Cambridge Journal of Economics Advance Access published February 5, 2016

Cambridge Journal of Economics 2016, 1 of 23

doi:10.1093/cje/bev067

International competitiveness in postKeynesian growth theory: controversies

and empirical evidence

A fundamental starting point for post-Keynesian theory concerning growth in

open economies is succinctly expressed in the following statement by Kaldor:
[T]he main autonomous factor governing both the level and the rate of growth
of effective demand of an industrial countryis the external demand for its exports:
and the main factor governing the latter is international competitiveness, which in
turn depends on the level of its industrial cost relatively to other industrial exporters
(Kaldor, 1971, p. 7; italics added). Moreover, thanks to increasing returns in
manufacturing, export expansion and international competitiveness would interact so as to create vicious or virtuous circles of cumulative causation. Afew years
later Kaldor, having found a positive correlation between the time changes of
the main industrial countries relative manufacturing export shares and that of
their relative unit costsa correlation that became known as the Kaldor paradoxdismissed his original cumulative causation theory and adopted a version
close to Harrods foreign trade multiplier. The purpose of this paper is to reaffirm the Kaldorian cumulative causation theory in its original version, by providing a firmer analytical basis and showing that, contrary to the Kaldor paradox,
time changes in export performance must be explained by levels rather than by
changes in unit costs.
Keywords: International competitiveness, Post-Keynesian growth theory, Kaldor
paradox, replicator equation
JEL classifications: F43, O10

Manuscript received 8 August 2013; final version received 13 January 2015.

Address for correspondence: Laura Barbieri, Universit Cattolica del Sacro Cuore, Dipartimento di Scienze
Economiche e Sociali, Via Emilia Parmense, 84 29122 Piacenza, Italy.
*Universit Cattolica del Sacro Cuore. Financial help by 2007 PRIN Heterogeneous Sectors, Growth
and Technical Change is gratefully acknowledged. The authors would like to thank the Final Conference
participants of that project, held in Naples (2729 April 2011), as well as Antonella Palumbo and other
participants of the S.I.E., Annual Scientific Meeting, held in Rome (1415 October 2011)for the stimulating discussions and, last but not least, they would like to thank Bruno Soro for his insightful comments on
Kaldors and Harrods views. This paper is based on the research that was being carried out by Professor
Boggio, who died unexpectedly as the work was coming to its conclusion. Laura Barbieri is deeply conscious
of the debt she owes him finishing the work in his absence and feels honoured to have collaborated with him
on the project. She would like to extend her gratitude to all those who assisted her in finishing it, in particular
Enrico Bellino, Antonella Palumbo and Mariacristina Piva.

The Author 2016. Published by Oxford University Press on behalf of the Cambridge Political Economy Society.
All rights reserved.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

LucianoBoggio and LauraBarbieri*

Page 2 of 23 L. Boggio and L.Barbieri

1.Introduction

1
Boyer and Petit (1981), while discussing Un modle dinspiration kaldorienne, summarized the essence
of their arguments in the diagram they presented on p.1128, which offers an immediate comprehension.
2
A useful way of visualising this relationship can be obtained from the first two columns of Table1 from
Fagerberg (1996), p.41.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

According to the mainstream theory of growth, based on the assumption of full

employment of factors of production, economic growth is simply the combined effect
of increases in the supply of factors and in their productivity. No room is left for an
autonomous role of effective demand, its pressure on productive capacity and factor
employment, nor its role in determining the rate of investment.
According to another view, proposed in particular by the post-Keynesian theory
of growth and shared by the present writers, the assumption of full employmentof
labour in particularwhen the time period considered is not a secular one is not the
most useful starting point for explaining differences in growth performances.
An extreme version of this view was given by Nicholas Kaldor in the following
terms: economic growth isalways demand-induced and not resource-constrained.
Resources, such as capital and labour do not determine growth, partly because they
are mobile between regions, and partly because they are never optimally allocated
(there are always economic sectors where labour is in surplus in the sense that its marginal productivity is zero or even negative, as e.g. in agriculture); and partly because
capital (in the sense of industrial capacity) is automatically generated as part of, and
in consequence of, the growth of demand (Kaldor, 1981, p.603; italics in original).
As long as post-Keynesian theory deals with open economies, the following statement by Kaldor (1971, p.7; italics added) offersin the writers opiniona fundamental starting point: [T]he main autonomous factor governing both the level and
the rate of growth of effective demand of an industrial country with a large share of
exports in its total production and of imports in its consumption is the external demand
for its exports: and the main factor governing the latter is international competitiveness,
which in turn depends on the level of its industrial cost relatively to other industrial
exporters. Moreover, thanks to increasing returns in manufacturingVerdoorns
lawexport expansion and international competitiveness would interact so as to create vicious or virtuous circles of cumulative causation (Myrdal, 1957; Kaldor, 1970).1
This position was partly modified a few years later in a paper where Kaldor (1978)
compared the time changesover periods longer than ten yearsof the main industrial countries relative manufacturing export shares with that of their relative unit
costs2 and found a positive correlation between the two variables.
This paradox has various explanations. The simplest is that [h]igher prices could
equally well reflect higher quality, which, in turn, might justify higher wages. From this
perspective higher growth in relative unit labour cost (RULC) could just as well be seen
as an indicator of growing quality relative to other countries or increasingrather than
deterioratingcompetitiveness (Fagerberg, 2002, p.1). Asecond (ambitious) interpretation, which has a clear Kaleckian flavour, is that lower unit labour costs (ULCS)
should not necessarily be interpreted as implying that an economy is more competitive, i.e., that it will grow faster, and vice versa. In wage-led growth economies, an
increase in the wage share leads to an increase in the equilibrium capacity utilization
rate, which leads to an increase in the growth rate of the capital stock. Hence it is possible to find that the countries with fast-growing ULCS are the ones registering faster

