Clive
Granger is Distinguished Research Professor and Professor Emeritus at the
The
Quotes from Clive Granger’s February 2005 lecture at
I am not sure that I
ever did become an economist. I started as a statistician and have ended as a
time series econometrician. I have picked up some economics on the way and the
field of econometrics has itself evolved to be closer to my interests as I have
moved closer to the core of econometrics. Thus there are two components to my
intellectual journey, from being a statistician to being something of an
economist, and within econometrics, from being purely a time series
econometrician to having greater appreciation for other components of the field
of econometrics.
It soon became clear to me that economists think differently
than mathematicians. Rather than dealing with carefully defined objects obeying
precise rules, economists considered large numbers of independent decision
makers who based their decisions on changing experiences including learning,
information, and institutions. These decision makers were assumed to be
rational, and sometimes super-rational to an impossible extent. When put into a
microeconomic framework their behavior could be satisfactorily described to a
mathematician, but I did not recognize my own economic behavior. In aggregate
these decision makers formed markets and became captured by mysterious forces
such as supply and demand, arbitration, and the invisible hand, which produced
charmingly simple rules but of dubious reality.
One area where we have
clearly improved is in knowing how to evaluate our forecasts, how to decide if
one method of forecasting is better than another, or whether some combination
of the two may be better than both, as is usually the case. Many of these
developments are possible because of the enormous increase in computing power,
both in speed and memory size, which has occurred during my career….We can also
attempt to forecast more things, very high frequency data, such as each trade
on the stock market, or over longer horizons, perhaps up to a quarter century
ahead and breaks, or sudden changes in the economy, such as a financial crisis.
These are all very difficult and we do not forecast them very well, but are
improving and by trying we are learning something about how we can approach
such problems.
I have also been involved
in a long term program which attempts to develop parts of time series analysis
and econometrics. In my early years as a researcher I observed that
econometrics did not emphasize the temporal aspects of economics, possibly
because the time series data available were rather short and thus difficult to
analyze. Slowly it was realized that time series methods were important,
particularly in macroeconomics and finance and coverage of them started to
appear in the major textbooks. From data analysis it also became clear that
economic time series did not obey the standard assumption that they were
stationary. Many series needed to be differenced to make them stationary. This
made them “unit root processes.” This observation implied that many standard
statistical procedures, such as regressions, could not be used without
interpretational problems. New procedures and models had to be devised, and
widely applied, and economic interpretations had to be found.
Throughout my career I have been involved with the important
concept of causality in economics, although to varying degrees at different
times. I entered the arena naively needing a definition in connection with an
interpretation of a technical concept know as the cross-spectrum. I was
directed to a paper written by a very eminent mathematician, Norbert Weiner.
There I found the definition that I later expanded….Initially the definition
was slow to be accepted, but later the application by Sims (1972) produced a
great deal of discussion. Soon many alternative tests became available and
applications appeared, although most writers did not quite accept the
definition of causality, saying that the definition used was not “real causality
but only Granger causality,” although no one would define “real causality” for
me.…I have since become involved in fairly heated debates about what is
causality and there are now various alternative definitions available to
applied economists. But I let demand for the product determine its current
worth and continue to maintain a belief that whatever the final definition that
we all agree on might be, it will contain my own as a component.
Selected links to Clive Granger’s
research:
Clive W.J. Granger, “Time Series Analysis,
Cointegration, and Applications,” Nobel Lecture, December 8,
2003 (pdf).
Nobel Committee, “Information
for the Public: Overview of Statistical Methods for Economic Time Series and
the Contributions of Granger and Engle,” 2003.
Nobel Committee, “Advanced
Information: Time Series Econometrics: Cointegration and Autoregressive
Conditional Heteroskedasticity,” 2003.
Clive W.J. Granger, “Modeling,
Evaluation and Methodology in the New Century,” Economic Inquiry,
January 2005 (pdf).
Additional resources on Clive
Granger are available at the Nobel web
site.