Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Econometric Theory, 19, 2003, 457–480+ Printed in the United States of America+

DOI: 10+10170S0266466603193061

J. DENIS SARGAN
AND THE ORIGINS OF
LSE ECONOMETRIC METHODOLOGY

DA V I D F. HE N D R Y
Oxford University

I review the part played by John Denis Sargan in the formation of the “LSE ap-
proach” to dynamic econometric modeling+ Despite his unassuming demeanor and
his location at LSE—which had earlier dismissed a substantive role for econo-
metric evidence—Sargan nevertheless radically altered the econometric approach
of a generation, establishing a powerful approach to empirical modeling of eco-
nomic time series+ His main contributions to econometric methodology, and the
subsequent research, are discussed as a complement to the other papers in this
memorial volume+

1. INTRODUCTION
This paper concerns the role of John Denis Sargan in the formation of the “LSE
approach” to dynamic econometric modeling+1 Denis’s work spanned many as-
pects of econometrics, analyzed in part here and also in the companion papers
by Peter Phillips and Peter Robinson+ Although Denis only published one pa-
per explicitly concerned with methodology, or the general procedures of econo-
metric modeling, much of his research had important implications for the
formulation of econometric models and their evaluation against empirical evi-
dence+ Histories of the development of econometrics from the mid-1950s on-
ward ascribe a major role to him in the formalization of dynamic econometric
modeling ~see, e+g+, Epstein, 1987; Qin, 1993!+ Denis’s accompanying paper
provides a brief earlier history of intellectual developments at the London School
of Economics ~LSE!+
The motivation for Denis’s research helps explain his general approach ~see
Phillips, 1985!+ To quote from Desai, Hendry, and Mizon ~1997, p+ 1121!:
+ + + it was not until he @Denis# visited a bookshop towards the end of the war and
found John Maynard Keynes’s General Theory that his enthusiasm for economics
was aroused+ He was so struck with Keynes’s ideas for tackling mass unemploy-

Financial support from the U+K+ Economic and Social Research Council under grant L138251009 is gratefully
acknowledged+ I am indebted to Alok Bhargava, Julia Campos, Meghnad Desai, Neil Ericsson, Toni Espasa,
Grayham Mizon, Peter Phillips, Timo Teräsvirta, and Ken Wallis for helpful comments and to Peter Phillips for
the invitation to contribute this paper to Econometric Theory+ Address correspondence to: David F+ Hendry, De-
partment of Economics, Oxford University, Manor Road Bldg+, Manor Road, Oxford OX1 3UQ, UK+

© 2003 Cambridge University Press 0266-4666003 $12+00 457


458 DAVID F. HENDRY

ment that he decided to use his knowledge of mathematics and statistics to help
tackle some of the pressing economic problems that faced society in the post-war
years+
It may come as a surprise that Denis first worked on economic theory and only
later turned to technical econometrics+ When he did so, however, his scientific
background enforced an empirical approach, and most of his theoretical re-
search was driven by the need to solve the resulting difficulties in formulating,
estimating, evaluating, and analyzing empirical econometric models+ This nat-
urally required an understanding of the modeling process itself—which is the
aspect of interest here—and also of the properties of the tools of econometri-
cians ~estimators and tests! and the tools of the toolmakers ~such as Monte Carlo
methods and analytic finite-sample investigations!+ Many of his technical pa-
pers are reprinted in Sargan ~1988a!+
The structure of the present paper is as follows+ Section 2 reviews earlier
“LSE approaches,” namely, those associated with Robbins ~1932! and with the
“positivist” strand of philosophy, before considering the precursors to Denis’s
contributions to methodology+ Section 3 focuses on the methodology embodied
in Sargan ~1964!, whereas Section 4 considers his formal statement of general-
to-specific ~Gets! modeling+ Section 5 turns to Denis’s contribution to under-
standing and counteracting “data mining,” in the sense of adventitiously
significant coefficients due to testing many effects+ Section 6 then considers
his contribution to model selection+ The papers in these two sections were widely
known and influential within LSE but have only recently been published+ Sec-
tion 7 sketches some of the consequential developments in Gets modeling that
have occurred more recently+ Section 8 concludes+

2. THREE “LSE APPROACHES”


2.1. Economics without Econometrics
The first powerfully expressed “LSE approach” was actually that enunciated
by Lionel Robbins+ Robbins ~1932! claimed that economic theory provided gen-
eral, formal explanations that were applicable always and everywhere that re-
source allocation impinged+ Since the assumptions of economics were “self-
evidently true” and the logic impeccable, the conclusions must be correct+
However, such insights as economics delivered could not be given quantitative
expression, since causes were nonuniform over both time and space+ Robbins
argued, therefore, that statistical relationships in economics would change when
the world did, so the discipline could never boast “statistical laws” like the
natural sciences ~see, e+g+, Hendry and Morgan, 1995!+ Robbins’s general ap-
proach was later supported at the LSE by the work of Friedrich von Hayek
~Hayek, 1963!+ More recently, some “freshwater” economists in the United States
seem to be the inheritors of the Robbins tradition ~see, e+g+, Kydland and Pres-
cott, 1991, 1996!+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 459

In principle, economic theory could indeed apply even when there were no
quantitative laws, although such explanations would be somewhat unsatis-
factory—like predicting calm seas as the norm after a storm, to borrow from
Keynes+ Moreover, Robbins’s positive argument for the power of economic
theory explanations only becomes a negative one for econometrics on substan-
tiating the claim that useful empirical regularities do not exist ~see, e+g+, Hen-
dry, 1995a!+ Certainly, Dr+ Blank’s nebulous demand function for herrings could
only last the test of time if many factors other than price and income were
included in his model, but such additional factors are not precluded a priori+
Indeed, when Hutchison ~1938! launched his critical attack on Robbins, he made
a similar argument and proposed instead a move toward “positive economics,”
a theme we resume in the next section ~for the background to the general argu-
ments, see, e+g+, Blaug, 1980!+ Nevertheless, since Robbins was one of LSE’s
most eminent economists, the LSE might seem an unpromising location for a
group of econometricians, particularly one with an interest in econometric
methodology+

2.2. Positive Economics


However, the postwar period saw a strong move within LSE economics toward
implementing a “positivist” philosophical standpoint ~for an explanation of the
variants of “positivism” see, e+g+, Caldwell, 1982!+ The LSE itself had a strong
tradition in the philosophy of science centered on Karl Popper, later supported
by others such as Imre Lakatos ~see, e+g+, Popper, 1959, 1963; Lakatos, 1974!+
Thus, “falsification” was much under discussion at the LSE+
In economics, activity focused on the Staff Seminar on Methodology, Mea-
surement and Testing ~M 2 T!, whose members included Chris Archibald, Kurt
Klappholz, and Richard Lipsey, all of whom published on economic methodol-
ogy ~see, e+g+, Archibald, 1959; Klappholz and Agassi, 1959; Lipsey, 1966!+
Here, for example, the famous Phillips curve was first presented and discussed
~Phillips, 1958; see, e+g+, Holt, 1969!+ The appointment of an engineer-economist
like A+W+H+ ~Bill! Phillips, who stressed empirical evidence, signaled that the
mood was less inimical to econometric modeling as the route ahead for eco-
nomics than the heritage of Robbins might suggest+ Indeed, important anteced-
ents were already in place+ Both Bill Phillips and Rex Bergstrom ~also then at
LSE! considered models that were essentially equilibrium-adjustment mecha-
nisms ~see, e+g+, Phillips, 1954, 1957; and also Bergstrom, 1962, 1967; and the
implementation in Bergstrom and Wymer, 1976!+2
In any case, Denis Sargan was first appointed in the Statistics Department
at the LSE in 1963, where Roy Allen played a major role+ Allen admired
Phillips’s earlier work on control theory ~Phillips, 1954, 1957!, and he in-
cluded an explanation of integral, derivative, and proportional control in his
textbook ~Allen, 1963, Ch+ 9!+ This strand also recurs subsequently, especially
in Section 3+2+
460 DAVID F. HENDRY