International competitiveness: controversies and empirical evidence Page 3 of 23

3
This Kaleckian explanation of Kaldors paradox essentially implies that growth is not really export led,
but rather domestically led. Actually, many post-Keynesians have expressed scepticism about the whole idea
that growth is always or typically export led, and favour an investment-driven approach instead (especially
for larger countries, and maybe for all countries). We thank one of the anonymous referees for drawing our
attention to that point.
4
This explanation of the paradox appears consistent with the sectoral investigations of the determinants
of export performance which began to appear in those years, and which gave an essential role to technological variables, in contrast to the poor or perverse (that is, positive) sign, in such investigations, of (the growth
of) unit labour cost or wages. See Table3 of Fagerberg (1996), summarizing five research projects of this
kind, Dosi and Soete (1983), and Dosi etal. (1990), ch.6.
5
In his 1981 paper, Kaldor asserts that the Harrod multiplier (Thirlwalls model) is dynamically stable
(we thank Bruno Soro for attracting our attention to this point). This assertion seems rather unwarranted,
but some kind of crude proof had been in Kaldor (1970), p.342. On this point, see also Palumbo (2009),
pp.35152. Both Palumbo (2009) and Soro (2011) show that Harrod himself did not share this assertion.
6
Data and analysis supporting this proposition can be found in Boggio (1995) for Italy and Boggio
(2003) for the Asian NICS.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

growth in exports or in GDP, [so that] Kaldors paradox ceases to be an anomalous

result (Felipe, 2005).3
A further explanation is that a third factor, like GDP growth, is positively linked to
growth in both exports and wages, hence in unit labour costs (GDP growth in turn
being strongly influencedaccording to the prevailing viewby technology4), thus
creating a positive link between the two time series.
Or, as Kaldor himself put it, several years after the previous sentence quoted: There
is only one important matter on which the events of the 1970s caused me to change my
mind. This concerns the relative importance of price (or cost?) competition, as against
other non price factors, such as superiority of design or quality, length and reliability
of delivery dates, after-sales service, etc. Exchange rate adjustments operate mainly on
cost and prices, and despite vast changes in relative exchange ratesin real, and not
just in nominal termsthere was little effect on the pattern of trade in manufacturing
(Kaldor, 1986, p.25).
The perverse (positive) relationship between export growth and price/unit labour
costs growth became known as the Kaldor paradox. This paradox was a damaging
blow for Kaldors thinking and led him to dismiss his own theory and endorse Harrods
foreign trade multiplier5 and Thirlwalls model of balance-of-payments-constrained
growth (Kaldor, 1981).
In that paper, cumulative causation was reduced to the fact that, because of increasing returns (in the production of goods and technical progress), success breeds further
success (ibid., p.596), whilst when increasing competitiveness and the widening of
international market shares were an essential feature of the cumulative causation processas in the earlier Kaldorian formulationit appeared a much more important
and powerful phenomenon.
We believe that the theory, put forward in the 1960s and 1970s by Kaldor (and
Beckerman before him) to explain the contrast between the fast growth of Japan and
some countries of continental EuropeWestern Germany and Italy in particular
and the slow growth of the United Kingdom and the United States, is still very useful
nowadays for understanding contemporary experiences, like those of the Asian NICs,
of fast growth with strong international competitiveness.6
The purpose of this paper is to reaffirm the Kaldorian cumulative causation theory
in its original version, by providing a firmer analytical base and showing, on the basis
of empirical evidence, that the Kaldor paradox is a weaker empirical regularity than
others fully favourable to that theory.

Page 4 of 23 L. Boggio and L.Barbieri

Later studies seem to confirm the paradox. For instance, Fagerberg (1996) shows
that, considering the period 197894, unit labour costs growth, a typical variable
expressing technological factors affecting competitiveness, has poorer explicative
power for export growth than changes in GDP and in R&D as a share ofGDP.
In effect, the consequence of the Kaldor paradox can be expressed by letting the
traditional multiplicative form of export demandbe

x = x ( p + p) + x y x , x > 0 , (1)

x = x ( p p) + x y + T . (2)

In the light of these premises, the aim of this paper is twofold. First, we theoretically
and empirically reassert the importance of cost competitiveness as a channel through
which cumulative growth dynamics can be sustained in open-economy growth models. Second, we show that in macrodynamics, the growth rates of some variables may
depend on the levels (rather than the growth rates) of other variables when replicator
dynamics governs out-of-equilibrium behaviour.
In Sections 2 and 3, we briefly examine the two main theoretical strands of Kaldorian
derivation, namely the Kaldorian models of balance-of-payments-constrained growth
(BPCG) and the Kaldorian models of export-led growth (ELG), in order to assess
their relationship with the Kaldorian cumulative causation theory. In Section 3, we
propose an alternative approach to that theory, based on the replicator equation and
Beckerman model. After a brief examination of the literature on the replicator equation, an empirical comparison between the ability of the Kaldor paradox and the replicator-like approach in explaining the growth of export is presented in Section4.
2. Kaldorian models of balance-of-payments-constrained growth(BPCG)
A problem with the original Kaldorian approach is that the possibility of balance-ofpayments problems is not considered. Thirlwall (1979) suggests a simple model for
avoiding the possibility based on traditional demand equations for both exports and
imports and their equality. It can be summarized using the standard multiplicative
form of export demand, equation (1), expounded above, and similarly also for import
demand:

m = m ( p p) + m y m , m > 0 , (3)

where m, p, p* and y are the growth rates of imports, home and foreign price indexes,
and domestic income, respectively, while m and m are the price and income elasticities of import demand.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

where x, p, p* and y* are the growth rates of, respectively, exports, home and foreign
price indexes measured in common monetary unit, and rest of the world income; and
x and x are the price and income elasticities of export demand. Hence, (p + p*) is
the relative price of foreigngoods.
Because of the paradox, there is a change of sign in the relative price effect and the
addition of other variables of technological factors expressed byT:

International competitiveness: controversies and empirical evidence Page 5 of 23

But if, as Thirlwall maintains, either relative prices did not vary in the longrun,

p p = 0, (4)

or, alternatively, x + m 1 0 , then weget

x y = m y y / y = x / m , (5)

y = x y /m . (6)

According to him (Thirlwall, 1979, pp.5253), and supported by Kaldors opinion

(Kaldor, 1981, p.603), the income elasticities of this model reflect the innovative ability and adaptive capacity of the producers in different countries.
Many empirical estimates have been produced in subsequent years about the two
crucial assumptions of the BPCG model, each eliminating the role of relative prices:
a) no long-run changes in relative prices; and b) the sum of price elasticities of export
and import demand approaching one. They tend to support the view that in the very
long period (50years or more)but not for shorter periods, ranging from one to a few
decadessuch assumptions are likely to be confirmed.7 What matters more from our
viewpoint is that here the Kaldorian cumulative causation processes are not considered and no
theoretical foundation is offered to them.
3. Kaldorian models of export-led growth(ELG)
The Kaldorian theory of cumulative causation circles in open economies during the
past decades has not ceased to attract the attention of economists.
The algebraic translation of a Kaldorian ELG model goes back to Dixon and
Thirlwall (1975), who established a tradition, with which we disagree on a fundamental point, namely the export equation. In this exposition, we shall use the more recent
versionsvery similar to the first oneby Setterfield and Cornwall (2002) (p.72) and
Blecker (2013). The ELG model can be written using the symbols already introduced:

x = x ( p + p) + x y x , x > 0 . (7)

It is interesting to see that the Kaldor paradox is completely neglected/forgotten.

By the small country assumption, foreign price p* is taken as exogenously given.
Setting w and q as changes in wages and labour productivity, changes in domestic price
are determined by changes in unit labour costs (w q) and the gross profit markup, :

p = w q + . (8)

7
For an accurate discussion of the literature on this point, see Blecker (2013) and Garcimartn et al.
(2010).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

which explains growth rate differentials in terms of income elasticity differentials.

Thirlwall showed that the growth rate of income of the main industrial countries during the period 196073 was approximated very well by the formula

Page 6 of 23 L. Boggio and L.Barbieri

By Verdoornslaw:

q = q0 + y , 0 < < 1; (9)

here, q0 is a parameter representing autonomous technical change while represents

the Verdoon effect.
Finally,
y = ( x x + A A), (10)

w = = 0 = w = (11)

p = w q + (12)

q = q 0 + y . (13)

Next, by combining these equations with equations (7), (8), and (10), onegets

y = fDR ( q ) + x x q (14)

[ A A x x q0 + x (x x ) y].

Equation (14) according to Setterfield and Cornwall (2002)following a tradition

established since Boyer and Petit (1988)may be called the demand regime (DR)
equation. This goes together with the twin notion of productivity regime (PR), defined
using (9) as follows:

y = f PR ( q ) ( q q0 ) / . (15)

Combining (14) and (15), weget

y = + x x ( q0 + y ). (16)

If

1 x x 0, (17)

the last equation gives a well-defined constant rate of growth for national income, y,

yE =

+ x x q0
1 x x

, (18)

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

where is an expenditure multiplier, x and A are, respectively, the share of exports

and that of exogenous domestic expenditures in total demand, and A is the rate of
change of the latter.
Let us denote by a star key * the foreign variables and assume:

International competitiveness: controversies and empirical evidence Page 7 of 23

generating a constant rate of growth also for all the other endogenous variables, p, q,
x.Since
f PR , fDR > 0,

sufficient conditions for yE > 0are

> 0 (is satisfied if the sum of the direct and indirect effects of y* on x is positive),
1 / > x x (is satisfied if the Verdoorn effect is small relative to the price effect
via exports on y, or f PR = 1 / > fDR = x x ).

1
qt +1 = f PR
( yt ) yt + q0 (19)

yt = fDR ( qt ) (20)

1
yt +1 = fDR ( qt +1 ) = fDR [ f PR
( yt )]. (21)

The stability will prevail if and onlyif

f PR / fDR = x x < 1, (22)

which is the condition for the existence of the constant growth ratepath!
A time lag between the two sides of the PR and DR equations seems to be taken
for granted by several authors, who then discuss the stability issue;8 whilst others, like
Blecker (2013), just mention it, to conclude that the mere existence of cumulative
causation is not sufficient to create a disequilibrium situation (p.14). For, as he correctly notes, it may seem contrary to the spirit of Kaldor (1972) to represent his ideas
using a model that has an equilibrium solution (ibid.). The suggestion by Setterfield
(2002) to retain the notion of equilibrium, therefore allowing for the existence of
point [weak] attractors and to build a model of Kaldorian traverse, is interesting.
However, we cannot but agree when in a footnote (p.230, fn. 12)he admits that In
fact, it is far from obvious that there is any place at all for equilibrium in Kaldors
growth schema.
Overall, our main critiques to the ELG approach are that a) the Kaldor paradox is
completely neglected/forgotten; and b) no place is given to Kaldors assertion, that the
rate of growth of effective demand of an industrial country depends on international
competitiveness, which in turn depends on the level of its industry costs relative to
other industrial exporters.
Cf. i.e. Castellacci (2001).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

There is a further problem with the above equation system: the behaviour of the economy outside the constant rate of growth cannot be established.
But if we introduce a lag in the system, as Dixon and Thirlwall (1975) did, we can
deal with the problem. We may, for instance, assume a slow Verdoorn effect, sothat

Page 8 of 23 L. Boggio and L.Barbieri

4. Beckerman model and the replicator equation
4.1. Replicator equation

x i = xi [ fi ( xi ) f ( x )], fi > 0, f ( x ) i xi fi ( xi ). (23)

This equation embodies the principle of natural selection: varieties with above-average
fitness will expand, those with below-average fitness will contract and the average fitness f ( x ) will increase. If the fitness functions fi ( xi ) are simple constants, the average
fitness will increase monotonically towards a state where only the varieties (or group of
varieties) with maximum fitness survive(s).
Notice that the last equation can be put in theform
x i / xi = fi ( xi ) f ( x ), (24)

which can be read as the proportional change in the share of the ith variety is an
increasing function of its fitness advantage.
But for the purpose of empirical analysis, a discrete version is necessary. It can be
obtained by adapting Verspagen (1993), p.191.
Let xit be the level of export in country i at time t, t = 1, 2, 3,..., i = 1, 2,...n and yit
the country is share of world export at time t:

yit

xit
xit
n
i

y it ( yit +1 yit ) / yit. (25)

Then the rate of change of yit, namely y it , is givenby

y it = mj j ( E jit / E jt 1), E jt in yit E jit , j = 1, 2,..., m , (26)

where Ejit is the competitiveness variable j of country i at time t, and j the corresponding (fixed) parameter to be estimated.
Equation (26) represents a typical R-like equation.

9
For a more detailed analysis, see Hofbauer and Sigmund (1997); for economic applications, see
Silverberg (1997) and Silverberg and Verspagen (1995).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

To capture Darwins notion of the survival of the fittest, Fisher (1930) introduced what
are now calledafter Richard Dawkins (Dawkins, 1976, pp.1321)replicator equations9, of which we give below a very simplified version.
Let us consider a population composed of n distinct competing varieties of a
given natural species and denote by xi , i = 1, 2,..., n , i xi = 1, the relative frequencies of the ith variety in the population. To each variety, let us associate its fitness
level expressed by function fi ( xi ) . In continuous time, their evolution might be
described by the following equations (as usual, dot over the variable denotes a time
derivative):

International competitiveness: controversies and empirical evidence Page 9 of 23

X ij = kij + aij PCij + bij IN ij + cij PTij . (27)

PC is a measure of unit labour cost, IN is (derived from) the ratio of investment to

production and PT is based on total patents for each sector. For each independent
variable, the following bell-shaped lag structure is adopted:

Z = ( Z ( 1))03 ( Z ( 2))04 ( Z ( 3))03 , Z = PC , IN , PT . (28)

The error-correction mechanism is built on the long-run value of the market share
(27),giving

X ij = ( + i )[ X ij X ij ( 1)], (29)

where i is a country dummy.

The unit labour cost variable on average performs very well, better than the other
two. Verspagen and Wakelin (1997) do not specify their model in the form of an exact
replicator, but base their functional specification upon earlier empirical models, such
as Amendola etal. (1993) and Amable and Verspagen (1995). Market shares of each
market are estimated for 10 countries and 15 manufacturing sectors. The average rate
of growth of a market share is regressed on the average of competitiveness variables for
the period 198085. These are patents and R&D, share of investment in output and
wages in dollars (significant in 16 or 22 cases out 45), share of investment in output
(significant in 29 cases) and wages (significant in 20 cases).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

4.1.1. The replicator-like equations in the literature. Empirical tests of R-like equations
found in the existing literature are very few, since the bulk of the literature on the effect
of international competitiveness on export growth is based on growth rather than
levels of competitiveness variables. On the whole, they show that unit cost/wage levels
are significant (with the expected sign) in the explanation of export shares, sometimes
more significant than technology variables.
In Verspagen (1993), regressions are performed for the whole sample of pooled
observations (196487) for each of the 18 industrial sectors. The equation employed
is a modified version of (26), and the whole exercise is explicitly under the heading of
the replicator equation.
In the various specifications tried, wage and labour productivity variables give good
results, much better than those for the specifications of the patent variables (p.215).
The paper of Amendola etal. (1993) deals with manufacturing export market shares of
16 OECD countries from 1967 to 1987. It is the only one that refers to the total manufacturing export of each country. Starting from an equation almost identical to equation (26),
a general autoregressive-distributed lag (ADL) is derived, giving rise also to long-run multiplier estimates. The independent variables are unit labour cost, investment in machinery
and total patents. However, in the reformulated ADL equations with long-run multipliers,
the independent variables are again time changes. The Kaldor paradoxthey assert
[a]s a long-term phenomenonseems the outcome of a spurious correlation (p.465).
Amable and Verspagen (1995) present an estimation of an empirical model of market share dynamics for five industrialized countries and 18 industries. In their paper, a
long-run export market share X* is built as follows:

Page 10 of 23 L. Boggio and L.Barbieri

4.2.Beckermanmodel

x = a + b(1 ), (30)

where is the price level of a given country relative to competitor countries and a can
represent the rate of increase in world trade (ibid., p.919). The meaning of is made
more clear by explicitly introducing the level of prices in the two countries: let pi be
the level of prices (in common units) in country i, i = 1, 2 , where 1 is the home country
and 2 a competitor country. Then p1 / p2 and x1 = a + b(1 ) .
Then wehave
10
Dixon and Thirlwall (1975) noticed that Beckermans model bears many similarities to Kaldors and
predates it (p.202, fn. l), but (mis)interpreted Beckermans export equation as a wrong specification of an
export demand (p.211). Cf. also Thirlwall and Dixon (1979) and Cornwall (1977).
11
In the writers opinion, Beckermans model could be seen as embodying the replicator equation principle even if the author does not consciously refer to it.
12
Models spelling out a plausible description of the causal link between across firms differences in unit
cost levels on one side and differences in profits, investments and productive capacity growth on the other
can be found in Boggio (1974, 1996, 2003).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

In this subsection, we shall consider the model published by W.Beckerman in 1962,

a few years before Kaldors papers mentioned above that anticipated some of the main
ideas, in particular the cumulative growth model in open economies.10 But, differently
from all later Kaldorian models, Beckerman implicitly adopts an R-equation (shorthand for
replicator equation) form for his export equation, therefore also anticipating the application of
that idea to economic problems.11
In this model, export changes are a decreasing function of the price of given country
relative to competitor countries (Beckerman, 1962, p.914). More precisely, the rate
of growth of the share of world exports held by a given country, hence the differences
in the growth of exports, depends not on time changes of relative priceas in the traditional approach to export determination, exemplified by the exports equation of the
BPCG and ELG modelsbut on cross-country price differences.
The first reason we adopt the latter approach is that, accordingly with the Kaldors
non-equilibrium thinking, we believe market shares are often far from their long-run
equilibrium value and that their rates of change over time can be approximated by a
replicator-type equation. This statement becomes clearer if we express the main reason
we stress cross-country differences in levels of unit costs or prices: normally, they also
imply cross-country differences in profit margins, hence in the ability of the single firm
to invest, expand its productive capacity, pursue technical progress and spend in nonprice competition. In this waycontrary to an export growthbased approach on the
demand side only, as in the model described in the previous sectionswe implicitly
introduce also the supply side of the firms.12
We shall now describe the Beckerman model (Beckerman, 1962, pp.91819), introducing slight modifications that will make further formal developments easier.
The symbols used are the same as those to be used in later years by the models
expounded above, but for the productivity growth rate, for which we shall maintain q
asabove.
The most interesting equations are the first ones, embodying the replicator equation
principle:

