Rubinomics Redux

I still don’t know what the word “redux” means, despite someone telling me
a while ago, but they use it all the time in Slate so it must still
be fashionable.

There has been much mud (and indeed shit) thrown around the economic policy
community over the question “Do deficits raise interest rates”? I thought
I’d throw in my couple of penn’orth. Recent visitors, do feel free to
summarise the following in a sarcastic one-liner; it may turn out long.

Brad DeLong’s argument is pretty simple; he tries to explain it in terms of
supply and demand. If you increase the deficit, you increase the supply of
bonds, so you drive down their price, and one minus the reciprocal of the
bond price is the interest rate. It’s a simple, plausible argument and I
hate it.

Why? OK, first up, a couple of economics concepts.

I admitted in comments a few stories below that although I act like I do, I
don’t actually have an advanced degree in economics. My degree is in
Finance, a closely related but separate discipline. To my way of thinking,
the chief difference between finance and economics is that in finance,
supply-and-demand arguments are generally treated with suspicion and
no-arbitrage arguments are the standard method of proof, while in
economics, things are the other way around.

Definition: A “no arbitrage” argument is a piece of economic reasoning
which proceeds by showing that, under assumptions, a financial claim must
have a particular price, because any other price would allow a speculator
to make unlimited riskless profit by following a particular trading
strategy. For example, the correct price of a 3-month S&P500 future is the
value of the S&P today, discounted by the 3 month interest rate. If the
future is any dearer than this, you short the future and borrow three month
money to buy the S&P today; if it’s any less, vice versa.

I’m assuming that you guys can work out what “supply and demand” arguments

Economists tend to distrust no-arbitrage reasoning because it looks too
much like an analytical free lunch, because they often make restrictive
assumptions which are not too bad when the context is a financial market
but ludicrous in other contexts, and because they are “ketchup reasoning” –
they tell you that a quart of ketchup costs the same as two pints, when you
want to know the price of ketchup given the cost of tomatoes.

Finance theorists tend to distrust supply-demand arguments because in the
first place, they are often used as a dodge to get out of some difficult
maths, and because the elasticity of substitution of financial assets is
extremely high. For example, if a big fund tries to sell a huge chunk of
GE stock, should it push the price down? Well, not really. If the price
drops as a result of the new supply, then somebody might as well sell some
of their Boeing to buy GE; since all they care about are cashflows and
exhypothesi the supply-induced drop in GE means that GE cashflows
are cheaper than anyone else’s, people would step in to iron out the

Obviously it’s horses for courses. The question is, are US Treasury bonds
more like GE stock, or are they more like the sort of commodities for which
supply-demand arguments are appropriate?

It depends how you model it. Implicit in Brad’s argument, and I wish he’d
made it explicit, is something like the following model of the bond market.

Bonds are claims on the productive resources of an uncertain future.
Longtime readers of this blog will recall that neoclassical economics
doesn’t deal with time and uncertainty very well at all; they deal with it
by flattening the future down into the past by means of “taking
expectations”. This is what Brad’s model is implicitly doing, and it’s the
source of much of the confusion over what Glenn Hubbard said and what the
evidence shows. Instead of thinking of the market as being one in which
the government decides every day on the supply of bonds and investors
decide every day on their demand, we have to pretend that tomorrow the
government is going to announce the net amount of bonds it wants to sell on
every future date until the end of time, and investors are going to have to
say what price they would be prepared to pay for each of those future bond
issues. Instead of a demand function for bonds, if we’re going to
use a supply-demand argument, we need a demand function over debt

This makes the argument a bit more sophisticated at the cost of not much
confusion, I think. A one-off big deficit this year wouldn’t have much
impact on the interest rate if we knew that it was matched at some future
date by an equally big net repayment of bonds (if you doubt this, think
what the effect on ten year interest rates would be if the government
announced that, just for fun, it was going to issue $1trn ten year bonds
today and use the proceeds to buy them back tomorrow). This is why there
is, to coin a phrase “no one-for-one relationship between deficits and
interest rates” in the empirical literature – if you are just trying to
match up today’s deficit with today’s interest rates, then you’re ignoring
expectations, and you shouldn’t be surprised if you don’t find a
relationship between an interest rate based on an entire future path of
deficit expectations, and today’s single time-slice of that path. Brad’s
supply and demand argument is clearly one about a static demand function
over debt paths, and the Bush deficits (and Clinton deficit reductions)
work on interest rates through their effect on market expectations of
future deficits out to the end of time (or in fact, out to thirty years, as
that is, noncoincedentally, both the longest maturity of treasury bond and
the point beyond which cash flows become really small when discounted back
to today).

That’s a supply-demand argument. I still don’t like it, because even in
its debt-path form, it assumes a constant, stable demand function over debt
paths. I don’t like this assumption, because I think that the willingness
of the population to trade off current consumption for future will depend
on a variety of factors which ought to change in historical time. I also
don’t like the use of the expectations operator because I think that we
live in a nonergodic world (what was your expectation in 1995 of the fiscal
stance of the Bush government? Well, if you were pricing ten year
government bonds at the beginning of Rubinomics, this model says that you
had to have had one!).