2.3. The Sargan “Revolution”


Denis himself was a somewhat unlikely candidate to revolutionize econometric
methodology+ His initial research had been on mathematical economics, fol-
lowed by econometric theory, particularly seeking to establish the same form
of rigorous inferential basis for its application to small samples that Student’s
famous t-distribution paper had done for statistics ~see Student, 1908!+ Indeed,
the majority of his publications were to be in what he called “advanced econo-
metric theory” ~see Sargan, 1988b!+
However, prior to his arrival at the LSE in 1963, Denis had written two pa-
pers on empirical modeling issues, published as Sargan ~1957! and Sargan
~1964!+ The former points to his later role in developing methodology; the lat-
ter provided several major advances therein+
Sargan ~1957! was a discussion of the pathbreaking analysis of the Oxford
Savings Survey by Malcolm Fisher ~1956!+3 It was concerned with three issues
that recurred in his later research on methodology: the abstract and constrained
form of economic-theory models relative to the complexities of the data under
analysis; the oversimplified nature of many estimated regression equations, ex-
cluding effects likely to be important; and the problems of interpretation when
testing large numbers of hypotheses+ The first two perhaps led to his interest in
estimating relatively general and unrestricted models and the third to his ideas
about “data mining,” considered in Section 5+
Sargan ~1964!, which was to become perhaps his most famous paper, was
prepared for the Colston Society conference on National Economic Planning
held at Bristol University in 1963 ~for an appraisal of its impact, see, e+g+, De-
sai et al+, 1997!+ The next section considers that paper in more detail, empha-
sizing that almost every aspect of the current “LSE approach” was present+

3. THE “COLSTON PAPER”


In this paper, Denis laid out the conceptual foundations of what has become
the “LSE approach+” The essential elements that he formalized included
~1! the use of “long-run” economic analysis to specify the equilibrium of the model;
~2! the introduction of “equilibrium-correction” mechanisms into behavioral dy-
namic econometric models;
~3! the development of a new interpretation of autoregressive errors in time-series
models;
~4! the construction of valid misspecification tests after estimating dynamic
models;
~5! the use of model comparison procedures for linear against logarithmic
specifications;
~6! the investigation of the impact of data transforms on the selection of models;
~7! a nonlinear in parameters instrumental variables estimator for measurement errors;
~8! the development of operational computer programs to implement the new econo-
metric methods;
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 461

~9! a proof that his iterative computations would converge with near certainty; and
~10! matching the econometric theory to the substantive empirical modeling problem+

Many of these topics arise in subsequent sections, but we first comment on


each of them in turn+

3.1. Long-Run Economic Theory

The substantive problem that Denis addressed was empirical modeling of wages
and prices ~described in more detail in Section 3+10!+ Previous models had con-
sidered the changes in the variables, namely, wage and price inflation+ Con-
sider a conditional bivariate model of a variable denoted yt ~the log of wages,
say! given another variable z t ~the log of prices!, written in first differences as

Dyt 5 b0 1 b1 Dz t 1 b2 Dyt21 1 et + (1)

Assuming that E@et # 5 0 and that the differenced variables are stationary with
means E@Dyt # 5 gy and E@Dz t # 5 gz , then the long-run steady-state solution to
~1! is

b0 1 b1 gz
gy 5 +
1 2 b2

As formulated, ~1! does not establish any relationship between the levels yt and
z t ; hence they could drift apart+ Denis argued that economic agents would be
concerned about the level of real wages, namely, yt 2 z t , not just price infla-
tion+ Thus, he formulated a model with a long-run levels equilibrium, denoted
~ y 2 z!e, t and given by

~ y 2 z!e, t 5 d0 1 d1 Dz t 1 d2' x t , (2)

where x t denotes a vector of additional variables, such as unemployment, real


productivity, and political factors+ A unit coefficient on productivity would make
the wage share a component of the left-hand side, but in practice, productivity
was approximated by a linear deterministic trend in his wage models+
The disequilibrium is given by

vt 5 yt 2 z t 2 d0 2 d1 Dz t 2 d2' x t + (3)

In modern parlance, the levels are I~1! when the differences are I~0!+ Conse-
quently, if such a relation as ~3! did exist, it would require that vt was I~0! and
so entail that yt was cointegrated with z t and x t ~see, among many others, Engle
and Granger, 1987; Phillips and Loretan, 1991; Banerjee, Dolado, Galbraith,
and Hendry, 1993!+
462 DAVID F. HENDRY

3.2. Equilibrium-Correction Mechanisms

Some of the antecedents to the formulation of equilibrium-correction mecha-


nisms were noted in Section 2+2; Section 3+11 comments on the terminology+ If
~2! is to be the long-run equilibrium, then a mechanism is needed in ~1! to
reestablish the relationship between the levels of yt , z t , and x t whenever they
drift apart+ To achieve ~2!, Denis postulated the explicit adjustment equation,

Dyt 5 a~ yt21 2 z t21 2 ~ y 2 z!e, t21 !, (4)

equivalent to including vt21 from ~3! in ~1!+ Doing so unrestrictedly then en-
tails adding the levels terms ~ y 2 z!t21 and x t21 + When contemporaneous vari-
ables are excluded, the resulting equation is

Dyt 5 p0 1 p1 Dz t21 1 p2 Dyt21 2 p3 ~ y 2 z!t21 1 p4' x t21 1 u t + (5)

Consequently, when p3 Þ 0, the long-run levels equilibrium solution to ~5!


becomes

E@ yt 2 z t 2 f1 Dz t 2 f4' x t # 5 f 0 ,

matching ~2!+
The model in ~5! has both derivative and proportional control ~e+g+, Dz t21
and ~ y 2 z!t21 !, following up Phillips ~1954, 1957! ~see Salmon, 1982!+ The
latter mechanism also ensures the disequilibrium adjustment, based on the ~de-
trended! log-ratio of two nominal levels ~see Bergstrom, 1962!+ Although De-
nis formulated ~5! without specifying any specific integration properties for the
data, the long-run relative stability of the “great ratios” in Klein ~1953! was
implicitly assumed to justify such transformations ~see, e+g+, the discussion of
Granger and Newbold, 1977, in Hendry, 1977!+ Granger ~1981! related the type
of model in ~5! to cointegration, and Granger ~1986! showed that one of Dyt or
Dz t must then depend on the equilibrium correction: the assumption in ~4! is
that yt adjusts to the disequilibrium+ If both variables yt and z t adjust to the
disequilibrium, then z t is not weakly exogenous for the $pi % ~see Phillips and
Loretan, 1991; Hendry, 1995b!+ It is primarily because of cointegration that
equilibrium-correction models like ~5! have proved a popular specification+
His empirical model is described in more detail in Section 3+10, but an inter-
esting feature of his two main equations for wages and prices, respectively,
was that they had different long-run solutions+ Thus, although the integrability
of data derives from a reduced-rank condition ~fewer corrections than vari-
ables; see, e+g+, Johansen, 1988!, the existence of equilibrium corrections in his
two main equations by itself did not preclude that the data were I~0! around a
deterministic trend+ Alternatively, the variables could have been I~1! by inher-
iting unit roots from either import prices ~from the price equation! or produc-
tivity ~wage equation!+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 463

3.3. Autoregressive Errors

Prior to Denis’s Colston paper, it was common in econometrics to test for re-
sidual autocorrelation ~e+g+, by the Durbin and Watson statistic; see Durbin and
Watson, 1950, 1951!, and if such was found, to estimate an autoregressive-
error process ~see Cochrane and Orcutt, 1949!+ Thus, autoregressive errors were
viewed as a “generalization” of white-noise errors+ Denis showed that auto-
regressive errors were a restriction on a more general dynamic model that, when
valid, allowed a more parsimonious representation, thereby reversing the con-
ventional interpretation+ Further developments on this topic are taken up in
Section 4+

3.4. Misspecification Tests

Denis stressed that empirical specifications should be stringently evaluated+


Drawing on Sargan ~1961!, he formulated a test for the validity of the instru-
mental variables used in estimation and derived its distribution under the null+
He estimated an autoregressive-error process for each equation by least squares
or instrumental variables+ To evaluate such nonlinear in parameters models, he
developed a test for higher order autoregressive errors based on the residuals
from the estimated equations, and he derived its asymptotic distribution+ Such
a test was, therefore, valid in dynamic equations in general+ Nevertheless, the
Durbin–Watson test—which did not follow its tabulated distribution if applied
to such models—continued to be widely used for many years, although it would
often fail to reveal significant misspecification ~see, e+g+, Nerlove and Wallis,
1966!+

3.5. Model Evaluation

Denis also developed a criterion for choice of functional form between log and
linear specifications+ In essence, he showed that one should simply convert the
respective equation standard errors to be proportions of the level of the depen-
dent variable and then select the smaller+ The log model automatically had an
equation standard error of that form, but the levels equation standard error had
to be scaled by the geometric mean of yt +

3.6. Data Transforms

Previous models of wages given prices, such as Klein, Ball, Hazlewood, and
Vandome ~1961!, modeled annual changes and introduced moving averages of
past inflation and unemployment+ Denis suggested that doing so would induce
autocorrelated errors and would distort measures such as equation standard
errors and R 2+ Consequently, he argued for selecting the data transformations
464 DAVID F. HENDRY

that led to an equation with “errors which are independent in different time
periods+”

