The Dynamic of International Competitiveness

Abstract

The Dynamics of International Competitiveness. — This paper focuses on the determinants of international competitiveness over the seventies and eighties. The theoretical framework adopted here is based on a “technology-gap” account of trade flows. The econometric analysis relies on a dynamic model estimated by pooling time-series across countries. The short- and long-term impacts of both technical change and labour costs on trade performance are investigated. It is found that technological variables (patents and investments) play a major role in shaping dynamics of exportshares, while labour costs asymmetries among countries appear to affect trade performance only in the short term.

Full text

The Dynamics of
International Competitiveness:
First Results from
an Analysis at the Industry Level
Paolo Guerrieri*
Paola Maggiolini**
Gennaro Zezza***
Presented at the International Workshop "The Mulltiple Linkages between
Technological Change and the Economy", Rome, 14th-16th March, 1996
*) Università di Roma "La Sapienza", via del Castro Laurenziano, 9 - 00165 Roma
**) Istituto di Ricerca sul Credito cooperativo dell'Economia Locale, via M.D'Azeglio, 33 -
00184 Roma
***) Università di Napoli "Ferderico II", Dipartimento di Scienze Economiche e Statistiche,
via G.Sanfelice, 47 - 80134 Napoli. e-mail: gzezza@unina.it

The Dynamics of International Competitiveness:
First Results from an Analysis at the Industry Level
Paolo Guerrieri - University of Rome
Paola Maggiolini - I.R.C.E.L.
Gennaro Zezza - University of Naples
°
Introduction
Over the past two decades a growing number of theoretical contributions and
empirical verifications have led to the recognition of the key role of technology in
determining trade flows and international competitiveness at the firm and country
level (for a survey see Dosi et al. 1988). In addition, they have shown that the
performance of firms depends not only on successfull management practices by
enterpreneurs but, to a large extent, also on the structural features of the countries
and sectors in which they operate. This is associated with the fact that the
competitive advantages of single countries are concentrated on given industries
and clusters of industries reflecting the systemic characteristics and the
interrelatedness of many technologies (Oecd, 1992; Porter, 1990; Guerrieri and
Tylecote, 1994).
These results are related both to the new trade and growth theories based on
imperfect competition models (Grossman and Helpman, 1991) and to the
evolutionary approaches to technological change and trade (for a survey see
Nelson, 1995).
One particular feature of technological innovation is its sectoral differentiation
(Pavitt, 1988). Numerous empirical studies confirm stable and similar differences
between sectors in the relative importance of both the various sources of new
knowledge (such as basic research, in-house, suppliers, etc.) and the industries in
which innovations are used (Pavitt, 1984; Scherer, 1986; Dosi, Pavitt, Soete,
1990). These trends reflect the fact that development of technological knowledge
follow cumulative and differentiated paths in individual sectors. In other words
°
This paper is the result of research made in collaboration by the authors. However, the
introduction and par. 2 have been written by P.Guerrieri, while G.Zezza wrote par.1 and
par.3, and P.Maggiolini constructed the data set.
Financial support from C.N.R. and M.U.R.S.T. is gratefully acknowledged.
1

processes of technological change tend to assume varying sectoral features, in
terms of differences in technological opportunities, cumulativeness and
appropriability conditions.
Altough the differences among sectors in their technological trajectories are
widely recognized, they have often been neglected by empirical studies on the
relationship between innovative activities and trade performance which have
prevalently provided aggregate analysis in the dynamic version (Amendola et al.,
1993; Amendola, Guerrieri, Padoan, 1992) or used cross-sectional data in the
static approaches (Soete, 1981; Fagerberg, 1988). That was also due to a lack of
an adequate set of comparable data on technology, trade and production of
different countries.
In the following we present a set of new data on trade, production, technology
and costs at industry level for a certain number of countries and use them by
trying to relate trade performance (international competitiviness) to a set of
different economic and technological factors accross countries and industrial
sectors since the early 1970s. The paper presents some preliminary results,
related to a certain number of different industrial sectors for the major
industrialized countries. A single model of trade specialization is applied to the
data in order to establish the impact of innovation, costs and country specific
factos to overall performance, both in the short and the long run, via a panel data
analysis. Model specification follows the now standard error-correction approach
and is discussed in section 2. The following section 1 presents the data set, while
the results of the estimates are discussed in section 3.
1 The theoretical model and the data set
Following a well-established approach1 we take competitiviness, and hence
international market shares, to be determined by:
- relative prices;
- relative technological capabitilies.
Moreover, the degree of importance of each factor depends upon the
characteristics of each specific industry: following Pavitt (1984, 1988) we expect
to find clusters of sectors responding in a different way to shocks, with respect to
industry technology, market conditions, degree of innovation etc.
1 See Amendola et al. (1993) among others.
2

Given the great amount of sectors and countries to analyze, we choose to
estimate a model as general as possible for each sector and country. The
theoretical function is
Xij = f(Tij, Cij) (1
where T represents the technological capabilities for country i in sector j, and C
stands for relative costs.
As a guide to model interpretation we have reported the data2 for two different
sectors in charts 1-12. We have choosen two sectors (Apparel, Drugs and
medicines) which should be consistently different with respect to technology,
costs and market structure.
Export market shares for Apparel (chart 1) clearly depicts the relevant questions
for the model-builder. We have to capture: a) the different specialization levels
accross countries, eg why Italy is more specialized than other countries, and; b)
the dynamics of market shares, eg why market shares tend to diminish for each
country and for the whole sample, and why (possibly) two countries experience
different performances through time.
The overall picture for all sectors shows a relative persistence of country
specialization through time.
The variables we choose to proxy for technology, given the available data set,
have been the share of the stock of US patents3 for each country in each sector4,
which is a proxy for the ability to innovate, and some measures of investment
intensity, to control for innovation embodied in the capital stock.
Patent share for Apparel and Drugs and medicines are reported in Charts 4 and 8,
respectively. Even though we normalized non-US share in patents, US share still
has a predominant role, in levels. Simple correlations of patents with export
market shares do not give any insight, since the increase in patent share of Japan,
say, is not always followed by an increase in overall competitiveness.
2 The data set comes from different sources (UN, Oecd, other data banks). All variables have
been appropriately allocated to the same definition of industry following the SITC
classification. A detailed description of the data set can be found in Guerrieri (1995).
3 The variable is Bijt = (Σs=0,t Pijs)/(ΣiΣs=0,t Pijs). We are aware of the several shortcomings
of patents as a proxy for technological effort in a given industry. For instance, the
propensity to patent may differ accross sectors and countries. However, the number of
patents in the United States was the best indicator available - at the time of writing - with
the required sectoral level of detail.
4 This variable obviously over-represents the share of the US. To partially adjust this effect,
the share of other countries is obtained with respect to the total number of non-US patents.
3

Given the well known problems in dealing with investment, we have choosen two
different indicators: the investment/product ratio, reported in charts 5 and 9, and
the “investment intensity”, defined as the ratio of investment in equipment in
sector i over total investment in equipment, both expressed as index numbers,
reported in charts 6 and 10.
The figures suggest that investment may help explain the dynamics of export
market shares, but there is no apparent correlation between investment intensity,
however measured, and trade specialization.
Cost competitiveness is proxied by average relative wages5, reported in charts 7
and 11. Again, there is no apparent (negative) correlation between better export
performance and lower labor costs.
2. Model specification.
Model specification has followed the now standard cointegration approach,
where a long run cointegrating vector is estimated along with the dynamic
adjustment process towards long-run equilibrium.
This procedure requires variables with the same order of integration, but it is not
feasible to run a proper integration analysis for each variable with such a large
data set. Even though export shares are expected to be I(0), since they have
upper and lower bounds, ADF tests did not reject the hypothesis of variables
being distributed as I(1). We therefore used first differences in the dynamic
specification, and we have appropriately transformed those variables which were
expected to be of higher order of integration. This implied, for instance,
normalizing average wages to average (world) unit wages.
The estimated model is, for each sector j:
∆Xijt = Σs=1,3 ρs ∆Xijt-s + Σs=0,3 βs ∆Iijt-s + Σs=0,3 γs ∆Bijt-s +
+ Σs=0,3 λs ∆Wijt-s - (ρ Xijt-1 - β Iijt-1 - γ Bijt-1 - λ Wijt-1) +
+ αi + Uijt(2
where:
X = share of country i exports for sector j;
I = An investment intensity measure for country i in sector j;
5 This variable is not entirely appropriate since it does not take into account the effects of
shifts in productivity. We are updating our database to get coherent estimates of output,
which will enable estimates of unit labor costs.
4