How would a no-arbitrage version of the same argument go? Basically, this
literature starts suprisingly recently in the 1970s with Vasicek’s
one-factor model of the yield curve. The idea is that, while GE stock
probably isn’t a good substitute for ten year Treasuries, nine year
treasuries probably are. And eight year treasuries are a decent substitute
for nine year, and so on … until, by the (arguable) transitivity of the
relation “is a good substitute for”, we have the overnight repo rate as a
good enough substitute for ten year treasuries to get an arbitrage argument
off the ground. Which isn’t as unrealistic as it sounds; there’s quite an
intuitive appeal to the concept that it ought not to make a difference
whether you buy a ten year bond and hold it to maturity on the one hand, or
on the other, you take your money and invest it at the overnight rate every
night for ten years.

Under Vasicek’s model, the no-arbitrage price for a ten year bond is
exactly that price at which the two strategies above have the same expected
return (note that finance doesn’t get away from using the expectations
operator; below I argue that it does so in a less pernicious way). If the
ten year is cheap relative to your expectation of the future of interest
rates over ten years, then you borrow overnight money today and buy a ten
year bond, then roll your debt over, every day borrowing overnight money to
pay back the previous day’s loan until you get your final payment on the
ten year bond, which compensates you for the interest costs you have
incurred over the previous ten years. Or equivalently, every ten year bond
price is implicitly a forecast of the ten year path of overnight interest
rates, and the price of government debt in general should, under conditions
of strict no-arbitrage, be equivalent to the market’s expectation of the
path of overnight rates until the end of time.

So, we’ve got a bit of a conundrum here. Under the supply-demand model,
the ten year interest rate is determined by the market’s preferences over
future debt paths. Under the Vasicek no-arbitrage model (there are more
complicated ones which incorporate time and risk premia, but give me a
break), the ten year interest rate has to be identically equal to
the market’s expectations of overnight interest rates. There is some quite
severe over-determination here. Are the market’s expectations of overnight
rates determined by the market’s expectations of future debt paths and
preferences of debt paths? We can’t rely on this being the case, because
the one thing that we do know are that the market for overnight money is
not composed of the same individuals as the market for ten year money —
one of these markets provides liquidity to the other, so the one thing we
know about the two investor bases is that they don’t have similar views
about liquidity, so why would their expectations and preferences
necessarily coincide?

I think that this antinomy has to be resolved by dropping the supply-demand
model of government deficits and the yield curve. I believe this for the
following reason; the use of the expectations operator in the two models is
not symmetrical. Note that in the supply-demand model, the market is being
asked to do two things; to form an expectation of the future debt path
based on today’s information, and to decide on the terms on which it is
prepared to provide ten year money given that debt path. In the Vasicek
model, the epistemological demands on the market aren’t so great.
Participants aren’t asked to make any guesses about the future of the world
at all; they’re just asked to give a snap judgement of what terms they see
themselves providing money going forward. Obviously, the objection to this
is that the terms on which they will provide money going forward will
depend on the state of the world in the future, so the position is
symmetrical, but I think that there is a difference. Basically, it all
comes back to ergodicity. The future debt path is nonergodic; there’s
destabilising feedback on all sides, and just isn’t possible to have
sensible expectations ten years out. But ten years out, the money market
will be about the same as it is today. The mean of the spread of market
expectations of the overnight rate ten years’ out is likely to be much more
informative than the mean of expectations of the deficit, because the
interest rate has much more structure; it’s anchored in the region 0-20%
for practical purposes, can’t be negative, etc. The market expectation of the overnight rate ten years out might not be right; it might not even be an unbiased or efficient predictor, but I don�t need that. All I�m saying is that it�s more plausible to assume that this future price quote exists, than that a coherent expectation of future debt paths exists.

So, given this interpretation, what effect do deficits have on interest rates? Well, do you know, I think it comes down to animal spirits in the end. The important thing about Rubinomics wasn�t really deficit reduction, it was that there was a coherent story to tell. People didn�t bid down ten year treasuries because they thought that ten year treasuries were going to be incredibly scarce in future; they did so because they thought that we�d entered a new environment of lower interest rates. And similarly, when Glenn Hubbard tells fibs about deficits, it�s the fibs that will have the effect on long term interest rates, not the deficits. This is the kind of economics that they don�t teach you in universities, but that you pick up quite naturally on the job. It�s an intrinsically nonmathematical way of looking at the fundamental drivers of a very mathematical model (try solving the Vasicek one day). It�s something we ought to learn about from the sociologists. It�s all about telling a story.

If you care about this sort of thing, Nicholas Dunbar’s Inventing Money is a very good book. It’s not quite such a thrilling account of the Long Term Capital Management disaster as Lowenstein’s “When Genius Failed”, but it doesn’t dumb down the finance theory.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: