CC/NUMBER 38
®
This Week's Citation Classic
SEPTEMBER 20, 1993
Engle R F III. Autoregressive conditional heteroscedasticity with estimates of the variance of
United Kingdom inflation. Econometrica 50:987-1008, 1982. [University of California, San
Diego, CA]
Autoregressive conditional heteroscedasticity is I sought a conditional model for
a statistical forecasting model for volatility. Vari- which autocorrelations of squared re-
ance forecasts are calculated conditional on the siduals were the LM test. The answer
past values of these and other random variables. was the ARCH model, and the test is
Unknown parameters are estimated by maxi-
mum likelihood; moments of such random vari-
now better known as the ARCH test.
®
ables are derived. [The SSCI and the SCI® David proposed the acronym and
indicate that this paper has been cited in more agreed to have Frank Srba program
than 430 publications.] the first version to be applied to UK
inflation data. We also began a Monte
Carlo evaluation of the procedure. I
ARCH presented the new paper at an LSE
workshop and at the Athens meetings
Robert F. Engle III of the Econometric Society.
Econometric Services Although the reception in Europe
La Jolla, CA 92037 was very promising, ARCH gained
and popularity only slowly in the US. The
Department of Economics paper was published in 1982 but was
University of California not initially picked up by other econo-
La Jolla, CA 92093 metricians. I had time to work out
The autoregressive conditional extensions with my graduate students.
heteroscedasticity (ARCH) model was Tim Boilerslev was a gifted gradu-
developed while I was on sabbatical ate student when David visited in about
at the London School of Economics 1985, and the idea for generalized
(LSE) in the spring of 1979. Having ARCH (GARCH) was born. Shortly
morning coffee, lunch, and afternoon thereafter we discovered the inte-
tea with David Hendry, as well as grated GARCH model, several nonlin-
Sargan, Durbin, Gorman, Mizon, and ear forms, and conditional fat-tailed
many others contributed greatly to distributions. The new formulation in
the process. My (hopeless) goal was Boilerslev1 was very attractive and
to improve rational expectations was used by Kenneth R. French, G.
macro models by introducing time William Schwert, and Robert F. Stam-
varying second moments. I was armed baugh,2 who along with Engle, David
with the ideas of conditioning in M. Lilien, and Russell P. Robins3 in-
Kalman filter models as developed by troduced the model to finance where
Adrian Pagan, with my own work on it has its greatest impact. Today, the
least means (LM) tests, and with a test ARCH model is widely used in finan-
my colleague Clive Granger had de- cial applications to forecast volatility,
veloped for bilinearity. In fact, Clive hedge risky positions, price options,
had suggested that I look at the auto- and optimize portfolios. A special is-
correlogram of the squared residuals sue of the Journal of Econometrics4
from a model I had on the computer and a new Handbook of Econometrics
one day, and I was amazed at how chapter5 provide hundreds of refer-
significant they were. ences to recent applications.
1. Bollerslev T. Generalized autoregressive conditional heteroskedasticity. J. Econometrics 31:307-27. 1986. (Cited 140 times.)
2. French K R, Schwert G W & Stambaugh R F. Expected stock returns and volatility. J. Finan. Econ. 19:3-30, 1987.
(Cited 110 times.)
3. Engle R F III, Lilien D M & Robins R P. Estimating time varying risk premia in the term structure: the AKCH-M model.
Econometrica 55:391-407, 1987. (Cited 95 times.)
4. Engle R F III & Rothschild M, eds. ARCH models in finance. (Whole issue.) J. Econometrics 52(1-2). 1992. 311 p.
5. Boilerslev T, Engle R F m & Nelson D. ARCH models. (Engle R F IE & McFadden D, eds.) Handbook of econometrics.
Volume IV. Amsterdam, The Netherlands: North Holland. (In press.)
Received August 10. 1993
18 CURRENT CONTENTS® ©1993 by ISI®
