INNOVATION AND COMPETITIVENESS: TRENDS IN
UNIT PRICES IN GLOBAL TRADE
Raphael Kaplinsky,
Institute of Development Studies,
University of Sussex,
Brighton, and
Centre for Research in Innovation Management,
University of Brighton
and
Amelia Santos Paulino
Institute of Development Studies,
University of Sussex,
Brighton
We are grateful to Shaun Gannon for his careful translation of various nomenclatures,
to John Humphrey for comments on an earlier draft, and to the UK Dept. for
International Development for supporting this research (Project R8244: Trends in
Developing Country Unit Export Prices, 1988-2001)
SUMMARY
This paper seeks to build on theory, to develop new methods for understanding the
nature and basis of sectoral and national competitive advantage, and to do so with a
temporal perspective.
Neo-Schumpeterian and evolutionary economics perspectives are in large part built
around the concepts of barriers to entry and core competences. Unless these are
established, individual firms, networks of firms and countries will be unable to
generate sustained income growth. There is no one measure which adequately reflects
these barriers to entry, and much of the research has been concerned to generate
proxies, each of which is in itself partial, but which together provide for a
comprehensive picture.
During the late 1970s, preliminary work was done on the unit price of UK trade as an
indicator of relative technological competence. However, this approach has largely
been neglected since then, receiving only sporadic attention in the US literature, and
at high levels of product aggregation. This paper utilises this approach to try and
reflect the dynamic process of shifting competitive advantage in the global economy.
Its distinctive feature is the level of detail – six-digit trade classifications – and its
breadth of coverage, being applied to seven sets of sectoral classifications involving
more than 12,000 product groups. The data-set relates to EU imports of manufactures
between 1988 and 2001.
Keywords
Technology and trade
Factor intensity and trade
Terms of trade
Global competitiveness
Fallacy of composition
1
1. Indicators of innovation
Schumpeter’s model of innovation is founded on the role which barriers of entry
provide in sustaining innovation rents. If barriers to entry are substantial, competition
is limited, and price-competition is held at bay and, if the innovation is attractive to
users, incomes are high and sustainable. Without innovation or if barriers to entry are
low, price competition drives producers out of production or, at least, in a Malthusian
race to the bottom in living standards.
How might we know if innovation is indeed being sustained? If we work at the plantor firm-level, this is a matter for empirical investigation, examining production
processes (reflected for example in factor productivity, quality, lead time and other
indicators), products (for example, the introduction of new or differentiated products)
and function (whether firms are involved in production, design, marketing or other
links in the value chain). 1 However, once the focus moves beyond the plant and the
firm, the measurement of innovation becomes more difficult.
Typically, in assessing rates of innovation in clusters, sectors or countries, innovative
activity is reflected by the use of one or both of two indicators – an input indicator
(for example, R&D expenditure, percentage of skills of different sorts in the labour
force) or an output indicator (notably patents). Clearly, none of these input or output
indicators of innovation are ideal. Input indicators are bedevilled by measurement
problems, the effectiveness with resources are utilised, and their lack of recognition of
processes of incremental technical change. Output indicators are hampered by the
differential patenting activity between sectors. We live in a world of the second-best
(or perhaps the fourth- or fifth-best!). So, at best, we need to apply a range of
innovation indicators, in each case interpreting the results with care.
A further indicator of innovation promoted during the early 1980s is the unit price of
output. To the best of our knowledge, this indicator was first used in 1977 in a study
by the then National Economic Development Office focusing on UK competitiveness
(Stout, et al, 1977). This was picked up by Pavitt Keith and his SPRU colleagues in
their 1980 volume entitled Technical Innovation in British Economic Performance. In
his Introduction Pavitt concluded that “[c]ompared with Germany and other major
competitors, Britain is producing unsophisticated machinery and consumer durable
goods, requiring relatively few innovative activities, and having relatively low unit
values and value to weight ratios” (emphasis added) (Pavitt, 1980: 7). This measure
was used by SPRU colleagues to assess the competitiveness of the UK defence sector
(Kaldor, 1980), textile machinery (Rothwell, 1980), and electrical power tools
(Walker and Gardiner, 1980). In recent years there has been renewed interest in the
use of unit prices as an indicator of trade specialisation in US imports (Schott, 2002),
quality and product innovation in EU imports (Aiginger, 2000), in analysing the
impact of trade on employment in Italy (Celi and Smith, 2003), and in models of
import prices in which exporters to the euro area set export prices through a
combination of a mark-up on their production costs (the degree of exchange rate pass1
Traditionally innovation has been thought of in relation to process and product. However,
recent work on global value chains has thrown the spotlight on two other categories of
innovation – functional upgrading (repositioning within the chain) and moving between
chains. See the various contributions in Gereffi and Kaplinsky (eds.) (2001).
2
through) and pricing to market (Anderton, 2003). None of these studies – both in the
1980s and those of a more recent nature - looked in any systematic way at trends in
unit prices, which is the focus of this paper.
The rationale for using unit prices as an indicator of competitiveness is that it harks
back to Schumpeter’s discussion of innovation – low barriers to entry allow
competitors into the market which has the effect of driving prices (and hence incomes
and margins) down. This indicator suffers from two key assumptions. The first is that
cost-reducing technical change is neutral across sectors, since if costs fall more than
prices, than a fall in unit prices may not necessarily be associated with a decline in
margins and incomes. Whilst this is notably a false assumption with regard to the
electronics sector, there is no empirical basis for arguing the validity of this
assumption across other sectors. The second assumption is that the degree of value
added in global trade is either unchanged, or that the changes are invariant across
sectors.. This assumption is also problematic, since the fracturing of global production
processes has meant a thinning of value added in many products (Feenstra, 1998,
IMF, 2002). But once again we have no basis for arguing in any systematic way that
this is uneven across sectors. One way around these objections is to link the
discussion of unit-prices to that of market shares – that is, a combination of rising unit
prices and rising market shares may indicate a virtuous path of innovation which
provides the product rents to sustain growing incomes; and vice versa for falling
prices and falling shares. This is a technique used in an embryonic form by Roy
Rothwell in his study of textile machinery (op cit, 1980), and developed further by
Jeff Readman and myself in a study of the global furniture industry (Kaplinsky and
Readman, 2004).