3.7. Instrumental Variables


The Colston paper commenced with a discussion of appropriate estimation meth-
ods for dynamic equations with autoregressive errors+ The analysis was based
on Sargan ~1959, 1961! and also described the nonlinear in parameters instru-
mental variables estimator, which allowed for possible data measurement er-
rors and autoregressive residuals+

3.8. Computer Implementation


Despite the somewhat primitive nature of the available computing facilities,
the development of an operational program to calculate the new econometric
methods was high on Denis’s agenda+ I learned later from discussions with him
how much effort was required in the programming, involving physically re-
wiring the computer, to ensure that the answers were correct+ Denis carefully
addressed the logic of the calculations to embed both estimators in a common
framework and to ensure as efficient an iterative procedure as possible, includ-
ing selection of the initial values and step lengths+ He tried both stepwise and
Newton–Raphson optimizers, analyzed their speeds of convergence—within and
between iterations—and also investigated possible multiple optima+ Overall, he
set a high standard of scholarship for the treatment of the computing aspect of
empirical research, in marked contrast to many later investigators, who do not
even cite the software they use+

3.9. Convergence of the Iterations


Denis provided a proof that stepwise iterative computations would converge
with near certainty to a local optimum when the criterion function was bounded
above and every step ensured that the function did not decrease+ Such a proof
includes, for example, the Cochrane and Orcutt ~1949! procedure—although
Denis noted that its convergence could be slow in dynamic equations—but does
not include many stepwise iterations that seek a “fix-point” solution rather than
the optimum of a criterion function+

3.10. The Empirical Wage-Price Model


Having developed the dynamic models, estimators, tests, algorithms, and com-
puter programs, Denis applied this new analytical apparatus to the near-
intractable problem of 1960s UK wage-price inflation+ The main data set he
analyzed was an extension up to 1960~4! of the quarterly seasonally unadjusted
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 465

time series of Klein et al+ ~1961! over 1948~1!–1957~4!+ His wage equation
was formulated in terms of real wages—as described in Sections 3+1 and 3+2—
unlike, say, the Phillips curve+ Denis used a data-based proxy for “inflation
expectations,” which he called “an extrapolation of past price movements into
the future+” His analysis highlighted the role of real-wage resistance in wage
bargains, interpreting the equilibrium correction—the deviation of real wages
from a productivity trend—as a “catch-up” mechanism for recouping losses in-
curred from unanticipated inflation+ Permanent and transitory effects were care-
fully distinguished in his policy discussion to ascertain which changes would
fade out ~such as devaluations!+ But first, Denis checked that his model ade-
quately described the evidence by testing for various misspecifications against
the hypothesis that the residuals were white noise and independent of the in-
struments, as discussed in Sections 3+3–3+6+
To illustrate his general findings, it seemed of interest to estimate, test, and
forecast with analogous equations on the closely related annual data docu-
mented in Hendry ~2001!+ Figure 1 records these annual time series for the
period 1945–1965+ As can be seen in the top-left panel, wage inflation was
above 5% per annum on average and higher than price inflation, and the top-
right panel confirms that real wages ~w 2 p! rose substantially over the period+
GDP growth was also high after 1947 ~bottom-left panel!, and correspond-
ingly, the unemployment rate ~Ur !was very low, particularly by recent stan-
dards ~bottom-right panel!+ In this notation, lowercase letters denote logs+

Figure 1. Inflation, real wages, GDP growth, and unemployment in the UK+
466 DAVID F. HENDRY

Using the annual data shown in Figure 1, an equation of the following form
can be obtained over 1946–1965 ~after lagging!:

Z t 5 20+78 1 0+38Dwt21 1 0+69Dpt 2 0+56Dpt21


Dw
~0+20! ~0+16! ~0+13! ~0+14!

2 0+46 ~w 2 p!t21 1 0+006 t,


~0+10! ~0+0016!
R 2 5 0+76 s[ 5 0+0114 SC 5 28+40 xnd
2
~2! 5 0+34,
Far~2,12! 5 0+17 Farch~1,12! 5 0+47
Fhet~10,3! 5 0+50 Freset~1,13! 5 0+00+ (6)
In ~6!, R 2 is the squared coefficient of multiple correlation, s[ is the residual
standard deviation, coefficient standard errors are shown in parentheses, and SC
is the Schwarz ~1978! criterion+ The diagnostic tests are of the form Fj~k, T 2 l !,
which denotes an approximate F-test against the alternative hypothesis j for
kth-order serial correlation ~Far; see Godfrey, 1978!, kth-order autoregressive
conditional heteroskedasticity ~Farch; see Engle, 1982!, heteroskedasticity ~Fhet;
see White, 1980!; the RESET test ~Freset; see Ramsey, 1969!; and a chi-square
test for normality ~ xnd
2
~2!; see Doornik and Hansen, 1994!+ None of these di-
agnostic tests is significant+
Relative to equation ~17! in Sargan ~1964!, price inflation plays a more im-
portant role in ~6! but is close to a second difference ~surprise inflation!+ How-
ever, unemployment is insignificant, with a small positive coefficient if entered,
and so it was omitted+ The equilibrium solution for real wages has an annual
trend ~approximating productivity growth! of about 1+4% p+a+, which is close
to Denis’s finding+ The feedback coefficient is important in ~6! and clearly worth
retaining given the small sample+ The overall fit at s[ 5 0+011 is poorer than
Denis reported ~ sI 5 0+006, not adjusted for degrees of freedom!, suggesting
that intrayear information is useful, a finding that is consistent with different
groups of workers agreeing on their annual wage settlements in different quarters+
Finally, expressing the equilibrium correction as
~w 2 p!e, t 5 ~w 2 p!t 1 1+71 2 0+014t,
and transforming the dependent variable to real wages, leads to the more par-
simonious representation

D~w Z 2 p!t 5 0+48 D~w 2 p!t21 2 0+37Dpt 2 0+44~w 2 p!e, t21 ,


~0+09! ~0+08! ~0+06!

s[ 5 1+1% R 5 0+82
2
SC 5 28+75
Far~2,15! 5 0+30 Farch~1,15! 5 0+06,
Fhet~6,10! 5 2+17 Freset~1,16! 5 4+04 xnd
2
~2! 5 0+13+ (7)
In ~7!, there is “short-run erosion” of real wages from price inflation, so an
“inflation surprise” formulation is in fact rejected+ However, the price inflation
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 467

effect was probably hard to notice at the inflation rates prevalent over most of
the sample against the annual growth of 1+5% from productivity gains+ The
disequilibrium feedback automatically offsets some of that erosion+ In any case,
the coefficient of Dpt is probably not constant in the face of persistent high
inflation but should decline to zero as inflation increases+ Indeed, the forecast
performance of ~7! over the next decade of high inflation is very poor, even for
one-step forecasts: testing parameter constancy to 1975 yields FChow~10,17! 5
8+07 ** ~see Chow, 1960!+ In fact, most of this forecast failure is due to location
shifts, as the equation in differences ~i+e+, dropping ~w 2 p!e, t21 ! delivers
FChow~10,18! 5 0+85+ At the time, and for many years afterward, it was not
known just how susceptible “equilibrium-correction models” were to shifts in
the coefficients of deterministic terms: changes to the intercept and0or trend
have a pernicious effect on their performance ~see, e+g+, Clements and Hendry,
1999!+ Income policies, wage freezes, and other related governmental interven-
tions induce such shifts, so it is unsurprising that the original specification fails
to characterize the changed process+ We resume this aspect of the story in
Section 5+

3.11. Retrospect
It is always difficult to understand why insights are so slow in coming+ The
argument from long-run homogeneity by which Denis justified the equilibrium-
correction term did not immediately prompt analogies or other applications,
perhaps because the Colston volume was an obscure source for economists—
and Denis was certainly not a propagandist+ Even closer to home, Hendry and
Anderson ~1977! did not notice that their theoretical derivation of the equilib-
rium behavior of building societies was an exemplar+ Indeed, it was not till
after rediscovering such feedback in Davidson, Hendry, Srba, and Yeo ~1978!
as “the vital ‘initial disequilibrium’ effect” that those authors realized they had
an example of Denis’s formulation ~see, e+g+, Hendry, @1993# 2000!+ Hendry
and von Ungern-Sternberg ~1981! named the class “error-correction” following
Phillips ~1957!, but its limitation to adjustment within equilibria ~and not be-
tween! was not foreseen+4
Starting from a general specification was at best implicit in the formulation
of the “transformed” model against which the autoregressive error form was to
be tested+ In more recent parlance, that restriction corresponded to a common
factor in the lag polynomials characterizing the dynamics of the unrestricted
model ~see Section 4!+ However, Denis’s “practical methodology” still in-
cluded trying a variety of specifications of variables, lag lengths, indicator vari-
ables, and data transformations, probably because the sample was too small to
sustain useful estimates of the nesting model and computer power was then
very limited+ Like many empirical investigators, Denis scrutinized the internal
consistency of his results and the relationships between them, an aspect even-
tually formalized by encompassing+
468 DAVID F. HENDRY

