable rates directly, and back out the implied bond price processes, not the other way around.
Finally, everyone agrees that what central banks really want to do is to control asset prices. But 1) the prices they want to influence are those of non-financial assets, and 2) the direct means they have to influence them are over interbank rates and the prices of government bonds. What I have against your post is the same thing I always have against Scott Sumner: the problem is not with the conception, but with the execution. The market is always going to pay attention to what you can actually do, period. What you say only has influence in so far as it suggests what you are going to do in the future. If you have no direct control over asset prices - no central bank does - then you aren't going to change them in the future any more than you changed them today.
Posted by: Phil Koop | August 26, 2010 at 12:49 PM
"Second, because the main proximate effect of monetary policy on AD is via Tobin's q -- when the price of existing assets rises relative to the marginal cost of producing new assets, firms will move along their MC curves and produce more new assets."
Does this apply to the production of currency denominated debt?
Does this apply to a stock price?
Posted by: Too Much Fed | August 26, 2010 at 03:15 PM
If lower and middle class workers were put into some kind of asset like a stock, how should that value be raised?
Posted by: Too Much Fed | August 26, 2010 at 03:20 PM
Phil Koop said: "If you have no direct control over asset prices - no central bank does - then you aren't going to change them in the future any more than you changed them today."
Agreed, nor should central banks. However, will "people" start saying the fed should buy stocks to "prop" them up, which mostly helps the rich?
Posted by: Too Much Fed | August 26, 2010 at 03:36 PM