However, in this paper our primary objective will be to focus on unit prices alone as
an indicator of innovation and competitiveness. We will work with the hypothesis that
there is a direct relationship between unit price performance and innovative
capabilities – rising unit prices are said to reflect growing product innovation and/or
margins protected by barriers to entry, and conversely, falling unit prices reflect the
inability to erect barriers to entry and/or to augment products. (More accurately, the
relative tendency for prices to rise or fall between different sectoral classifications). In
doing so we are fully aware of the dangers of using unit-prices as a measure of
innovation, but we do so in the belief that used in conjunction with other innovation
indicators, it does offer the possibility of enhancing our understanding of the outcome
of innovation processes. This is, therefore, one arrow in an arsenal targeting a fuller
reflection of innovation processes, and the paper should be read with this healthwarning in mind. Our secondary objective is to produce detailed sectoral taxonomies
which others can use in sectoral analysis. Here we have been guided by the need to
disaggregate data to the maximum extent, since our complementary work shows that
the greater the degree of disaggregation, the greater the utility of the data; 2 these
detailed taxonomies can be drawn down from the website we will be developing (see
www.ids.ac.uk/global).
2
Briefly, the problem with much macro-economic analysis (for example, in the relationship
between trade and employment and on the terms of trade) is that it is conducted at a two/three-digit level of disaggregation. Our data shows that it is necessary to go to a much higher
degree of disaggregation if price and industry trends are to become visible. Similar
conclusions are reached by Celi and Smith (2003) in their analysis of the employment effect
of Italian imports from low-wage economies.
3
The dataset we use is the EUROSTAT COMEXT database, which provides monthly
data at the eight-digit product level of European imports and exports in value and
volume from 1988. With the exception of a single classificatory system (see below)
we confine our analysis to the unit prices of manufactured products. We have also
expanded all of the examined sectoral classifications to the six-digit HS level.
In determining the trend in unit prices we began by using the augmented DickeyFuller unit root tests (the ADF test) and the Kalman Filter methodology. The ADF test
T
is based on a regression of the form: Δy t = α + φy t −1 + ∑ ΘΔy t −i + δt + ε t , where εt is
i =1
a random error term, and α and t are a constant and time trend, respectively. The ADF
test corresponds to the value of the t-ratio of the coefficient φ. The null hypothesis of
the ADF test is that yt is a non-stationary series, which is rejected when φ is
significantly negative. Twelve lags, a constant, and a time trend were included in the
ADF regressions of the levels of the variables. For the level variables, the sample is
1988-2001 monthly data. The ADF test determines whether price trends are indeed to
be found. We then used a subsidiary t-test to determine the significance of the slope of
these lines - a minus result indicates a falling trend in prices (the larger the magnitude
the greater the fall in prices) and conversely for rising prices; these are characterised
by various levels of statistical significance.
Unfortunately, the limitations imposed by our data base – only 13 years’ data –
diminished the likelihood of our finding statistically significant price trends over the
whole period, and particularly in determining whether there have been breaks in trend
(for example after the 1997 East Asian crisis). There is no way of getting around these
data limitations and to the best of our knowledge there is no other statistical method
which has the rigour to allow us to conclude whether price trends do exist over a 13
year period. 3 We have therefore estimated unit prices by using two-year moving
average prices – for the initial period (1988/9) and for the final period (2001/2)
In summary, then, our methodology involves an examination of unit-price trends of
EU imports between 1988 and 2001. Our indicator in the analysis of sectoral
classifications will be the proportion of sectors showing falling relative prices.
2. Sectoral differentiation and unit price trends
How can we put this price data to use in illuminating the innovation content in global
production and trade? One way is to examine the unit-price behaviour of products
emanating from different countries and groups of countries – the function of this is to
highlight the inter-country distributional impact of innovation and the robustness of
national systems of innovation. We have undertaken this in a complementary paper
with results which show a clear inverse relationship between per-capita incomes of
country groups and the unit-price of their exports to the EU (Kaplinsky and SantosPaulino, 2004). But resources are more meaningfully allocated at the sectoral level
3
In reaching this decision we consulted widely within the profession. One gathering at which
we discussed methodology was focused on the terms of trade of primary product producers. It
concluded that their time series also was too short to make meaningful use of the ADF and KF
techniques – in their case, their data-set went back to 1926!
4
and it is at this level that our analysis is pitched in this paper. Our question is: to what
extent do changes in unit prices reflect sectoral characteristics? From this it may be
possible to make judgements about the degree of innovation involved in different
sectors and the extent to which, by erecting barriers to entry, this determines the
distribution of returns from production.
In distinguishing sectoral taxonomies to be used in this analysis we draw on the
taxonomies which have been identified by other researchers interested in these
issues. 4 Table 1 lists the characteristics of 23 studies which we were able to identify
and which generated sectoral taxonomies. The elements of these various studies
which we highlight (and use to select categories for sectoral price analysis) are:
4
•
the purpose for which the taxonomy is constructed
•
whether they focus on product or process characteristics
•
whether they use ordinal or cardinal measures (this has an important bearing
on statistical and econometric analysis)
•
whether they use single criteria (for example, R&D intensity) or multiple
criteria (for example, R&D and advertising intensity)
•
the type of data which is involved (for example, trade data, industrial statistics,
innovation data)
•
the level of detail and the number of sectors involved
•
the sectoral categories identified
•
the basis for allocating individual sectors into these categories
•
the time period of this data
•
the source of this data
Peneder (undated) provides a helpful review of many sector taxonomies, including many of
those included in Table 1. We agree with his observation that “in contrast to the prominent
attention it is given in various sciences such as biology, psychology and sociology, the proper
construction and use of classification has remained highly under-researched in the realm of
economics. We still find little or no methodological debate and a striking lack of awareness
for the different approaches pursued (p. 6).
5
Table 1: Summary of Sector Taxonomy Studies
Study
Purpose
Indicator
Process/
product
Type
Mayer, Butkevicius
and Kadri (2002)
Identify sectors of
dynamism in trade
Jaffee, Steven and
Gordon (1993);
World Bank (1994)
Identify high
margin export
sectors
Pavitt (1984)
Identify sectors of
technological
intensity and their
links with firmsize
Test the
Heckscher-Ohlin
theorem on the
determinants of
international trade.