It also remained unclear how to generalize the treatment of autoregressive


errors to higher orders—the theme of the next section—perhaps because the
unrestricted model was derived, rather than postulated initially+ Finally, although
a variety of tests for model misspecification was developed and implemented,
it was not obvious whether these provided a comprehensive evaluation—or were
even well behaved ~see, e+g+, Kiviet, 1986!; Section 7 provides an update+ Thus,
many of the ingredients of the “LSE methodology” were in place, although a
great deal of research still remained to be done+

4. STOCHASTIC AND SYSTEMATIC DYNAMICS

Much of Denis’s analysis in the Colston paper concerned the role of autocorre-
lated errors, which he regarded as likely in models estimated from aggregate
quarterly data+ Such concerns were prominent in the interests of his colleague
Bill Phillips ~see, e+g+, Leeson, 2000! and were also reflected in the research
of his doctoral students ~see, e+g+, Trivedi, 1970; Hendry and Trivedi, 1972;
Hendry, 1974!+ Both autoregressive and moving-average error processes were
of interest, as were vector generalizations thereof+ Two issues were involved:
selecting the longest lag needed to induce white-noise errors; and “allocating”
the dynamics between systematic components ~which affected the mean lag!
and error dynamics ~which did not!+ The Colston paper had addressed both but
essentially in the context of first-order autoregressive errors+

4.1. Vector Generalizations

Sargan ~1959, 1961! discussed estimating systems of dynamic equations with


vector autoregressive errors+ When the system was closed, there was an identi-
fication problem in extracting the error representation from the endogenous dy-
namics+ Consider the n 3 1 system in the endogenous variable y t :

y t 5 Pyt21 1 u t where u t 5 Ru t21 1 et

with et ; INn @0, V# + The “solved” form is

y t 5 ~P 1 R!y t21 2 RPyt22 1 et 5 C1 y t21 1 C2 y t22 1 et +

Then, P and R cannot be uniquely separated from information on C1 and C2 ,


as they correspond to the roots—which can be arranged in many ways+ Notice
that an unrestricted vector autoregression ~VAR! was, therefore, considered as
the general model+5 Denis had not solved the closed system case at that time ~it
was the first topic of this author’s thesis research; see Hendry, 1970!; however,
Denis clearly knew that sufficient additional “exogenous” variables would al-
low unique identification, and he guided me accordingly+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 469

4.2. Common Factors in Lag Polynomials


In 1975, Denis generalized the estimation of scalar dynamic equations with first-
order autoregressive errors to higher order errors on equations with longer lags+
Espasa ~1993! describes the circumstances surrounding this development+ De-
nis read widely, so he almost certainly knew the results in Wald ~1947! and
also in Anderson ~1962! on the optimality of a general-to-simple approach for
an ordered sequence of nested hypotheses; he cited the analysis in Anderson
~1971! supporting Gets for selecting lag length+
A maximum lag length must first be carefully chosen to reflect the context
~e+g+, data frequency!, data properties, and economic considerations, perhaps
restricted by a lack of degrees of freedom+ Once that has been done, consider
the least restricted autoregressive model:
k
yt 5 ( ui yt21 1 et +
i51

The Anderson procedure commenced by first testing uk 5 0, and if that was not
rejected, uk21 5 0, and so on, assuming that if um 5 0, then ui 5 0 for all i . m+
The reduction, therefore, ceased with the first rejection+
An autoregressive-distributed lag equation with a maximum lag length of k
and n “exogenous” regressors can be written as
k n k

( b0, j yt2j 5 g 1 (
j51
( bi, j x i, t2j 1 et +
i51 j50
(8)

Similar considerations could be applied for some aspects of the problem, such
as lag length+ Denis noted that such an approach also applied to selecting
the maximal common factor+ Writing ~8! in an obvious notation using lag poly-
nomials as
n
b0 ~L!yt 5 g 1 ( bi ~L!x i, t 1 et , (9)
i51

then an autoregressive error of order r # k can be extracted if ~9! can be ex-


pressed as
n
b0* ~L!r~L!yt 5 g 1 ( bi* ~L!r~L!x i, t 1 et , (10)
i51

where all the bi* ~L! ~i 5 0, + + + , n! are now polynomials of maximum order
k 2 r $ 0+ Since
r
r~L! 5 ) ~1 2 l j L!,
j51
470 DAVID F. HENDRY

the terms ~1 2 l j L! are the common factors in the original lag polynomials
bi ~L! ~i 5 0, + + + , n!; hence the name COMFAC+ Thus, an ordered sequence of
hypotheses is postulated, where the existence of a common factor of order m
~for m # r! clearly entails the existence of all orders m 2 l for l . 0+ Con-
versely, rejecting order m # r entails rejecting all higher orders+
The central achievement was developing an operational algorithm to deter-
mine the largest r that could be extracted, taking account of the possibility of
complex roots in the polynomials ~see Sargan and Sylwestrowicz, 1976!+ The
resulting formulation was published as Sargan ~1980!, although an exposition
appeared in advance in Hendry and Mizon ~1978!+
The most obvious feature of the COMFAC viewpoint is that the initial model
~9! must be the most general needed to characterize the dynamics+ This is es-
sential, both to allow extraction of the appropriate factors and so that conven-
tional inference procedures will apply validly+ Mizon ~1977b! ~then at the LSE!
explicitly investigated model simplification routes for testing the dynamic spec-
ification of an econometric model+ He considered multipath searches and the
nonordered and nonsequential nature of many of the hypotheses to be tested+
He stressed the need to commence from the minimally most-general specifica-
tion that embedded the theory-based model+ Thus, Gets was becoming the norm
at LSE+

5. “DATA MINING”
Sargan ~1957! was noted previously as one precursor to Denis’s interest in the
problem of interpreting large numbers of hypothesis tests+ There are three as-
pects to his comments:
~1! the large number of coefficients being compared;
~2! the choice of critical values ~or significance levels! for each individual test; and
~3! the role of nonnormal distributions+

These interacted to make him remark that


Having regard to the non-normality of the distributions of the variables and to the
large number of regression coefficients that are being compared there is a proba-
bility of order 1 per cent that one or more of these coefficients should differ from
their population values by more than four times their standard error+ ~p+ 177!

Which nonnormal distribution Denis had in mind is not specified; presumably


one with fatter tails than the normal+ A simple approximation that remains sym-
metric but with potentially much fatter tails is a t-distribution with few degrees
of freedom+ For example:
P~6t10 6 $ 4! . 0+0025
satisfies his statement when even four tests are undertaken ~t5 would do so with
one test!+ By way of contrast, the corresponding probability for a normal is
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 471

0+0001, so about 100 tests would be needed to have a probability of 1% of


observing a value in excess of 4+
Denis did not explicitly return to the problem of data mining until Sargan
~1973!+6 First, he noted that forecast failure does not uniquely entail previous
“data mining” because ~e+g+! the stochastic behavior of the variables might have
changed+ As shown previously, such a fate awaited the annual analogue of his
wage equation+ He emphasized possible changes in the data measurement sys-
tem and commented on both “optimal” and second-best methods for achieving
a constant representation+
Second, he considered the possibility that the rejection is due to “over-
fitting” or “data mining”—to quote:
By conducting numerous tests, all at a conventional 5% confidence level, the econ-
ometrician may be biasing his predictions, or alternatively understating the stan-
dard errors of his predictions+ If the number of parameters estimated is of the
same order of magnitude as the sample size, there is indeed a considerable prob-
lem if the distribution of the prediction errors is derived from asymptotic sam-
pling theory+ Within the limitations imposed by this, it would seem better to
formulate as general a model as is required to embrace all possible versions of
the relevant economic theories that the economist is prepared to consider+ ~p+ 160!

