Dynamics of Total Factor Productivities - article ; n°2 ; vol.44, pg 389-418

Klaus Neusser

Description

Revue économique - Année 1993 - Volume 44 - Numéro 2 - Pages 389-418
Dynamics of Total Factor Productivities
Cet article examine la dynamique de la productivité globale des facteurs sur une base sectorielle pour quatorze pays de l'OCDE. Après correction pour les biais de petit échantillon dans l'approche univariée, et en combinant la dimension en coupe et la dimension temporelle, il est établi que pratiquement toutes les séries suivent un processus intégré. Les interrelations sectorielles des productivités globales sont étudiées à l'aide d'une spécification à correction d'erreur. On trouve qu'une part importante de l'interaction transite par les niveaux de productivité, ce qui suggère l'hypothèse de convergence de la productivité globale entre les différents secteurs.
De plus, les interactions entre secteurs sont dominées par les évolutions des secteurs des services et des administrations publiques.
Dynamics of total factor productivities
The paper examines the dynamics of total factor productivities (TFP's) for 14 OECD countries on a sectoral basis. Correcting for small sample bias in the univariate approach and combining the cross section and time series dimension in a random fields approach, it is established that practically all series are better characterized as integrated processes. The interrelation between TFP's across sectors is analyzed through an error correction speci­fication. It is found that an important part of the interaction works through level effects suggesting convergence of TFP's across sectors. Moreover, interactions across sectors are partly driven by developments in the service and government sectors.
30 pages
Source : Persée ; Ministère de la jeunesse, de l’éducation nationale et de la recherche, Direction de l’enseignement supérieur, Sous-direction des bibliothèques et de la documentation.