Leamer (1984)
UNIDO (1988)
Explore LDC
capability in
capital goods
production
Products
Single or
multiple
criteria
Ordinal or
cardinal
measurement
Multiple
Dynamic
products
Product
Cardinal
Income
elasticity of
demand
Process
Cardinal
Nature of
innovation
Ordinal
Process
Multiple
Factor
intensity
Cardinal
Process
Multiple
103 indicators
Cardinal
Single
Multiple
Type of data
Level
detail
#
sector
categories/
subsectors
Categories
Basis of
allocation
Time
period
Source of data
Trade data
3-digit
20/20
Dynamic products
Authors’ use of
analytical criteria
1980-1988
COMTRADE – SITC
(REV3@?)
Trade data
3-digit
7/17
Income elastic products
Authors’ use of
analytical criteria
Late 1980searly 1990s
SITC REV2
Database of
innovations
11 2-digit,
and 26 3and 4digit ISIC
4 (37)
Supplier-dominated;
Production intensive (ScaleIntensive and Specialised
Suppliers); science-based
Judgement of
engineering
experts
1945-1980
Significant UK
Innovations – SPRU
database
2-digit
3-digit
10 (61)
2 primary products
(petroleum and raw
materials), 4 crops (forest
products, tropical/
Mediterranean agricultural
products, animal products)
and 4 manufactures (labourintensive, capital-intensive,
machinery, chemicals).
Author’s
judgement,
secondary
sources, use of
analytical
techniques
1948-1973
High
4
(1,100)
capital goods used: to
produce other capital goods;
used to produce intermediate
goods; used to produce
consumer goods; used
across sectors.
Engineering
experts using
analytical criteria
Unspecified
– but 10
year period
of
allocation
Trade data from UN
sources; capital from
national accounts,
resources from various
sources, skills from ILO
Yearbook. Sectoral
definitions of capital
and labour intensity
based on 1963 US data,
and drawn from
Hufbauer 1970.
UNIDO database
Trade data,
skills, factor
inputs
Unspecified,
but focus on
manufacturing
process
6
Table 1: Summary of Sector Taxonomy Studies (cont.)
Study
Purpose
Indicator
Process/
product
Type
Forstner and
Ballance (1990)
Wood (1994)
Wood, and Berge
(1997); Wood and
Mayer (2002)
Identify
determinants of
global trade
Explain
distribution of
income and
employment
Explain
differential LDC
exports of
manufactures
Wood and Mayer
(1998)
Explain
differential LDC
exports of
processed and
unprocessed
primary products
UNCTAD 1996
Explain source of
upgrading in NIEs
Process and
product
Single or
multiple
criteria
Ordinal or
cardinal
measurement
Multiple
Type of data
Level
detail
#
sector
categories/
subsectors
Categories
Basis of
allocation
Time
period
Source of data
Capital, labour
(skilled and
unskilled);
trade data
3-digit
4/25
High- and low- growth,
labour- and capital-intensive
Authors’ use of
analytical
criteria;
secondary
sources
1970 and
1985
ISIC (national
accounts), ISCO (ILO)
with concordance to
SITC
3/147
Ricardian, H-O, product
cycle
Factor
endowments;
product cycle
goods
Cardinal
Process
Multiple
Cardinal
Labour (skilled
and unskilled);
trade data
3-digit
3 (NA)
Primary, processed primary,
narrow manufactures
Multiple
Cardinal
Skills and trade
data
2- and
occasional
3-digit
4 (11)
Unprocessed- and
processed- primary, labourand skill-intensive
manufactures
Multiple
Cardinal
Trade data,
educational data
and resource
data
3-,4- and
occasional
5-digit
6 (188)
Process
Multiple
Trade data
5 (38)
Skill-, capitaltechnology
and scaleintensity
Ordinal
2- and
occasional
3-digit
Processed/unprocessed
minerals, metals, fuels;
Processed/unprocessed
dynamic agricultural
products; Processed/
unprocessed static
agricultural products
Non-fuel primary; labourand resource-intensive; lowskill, low-capital and lowtechnology; medium- skill,
medium-capital and medium
technology; high- skill,
high-capital and high
technology;
Factor
intensity
US
Authors’ use of
analytical
criteria;
secondary
sources
1960s,
1970s and
early 1980s
US 1981,
UNIDO
early 1990s
Skill levels from Barro
and Lee; SIC
concordance to SITC;
SITC COMTRADE
Judgement of
author
Authors’ use of
analytical
criteria;
secondary
sources
ISIC with concordance
to SITC COMTRADE
1965, 1975,
1985 and
1994
Skill levels from Barro
and Lee; SIC
concordance to SITC;
SITC COMTRADE ;
dynamic income elastic
trade data from
unpublished sources
SITC COMTRADE
7
Table 1: Summary of Sector Taxonomy Studies (cont.)
Study
Purpose
Indicator
Process/
product
Type
Marsili (2001)
Choudhri and
Hakura (2000
OECD (1992)
OECD (1994),
updated 2003.
Hatzichronoglou,
(1997)
Identify sectors of
dynamic
comparative
advantage
Identify
manufacturing
sectors with rapid
productivity
growth
Process
Various, based
on limitations
of R&D
(input) and
patent (output)
statistics
Process
Single or
multiple
criteria
Oo
rdinal or
cardinal
measurement
Single
Cardinal
Single
Type of data
Level
detail
# sector
categories/
subsectors
Categories
Basis of
allocation
Time
period
Source of data
Various – incl
patents, R&D,
skills, citations
Mostly
2-digit
SIC,
some 4digit
Various
but key
is
5 (18)
Learning-regimes: - sciencebased; fundamental process;
complex systems; product
engineering; continuous
process.
Author’s use of
analytical criteria
Various –
mostly
1990s
US National Science
Federation
SPRU database on
patents and global firms
PACE database on
European innovation
Input and
output data
2-digit
4 (9)
Non-Manufacturing; High-,
medium- and low-TFP
growth
Use of analytical
criteria
1970-1993
R&D data
2- and
occasion
al 3-digit
6 (36)
Non-fuel primary; labourintensive manufactures;
differentiated products
requiring specialised
suppliers; scale-intensive
manufactures; science-based
manufactures
High-tech; medium-high
etch; medium-low tech; lowtech
Authors’ use of
analytical criteria
Late 1980s
OECD International
Sectoral Database.;
UNIDO Industrial
Statistics Database
(Indstat3) UN SNA;
Feenstra, Lipsey and
Bowen (1997)
US R&D data converted
to SITC data
Aerospace; computersoffice; electronicstelecomms; pharmacy;
scientific instruments;
electrical machinery;
chemistry; non-electrical
machinery; armaments
Judgement of
engineering
experts
Total Factor
Productivity
Cardinal
Identify hightechnology sectors
to promote
industrial
development
Process
Multiple
R&D intensity
in production
(direct and
indirect)
Ordinal
Identify hightechnology sectors
to promote
industrial
development
Identify hightechnology sectors
to promote
industrial
development
R&D intensity
in production
(direct and
indirect)
R&D data
3- and
occasion
al 4-digit
4 (27)
R&D and
innovation
intensity of
products
R&D and
production data
4-digit
9 (76)
Authors’ use of
analytical criteria
Early 2000s
R&D data from 10
OECD countries
converted to SITC data
1988-1995
R&D data from 6
OECD countries,
converted to SITC data
8
Table 1: Summary of Sector Taxonomy Studies (cont.)
Study
Purpose
Indicator
Process/
product
Single or
multiple
criteria
EUROSTAT
(1995), cited in
Pearson and Jagger
(2003).
Identify sectors
with technological
intensity
Process
Ordinal or
cardinal
measurement
NA
Skills
Ordinal
Lall (2000)
Identify export
sectors which
promote dynamic
comparative
advantage
Process
Multiple
Technologyintensive and
capability
building
criteria
Process
Ordinal
Unit costs,
volumes,
customisation,
design variety,
diversity of
knowledge;
number
components/s
ubsystems,
interaction
with users
Process
Ordinal
Type
Acha et al (2002)
Schmoch et al
(2003)
Identify complex
production system
products – “high
cost, engineeringintensive products,
systems, networks
and constructs”
Link between
technology and
economic
performance
Innovation
and
production
indicators
Multiple
Single
Ordinal
Type of data
Level
detail
#
sector
categories/
subsectors
Categories
Basis of
allocation
Time
period
Source of data
NA
2-digit
SIC
6 (78)
NA
NA
NA
Not specified
3-digit
SITC
Rev2
5 (9)
(230)
Primary production; Hightech Manufacturing;
Medium-high-techmanufacturing; low-techmanufacturing; Knowledgeintensive services; Other
services
Primary products; Resourcebased products; Low-tech
products; Medium-tech
products; High-tech
products
Judgement of
researcher
Late 1990s
UN COMTRADE
Gross wages/
employee;
purchases of IT
services;
purchases of
telecoms;
expenditure on
branding and
advertising
3-digit
and 4digit
SIC92
1 (503
4-digit
and
253 5digit)
Complex production system
products
Judgement of
authors and use
of analytical
criteria
1997-1999
UK Annual Business
Inquiry
Patent statistics
and SIC
categories
2-digit
SIC and
65 IPC
patent
classes
44 SIC
65 IPC
NA
Use of analytical
criteria
1997
European Patent Office
9
Study
Purpose
Indicator
Process/
product
Type
Neven (1994)
Davies
and Lyons et al
(1996)
Aiginger (2000)
Sutton (1998)
Single or
multiple
criteria
Ordinal or
cardinal
measurement
Multiple
Identify factor
content of trade in
order to asses
welfare impact of
trade between EU
and E. Europe.
Process
Labourintensity,
capitalintensity,
wage levels,
skills
Cardinal
(i) assemble a
Europe-wide
industrial database
(ii) develop new
taxonomies of
industrial structure
Identify sectors
where quality
rather than price is
significant factor
To explore the link
between R&D
intensity and
concentration
Process
Multiple
Innovation
Ordinal (binary
category)
Product
Single
Qualityelasticity
Process and
product
Ordinal
Multiple
Type of data
Level
detail
#
sector
categories/
subsectors
Categories
Basis of
allocation
Time
period
Source of data
Wages, value
added,
investment,
skills
NACE 3
and some
4 digit
5 (140)
(i) High-tech, high human
capital (high wages/VA, high
avg wage, high white collar)
(ii) High human capital, low
invest (low invest/VA, high avg
wages, high wage/VA)
(iii) Lab intensive (low avg
wage, high wage/VA, low
invest/VA)
(iv) Labour and capital intensive
(high invest/VA, low avg wage,
low white collar, intermediate
wage/VA)
(v) Human capital and invest
intensive (high avg wages,
intermediate wages/VA, high
invest/VA, high white collar)
Cluster analysis
1985-1990
SIC - Germany
(triangulated with other
11 EU countries)
Scale, R&D,
advertising,
ownership
3-digit
NACE
100 (4)
Based on advertising and
R&D
Use of analytical
criteria (R&D
and advertising
intensity)
1987
Advertising – UK
commercial agency;
R&D from UK and Italy
census of production
Trade – unit
values and trade
balance
3 digit
SIC
3 (93)
High, medium and low
“Revealed Quality
Elasticity”
1988-1998
EUROSTAT
R&D,
advertising
intensity,
product
homogeneity
4- and
some 5digit SIC
2 (34
and
119)*
R%D Intensive and Low
R&D, low-advertising
intensive
Original
indicator using
trade (price and
volume) data
Use of analytical
criteria
1977
US Census of
Manufacturing and Fair
Trade Commission
Innovation
Ordinal
intensity and
product
homogeneity
* Sutton’s analysis uses 34 R&D intensive sectors (R&D/sales ratio of >4%) and bottom 50 sectors with low R&D and advertising intensity control group. However, the 50 low-innovation control group is never indentified
so we use the 119 sector population of low-innovation intensive firms from which the 50-sector sample was constructed
.
10
Each of these elements is relevant for different uses. However, in choosing a set of
classifications for price analysis in this innovation-focused paper, we have taken
account of the following issues:
•
Loosely-speaking it is possible to distinguish three types of sectoral
classifications – those focusing on product characteristics (income elasticity,
for example), those on factor content (notably capital and labour intensity),
and those targeted at innovation- and technology-intensity; clearly it is the
latter focus which will inform this unit-price analysis
•
Many of the sectoral classifications which have been developed use very old
data. The problem is not just with the age of the data, but also that where they
involve structural relationships (for example, factor intensity) the nature of
these input-output relationships might have changed significantly over time.
(This is particularly true of the classic study by Leamer which was published
in 1984 using data from the 1970s and which is still widely used in the
definition of factor-intensity – Leamer, 1984).
•
We have striven to achieve as much details as possible and have therefore
tried to go for maximum sectoral disaggregation, in all cases extending the
initial two- and three-digit level classifications of the original sector
classifications to the six-digit level. The reason for this is that our
complementary analysis has shown that the incidence of unit-price trends is
directly related to the degree of disaggregation (Kaplinsky and Santos Paulino,
2004) 5
•
With the exception of a group of resource-based industries identified by Lall,
we have confined the analysis to manufacturing sectors. Resource-based
sectors have already received extensive price-analysis (notably by the terms of
trade literature – see, for example, the classic by Singer, 1950) and our
ultimate objective is to chart the growing competitiveness in the
manufacturing sector and changes in the intra-manufacturing sector’s terms of
trade. Services are excluded because they are not covered in the EU COMEXT
database.
Based on these criteria, we have tested unit-price trends for the following taxonomies:
•
5
Davies and Lyons’ distinction between sectors with no quality focus, R&D
intensive sectors and R&D+advertising intensive sectors at the two-digit level
(we have extended this to the four-, and six-digit level). This has the
advantage of recognising both formal R&D inputs and firms’ investment in
market-based and product-oriented intangibles. It is also based on the
application of criteria (the share of R&D and advertising in sales). The
downside is that these sectors are defined on the basis of 1987 data.
As Schott observes, “using [aggregated] industry-level data [is] problematic because much of
the factor proportions action occurs at a level that is hidden from researchers” (Schott, 2002:
3).
11
•
Neven’s distinction between high-tech/high human capital, high human
capital/low invest, labour intensive, labour/capital intensive, and human
capital/investment intensive sectors has two primary strengths. First it is based
on multiple criteria, and is trade-focused. Secondly the sector categorisation is
derived from cluster analysis which is an inductive approach which arguably
better reflects sector characteristics than the didactic and often personal
methodologies used by other authors. The downside is that it reflects (West)
German economic structure (albeit triangulated with other industrially
advanced countries) and is dated (1985-1990). We have extended his fourdigit level taxonomy to thesix6-digit level.
•
UNCTAD’s categorisation of labour/resource intensive, low-skill/lowtech/low capital intensive, medium-skill/medium-tech/medium-capital
intensive, and high-skill/high-tech/high-capital intensive sectors at the threedigit level (we have extended this to the six-digit level). The strengths of this
nomenclature are that it is linked to an analysis of inter-country technological
capabilities and is based on multiple criteria (more closely reflecting the
complexity of factors affecting competitiveness). On the other hand, much of
the data on which these judgements were made – based on an assessment of
individual UN desk-officers rather than the application of criteria – is dated.
•
The OECD process categorisation is based on R&D inputs into production and
distinguishes low technology, medium-low technology, medium-high
technology and high-technology sectors. It uses data from the second half of
the 1990s, but is only based on a single criterion (which we know from the
literature provides only a partial perspective on innovation) and is defined at a
high level of aggregation. We have extended their two-digit level taxonomy to
the six-digit level.
•
Lall’s distinction between resource-based, low technology, medium
technology, engineering and high technology sectors at a three-digit level
(although we have decomposed this to the six-digit level). This categorisation
has the distinction of being recent (late 1990s) and detailed; however, the
downside is that the allocation of sectors reflects the judgement of the author,
which is inevitably based on partial knowledge.
•
The COPS classification of sectors provides a new and stimulating taxonomy
of a specific category of sectors. We have extended their three-digit
classification to the four-digit level.
•
UNCTAD recently produced an analysis of the 20 most rapidly growing
products in global trade (Mayer, Butkevicius and Kadri, 2003). We have
excluded resource and primary products, and have expanded the 13 three-digit
manufacturing classification to 237 six-digit sectors.
•
Sutton’s classification of R&D-intensive sectors.
In total we therefore examined the unit price behaviour of 12,4390 sub-sectors. In
drawing on these technological trajectories we have had to undertake a great deal of
work in translating the various nomenclatures used - SITC (various Revisions), ISIC
12
(various revisions), NACE (various revisions) and ISCO – into the HS nomenclature
utilised in the EU COMEXT database. In deepening the detail of the analysis we have
also extended the two- and three-digit classifications to four- and six-digit levels.
Inevitably there are also some cases where the translation between the processoriented ISIC production taxonomies are not adequately captured by the productoriented trade classifications (HS/SITC), although we have utilised the established
protocols for this translation.
3. Results
So what emerges from the analysis of price trends of the various sectoral taxonomies?
It is possible to conclude from Table 2 that the median proportion of subs-sectors
showing a negative unit-price performance in the 1988-2001 period is around twothirds. For example, the proportion of sectors with falling prices in the four largest
sectoral investigations (OECD, UNCTAD, Lall and Neven) is 66 percent, 65 percent,
63 percent and 62 percent respectively. We will therefore use this figure of 60-65
percent of sectors displaying falling unit prices as a benchmark for the “average” unitprice performance of sectoral classifications selling into the EU.
The data can be interpreted in two ways. The first is in pursuit of the hypothesis that
price reduction is a reflection of low barriers to entry and that technology, innovation
and knowledge are important barriers to entry. Therefore, the greater the technology-,
innovation- and knowledge-intensive the product, the less likely that its prices will
fall. An alternative use of the data is to assume that the proportion of sectors with
falling prices will be lower for technology- and innovation-intensive products, and
therefore that the data can be used to test the extent to which different sector
classifications reflect these characteristics.
3.1. Price performance reflects technology-, innovation- and knowledge-intensity.
Broadly speaking, the results of the sectoral investigation is as follows:
•
In the case of Davies and Lyons’ classification, there is a substantive
difference between those sectors which are advertising and R&D intensive and
other sectors which either have no quality focus, or are only R&D intensive.
This would seem to confirm the basic hypothesis that price behaviour reflects
innovation/knowledge intensity.
•
Neven provides a complex, multi-faceted classificatory system. The two
sectoral categories which stand out as being less susceptible to price declines
are the high-tech/high human capital and the human capital/investment
intensive sectors. An interesting result of this analysis is that there was little
difference in the susceptibility towards price-decline between the labourintensive and the labour-/capital-intensive sectors; as Neven points out, in the
contemporary world of high liquidity, capital is almost as widely available as
labour (a similar assumption underlies Wood’s widely-cited analysis of the
employment effects of global trade – Wood, 1994)
13
•
The UNCTAD technology-schema shows little differences between the
different classifications. However its more recent category of dynamicallytraded products does not suggest that these products are less susceptible to
falling unit prices. (In fact the opposite might be expected, with growing trade
being associated with falling prices; but there is no evidence of this either).
•
Using the OECD classificatory system, there is evidence to suggest that the
medium-high and the high tech sectors are least susceptible to price fall.
•
Lall’s classificatory system shows that the high-tech sectors are relatively
unaffected by price-pressures; what is interesting from his analysis is that the
low- tech and engineering sectors are more affected by falling prices, no doubt
reflecting the growing participation of new entrants (notably China) during the
1990s.
•
Perhaps surprisingly, the Complex Production System products show no
particular tendency to resist pricing pressures; however the one-off,
customised nature of these products makes it less likely that the products
traded during this period will be easily comparable.
•
Sutton’s R&D intensive sub-sectors are relatively unaffected by falling prices,
confirming the conclusions arising from Davies and Lyons’ related categories.
14
Table 2: Unit price behaviour, 1988-2001*
Sector
Davies and Lyons
Total
No quality focus
R&D Intensive
Advertising Intensive
R&D and Advertising Intensive
Neven
Total
High-tech, high human capital
High human capital low invest
Labour intensive
Labour and capital intensive
Human capital and invest intensive
UNCTAD
Total
Labour/resource intensive
Low-skill/low-tech/low capital intensive
Medium-skill/medium-tech/medium capital
intensive
High-skill/high-tech/high capital intensive
OECD
Total
Low
Medium low
Medium high
High
Lall
Total
Resource-based
Low technology
Medium technology
Engineering
High technology
COPS (4-digit)
Manufacturing
UNCTAD
Dynamic products
SUTTON
R&D Intensive
Total
Positive slopes
Negative slopes
Number
%
Number
%
571
297
275
6
84
235
93
112
4
44
41
31
41
67
52
336
204
163
2
40
59
69
59
33
48
1,904
585
907
406
424
6
719
263
357
136
139
4
38
45
39
33
33
67
1,185
322
550
270
285
2
62
55
61
67
67
33
3,632
1,118
430
738
1,287
343
142
264
35
31
33
36
2,345
775
288
474
65
69
67
64
1,043
432
41
611
59
3,816
1,215
767
1,451
384
1,297
362
204
544
188
34
30
27
37
49
2,519
853
563
907
196
66
70
73
63
51
2,006
472
674
295
336
245
737
185
196
120
111
119
37
39
29
41
33
49
1,269
287
478
175
225
126
63
61
71
59
67
51
69
29
42
40
58
322
141
44
181
56
144
71
49
73
51
In summary, we believe that it is fair to conclude that despite the differences in the
classificatory systems utilised and despite ambiguities within and between these
classifications, the evidence would seem to bear out the hypothesis that the more
technology- and knowledge-intensive the sector, the less likely it is that unit-prices
will fall.
15
3.2. How does the data illuminate sectoral classifications?
If it is assumed that price behaviour reflects technology-, innovation- and knowledgeintensity, the results in Table 2 can be sued to interrogate the robustness of various
sectoral classifications aiming to address these elements of factor-intensity. The main
conclusions which can be drawn are as follows:
•
Davies and Lyons show the importance of a combination of process (R&D)
and product (advertising) intensity. The small number of advertising- (but not
R&D-) intensive sectors makes it difficult to support what looks likely to be
an especially strong association between branding and price performance,
although there are strong a priori reasons to suppose that brand-intensive
products are relatively immune from price pressure.
•
Neven presents a complex amalgam of sectors. The only category which
seems relatively immune from price pressure is one with very low sectoral
representation, that is human capital/investment intensive; however, the fourdigit analysis (which due to data unevenness in the COMEXT dataset)
provides data on 13 sub-sectors in this category (out of a total of 746 subsectors) and this, too, shows a lower incidence of price decline (46 percent).
•
The UNCTAD classification shows little difference between the price
performance of the individual categories, bar that for the high-skill/hightech/high-capital intensive group, a similar conclusion to that of Neven.
•
The OECD classification seems to provide supportive results for the mediumhigh and high-tech categories; but the medium-low technology group performs
in a similar nature to the low-tech group.
•
Lall’s classification is probably most clearly supported by the data. There is a
clearly a smaller tendency for prices to fall the greater the technologyintensity; the engineering industries reflect a particular subset of
competitively-traded goods and do not necessarily align with technology
intensity. Interestingly, the data suggests that the resource-intensive sectors are
subject to a somewhat lower degree of price competition.
•
Although the COPS category is made-up of a number of knowledge-intensive
one-off products, this does not appear to be reflected in this sector’s price
performance
•
Sutton’s widely-used R&D intensive classification is supported by the data;
however it does not appear to corroborate the R&D intensive classification
provided by Davies and Lyons.
16
4. Discussion and Conclusions
In this paper we have examined the links between price performance in globallytraded goods and technology- and innovation-intensity. Assuming away the problem
of cost-reducing productivity change and differential “vertical disintegration of trade”
(Hummels et. al., 1998) (heroic assumptions perhaps, but no more heroic than those
made in studies which measure innovation through either input or output indicators),
this is a logical step in neo-Schumpeterian analysis. The results arising from the
examination of unit-price trends in more than 12,000 different sub-sectors provides
qualified support both for the primary hypothesis underlying this paper and for some
of the received classifications of technology-innovation and knowledge-intensity. We
can therefore conclude that the analysis of unit-price performance is a valid technique
to be used in innovation studies, in concert with other similarly-flawed measures such
as R&D intensity and patenting activity. (We have placed the data on our website –
www.ids.ac.uk/global/ - so that they can be more widely used).