Personally, I imagine Denis in the LSE coffee bar, sketching his analysis on the
proverbial back of an envelope—students have recovered complete lectures
sketched that way—with most of the following steps missing+ Here they are
included, mainly to define notation+ Let
y 5 Xb 1 u where u ; INT @0, su2 I# ,
where there are n regressors; then
bZ 5 ~X ' X!21 X ' y Z 5 T 21 su2 ~T 21 X ' X!21+
with V@ b#
Let y[ T11 5 bZ ' x T11 with u[ T11 5 yT11 2 y[ T11 + Then

V@ u[ T11 # 5 su2 ~1 1 T 21 x 'T11 ~T 21 X ' X!21 x T11 ! r su2 +


Tr`

However, he probably began ~and ended!! by using an approximation like


T 21 x 'T11 ~T 21 X ' X!21 x T11 5 T 21 tr ~~T 21 X ' X!21 x T11 x 'T11 ! . T 21 tr~I n !
n
5 5 q+ (11)
T
Thus, the asymptotic formula su2 understates the standard error of prediction
by 100q%+ One suspects he had a value like q 5 0+5 in mind+ Nevertheless, he
still argued for commencing from the general model+
Denis then proceeded to discuss the “repeated testing” problem, noting the
result in Dunn ~1959! that the impact of chance significance in repeated testing
from a t-distribution with ~say! 120 degrees of freedom and 50 additional irrel-
472 DAVID F. HENDRY

evant variables could be controlled at 5% by setting a critical value of 3+38


~roughly the 0+1% level for each test, as ~1 2 0+001! 50 . 0+95!+ Denis appears
to be trying to design an “optimal procedure” for estimating a model when a
subset of variables may not matter+ He proposes a two-step procedure in which
a pretest estimator is first calculated but the decision on whether to accept it is
based on a loss function+ He notes that the estimator associated with ~say! Dunn’s
critical values need not minimize a quadratic loss function+ He also considers a
Bayesian approach with informative priors, but ~from a missing appendix! he
appears to conclude that the resulting estimators may be highly affected by
changes in the prior distributions+ He relates decisions to possible loss func-
tions that might sustain alternative approaches+ The issue of model size is
noted—this comment suggests that Denis was already at work on Sargan ~1975!+
Referring back to Sargan ~1957!, the suggestion in Dunn ~1959! necessitated
that the errors on the fitted model were indeed approximately normal; as seen
above, if fat tails were common, then so would be larger “t”-values ~as they
would, of course, not be drawings from the Student t-distribution with the an-
ticipated degrees of freedom!+ This argues in favor of transformations that al-
low normality to be a useful approximation, so that a “tight-tailed” distribution
can be used, such that reasonable critical values entail stringent significance
levels; that was certainly part of the oral tradition among LSE econometricians+

6. MODEL SELECTION
As noted previously, Sargan ~1964! viewed autoregressive errors as a restric-
tion on a more general dynamic equation, but he nevertheless estimated a con-
siderable number of separate specifications+ He carefully considered the model
choice between log and linear representations, and between systematic and er-
ror dynamics, but did not formally address other selection issues+ However, his
continued interest in model selection was reflected in Grayham Mizon’s thesis,
some of which was published as Mizon ~1974, 1977a!+
Denis presented an overview paper on model selection to the LSE econo-
metrics workshop in 1981 ~Sargan, 1981! that discussed model-selection cri-
teria, focusing on that proposed by Schwarz ~1978!+ The use of V@ u[ T11 # .
su2 ~1 1 q! deriving from ~11! is reminiscent of the Akaike final prediction er-
ror criterion ~see, e+g+, Akaike, 1969!, “penalizing” each additional parameter+
Another criterion also bearing a close resemblance is the ~probably oldest! model
selection criterion, namely, the unbiased residual variance criterion that Theil
~1961! proposed—so that aspect was well understood+ Denis’s decision to in-
vestigate this issue even when decisions could not be ordered may have de-
rived from Hannan and Quinn’s ~1979! ability to establish that the consistency
conditions for selection rules applied to autoregressions, although there is no
explicit reference to their paper+ Denis shows that the Schwarz criterion is not
the weakest that ensures a consistent model selection, and his analysis finds no
reasons for setting the constant of proportionality ~c in Hannan and Quinn, 1979!
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 473

at unity+ Moreover, he references neither of the closely related papers on this


issue by Shibata ~1981! and Atkinson ~1981!, which reached similar conclusions+7

7. LATER DEVELOPMENTS IN Gets MODELING


Early research on Gets sought to establish optimality properties of selection
procedures in structured situations, such as ordered sequences of nested hypoth-
eses ~see, e+g+, Anderson, 1962, 1971!+ This was generalized to autoregressive-
error extraction in COMFAC, but as Mizon ~1977b! stressed, the optimality of
simplification approaches could not be shown for unordered sequences+ Equally,
the nonviability in most settings of ~say! simple to general modeling was clear,
since it could fail even in cases where Gets was optimal+
The theory of reduction, elucidating the information losses that could arise
by ~e+g+! conditioning and marginalizing, was being carefully investigated by a
group at CORE, Louvain-la-Neuve, Belgium ~see, inter alia, Florens, Mouchart,
and Richard, 1974, 1976; Florens and Mouchart, 1980, 1985, consolidated in
Florens, Mouchart, and Rolin, 1990!+ Jean-François Richard ~from CORE! vis-
ited LSE in 1973–1974 and added these ideas to the developing “LSE method-
ology+” The integrated strand that evolved led to a series of papers on exogeneity
~Richard, 1980; Engle, Hendry, and Richard, 1983!, model formulation and a
taxonomy of evaluation information ~Hendry and Richard, 1982, 1983!, and
the theory of reduction as explaining the origin of empirical models ~Hendry,
1987!+ Another visitor to both institutions, Adrian Pagan, who had earlier also
worked on estimating models with autocorrelated errors ~see, e+g+, Pagan and
Nicholls, 1976! and model evaluation, contributed to formalizing the model types
and their properties ~see Hendry, Pagan, and Sargan, 1984!+ Useful reviews of
this methodology prior to the implementation of computer-search methods are
provided by Gilbert ~1986! and Ericsson, Campos, and Tran ~1990!; more re-
cently, see Mizon ~1995!+
However, the computer automation of simplification procedures did not orig-
inate at LSE, perhaps because the approach in Coen, Gomme, and Kendall ~1969!
was not received very enthusiastically+ To my knowledge no one at the LSE
attempted to develop an automated Gets approach, although the complete im-
plementation of COMFAC did represent a special case+8
Two strands of research have since established the fruitfulness of automatic
selection+ Following up on Phillips ~1994!, Phillips ~1996! enunciated the prin-
ciples for an automated approach to model reduction with evaluative testing
that allows for parameter evolution and ~e+g+! reduced rank ~equilibrium cor-
rection! in systems, in addition to single equation models+ His approach has
been successfully implemented in empirical models for forecasting the U+S+ and
various Asia–Pacific economies ~see e+g+, Phillips, 1995; Schiff and Phillips,
2002, a recent overview is provided in Phillips, 2002!+
The other stimulus was the work of Hoover and Perez ~1999!, who demon-
strated the properties of their search algorithm by Monte Carlo simulation+ Ex-
474 DAVID F. HENDRY

tensions of their approach are discussed in Hendry and Krolzig ~1999, 2001,
2002! and Hoover and Perez ~2000!+ The main finding of these Monte Carlo
studies is that the null rejection frequency of the Gets simplification procedure
is close to the nominal size selected, and the power is close to that obtained
when commencing from the correct specification with no irrelevant variables+
Whether the selected model is too large or too small depends on the critical
values set by the user and the nature of the correct model, but neither is intrin-
sic to Gets+ The estimated fit ~equation standard error! is approximately un-
biased ~within 65% depending on the nominal size selected!+
One of the issues raised in Sargan ~1957! was the interpretation of models
after many hypotheses had been tested, representing Denis’s worry about the
adventitious significance of coefficients+ Formally, the overall “size” of a re-
peated testing exercise with n independent tests each at significance level p
when the null is valid is 1 2 ~1 2 p! n 5 pn , which rapidly escalates in n and p+
A well-used example is p 5 0+01 and n 5 40 leading to pn 5 0+33+ This seems
an excessive size, especially relative to the simple expedient of first testing the
overall null of no effect before conducting individual tests+ Alternatively, from
the perspective of selecting a model, a more useful interpretation may be the
average number of irrelevant variables retained from the large candidate set
initially considered, namely, p 3 n+ For our example, p 3 n 5 0+40, interpreted
as one variable being retained on two occasions out of five that the general
model contains 40 irrelevant regressors, with none retained on the remainder+
Thus, dramatic reductions are achieved in both cases+ In view of the potentially
large costs of omitting relevant variables, Denis’s advocacy of Gets seems
supported+

8. CONCLUSION

Given his quiet, unassuming approach, and his location at what had earlier been
a central location for the denigration of econometric evidence, Denis Sargan at
the LSE was an unlikely candidate to radically alter the econometric approach
of a generation+ Yet without doubt he did so, establishing a powerful approach
to empirical modeling of economic time series+ This paper has outlined the main
developments in econometric methodology that he both initiated and under-
took, and also some of the subsequent research+ But, as the other papers in this
volume demonstrate, that represents merely one of his many research strands+
Denis has left a splendid legacy of intellectual achievement, in addition to co-
horts of well-trained students, many of whom have continued to advance the
discipline across, and beyond, the range of topics on which he left his mark+