International competitiveness:controversies and empirical evidence Page 11 of 23

productivity (output per head) equation13
q1 = c + dx1 , (31)

wage equation

w1 = z + vq1 , n < 1, (32)

where b, c, d, z, v are positive parameters.

price equation

from (32) and(33):

p1 = z + q1( v 1), and (34)

from (30) and (31) for the given country:

p1 = z + q1( v 1) = z + ( v 1)[( c + ad ) + bd (1 )], (35)

Following Beckerman (1962), we assume parameters are the same in all countries, but
for a competitor country, wechoose

x2 = a + b(1 1 / ) (36)

instead of Beckermans x2 = a b(1 ) .

Along these lines, for the competitor country, equation (35) becomes

p2 = z + q2 ( v 1) = z + ( v 1)[( c + ad ) + bd (1 1 / )], (37)

Hence, assuming the parameters in the basic equations are the same in all countries,
a country having a competitive advantage will have a faster than average rate of growth
of productivity. Hence prices will rise less (or fall more) in the country starting with the
competitive advantage. Thus, an initial price disparity will tend to be further accentuated and to be accompanied by a growing disparity in growth rates (ibid.).
In this way, the substance of the Kaldorian cumulative causation process in open
economies was clearly expounded some years in advance.
Since the replicator principle can be summarized in the sentence the proportional
change in the share of the ith variety is an increasing function of its fitness advantage,
the same is true in Beckermans first equation, if we remember that a can represent
the rate of increase in world trade and notice that, when equation (30) is writtenas
13
[T]he rate of increase of productivity is positively correlated with the rate of increase of output, which
is in turn correlated with the rate of increase of exports (ibid.). No further justification of this equation is
given, nor mention to Verdoorns papers appears, but in its twin paper (Beckerman, 1965, pp.38081), all
features supporting Verdoorn Law are included.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

p1 = w q1 , (33)

Page 12 of 23 L. Boggio and L.Barbieri

x a = b(1 ), (38)

the left-hand side is the rate of growth of the share of the countrys exports on the
world total, which is made dependent on the competitiveness factor .
Let us further develop this model (see also Boggio and Seravalli, 2003).
If we maintain the assumption that the parameters in the basic equations are the
same in all countries (Beckerman, 1962),then

d log
= p1 p2 = ( 1 / ), (1 v )bd > 0 (39)
dt

d
= ( 2 1), (40)
dt

so that the model gives rise to a differential equation and assumes the following graphic
representation seen in Figure1.
Within the open interval (0, ) , only one equilibrium, =1, exists. Outside this equilibrium, moves away from it, entering either a virtuous circle of cumulative causation or a
vicious one. In the former case, competitiveness, the growth of exports, output, productivity and wages all increase and, for a while, at an increasing speed. In the latter, competitiveness, the growth of exports, output, productivity and wages decrease at an increasingspeed.

Fig.1. Beckerman models representation in ( , ).

14
Beckerman, on the basis of his observation of European countries, rejected the suggestionthat differential rates of growth of the labour force may be an important explanation for differential growth rates
simply does not stand up to empirical verification (ibid., 921, fn. 2). See also his Reply (Beckerman, 1963)
to Balassa (1963).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

International competitiveness:controversies and empirical evidence Page 13 of 23

Obviously, only proximity to the equilibrium should be considered as economically
meaningful, because, as long as a less extreme version than Kaldors of the premises
of the post-Keynesian theory of growth is adopted, it should be stressed that many
supply-side factors concur in dampening the divergence brought out by the model.14
Similarly, one should remember that in the real world, technical progress not amenable
to Verdoorns law may turn out to be most important.
As with all models, ours is an abstraction, a glimpse of reality, that, if accepted,
catches a true aspect of reality.
Ours is one of the divergent circles of cumulative causation based on national competitiveness as in the original Kaldorian theory.

As mentioned before, empirical tests of R-like equations found in the existing literature
show that unit cost/wage levels are significant (with the expected sign) in the explanation of
export shares, sometimes more significant than technology variables.
Therefore, we decided to try some preliminary tests, in order to begin to assess the
relevance of Kaldors paradox relative to that of R-like equations.15
5.1. Our firsttests
Very simple regression equations were performed, starting from OECD data for 33
countriesincluding India, China and Brazilon aggregate good exports in US dollars (EXP) and manufacturing unit labour costs (ULC) for the 15-year period 1993
2007. The regressions are based on equation (26) (replicator equation): the dependent
variable y it is the average growth rate of export share (EXPGR)16 of country i over the
15years, while Ejit is the value of the jth competitiveness indicator for country i; the
independent variables are the average unit labour costs (ULCAV) and their average
growth rate (ULCGR) over the 15years (see Appendix Afor more details).
According to the Kaldor paradox, EXPGR should mainly be determined by ULCGR
(with positive coefficient) and according to replicator-like equations by ULCAV (with
negative coefficient). The results of the cross-section analysis are in Table1.
The ULCAV variable is always significant, also when coupled with ULCGR. The latter
never reaches a standard level of statistical significance. Hence, the replicator-like specification turns out very relevant, whilst the Kaldor paradox specification is not significant.
In order to take advantage of the time-series dimension of the disposable data, which
can be relevant for the phenomenon under consideration, we also implemented a panel
data analysis, which fully confirms the previous results about the relationship between
EXPGR, ULCAV and ULCGR.
First, the three regressions presented in Table1 have been re-estimated in pooled
data (see Appendix B for details on the variables construction and some descriptive
15
Alonso and Garcimartn (1998) and Len-Ledesma (2002), starting from a BPCG and an ELG model,
respectively, developed additional refinements based of the theory here illustrated and empirically tested it
with econometric estimates. Nevertheless, their models differ from our design, which focuses on the use of
replicator equations and avoids lag complexity referring to a different and less monetary instable historical
period (19932007 in our case).
16
Following the original specification proposed by Kaldor (cf. Kaldor, 1978, fn. 2, p.102), our design is
based on export share instead of on the quantity of export as in Len-Ledesma (2002).

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

5. Empirical evidence

Page 14 of 23 L. Boggio and L.Barbieri

Table1. OLS estimation of the replicator equation. Cross-section analysis
Dependent variable: export share growth - EXPGR
Independent Variables

(1)

ULCAV

-0.137***
(0.042)

ULCGR
const

(3)

-0.001
(0.001)
0.010
(0.007)
33
0.047
3.449
0.043
2.216

-0.137**
(0.057)
1.60E-05
(0.001)
0.095**
(0.036)
33
0.202
2.119
1.191
7.551

1.Data refer to 33 OECD countries.

2.Standard errors are in parenthesis. In the first equation, they are corrected for the presence of heteroskedasticity, given the results of the White test.
3.* Statistically significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
4.JB stays for the Jarque-Bera statistic for normality; Reset is the Ramsey test for specification errors;
White is the homoskedasticitys test.

statistics). The results are reported in the first three columns of Table2; it is easy to
note that they confirm the conclusion of the cross-section analysis.
Specifically, as regards the first regression, the ULC variable seems to have a significantly
negative impact on the dependent variable, while the second regression shows that the
ULCGR variable is not significant. Finally, when we insert both independent variables in
the regression, the significance of the ULC variable is confirmed with a negative coefficient.
Moreover, in order to control for potential endogeneity of the unit labour cost
growth rate regressor (i.e. not only slower growth of relative unit costs promotes more
rapid export growth, but also successful export growth leads to more rapid increases
in relative unit costs), the significance of the lagged regressors has been verified. The
results, reported in column (4) of Table2, prove the absence of reverse causality.
We reach the same results considering the random-effects model,17 for which we
obtain the results in Table3.
5.2. Introducing technology variables
As amply discussed in the first part of this paper, several authors, while highlighting
the Kaldor paradox, stressed the role of technical progress as an alternative explanation of trade flows. Therefore, we would like to introduce in our equations a variable
expressing such arole.
To this end, after having considered the measures of technological capabilities proposed by Archibugi and Coco (2005) (see Appendix C for more details), we choose the
average gross domestic expenditure in R&D of the 19952000 period (GERDAV) and,
17
We chose a random-effects model after considering alternatives from fixed-effects and random-effects
models and implementing the Hausman test, which did not reject the null hypothesis, suggesting we use
random-effects model.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Number of obs.
R-squared
JB(2)
Reset(2)
White(2)

0.094***
(0.028)
33
0.202
2.114
1.196
6.037**

(2)

International competitiveness:controversies and empirical evidence Page 15 of 23

Table2. OLS estimation of the replicator equation. Pooled regressions
Dependent variable: export share growth - EXPGR
Independent Variables

(1)

ULC

-0.111***
(0.025)

ULCGR
ULC(t1)

(2)

(3)

0.0003
(0.0003)

-0.111***
(0.028)
0.0002
(0.0003)

const
Number of obs.
R-squared
JB(2)
Reset(2)
White(2)

0.079***
(0.016)
448
0.036
0.456
0.034
5.121*

0.014***
(0.004)
440
0.002
0.476
0.346
0.151

0.079***
(0.017)
440
0.038
0.439
0.028
5.228

-0.112***
(0.029)
0.0002
(0.0003)
0.079***
(0.017)
411
0.038
0.210
1.196
5.022

1.Data refer to the 33 OECD countries over the period 19932007.

2.Standard errors are in parenthesis. In the first equation, they are corrected for the presence of
heteroskedasticity.
3.* Statistically significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
4.JB stays for the Jarque-Bera statistic for normality; Reset is the Ramsey test for specification errors;
White is the homoskedasticitys test.

alternatively, the average gross domestic product of the same period (GDPAV) as the
main variables that should capture the innovation capability of a country.
By inserting the GERDAV variable into our third equation, we obtain the results
reported in Table4.
The GERDAV variable is not significant, and its effect on the equation results is negligible. The ULCAV variable is significant at the 5% level, while ULCGR is not. This
seems to confirm the validity of the replicator approach against the Kaldor paradox
and technological capabilities.
However, things are not so simple: if we take the GDPAV variable as expressing
the technological capabilities and introduce it into our equations, in spite of the high
Spearman correlation between GDPAV and GERDAV, the results turn out very differently. If we replace the latter by the former variable in the last three equations, the only
significant variable turns out to be GDPAV itself (see model (2), Table4).
Various specificationsintroducing dummies for the UE or the Eurozone, as well as
weights based on GDP, distinguishing countries according to an income threshold
did not significantly change the results.18
However, GDPAV not only expresses the technological level of the country, but it
approximates the manufacturing productivity, which in turn is the denominator of
the ULCAV. This suggests, following Verspagen (1993) and Verspagen and Wakelin
(1997), that the latter could be conveniently replaced by the wage rate. To this aim,
we introduced the WAGESAV independent variable, computed for each country as
Results are available upon request.

18

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

ULCGR(t1)

(4)

Page 16 of 23 L. Boggio and L.Barbieri

Table3. Random effects GLS regression for panel data
Dependent variable: export share growth - EXPGR
Independent Variables

(1)

ULCAV

-0.088**
(0.038)

ULCGR
const

(3)

0.065***
(0.023)
440

0.0003
(0.0003)
0.014**
(0.006)
440

-0.084**
(0.039)
0.0002
(0.0003)
0.064***
(0.024)
440

0.001
0.175
0.036
5.320**

0.002
0.013
0.002
0.750

0.001
0.191
0.038
5.510*

1.Data refer to the 33 OECD countries over the period 19932007.

2.Standard errors are in parenthesis. In the first equation, they are corrected for the presence of
heteroskedasticity.
3.* Statistically significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.