In undertaking this analysis we have surveyed 23 different sectoral classifications. We
have two primary concerns about these received taxonomies. First, in most cases they
are based on dated economic structures even though we have deliberately excluded
some of those such as the classic and frequently-cited study by Leamer which relies
on pre-1973 data on economic structures (and which to our surprise – shock? - is still
used in contemporary classifications). And, second, they are almost all based on
aggregative two-digit and three-digit data.
In response to these weaknesses, rather than replicating the static price analysis found
in much of the literature (Celi and Smith, 2003; Schott, 2002; Aiginger, 2000), we
have focused on changes in prices. However, the short duration of the COMEXT
data-base (1988-2002, albeit with monthly data) makes it difficult for any acceptable
statistical technique to verify price-trends, and for this reason we have only calculated
average unit price trends. In addition, we have given primacy to sectoral
disaggregation, widening the two- and three-digit classifications used in received
studies to six-digits. Our complementary analysis of unit-price trends (using a larger
number of sectors than those involved in this paper on innovation-intensive sectors)
shows that the higher the degree of disaggregation, the greater the incidence of price
trends (Kaplinsky and Santos-Paulino, 2004), a conclusion corroborated by Celi and
Smith (Celi and Smith, 2003).
The major analytical conclusion which arises from this research is the need to push
forward classificatory systems to both embody greater detail and more recent
structural relationships. Of all the 23 classifications which we have examined, only
Lall’s begins to meet these challenges. However, as can be seen from Table 1, Lall’s
classification is based on the author’s judgement, and whilst his expertise is
considerable, it is a poor substitute for the use of measured structural relationships to
define different sectors. Can we make further progress on this front? There are two
possible data-sets which we have identified which hold promise, and which provide
the capacity for the integration required to provide a comprehensive picture of
sectoral dynamics. The first is Office of National Statistics Annual Business Survey
covering 78,500 enterprises and conducted most recently to cover the years 1997-
17
2001. 6 The second are the various Community Innovation Surveys conducted in
various EU economies which are based on the Oslo Manual. These data-sets are not
ideal, since the level of detail they provide only allows two- and perhaps three-digit
data-analysis. However, their recent vintage means that they will provide an
opportunity to update the structural relationships in the various received classificatory
systems reviewed in Table 1. They also provide a combination of input- and outputbased innovation indicators, including the use of ICT and advertising and marketing
intensity.
But what of the policy conclusions which stem from our analysis? Again, assuming
away the problem of differential productivity change and value-added thinning
between sectors, and therefore assuming that unit-price performance reflects the
income streams associated with global trade, there are clear conclusions which arise
from the data. First, brand-intensity – despite the small sample in the Davies and
Lyons analysis – is probably an important signifier of sustainable income, a
conclusion which is corroborated in the growing volume of value chain literature
(Gereffi and Kaplinsky, 2001; Gereffi, Humphrey and Sturgeon, 2004). Second, the
lower the technological content in products – reflected in a range of measures – the
more likely that price pressure will be felt. And, third, perhaps surprisingly to some,
resource-intensive sectors may not be under as much relative price-pressure as has
been assumed in much of the literature on terms of trade and industrialisation, since
the growing presence of China in world trade of manufactures has become manifest.
These conclusions are relevant to the corporate sector, although for some time the
corporate sector has been implementing appropriate strategies, particularly those firms
based in the high-wage economies. But more pertinently, the conclusions need to be
absorbed by public policy actors in low-wage economies, and by the industry of
consultants and multi-lateral and bi-lateral agencies who advise (and perhaps more
often “guide”) resource allocation in low wage economies. For example, the World
Bank’s definitive statement of its position on globalisation in 2002, concluded that the
exports of labour-intensive and low-technology manufactures has been the major
factor in allowing developing countries to grow rapidly and to alleviate poverty
(World Bank, 2002). Whilst the analysis of export prices alone does not allow this
conclusion to be definitively questioned, it does suggest that there may be a fallacy of
composition in this policy prescription (Mayer, 2002). Maizels’ two-digit studies of
the falling terms of trade between developing countries and the EU, the US and Japan
(Maizels et al, 1998 and 1999; Maizels, 2003) supports this cautionary conclusion.
6
For a discussion of the ABI, see Jones 1990
18
REFERENCES
Anderton, R. (2003), ‘Extra-Euro Area Manufacturing Import Prices and Exchange
Pass-Through’, Working Paper No. 219, European Central Bank.
Aiginger, K. (2001), “Europe’s Position in Quality Competition: Background Report
for the European Competitiveness Report 2000”, Enterprise Papers No. 4.
Luxembourg: European Communities.
Celi, G. and A. Smith (2003), “Quality differentiation and the labour market effects of
international trade”, mimeo, Brighton: University of Sussex
Choudhri, E. U. and D. S. Hakura (2000), “International Trade and Productivity
Growth: Exploring the Sectoral Effects for Developing Countries”, IMF Staff
Papers, Vol. 47, No. 1, pp 30-53.
Davies, S. and B. Lyons et. al. (1996), Industrial Organization in the European Union:
Structure, Strategy, and the Competitive Mechanism, Oxford: Clarendon
Press.
DTI (20020, Community Innovation Survey, London: DTI
Feenstra R. C. (1998), ‘Integration of Trade and Disintegration of Production in the
Global Economy’, Journal of Economic Perspectives, Vol. 12, No. 4, pp.31–
50.
Forstner H and Ballance R (1990), Competing in a Global Economy, London, Unwin
Hyman/UNIDO.
Gereffi, G. and R. Kaplinsky (eds.) (2001), IDS Bulletin Special Issue on The Value
of Value Chains, Vol. 32, No. 3.
Gereffi G., T. Sturgeon and J. Humphrey (2004, forthcoming), The Governance of
Global Value Chains, Review of International Political Economy.
