For some time, heteroskedasticity has been the code-word
around my office for the more opaque concerns of econometrics.
Who beyond the relatively small community of professionals who
specialize in statistical inference cares about T ratios and
P values and chi squares anyway? I've always believed that
clever applications of economic theory, expressed in models,
and illustrated by dominating examples, are the path to enlightenment.
Sufficiently strongly about this did I feel last week, when,
running down a list of Old Salts and Not-So-Old-Salts in economics
who I thought might be particularly interesting to readers,
I put Clive Granger on the list and then took him off again,
settling for closer-to-home Christopher Sims instead.
Clearly Granger was important to other econometricians, I
thought, and to some macroeconomists who used his tools, but
not to the kinds of people who read this weekly. If the Swedes
got around to him, I would write the story when they did.
And Robert Engle? A lengthy and enthusiastic article last spring in The
New York Times Magazine told how New York University was "beefing up
its roster [in economics] to make a run at the pennant by adding a core
of solid role players and a Hall of Fame pitcher [Thomas Sargent] in
the prime of his career" -- but failed to so much as mention Engle,
even though the trophy-scholar had been expensively lured to NYU's
business school from the University of California at San Diego four
So even though Granger has been working for years to get a
handle on the dynamics of the deforestation of the Amazon rain
forest (and Engle is a big noise on Wall Street), these are
the kinds of guys who, when they win Academy Awards, receive
their Oscars at a separate dinner a few weeks before the big
night. Or so I thought.
Score another for the Swedes. Last week Granger and Engle
were named winners of the 36th Nobel Award in Economics
for the tools they developed for interpreting "time series"
data -- cointegration in Granger's case, autoregressive
conditional heteroskedasticity (ARCH) in Engle's case.
More than ever, it came clear what the committee that nominates
the prizewinners recently has been trying to say.
It was the third time in four years that the award was given for
contributions to the tool-kit of empirical economists -- applied
microeconomic analysis one year, experimental techniques the next,
forecasting methods after that. In between, a prize for "lemons" --
three economists who were among the first to successfully capture
market structure with price theory by concentrating on the role of
The committee seems to be buttressing the case for the Nobel
It is, of course, always possible that the timing of the award -- just three years after another prize, to James Heckman
and Daniel McFadden, for the making of econometric tools for
microeconomics -- resulted from an inability to agree on some
other course of action.
But coming so quickly on the heels of the earlier award, this
year's prize may be directed less at the lay public, which
is always hoping to understand what is going on in economics,
than at the award's real constituency -- the scientists
of the Royal Swedish Academy of Sciences, mainly physical scientists,
who actually vote the award.
Remember, economics was riding pretty high in popular esteem when the
Bank of Sweden Prize in Economic Sciences in Memory of Alfred Nobel was
established in 1969 (that is its official name). The economics award is
the only addition to the roster of five prizes established in 1895 by
the will of Alfred Nobel -- physics, chemistry, medicine or physiology,
literature and peace
established to commemorate the 300th anniversary of the founding of the
world's first central bank -- an invention by practical men that slowly
evolved in the direction of successful theory, rather like, say, the
study of electricity. The first few prizes established clearly what the
Swedish Academy meant when it said "economics." And those awards
confirmed, at least in many minds, the authority of the prize (if not
necessarily all the fine points of its emphases).
Ragnar Frisch of Norway and Jan Tinbergen of the Netherlands
were honored first in 1969, as pioneers of the newfangled combination
of economic theory, statistics and mathematics known as econometrics.
Then in 1970 came polymath Paul Samuelson; measurement economist
Simon Kuznets in 1971; theorists Kenneth Arrow and John Hicks
In the 1970s, the prestige of economics began to slip somewhat,
and with it the reputation of the prize, as the world economy
entered what seemed to be a period of dangerous instability.
In 1974 the prize was shared equally by Friedrich Hayek and
Gunnar Myrdal -- two remarkable men of generally opposing
political points of view, more nearly philosophers of economics
than exemplars of the slice-it-thin and nail-it-down style of
scientific advance. Yet when the Swedish Academy did reward
tool-makers or successful theorists, their work was dismissed,
at least in some quarters, as obvious or trivial.
When, in the late 1980s, the prize was awarded consecutively
to two European thinkers (Trgyve Haavelmo and Maurice Allais)
for contributions they had made more than 40 years before, some
worried that the committee was losing its touch.
Then in 1993, leadership among the economists in the Swedish
Academy passed to a new generation, and within a year the Nobel
Award in Economics was painting a very different history of
the field. In 1994 came the first prize ever for game theory.
The electrifying decision to honor John Nash not only changed
the way the history of economic thought was written (though
it hardly wrote the final draft); it sparked an eventual hit
movie as well.
Next came Robert Lucas, whose successful incorporation of
expectations into dynamic models brought psychology back to
economics, and served as the hinge on which all subsequent developments
in macroeconomics have turned. Recognized next were James Mirrlees
and William Vickrey, economists whose models of asymmetric information
laid the groundwork for any number of new auction strategies.
They were followed by Robert Merton and Myron Scholes, whose
option-pricing formula formed the basis for vast new global
markets to diminish volatility and manage risk.
Then came slightly more philosophical nods to left and right, before
the return to the emphasis on tools and their uses -- Amartya Sen in
1998 for his highly practical theoretical work on ameliorating poverty,
Robert Mundell in 1999 for the ideas that eventually informed the
decision to create a common European currency.