Table4. OLS estimation of the replicator equation with technological variables. Cross-section
analysis
Dependent variable: export share growth - EXPGR
Independent Variables

(1)

(2)

ULCAV

-0.136**
(0.058)

-0.090
(0.064)

WAGESAV
ULCGR
GERDAV
GDPAV
const
Number of obs.
R-squared
JB(2)
Reset(2)
White(2)

3.46E-05
(0.001)
-6.37E-06
(1.94E-05)
0.097***
(0.037)
31
0.205
2.270
1.581
10.128

(3)

0.001
(0.005)

-3.08E-06***
(1.07E-06)
-4.84E-04
(0.001)

-1.49E-06**
(5.92E-07)
0.104**
(0.038)
31
0.274
0.941
10.078
7.519

1.21E-06
(1.02E-06)
0.071***
(0.019)
27
0.484
2.450
1.259
7.203

1.Data refer to 33 OECD countries.

2.Standard errors are in parenthesis. In the first equation, they are corrected for the presence of heteroskedasticity, given the results of the White test.
3.* Statistically significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.
4.JB stays for the Jarque-Bera statistic for normality; Reset is the Ramsey test for specification errors;
White is the homoskedasticitys test.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Number of obs.
R-squared
within
between
overall
Wald(2)

(2)

International competitiveness:controversies and empirical evidence Page 17 of 23

the average wage of the 15 disposable annual data (see Appendix Afor more details),
obtaining the results shown in the last column of Table4.
The main result is that the level of the wage rate, expressed by WAGESAV, when
per capita GDP (GDPAV) is included in the regression, is very significant, whilst the
growth of unit costs (ULCGR) has a coefficient smaller than the standard error. Similar
results are obtained when using GERDAV instead of GDPAV, or when the logarithm of
GDP or the above described DUMMIES are introduced.
6. Concluding remarks

Conflict of interest statement. None declared.

Bibliography
Alonso, J. and Garcimartn, C. (1998). A new approach to balance-of-payments constraint:
some empirical evidence, Journal of Post Keynesian Economics, 25982
Amable, B. and Verspagen, B. (1995). The role of technology in market shares dynamics, Applied
Economics, vol. 27, no. 2, 197204
Amendola, G., Dosi, G. and Papagni, E. (1993). The dynamics of international competitiveness,
Review of World Economics (Weltwirtschaftliches Archiv), vol. 129, no. 3, 45171
Archibugi, D. and Coco, D. (2005). Measuring technological capabilities at the country level: a
survey and a menu for choice, Research Policy, vol. 34, 17594
Balassa, B. (1963). Some observations on Mr. Beckermans `export-propelled growth model,
Economic Journal, vol. 73, 76384
Beckerman, W. (1962). Projecting Europes growth, Economic Journal, vol. 72, 91225
Beckerman, W. (1963). Some observations on Mr. Beckermans `export-propelled growth
model: a reply, Economic Journal, vol. 73, 78587

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Since we consider the original Kaldorian cumulative causation theory as the most
important tool of analysis to understand the historical experiences of fast-growing
open economies, the main purpose of this paperas stated at the beginningis to
reaffirm such a theory, by providing a firmer analytical basis and showing that Kaldors
paradox cannot impair thatbasis.
We believe that our discussion of the replicator equation and the developments we
presented of Beckermans model may provide these firmer theoretical bases. Also, the
results of the quantitative work done along the lines of the replicator equation by various authors and by ourselves may support the view that export growth depends on levels of competitiveness factorslabour cost in particularrather than on their growth.
Hence, the Kaldor paradox, though statistically significant in some papers discussed in
the introduction, in recent years is a weaker hypothesis than the replicator-type equations, and thus cannot diminish the importance of the original Kaldorian cumulative
causation theory as interpreted in light of Beckermansmodel.
The results we obtained when per capita GERD and GDP are introduced could
suggest that, even with variable expressing technological capabilities, the fundamental
results of this paper are confirmed: namely, the validity of the replicator approach vis-vis the Kaldor paradox.
This shows that replicator dynamics are a potential boon to Kaldorian growth theory (with its emphasis on growth as an out-of-equilibrium or non-equilibrium process), and may assist in both the further development and advancement of this theory
and reconciliation of Kaldorian growth theory with empirical evidence.