Hartzichronoglou, T. (1997), “Revision of the High-Technology Sector and Product
Classification”, STI Working Paper 997/2, OCDE/GD(97)216, Paris: OECD
Hummels D., Jun Ishii and Kei-Mu Yi (1998), “Vertical Specialization and the
Changing Nature of World Trade”, Federal Reserve Bank New York
Economic Policy Review, 79-99.
IMF (2002), World Economic Outlook, Washington: IMF.
Jaffee, S. and P. Gordon (1993), “Exporting high-value food commodities: success
stories from developing countries”, World Bank Discussion Paper, 198,
Washington, D.C.;
Jones, G. (2000), “The Development of the Annual Business Enquiry”, Economic
Trends, No. 564, 49-57.
Kaldor, M. (1980), “Innovation and Competitiveness in the Military Sector”, in K.
Pavitt (ed.), Technical Innovation in British Economic Performance, London:
Macmillan
Kaplinsky, R. and J. Readman (2004), “Globalisation and Upgrading: What can (and
cannot) be Learnt from International Trade Statistics in the Wood Furniture
Sector?”, mimeo, Brighton, Centre for Research in Innovation Management,
University of Brighton and Institute of Development Studies, University of
Sussex.
Kaplinsky, R. and A. Santos-Paulino (2004), “The unit-price performance of EU
imports of manufactures, 1988-2001 and the implications for the analysis of
trade and technology patterns” mimeo, Brighton: Institute of Development
Studies, University of Sussex.
Lall, S. (2000), “The Technological Structure and Performance of Developing
Country Manufactured Exports, 1995-1998, Oxford Development Studies,
19
Vol. 28, No. 3, pp 337-369.
Leamer, E. E. (1984), Sources of International Comparative Advantage, Cambridge,
Mass: MIT Press.
Maizels, A., T. Palaskas and T. Crowe (1998), “The Prebisch-Singer Hypothesis
Revisited”, in Sapsford, D. and J. Chen (eds.) (1998), Development
Economics and Policy: The Conference Volume to Celebrate the 85th
Birthday of Professor Sir Hans Singer, Basingstoke: Macmillan.
Maizels, A., K. Berge, T. Crowe and T. B. Palaskas (1999), ‘The Manufactures Terms
of Trade of Developing Countries with the United States, 1981–97’, mimeo,
Oxford: Finance and Trade Policy Centre, Queen Elizabeth House.
Maizels, A. (2003), The Manufactures terms of Trade of developing and developed
countries with Japan, 1981-2000, mimeo, Oxford: Queen Elizabeth House.
Marsili, O. (2001). The Anatomy and Evolution of Industries: Technological Change
and Industrial Dynamics, Cheltenham: Edward Elgar.
Mayer, J., A. (2002), “The Fallacy of Composition: A Review of the Literature”, The
World Economy, Vol. 25, No. 6, pp 875-894.
Mayer, J., A. Butkevicius and A. Kadri (2002), “Dynamic Products in World
Exports”, Discussion Paper No. 159, Geneva: UNCTAD.
Neven, D. (1994), “Trade Liberalisation with Eastern Nations: How Sensitive?”, in R.
Faini, and R. Portes (eds), European Union Trade with Eastern Europe:
Adjustment and Opportunities, London: CEPR.
OECD (1992), Industrial Policies in OECD Countries, Annual Review 1992, Paris:
OECD.
OECD (1994 ), Industrial Policies in OECD Countries, Annual Review 1994, Paris:
OECD.
Office of National Statistics (2002), Annual Business Survey, London: ONS
Patel P and Pavitt K (1988), "The International Distribution and Determinants of
Technological Activities", Oxford Review of Economic Policy, Vol 4, pp 3555.
Pavitt, K. (1980), Introduction and Summary in K. Pavitt (ed.), Technical Innovation
in British Economic Performance, London: Macmillan.
Pavitt, K. (ed.) (1980), Technical Innovation in British Economic Performance,
London: Macmillan.
Pavitt, K. (1984), “Sectoral patterns of technical change: Towards a taxonomy and a
theory”, Research Policy, 13, pp. 343-373.
Peneder, M. (2003), “Industry Classifications: Aim, Scope and Techniques”, Journal
of Industry, Competition and Trade, 3:1/2, 109-129, 2003
Rothwell, R. (1980), “Innovation in Textile Machinery”, in K. Pavitt (ed.), Technical
Innovation in British Economic Performance, London: Macmillan
Schmoch, U., F. Laville, P. Patel and R. Frietsch (2002), “Linking Technology Areas
to Industrial Sectors”, mimeo, SPRU, University of Sussex.
Schott, P. K. (2002), Across-Product versus Within-Product Specialization in
International Trade”, mimeo, Yale School of Management,
Singer H W (1950), “The Distribution of Gains between Investing and Borrowing
Countries”, American Economic Review, 15, pp. 473-85.
Stout, D. et al (1977), International Price Competitiveness, Non-Price Factors and
Economic Performance, London: National Economic Development Office.
Sutton, J. (1998), Technology and Market Structure: Theory and Practice, Cambridge,
Mass: MIT Press.
UNCTAD (1996), Trade and Development Report, N York, United Nations.
20
UNIDO (1988), Industry and Development, Vienna: UNIDO.
Walker, W. and J. P. Gardiner (1980), “Innovation and Competitiveness in Portable
Power Tools”, in K. Pavitt (ed.), Technical Innovation in British Economic
Performance, London: Macmillan
Wood Adrian (1994), North-South Trade, Employment and Inequality: Changing
Fortunes in a Skill-Driven World, Oxford: Clarendon Press.
Wood, A. and J. Mayer (1998), “Africa’s export structure in a comparative
perspective”, Study No. 4, African Development in a Comparative
Perspective, Geneva: UNCTAD.
Wood, A. and K. Berge (1997), “Exporting manufactures: human resources, natural
resources, and trade policy”, Journal of Development Studies, 34, pp. 35-59.
World Bank (1994), Adjustment in Africa. Reforms, Results, and the Road Ahead,
Oxford and New York: Oxford University Press.
World Bank (2002), Globalization, growth, and poverty: building an inclusive world
economy, Policy Research report, Washington: World Bank and Oxford:
Oxford University Press.