NOTES

1+ It draws on material in Hendry and Wallis ~1984!, Phillips ~1985!, Desai et al+ ~1997!,
Ericsson, Maasoumi, and Mizon ~2001!, and Hendry and Pesaran ~2001!+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 475

2+ Phillips and Loretan ~1991! discuss these precursors, and Leeson ~2000! provides a detailed
analysis of Bill Phillips’s contributions, where many of his papers are reprinted+
3+ Fisher’s paper was also discussed by Franco Modigliani, Albert Ando, Milton Friedman, Trygve
Haavelmo, Lawrence Klein, and James Tobin+
4+ Ken Wallis has pointed out that another precursor is Marglin ~1975!, who called his con-
sumption function a “disequilibrium hypothesis+” As noted earlier, in Rex Bergstrom’s empirical
disequilibrium models of economic growth, most of the equations also involved adjustment of an
“equilibrium-correction” form around a long-run steady-state growth path, but this author for one
did not perceive the generality of the formulation+
5+ Antecedents again occur much earlier ~see, e+g+, Phillips and Quenouille, 1960!+
6+ My mimeo copy from March 1973 is actually entitled “Model Building and Data Using,” an
interesting difference of emphasis+
7+ Timo Teräsvirta pointed out that the Schwarz criterion was independently derived by Ris-
sanen ~1978! using coding theory, but it is unlikely that Denis had seen that paper in 1981 because
Rissanen published in Automatica+
8+ Ken Wallis reminded me that Mike Garside worked at LSE on developing efficient algo-
rithms for computing all possible 2 n 2 1 regressions+

REFERENCES

Akaike, H+ ~1969! Fitting autoregressive models for prediction+ Annals of the Institute of Statistical
Mathematics 21, 243–247+
Allen, R+G+D+ ~1963! Mathematical Economics, 2nd ed+ London: Macmillan+
Anderson, T+W+ ~1962! The choice of the degree of a polynomial regression as a multiple-decision
problem+ Annals of Mathematical Statistics 33, 255–265+
Anderson, T+W+ ~1971! The Statistical Analysis of Time Series+ New York: Wiley+
Archibald, G+C+ ~1959! The state of economic science+ British Journal for the Philosophy of Sci-
ence 10, 58– 69+
Atkinson, A+C+ ~1981! Likelihood ratios, posterior odds, and information criteria+ Journal of Econo-
metrics 16, 15–20+
Banerjee, A+, J+J+ Dolado, J+W+ Galbraith, & D+F+ Hendry ~1993! Co-integration, Error Correction,
and the Econometric Analysis of Non-Stationary Data+ Oxford: Oxford University Press+
Bergstrom, A+R+ ~1962! A model of technical progress: The production function and cyclical growth+
Economica 29, 357–370+ Reprinted in A+R+ Bergstrom, Continuous Time Modelling, pp+ 23–37
~Oxford: Oxford University Press, 1990!+
Bergstrom, A+R+ ~1967! The Construction and Use of Economic Models+ London: English Univer-
sities Press+
Bergstrom, A+R+, & C+R+ Wymer ~1976! A model of disequilibrium neoclassical growth and its
application to the United Kingdom+ In A+R+ Bergstrom ~ed+!, Statistical Inference in Continuous
Time Econometric Models, pp+ 267–327+ Amsterdam: North-Holland+
Blaug, M+ ~1980! The Methodology of Economics+ Cambridge: Cambridge University Press+
Caldwell, B+ ~1982! Beyond Positivism: Economic Methodology in the Twentieth Century+ London:
Allen and Unwin+
Chow, G+C+ ~1960! Tests of equality between sets of coefficients in two linear regressions+ Econo-
metrica 28, 591– 605+
Clements, M+P+, & D+F+ Hendry ~1999! Forecasting Non-stationary Economic Time Series+ Cam-
bridge, Massachusetts: MIT Press+
Cochrane, D+, & G+H+ Orcutt ~1949! Application of least squares regression to relationships con-
taining auto-correlated error terms+ Journal of the American Statistical Association 44, 32– 61+
Coen, P+G+, E+D+ Gomme, & M+G+ Kendall ~1969! Lagged relationships in economic forecasting+
Journal of the Royal Statistical Society, Series A 132, 133–163+ Reprinted in T+C+ Mills ~ed+!,
Economic Forecasting ~Edward Elgar, 1999!+
476 DAVID F. HENDRY

Davidson, J+E+H+, D+F+ Hendry, F+ Srba, & J+S+ Yeo ~1978! Econometric modelling of the aggregate
time-series relationship between consumers’ expenditure and income in the United Kingdom+
Economic Journal 88, 661– 692+ Reprinted in D+F+ Hendry, Econometrics: Alchemy or Science?
~Oxford: Blackwell Publishers, 1993; Oxford: Oxford University Press, 2000!+
Desai, M+J+, D+F+ Hendry, & G+E+ Mizon ~1997! John Denis Sargan: An obituary+ Economic Jour-
nal 107, 1121–1125+
Doornik, J+A+, & H+ Hansen ~1994! A Practical Test for Univariate and Multivariate Normality+
Discussion paper, Nuffield College+
Dunn, O+J+ ~1959! Confidence intervals for the means of dependent, normally distributed variables+
Journal of the American Statistical Association 54, 613– 621+
Durbin, J+, & G+S+ Watson ~1950! Testing for serial correlation in least squares regression, part I+
Biometrika 37, 409– 428+
Durbin, J+, & G+S+ Watson ~1951! Testing for serial correlation in least squares regression, part II+
Biometrika 38, 159–178+
Engle, R+F+ ~1982! Autoregressive conditional heteroscedasticity, with estimates of the variance of
United Kingdom inflation+ Econometrica 50, 987–1007+
Engle, R+F+, & C+W+J+ Granger ~1987! Cointegration and error correction: Representation, estima-
tion, and testing+ Econometrica 55, 251–276+
Engle, R+F+, D+F+ Hendry, & J+-F+ Richard ~1983! Exogeneity+ Econometrica 51, 277–304+ Re-
printed in D+F+ Hendry, Econometrics: Alchemy or Science? ~Oxford: Blackwell Publishers,
1993; Oxford: Oxford University Press, 2000!; and in N+R+ Ericsson & J+S+ Irons ~eds+!, Testing
Exogeneity ~Oxford: Oxford University Press, 1994!+
Epstein, R+ ~1987! A History of Econometrics+ Amsterdam: North-Holland+
Ericsson, N+R+, J+ Campos, & H+-A+ Tran ~1990! PC-GIVE and David Hendry’s econometric meth-
odology+ Revista de Econometria 10, 7–117+
Ericsson, N+R+, E+ Maasoumi, & G+E+ Mizon ~2001! A retrospective on J+ Denis Sargan+ Econo-
metric Reviews 20, 133–157+
Espasa, A+ ~1993! Laudatio on the occasion of the investiture of Professor John Denis Sargan with
the Degree of Doctor Honoris Causa+ Mimeo, Universidad Carlos III, Madrid+
Fisher, M+R+ ~1956! Exploration in saving behaviour+ Oxford Bulletin of Economics and Statistics
18, 201–278+
Florens, J+-P+, & M+ Mouchart ~1980! Initial and Sequential Reduction of Bayesian Experiments+
Discussion paper 8015, CORE, Louvain-La-Neuve, Belgium+
Florens, J+-P+, & M+ Mouchart ~1985! Conditioning in dynamic models+ Journal of Time Series
Analysis 6, 15–34+
Florens, J+-P+, M+ Mouchart, & J+-F+ Richard ~1974! Bayesian inference in errors-in-variables
models+ Journal of Multivariate Analysis 4, 419– 452+
Florens, J+-P+, M+ Mouchart, & J+-F+ Richard ~1976! Likelihood Analysis of Linear Models+ Dis-
cussion paper 7619, CORE, Louvain-La-Neuve, Belgium+
Florens, J+-P+, M+ Mouchart, & J+-M+ Rolin ~1990! Elements of Bayesian Statistics+ New York: Mar-
cel Dekker+
Gilbert, C+L+ ~1986! Professor Hendry’s econometric methodology+ Oxford Bulletin of Economics
and Statistics 48, 283–307+ Reprinted in C+W+J+ Granger ~ed+!, Modelling Economic Series
~Oxford: Clarendon Press, 1990!+
Godfrey, L+G+ ~1978! Testing for higher order serial correlation in regression equations when the
regressors include lagged dependent variables+ Econometrica 46, 1303–1313+
Granger, C+W+J+ ~1981! Some properties of time series data and their use in econometric model
specification+ Journal of Econometrics 16, 121–130+
Granger, C+W+J+ ~1986! Developments in the study of cointegrated economic variables+ Oxford Bul-
letin of Economics and Statistics 48, 213–228+
Granger, C+W+J+, & P+ Newbold ~1977! The time series approach to econometric model building+ In
C+A+ Sims ~ed+!, New Methods in Business Cycle Research, pp+ 7–21+ Minneapolis: Federal Re-
serve Bank of Minneapolis+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 477