B = Share of US patents stock for country i in sector j;
W = unit wages for county i in sector j with respect to average unit wage in sector
j;
U is a random disturbance, and ∆x stands for xt - xt-16.
The terms in parenthesis represent the long-term relation among all variables:
X = (β I + γ B + λ W)/ρ
with a speed of adjustment of ρ.
We want to use model (2) to test several hyopthesis:
a) exploring the determinants of competitiviness in both the short and the long
run, eg test for b=0 where b is the vector of all parameters;
b) explore differences accross countries and sectors
Model (2) has been estimated via a panel data analysis, using 17 observations
(1974-1990) for eleven OECD countries. The two different measures of
investment intensity have been alternatively used.
3. Model estimates.
Table 1 reports the estimates of model (2), using the investment/GDP ratio as the
measure of innovation embodied in the capital stock. The last coluumn reports a
panel estimate for all sectors pooled together.
Table 1 gives a number of interesting suggestions, most of which are in line with
previous research at the aggregate level7:
a) the evidence of a long-run relation among the choosen variable is weak, even
though some sectors specificities seem to arise;
b) the effects of technology, as measured by the share of US patents, tend to be
important in the long run but of secondary importance in the dynamic
adjustment. Anyway, patent indicators contributes to differentiate sectoral
6 First differences in market shares are expected to exhibit heteroscedasticity, since countries
with a higher market share in level will experience larger fluctuations (see chart 2). We
estimated model (2) using percent changes in market shares (see chart 3) instead of first
differences. The overall results are not affected, but some variables exhibit erratic
behaviours, given by absolute values close to zero, generating spurious outliers in the
transformed variable. We therefore choose to use first differences, and control for
heteroscedasticity.
7 See Amendola et al., 1993.
5

behaviours, as in the case of Chemicals, Metal Working and Machinery,
Machinery n.e.c.;
c) the investment to output ratio seems to have little explanatory power. Since
this result is strikingly different from what we expected, we estimated the
model with a different measure of investment intensity, namely the ratio of
sectoral investment in equipment to total investment in equipment, for each
country. Results are reported in table 2, but only for the Apparel sector the
new measure of investment is significant in the long run;
d) an increase in relative costs is directly related to export performance in the
short run. This result could be interpreted as a sectoral J-curve effect, with
demand adjusting slowly to changes in the exchange rate. Alternatively, there
could be an inverse relation, where good export performance signals higher
quality or higher productivity, which is reflected on salaries above the average.
The long-run effects of relative costs on competitiveness are negligible.
e) fixed effects play a dominant role in some sectors, such as “Rubber and
plastics” and “Instruments”, perhaps offsetting technology variables, as the
high coefficients for Japan and West Germany suggest;
f) structural fixed effects determine significant advantages for some countries in
some sectors, confirming persistent sectoral specilization in the long rn. These
are the case of West Germany and Italy for “Metal working and machinery”
and “Apparel”, or France, Great Britain and the Netherlands in “Chemicals”.
These effects should be further examined, incorporating other structural
variables in the model at the country level.
Overall, model estimates tend to stress some pecularities in the data which were
already apparent from graphical analysis. Export specialization tends to be sticky
with respect both to technology and cost shocks. This could be in line with recent
theoretical research, which assigns a relevant role to endogenous accumulation of
knowledge, so that when a country specializes, learning-by-learning tends to
predominate over adverse shocks in overall competitiveness.
We believe, however, that more definite conclusions can be drawn only with
further research, since the quality of some of our technology indicators and
6

sectoral data has to be further tested8. Costs indicators could be refined including
some measure of relative productivity, which is not available at present.
We aim also to extend and refine our analysis with multivariate techniques,
measuring the distance of sectors and countries in the variables space, in order to
test hypothesis about dissimilarities accross countries and sectors.
8 We have tested the model excluding the US, since the bias towards this country in the patent
indicator could be misleading. The results did not show any improvement with regard to the
effects of technology on trade performance.
7