Sujets

• Sociologie, société et politique

Informations

• Publié par: REVUE_ECONOMIQUE
• Publié le: 01 janvier 1993
• Nombre de lectures: 29
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Monsieur Klaus Neusser
Dynamics of Total Factor Productivities
In: Revue économique. Volume 44, n°2, 1993. pp. 389-418.
Résumé
Cet article examine la dynamique de la productivité globale des facteurs sur une base sectorielle pour quatorze pays de l'OCDE.
Après correction pour les biais de petit échantillon dans l'approche univariée, et en combinant la dimension en coupe et la
dimension temporelle, il est établi que pratiquement toutes les séries suivent un processus intégré. Les interrelations sectorielles
des productivités globales sont étudiées à l'aide d'une spécification à correction d'erreur. On trouve qu'une part importante de
l'interaction transite par les niveaux de productivité, ce qui suggère l'hypothèse de convergence de la productivité globale entre
les différents secteurs.
De plus, les interactions entre secteurs sont dominées par les évolutions des secteurs des services et des administrations
publiques.
Abstract
Dynamics of total factor productivities
The paper examines the dynamics of total factor productivities (TFP's) for 14 OECD countries on a sectoral basis. Correcting for
small sample bias in the univariate approach and combining the cross section and time series dimension in a random fields
approach, it is established that practically all series are better characterized as integrated processes. The interrelation between
TFP's across sectors is analyzed through an error correction speci-fication. It is found that an important part of the interaction
works through level effects suggesting convergence of TFP's across sectors. Moreover, interactions across sectors are partly
driven by developments in the service and government sectors.
Citer ce document / Cite this document :
Neusser Klaus. Dynamics of Total Factor Productivities. In: Revue économique. Volume 44, n°2, 1993. pp. 389-418.
http://www.persee.fr/web/revues/home/prescript/article/reco_0035-2764_1993_num_44_2_409455of Total Factor Dynamics
Productivities
Klaus Neusser
Cet article examine la dynamique de la productivité globale des facteurs sur une
base sectorielle pour quatorze pays de l'OCDE. Après correction pour les biais de
petit échantillon dans l'approche univariée, et en combinant la dimension en coupe
et la dimension temporelle, il est établi que pratiquement toutes les séries suivent
un processus intégré. Les interrelations sectorielles des productivités globales sont
étudiées à l'aide d'une spécification à correction d'erreur. On trouve qu'une part
importante de l'interaction transite par les niveaux de productivité, ce qui suggère
l'hypothèse de convergence de la productivité globale entre les différents secteurs.
De plus, les interactions entre secteurs sont dominées par les évolutions des
secteurs des services et des administrations publiques.
1. INTRODUCTION
Recent theoretical models view increasing returns to scale, external effects,
strong strategic complementarities and imperfect competition as instrumental
in understanding economic growth. Although the results of this paper are
relevant for many of them, it seems convenient for this investigation to take the
model of Romer [1986] as a reference. Romer combines competitive behavior
and constant returns to scale on the firm level with increasing returns on the level
of the branch, the manufacturing sector, or even the whole industry. The
increasing return stems from a positive externality which is external to the indi
vidual firm and which can be justified, for example, through whithin sectoral divi
sion of labor, specialization, or the provision of public intermediate inputs.
Although this formulation presents some conceptual difficulties , Romer 's
* The author tanks John Tatom, the conference and seminar participants in Marrackech
and Konstanz for helpful comments and suggestions. Financial support by the Viennese
Chamber of Trade is greatfuly acknowledged.
1 . Typically increased division of labor or specialization works trough creation of new
goods respectively new markets and are thus not externalities in a strict sense. Moreover,
the presence of externalities provides economic incentives to "internalize" these external
ities through mergers and acquisitions.
389
Revue économique — N° 2, mars 1993, p. 389-418. économique Revue
formulation, considered as a "reduced form", represents a useful framework for
organizing the empirical analysis.
This empirical analysis is still slow to develop. Many studies relied on the
Penn World Table (Summers and Heston [1991]) to detect and quantify these
external economies (Backus, Kehoe, and Kehoe[1990] ; De Long and Summers
[1991] among others), but this data set is probably too highly aggregated and too
heterogeneous for this purpose. In contrast, this paper relies on sectorial data for
a relatively homogeneous group of OECD countries. The methodology initiated
by Hall [1988, 1991] and further extended by Caballero and Lyons [1989, 1990]
examines the so-called Solow residuals or total factor productivities (TFPs).
Their analysis relies on a regression of the growth rate of output on the growth
rates of an index of factor inputs and of total factor inputs for the whole industry
to quantify the externalities.
This particular specification has several shortcomings. First, the externality
enters the regression as the sum of total output or an index of total inputs of the
industry. This formulation is too simplistic because externalities, especially in
connection with R&D, relate firms between and within industries in very complex
ways. Thus one needs a notion of "distance" or weighting scheme to compute
the appropriate externality variable (see the discussion in Griliches [1991]). Bar-
telsman, Caballero and Lyons [1991] approach this difficulty by weighting the
other industries activities using information from input-output tables. Second,
for many especially in small open economies the relevant spillover
may occur between countries rather than within countries. This becomes immed
iately relevant when other than the U.S . are considered. 1 Third, the stu
dies referred to above allow only contemporaneous interactions between
productivity growth rates and disregard any dynamics. It seems, however,
implausible that externalities should only work through a contemporaneous
effect, typically it takes some time until they have reached other sectors of the
economy. Furthermore, one would expect that the interaction depends not only
on current or past growth rates but also on differences in productivity levels across
industries or countries.
Indeed in a world of capital mobility and free trade large differences in pro
ductivity levels should not persist over extended periods. Thus even if product
ivities in different sectors and countries move around extensively, they should
not deviate too much from a common underlying stochastic trend. This notion
underlies the concept of cointegration introduced by Engle and Granger [1987]
in the context of integrated time series.
The paper is primarily concerned with this third aspect. The investigation of
productivity levels proceeds in several steps. First the issue of nonstationarity is
analyzed by a univariate analysis. The results rely on the conventional Dickey-
Fuller test (Dickey and Fuller [1979]) and.on a small sample method proposed
1 . Stockman [1988] reports evidence that growth rates of industrial production are dr
iven both by industry and country specific disturbances.
390 Klaus Neusser
by Andrews [1991]. As the sample period is very short, the power of the test is
rather low. This can be overcome by pooling the time series across countries and/
or industries in the second step. The enlarged data set can then be viewed as a
random field.1 As demonstrated by Quah [1990b], the conventional unit root tests
of Dickey and Fuller [1979] and Phillips [1987] can be extended to this type of
data. The third step consists in the analysis of the interaction between productiv
ities. The long-run relations between TFPs are estimated using the single equa
tion nonlinear error correction model proposed by Phillips and Loretan [1991].
Under the assumption that a cointegrating relation exists, this method yields fully
efficient and median unbiased estimates of long-run relations.
The data for this investigation are compiled from the International Sectorial
Databank made available by the OECD (Meyer-zu-Schlochtern [1988]). This
databank allows to compute TFP's for 14 country and 32 industries.2 Although
the data start only in 1970 for most countries and is thus rather short from a time
series perspective, the pooling of times series and cross-section presents a wealth
of information. The major difference between the data analyzed by Hall [1988]
and Caballero and Lyons [1989, 1990] is that constant returns to scale are ass
umed at the sectoral level. This seems, however, not such a big problem because
Caballero and Lyons' estimates do not reject this assumption most of