Hannan, E+J+, & B+G+ Quinn ~1979! The determination of the order of an autoregression+ Journal of
the Royal Statistical Society, Series B 41, 190–195+
Hayek, F+A+ ~1963! Studies in Philosophy, Politics, and Economics+ London: Routledge and Kegan
Paul+
Hendry, D+F+ ~1970! The Estimation of Economic Models with Autoregressive Errors+ Ph+D+ Thesis,
London University+
Hendry, D+F+ ~1974! Stochastic specification in an aggregate demand model of the United King-
dom+ Econometrica 42, 559–578+ Reprinted in D+F+ Hendry, Econometrics: Alchemy or Science?
~Oxford: Blackwell Publishers, 1993; Oxford: Oxford University Press, 2000!+
Hendry, D+F+ ~1977! On the time series approach to econometric model building+ In C+A+ Sims
~ed+!, New Methods in Business Cycle Research, pp+ 183–202+ Minneapolis: Federal Reserve
Bank of Minneapolis+ Reprinted in D+F+ Hendry, Econometrics: Alchemy or Science? ~Oxford:
Blackwell Publishers, 1993; Oxford: Oxford University Press, 2000!+
Hendry, D+F+ ~1987! Econometric methodology: A personal perspective+ In T+F+ Bewley ~ed+!,
Advances in Econometrics, Ch+ 10+ Cambridge: Cambridge University Press+
Hendry, D+F+ @1993# ~2000! Econometrics: Alchemy or Science? Oxford: Blackwell Publishers+
Reprint, Oxford: Oxford University Press+
Hendry, D+F+ ~1995a! Econometrics and business cycle empirics+ Economic Journal 105, 1622–
1636+
Hendry, D+F+ ~1995b! On the interactions of unit roots and exogeneity+ Econometric Reviews 14,
383– 419+
Hendry, D+F+ ~2001! Modelling UK inflation, 1875–1991+ Journal of Applied Econometrics 16,
255–275+
Hendry, D+F+, & G+J+ Anderson ~1977! Testing dynamic specification in small simultaneous sys-
tems: An application to a model of building society behaviour in the United Kingdom+ In M+D+
Intriligator ~ed+!, Frontiers in Quantitative Economics, vol+ 3, pp+ 361–383+ Amsterdam: North-
Holland+ Reprinted in D+F+ Hendry, Econometrics: Alchemy or Science? ~Oxford: Blackwell Pub-
lishers, 1993; Oxford: Oxford University Press, 2000!+
Hendry, D+F+, & H+-M+ Krolzig ~1999! Improving on “Data mining reconsidered” by K+D+ Hoover
& S+J+ Perez+ Econometrics Journal 2, 202–219+
Hendry, D+F+, & H+-M+ Krolzig ~2001! Automatic Econometric Model Selection+ London: Timber-
lake Consultants Press+
Hendry, D+F+, & H+-M+ Krolzig ~2002! New developments in automatic general-to-specific model-
ling+ In B+P+ Stigum ~ed+!, Econometrics and the Philosophy of Economics+ Princeton: Princeton
University Press, forthcoming+
Hendry, D+F+, & G+E+ Mizon ~1978! Serial correlation as a convenient simplification, not a nui-
sance: A comment on a study of the demand for money by the Bank of England+ Economic
Journal 88, 549–563+ Reprinted in D+F+ Hendry, Econometrics: Alchemy or Science? ~Oxford:
Blackwell Publishers, 1993; Oxford: Oxford University Press, 2000!+
Hendry, D+F+, & M+S+ Morgan ~1995! The Foundations of Econometric Analysis+ Cambridge: Cam-
bridge University Press+
Hendry, D+F+, A+R+ Pagan, & J+D+ Sargan ~1984! Dynamic specification+ In Z+ Griliches and M+D+
Intriligator ~eds+!, Handbook of Econometrics, vol+ 2, Ch+ 18+ Amsterdam: North-Holland+ Re-
printed in D+F+ Hendry, Econometrics: Alchemy or Science? ~Oxford: Blackwell Publishers, 1993;
Oxford: Oxford University Press, 2000!+
Hendry, D+F+, & M+H+ Pesaran ~2001! Editors’ introduction+ Journal of Applied Econometrics 16,
197–202+
Hendry, D+F+, & J+-F+ Richard ~1982! On the formulation of empirical models in dynamic econo-
metrics+ Journal of Econometrics 20, 3–33+ Reprinted in C+W+J+ Granger ~ed+!, Modelling Eco-
nomic Series ~Oxford: Clarendon Press, 1990!; and in D+F+ Hendry, Econometrics: Alchemy or
Science? ~Oxford: Blackwell Publishers, 1993; Oxford: Oxford University Press, 2000!+
Hendry, D+F+, & J+-F+ Richard ~1983! The econometric analysis of economic time series ~with dis-
cussion!+ International Statistical Review 51, 111–163+ Reprinted in D+F+ Hendry, Econometrics:
478 DAVID F. HENDRY

Alchemy or Science? ~Oxford: Blackwell Publishers, 1993; Oxford: Oxford University Press,
2000!+
Hendry, D+F+, & P+K+ Trivedi ~1972! Maximum likelihood estimation of difference equations with
moving-average errors: A simulation study+ Review of Economic Studies 32, 117–145+
Hendry, D+F+, & T+ von Ungern-Sternberg ~1981! Liquidity and inflation effects on consumers’
expenditure+ In A+S+ Deaton ~ed+!, Essays in the Theory and Measurement of Consumers’ Behav-
iour, pp+ 237–261+ Cambridge: Cambridge University Press+ Reprinted in D+F+ Hendry, Econo-
metrics: Alchemy or Science? ~Oxford: Blackwell Publishers, 1993; Oxford: Oxford University
Press, 2000!+
Hendry, D+F+, & K+F+ Wallis ~eds+! ~1984! Econometrics and Quantitative Economics+ Oxford:
Basil Blackwell+
Holt, C+C+ ~1969! Interactions with a fellow research engineer-economist+ Reprinted in R+ Leeson
~ed+!, A.W.H. Phillips: Collected Works in Contemporary Perspective, pp+ 308–314 ~Cambridge:
Cambridge University Press, 2000!+
Hoover, K+D+, & S+J+ Perez ~1999! Data mining reconsidered: Encompassing and the general-to-
specific approach to specification search+ Econometrics Journal 2, 167–191+
Hoover, K+D+, & S+J+ Perez ~2000! Truth and Robustness in Cross-Country Growth Regressions+
Working paper, Economics Department, University of California, Davis+
Hutchison, T+W+ ~1938! The Significance and Basic Postulates of Economic Theory+ London:
Macmillan+
Johansen, S+ ~1988! Statistical analysis of cointegration vectors+ Journal of Economic Dynamics
and Control 12, 231–254+ Reprinted in R+F+ Engle & C+W+J+ Granger ~eds+!, Long-Run Economic
Relationships, pp+ 131–152 ~Oxford: Oxford University Press, 1991!+
Kiviet, J+F+ ~1986! On the rigor of some mis-specification tests for modelling dynamic relation-
ships+ Review of Economic Studies 53, 241–261+
Klappholz, K+, & J+ Agassi ~1959! Methodological prescriptions in economics+ Economica 26, 23–39+
Klein, L+R+ ~1953! A Textbook of Econometrics+ Evanston, Illinois: Row, Peterson+
Klein, L+R+, R+J+ Ball, A+ Hazlewood, & P+ Vandome ~1961! An Econometric Model of the UK+
Oxford: Oxford University Press+
Kydland, F+E+, & E+C+ Prescott ~1991! The econometrics of the general equilibrium approach to
business cycles+ Scandinavian Journal of Economics 93, 161–178+
Kydland, F+E+, & E+C+ Prescott ~1996! The computational experiment: An econometric tool+ Jour-
nal of Economic Perspectives 69–85+
Lakatos, I+ ~1974! Falsification and the methodology of scientific research programmes+ In I+ Laka-
tos & A+ Musgrave ~eds+!, Criticism and the Growth of Knowledge, pp+ 91–196+ Cambridge:
Cambridge University Press+
Leeson, R+ ~ed+! ~2000! A. W. H. Phillips: Collected Works in Contemporary Perspective+ Cam-
bridge: Cambridge University Press+
Lipsey, R+G+ ~1966! An Introduction to Positive Economics, 2nd ed+ London: Weidenfeld and
Nicholson+
Marglin, S+A+ ~1975! What do bosses do? part II+ Review of Radical Political Economics 7, 20–37+
Mizon, G+E+ ~1974! The estimation of non-linear econometric equations: An application to the spec-
ification and estimation of an aggregate putty-clay relation for the U+K+ Review of Economic
Studies 41, 253–270+
Mizon, G+E+ ~1977a! Inferential procedures in nonlinear models: An application in a UK industrial
cross section study of factor substitution and returns to scale+ Econometrica 45, 1221–1242+
Mizon, G+E+ ~1977b! Model selection procedures+ In M+J+ Artis & A+R+ Nobay ~eds+!, Studies in
Modern Economic Analysis, pp+ 97–120+ Oxford: Basil Blackwell+
Mizon, G+E+ ~1995! Progressive modelling of macroeconomic time series: The LSE methodology+
In K+D+ Hoover ~ed+!, Macroeconometrics: Developments, Tensions, and Prospects, pp+ 107–
169+ Dordrecht: Kluwer Academic Press+
Nerlove, M+, & K+F+ Wallis ~1966! Use of the Durbin-Watson statistic in inappropriate situations+
Econometrica 34, 235–238+
J. DENIS SARGAN AND LSE ECONOMETRIC METHODOLOGY 479