References
Amendola, G., Guerrieri, P., Padoan, P.C., (1992), International Patterns of
Technological Accumulation and Trade, JOICE, 1.
Amendola, G., Dosi, G., Papagni, E., (1993), The Dynamics of International
Competitiveness, Weltwirtschaftiches Archiv, 3.
Dosi, G., Pavitt, K., Soete, L. (1990), The Economics of Technical Change and
International Trade, Wheatsheaf, Brighton.
Dosi, G. et al. (1988), Technical Change and Economic Theory, Frances Pinter,
London.
Fagerberg J. (1988), International Competitiveness, The Economic Journal, vol.
98, pp. 355-374.
Grossman, G. and Helpman, E. (1991), Innovation and growth in the Global
Economy, Cambridge, Mass.
Guerrieri, P. (1995), Interdipendenze tecnologiche e mutamenti strutturali,
risultati dell’unità operativa del progetto ISPE-CNR Technological change
and economic growth, Rome.
Guerrieri, P. and Tylecote, A. (1994), National Competitive Advantages and
Microeconomic Behaviour, Economics of Innovation and New Technology,
vol.3.
Nelson, R. (199.), Recent Evolutionary Theorizing about Economic Change,
JEL, vol. XXXIII, pp.48-90.
OECD (1992), Technology and the Economy: the Key Relationships, Paris.
Pavitt, K. (1984), Sectorals Patterns of Technical Change: Towards a Taxonomy
and a Theory, Research Policy, 13, pp.343-373.
Pavitt, K. (1988), International Patterns of Technological Accumulation, in Hood,
N. and Vahlne, J.E. (eds.), Strategies in Global Competition, Croom Helm,
London.
Porter, R. (1990), The Competitive Advantage of Nations, Macmillan, London-
New York.
Soete, L. (1981), A General Test of Technological Gap Trade Theory,
Weltwirtschaftiches Archiv, vol. 117.
Rosenberg, N. (1982), Inside the Black Box, Cambridge U.P., Cambridge.
8

Scherer, F.M. (1986), Innovation and Growth. Schumpeterian Perspectives,
MIT Press, Cambridge (Mass.).
9

Table 1. Sector estimates of model 2 (I = Investment/GDP)
CHEM MWM APP RUB DRU INS NEC ALL
DX(-1) -0.123 0.127 -0.055
DX(-2) -0.125 -0.219 -0.104
DX(-3) 0.179
DB 2.535
DB(-1) -1.345 0.098 -2.336
DB(-2) 1.775 1.360 -0.156
DB(-3) -2.707 -0.996
DI 0.294
DI(-1) -0.042 -0.190 -0.224 -0.189 -0.038
DI(-2) -0.054 0.125
DI(-3) 0.035
DW 0.021 0.027 0.014 0.028 0.025 0.022 0.040 0.025
DW(-1) -0.020 -0.009
DW(-2) -0.012 -0.007 0.017 -0.009
DW(-3) -0.013
X(-1) -0.448 -0.233 -0.197 -0.332 -0.298 -0.504 -0.301 -0.026
B(-1) 0.149 0.107 0.146
I(-1) 0.040 -0.007
W(-1)
Country effects
AUT 0.016 0.007
CAN 0.020 0.009
DEU 0.025 0.014 0.055 0.038 0.079 0.022 0.003
ESP 0.005 0.019
FRA 0.023 0.053 0.026 0.031
GBR 0.016 0.024 0.029 0.038
ITA 0.019 0.025 0.034 0.013 0.015 0.020
JPN 0.022 0.074 0.109 0.018 0.005
NLD 0.025 0.024 0.014 0.016
SWE 0.017 0.009 0.010
USA 0.104 -0.055 0.006
R0.352 0.414 0.239 0.291 0.202 0.378 0.338 0.155
DW 2.029 2.079 2.089 2.070 2.020 2.036 2.048 2.013
Notes: (chem) Chemicals; (mwm) Metal working machinery and equipment; (app) Apparel;
(rub) Plastics and Rubber; (dru) Drugs and medicines; (ins) Instruments; (nec) Machinery
NEC; (all) the seven sectors pooled together. (aut) Austria; (can) Canada; (deu) West
Germany; (esp) Spain; (fra) France; (gbr) United Kingdom; (ita) Italy; (jpn) Japan; (nld)
Nederland; (swe) Sweden; (usa) United States
Parameter estimates not significantly different from zero are not reported.
10