Against this background, then, what about cointegration and
autoregressive conditional heteroskedasticity? One sign of the
renewed ascendancy of economics is the fact The Financial Times,
The Wall Street, The New York times and The Washington Post
had skilled and knowledgeable beat reporters available to write
to the story.
The Journal's Jon Hilsenrath, for example, got Harvard
economist James Stock to illustrate the cointegration technique
of analyzing relationships between variables this way: "…(F)or
hundreds of years, Mars has been moving closer to Earth. During
the same period, national income in the U.S. has been rising.
A simple statistical regression model might show that as Mars
gets closer to Earth, incomes rise. The Engle-Granger test [of
causation] helps to prove statistically that there is no connection
between these movements, while there might be connections in
other movements, such as incomes and consumer spending, or long-term
interest rates and short-term interest rates."
Part of the significance of the cointegration technique, economists
say, is that it originated in economics and now is being
adopted by scientists in other fields, including weather forecasting.
It is a claim that the citation
In his Marshall
Lectures 1998, Granger dwelt on the central problem of the
evaluation of econometric models -- a problem that, as he
noted, has received almost no attention from economists themselves.
"Econometricians and empirical modelers in economic fields
spend a great deal of effort in constructing their models. Appropriate
data are gathered, alternative specifications considered, a
good dose of economic theory inserted and the model is carefully
estimated. The final model is then ready to be presented to
the public like some exotic dish in an expensive restaurant.
Just looking at the model, the natural question arises -- is it any good?"
Any realistic evaluation must take account of the use to which
the model is put, said Granger -- not just to the quality
of its inputs, but to the quality of its output, meaning its
usefulness to the decision-maker. Does it really matter? Of
course it does, he continued. Take the question of the deforestation
rate in the Amazon region of Brazil.
Suppose this were framed as a matter of the supply and demand
for wood. An elegant model could be specified and estimated.
It would fit the data well. But pity the policy-maker who depended
on it, since the driving force behind deforestation is not the
demand for wood but rather the demand for land on which the
And heteroskedasticity? True, the word shares a Greek root
with skedaddle -- skedannumi, meaning to scatter. But
rather than describing the tendency of a mixed group of individuals
to leave a single location all at once, heteroskedasticity refers
to unequal variance among regression errors.
Think that variance doesn't matter in everyday life?
Consider the asymmetry of costs, says Granger -- the difference
between being ten minutes early for a flight vs. being ten minutes
late. In Engle's hands, computer-driven models of share-price
volatility permitted financial economists to relax the traditional
assumption that volatility was constant -- with highly rewarding
Embodying more realistic assumptions (that volatility was
related to its own past pattern, for example), ARCH models have
provided guidance for risk managers of all sorts -- none
more than the Bank for International Settlements' Committee
on Banking Supervision, the Nobel citation pointed out, whose
capital requirements for banks have themselves been calculated
with ARCH models since 1996.
* * *
For all the novelty of what happened in California last week,
it was not much different from what happened last year -- for
the fourth time, no less -- in Massachusetts. Voters in a state
in which the Democratic Party had become overwhelmingly dominant
in the legislature chose to elect a Republican governor (three
candidates have won a total of four terms in Massachusetts since
1990), hoping that divided government will prove an adequate
defense against the tendency to runaway spending and self-dealing.
Indeed, for all his cartoonish ways and loathsome behavior,
Arnold Schwarzenegger shares certain important characteristics
with Massachusetts Gov. Mitt Romney. He was an accomplished
and wealthy practitioner in one of the leading industries of
his state -- film, as opposed to venture capital. He is
an outsider to the local culture -- an Austrian immigrant
instead of a Mormon.
He is a liberal on most social issues (though Romney has hopes
of bring back the death penalty to Massachusetts). And he is
well-advised, at least so far -- the Hoover Institution's
George Shultz and moneyman Warren Buffet in California, Fidelity
Management executive Robert Pozen and Conservation Law Foundation
founder Douglas Foy in Massachusetts (local celebrities for
a smaller state).
There are two key differences that may affect the outcome.
Romney was raised in a political family. (His father George
was governor of Michigan and a presidential aspirant.) Schwarzenegger
merely married into one (the Kennedys). He may lack some of
the fundamental instincts that are vital to politicians'
Moreover, the fiscal situation in California is immeasurably
worse than in Massachusetts. As Kent Smetters of the Wharton
School (and former deputy assistant secretary for economic policy
at the U.S. Treasury Department) pointed out in the Financial
Times Friday, soon-to-be former-Gov. Gray Davis and the Democratic
legislature resorted to a remarkable array of accounting tricks
to produce a budget that was "only" $8 billion in
Almost $20 billion in current spending has been shifted to
next year through a variety of gimmicks, Smetters says, including
a $10.7 billion bond issue to be paid off by "existing
resources" that don't exist.
But then that's presumably why California voters elected
the Republican movie star in the first place. He is a beneficiary,
at least for the moment, of the fiscal rip tide that soon will
be the dominant feature of American politics.
Perhaps the new governor will follow Smetters' suggestion
and make his top priority the creation of sound and transparent
accounts for the state, which, after all, constitutes an economy
the size of that of Germany. Good luck Arnold Schwarzenegger!