Pagan, A+R+, & D+F+ Nicholls ~1976! Exact maximum likelihood estimation of regression models
with finite order moving average errors+ Review of Economic Studies 43, 383–387+
Phillips, A+W+H+ ~1954! Stabilization policy in a closed economy+ Economic Journal 64, 290–333+
Reprinted in R+ Leeson, A.W.H. Phillips: Collected Works in Contemporary Perspective, pp+ 133–
168 ~Cambridge: Cambridge University Press, 2000!+
Phillips, A+W+H+ ~1957! Stabilization policy and the time form of lagged response+ Economic Jour-
nal 67, 265–277+ Reprinted in R+ Leeson, A.W.H. Phillips: Collected Works in Contemporary
Perspective, pp+ 169–183 ~Cambridge: Cambridge University Press, 2000!+
Phillips, A+W+H+ ~1958! The relation between unemployment and the rate of change of money wage
rates in the United Kingdom, 1861–1957+ Economica 25, 283–299+ Reprinted in R+ Leeson, A.W.H.
Phillips: Collected Works in Contemporary Perspective, pp+ 243–260 ~Cambridge: Cambridge
University Press, 2000!+
Phillips, A+W+H+, & M+H+ Quenouille ~1960! Estimation, regulation, and prediction in interdepen-
dent dynamic systems+ Bulletin de l’Institute de Statistique 37, 335–343+ Reprinted in R+ Lee-
son, A.W.H. Phillips: Collected Works in Contemporary Perspective, pp+ 408– 417 ~Cambridge:
Cambridge University Press, 2000!+
Phillips, P+C+B+ ~1985! The ET interview: Professor J+D+ Sargan+ Econometric Theory 2, 119–139+
Phillips, P+C+B+ ~1994! Bayes models and forecasts of Australian macroeconomic time series+ In C+
Hargreaves ~ed+!, Non-stationary Time-Series Analyses and Cointegration+ Oxford: Oxford Uni-
versity Press+
Phillips, P+C+B+ ~1995! Automated forecasts of Asia-Pacific economic activity+ Asia-Pacific Eco-
nomic Review 1, 92–102+
Phillips, P+C+B+ ~1996! Econometric model determination+ Econometrica 64, 763–812+
Phillips, P+C+B+ ~2002! Laws and limits of econometrics+ Sargan lecture, Royal Economic Society
Conference, Warwick University+
Phillips, P+C+B+, & M+ Loretan ~1991! Estimating long-run economic equilibria+ Review of Eco-
nomic Studies 58, 407– 436+
Popper, K+R+ ~1959! The Logic of Scientific Discovery+ New York: Basic Books+
Popper, K+R+ ~1963! Conjectures and Refutations+ New York: Basic Books+
Qin, D+ ~1993! The Formation of Econometrics: A Historical Perspective+ Oxford: Clarendon Press+
Ramsey, J+B+ ~1969! Tests for specification errors in classical linear least squares regression analy-
sis+ Journal of the Royal Statistical Society, Series B 31, 350–371+
Richard, J+-F+ ~1980! Models with several regimes and changes in exogeneity+ Review of Economic
Studies 47, 1–20+
Rissanen, J+ ~1978! Modeling by shortest data description+ Automatica 14, 465– 471+
Robbins, L+ ~1932! An Essay on the Nature and Significance of Economic Science+ London:
Macmillan+
Salmon, M+ ~1982! Error correction mechanisms+ Economic Journal 92, 615– 629+
Sargan, J+D+ ~1957! The danger of over-simplification+ Oxford Bulletin of Economics and Statistics
19, 171–178+
Sargan, J+D+ ~1959! The estimation of relationships with autocorrelated residuals by the use of in-
strumental variables+ Journal of the Royal Statistical Society, Series B 21, 91–105+ Reprinted in
J+D+ Sargan, Contributions to Economics, E+ Maasoumi ~ed+!, vol+ 1, pp+ 87–104 ~Cambridge:
Cambridge University Press, 1988!+
Sargan, J+D+ ~1961! The maximum likelihood estimation of economic relationships with autoregres-
sive residuals+ Econometrica 29, 414– 426+ Reprinted in J+D+ Sargan, Contributions to Econom-
ics, E+ Maasoumi ~ed+!, vol+ 1, pp+ 105–117 ~Cambridge: Cambridge University Press, 1988!+
Sargan, J+D+ ~1964! Wages and prices in the United Kingdom: A study in econometric methodology
~with discussion!+ In P+E+ Hart, G+ Mills, & J+K+ Whitaker ~eds+!, Econometric Analysis for Na-
tional Economic Planning, vol+ 16 of Colston Papers, pp+ 25– 63+ London: Butterworth+ Reprinted
in D+F+ Hendry & K+F+ Wallis ~eds+!, Econometrics and Quantitative Economics, pp+ 275–314
~Oxford: Basil Blackwell, 1984!; and in J+D+ Sargan, Contributions to Econometrics, E+ Maa-
soumi ~ed+!, vol+ 1, pp+ 124–169 ~Cambridge: Cambridge University Press, 1988!+
480 DAVID F. HENDRY

Sargan, J+D+ ~1973! Model building and data mining+ Paper presented to the Association of Univer-
sity Teachers of Economics Meeting, Manchester+ Published in Econometric Reviews ~2001!, 20,
159–170+
Sargan, J+D+ ~1975! Asymptotic theory and large models+ International Economic Review 16, 75–91+
Sargan, J+D+ ~1980! Some tests of dynamic specification for a single equation+ Econometrica 48,
879–897+ Reprinted in J+D+ Sargan, Contributions to Econometrics, E+ Maasoumi ~ed+!, vol+ 1,
pp+ 191–212+
Sargan, J+D+ ~1981! The choice between sets of regressors+ Mimeo, Economics Department, Lon-
don School of Economics+ Published in Econometric Reviews ~2001!, 20, 171–186+
Sargan, J+D+ ~1988a! Contributions to Econometrics, E+ Maasoumi ~ed+!, vols+ 1 and 2+ Cambridge:
Cambridge University Press+
Sargan, J+D+ ~1988b! Lectures on Advanced Econometric Theory+ Oxford: Basil Blackwell+
Sargan, J+D+, & J+D+ Sylwestrowicz ~1976! COMFAC: Algorithm for Wald tests of common factors
in lag polynomials+ Users’ manual, London School of Economics+
Schiff, A+F+, & P+C+B+ Phillips ~2002! Forecasting New Zealand’s real GDP+ Zealand Economic
Papers 34~2!, 159–182+
Schwarz, G+ ~1978! Estimating the dimension of a model+ Annals of Statistics 6, 461– 464+
Shibata, R+ ~1981! An optimal autoregressive spectral estimate+ Annals of Statistics 9, 300–306+
Student ~1908! On the probable error of the mean+ Biometrika 6, 1–25+
Theil, H+ ~1961! Economic Forecasts and Policy, 2nd ed+ Amsterdam: North-Holland+
Trivedi, P+K+ ~1970! The relation between the order-delivery lag and the rate of capacity utilization
in the engineering industry in the United Kingdom, 1958–1967+ Economica 37, 54– 67+
Wald, A+ ~1947! Sequential Analysis+ New York: Wiley+
White, H+ ~1980! A heteroskedastic-consistent covariance matrix estimator and a direct test for
heteroskedasticity+ Econometrica 48, 817–838+

You might also like