Table 2. Sector estimates of model 2 (I = Investment intensity)
CHEM MWM APP RUB DRU INS NEC ALL
DX(-1) 0.120 0.132 -0.058
DX(-2) -0.128 -0.228 -0.102
DX(-3) 0.160
DB 2.417
DB(-1) -1.554 0.109 -2.341
DB(-2) 1.794 1.442 -0.156
DB(-3) -2.601 -1.042 1.267
DI 0.020 0.028 0.008
DI(-1) -0.010 -0.021 -0.013
DI(-2) -0.012 0.009
DI(-3) 0.007
DW 0.017 0.027 0.011 0.028 0.024 0.022 0.039 0.026
DW(-1) -0.017 -0.009
DW(-2) -0.011 -0.007 0.015 -0.008
DW(-3)
X(-1) -0.456 -0.213 -0.238 -0.299 -0.280 -0.512 -0.290 -0.027
B(-1) 0.155 0.138
I(-1) 0.011 0.009 -0.007
W(-1)
Country effects
AUT 0.013
CAN 0.017
DEU 0.010 0.048 0.035 0.086 0.003
ESP 0.008
FRA 0.023 0.043 0.024 0.039
GBR 0.016 0.018 0.027 0.043
ITA 0.025 0.025 0.014 0.022
JPN 0.066 0.117 0.005
NLD 0.022 0.014 0.024
SWE 0.019
USA 0.114 -0.054 0.006
R0.323 0.426 0.257 0.292 0.195 0.382 0.336 0.160
DW 2.024 2.086 2.120 2.055 1.999 2.023 2.049 2.011
Notes: (chem) Chemicals; (mwm) Metal working machinery and equipment; (app) Apparel;
(rub) Plastics and Rubber; (dru) Drugs and medicines; (ins) Instruments; (nec) Machinery
NEC; (all) the seven sectors pooled together. (aut) Austria; (can) Canada; (deu) West
Germany; (esp) Spain; (fra) France; (gbr) United Kingdom; (ita) Italy; (jpn) Japan; (nld)
Nederland; (swe) Sweden; (usa) United States
Parameter estimates not significantly different from zero are not reported.
11

0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Market shares.
Chart 1
-0,03
-0,02
-0,02
-0,01
-0,01
0,00
0,01
0,01
0,02
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Annual change in market shares.
Chart 2
12

-0,50
-0,40
-0,30
-0,20
-0,10
0,00
0,10
0,20
0,30
0,40
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Annual percent change in market shares.
Chart 3
0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
0,80
0,90
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Patent shares.
Chart 4
13

0,00
0,01
0,01
0,02
0,02
0,03
0,03
0,04
0,04
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Investment ratio.
Chart 5
0,50
0,60
0,70
0,80
0,90
1,00
1,10
1,20
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Investment intensity.
Chart 6
14

0,30
0,50
0,70
0,90
1,10
1,30
1,50
1,70
1,90
2,10
2,30
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Apparel. 1974-90. Relative average wage.
Chart 7
0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
0,16
0,18
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Drugs and medicines. 1974-90. Market shares.
Chart 8
15

0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Drugs and medicines. 1974-90. Patent shares.
Chart 9
0,00
0,01
0,02
0,03
0,04
0,05
0,06
0,07
0,08
0,09
0,10
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Drugs and medicines. 1974-90. Investment ratio.
Chart 10
16

0,30
0,50
0,70
0,90
1,10
1,30
1,50
1,70
1,90
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Drugs and medicines. 1974-90. Investment intensity.
Chart 11
0,40
0,60
0,80
1,00
1,20
1,40
1,60
aut
can
deu
esp
fra
gbr
ita
jpn
nld
swe
usa
Drugs and medicines. 1974-90. Relative average wage.
Chart 12
17