Page 18 of 23 L. Boggio and L.Barbieri

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Beckerman, W. (1965). Demand, exports and growth, pp. 4472 in Beckerman, W. and Associates
(eds.), The British Economy in 1975, National Institute of Economic and Social Research.
Reprinted in Feinstein, C.(ed.) (1997), The Economic Development of the United Kingdom Since
1870:Volume II (pp.379411), Cheltenham, UK, Edward Elgar Publishing
Blecker, R. (2013). Long-run growth in open economies: export-led cumulative causation or
a balance-of-payments constraint? pp. 390414 in Harcourt, G. and Krielser, P. (eds.), The
Oxford Handbook of Post-Keynesian Economics, vol. 1, Oxford, Oxford University Press
Boggio, L. (1974). Crescita e specializzazione produttiva in uneconomia aperta, Milano, Vita e
Pensiero
Boggio, L. (1995). Italy: from semi-industrialized economy to the sixth industrial power, Rivista
Internazionale di Scienze Sociali (also at Universidad Catolica Argentina, Facultad de Ciencias
Sociales y Economicas), vol. 53, no. 2, 22352
Boggio, L. (1996). Growth and international competitiveness in a Kaldorian perspective,
Structural Change and Economic Dynamics, vol. 7, 299320
Boggio, L. (2003). A model of take-off and fast growth in open economies, Metroeconomica, vol.
54, no. 23, 30125
Boggio, L. and Seravalli, G. (2003). Lo Sviluppo Economico: fatti, teorie, politiche, Bologna, Italy,
Il Mulino
Boyer, R. and Petit, P. (1981). Progres technique, croissance et emploi: un modle dinspiration
kaldorienne pour six industries europennes, Revue conomique, vol. 32, no. 6, 11131153
Boyer, R. and Petit, P. (1988). The cumulative growth model revisited, political economy, Studies
in the Surplus Approach, vol. 4, no. 1, 2344
Castellacci, F. (2001). A technology-gap approach to cumulative growth toward an integrated
model: empirical evidence for Spain, 19601997, DRUID Working Paper, 01-04
Cornwall, J. (1977). Modern capitalism: its growth and transformation, New York, St Martins Press
Dawkins, R. (1976). The Selfish Gene, Oxford, Oxford University Press
Dixon, R. and Thirlwall, A. (1975). A model of regional growth-rate differences on Kaldorian
lines, Oxford Economic Papers, vol. 27, 20114
Dosi, G., Pavitt, K. and Soete, L. (1990). The Economics of Technical Change and International
Trade, Brighton, Wheatsheaf and New York, New York University Press
Dosi, G. and Soete, L. (1983). Technology gaps and cost-based adjustment: some explorations
on the determinants of international competitiveness, Metroeconomica, vol. 35, no. 3, 197322
Fagerberg, J. (1996). Technology and competitiveness, Oxford Review of Economic Policy, vol. 12,
no. 3
Fagerberg, J. (2002). Technology, Growth and Competitiveness: Selected Essays, Cheltenham, UK,
Edward Elgar
Felipe, J. (2005). A note on competitiveness, unit labor costs and growth: is Kaldors paradox a figment
of interpretation? Australian National University, CAMA Working Paper 6/2005
Fisher, R. (1930). The Genetical Theory of Natural Selection, Oxford, Clarendon Press
Garcimartn, C., Rivas, L. and Martnez, P. (2010). On the role of relative prices and capital
flows in balance-of-payments-constrained growth: the experiences of Portugal and Spain in
the Euro area, Journal of Post Keynesian Economics, vol. 33, no. 2, 281306
Hofbauer, J. and Sigmund, K. (1997). The Theory of Evolution and Dynamical Systems, Cambridge,
Cambridge University Press
Kaldor, N. (1970). The case for regional policies, Scottish Journal of Political Economy, vol. 18,
33747
Kaldor, N. (1971). Conflicts in national economic objectives, Economic Journal, vol. 81, 116
Kaldor, N. (1972). The irrelevance of equilibrium economics, Economic Journal, vol. 82,
December, 12371255
Kaldor, N. (1978). The effect of devaluations on trade in manufactures, pp. 99118 in Further
Essays on Applied Economics, London, Duckworth
Kaldor, N. (1981). The role of increasing returns, technical progress and cumulative causation
in the theory of international trade and economic growth, Economie Applique, vol. 34, no. 4,
593617
Kaldor, N. (1986). Recollections of an economist, Banca Nazionale del Lavoro Quarterly Review,
vol. 39, no. 154, 326

International competitiveness:controversies and empirical evidence Page 19 of 23

A. Data used and main variables

The source of the data used is OECD. Dataused:
export data of the International Trade (MEI) data set measured in billions of US
dollars (EXPORT);
unit labour costs, which measure the average cost of labour per unit of output (ULC);
gross domestic product (expenditure approach) per head, US dollars, constant
prices, constant PPPs, OECD base year (GDP);
gross domestic expenditure on R&D per capita population (current PPP $) from the
OECD, Main Science and Technology Indicators (GERD);
average annual wages in 2009 USD PPPs and the 2009 constant pricesaverage
annual wages per full-time, full-year equivalent, dependent employee(WAGES).
As regards the cross-section analysis, the dependent variable EXPGR is the average
growth rate of the export shares (EXPshare) over the 19932007 period for 33 OECD
countries.19 The export shares are computed as the ratio of the exports of country i
in year t and the sum of the exports of all the countries under consideration in year t;
that is,
19
The sample includes Australia, Austria, Belgium, Brazil, Canada, China, Czech Republic, Denmark,
Estonia, Luxembourg, Mexico, Finland, France, Germany, Greece, Hungary, India, Ireland, Israel, Italy,
Japan, Korea, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia,
Spain, Sweden, United Kingdom, and the United States.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Len-Ledesma, M. (2002). Accumulation, innovation and catching up: an extended cumulative

growth model, Cambridge Journal of Economics, vol. 26, no. 2, 20116
Myrdal, G. (1957). Economic Theory and Underdeveloped Regions, London, University Paperbacks,
Methuen
Palumbo, A. (2009). Adjusting theory to reality: the role of aggregate demand in Kaldors late
contributions on economic growth, Review of Political Economy, vol. 21, no. 3, 34168
Setterfield, M. (2002). A model of Kaldorian traverse: cumulative causation, structural change
and evolutionary hysteresis, in Setterfield, M. (ed.), The Economics of Demand-Led Growth:
Challenging the Supply-Side Vision of the Long Run, Cheltenham, UK, Edward Elgar
Setterfield, M. and Cornwall, J. (2002). A neo-Kaldorian perspective on the rise and decline of
the golden age, in Setterfield, M. (ed.), The Economics of Demand-Led Growth: Challenging the
Supply-Side Vision of the Long Run, Cheltenham, UK, Edward Elgar
Silverberg, G. (1997). Evolutionary Modeling in Economics: Recent History and Immediate Prospects,
University of Maastricht, MERIT
Silverberg, G. and Verspagen, B. (1995). Evolutionary theorizing on economic growth, in Dopfer,
K. (ed.), The Evolutionary Principles of Economics, Dordrecht, Kluwer Academic Publishers
Soro, B. (2011). Harrod, Kaldor, and the idea of the foreign trade multiplier, Economia internazionale, vol. 65, no. 1, 17387
Thirlwall, A. (1979). The balance of payments constraint as an explanation of international
growth rate differences, Banca Nazionale del Lavoro Quarterly Review, vol. 32, no. 1,
4553
Thirlwall, A. and Dixon, R. (1979). A model of export-led growth with a balance of payments
constraint, in Bowers, J. (ed.), Inflation Development and Integration, Leeds, Leeds University
Press
Verspagen, B. (1993). Uneven Growth between Interdependent Economies: An Evolutionary View on
Technology Gaps, Trade and Growth, Aldershot Brookfield USA, Avebury
Verspagen, B. and Wakelin, K. (1997). Technology, employment and trade: perspectives on
European integration, pp. 5674 in Fagerberg, J., Hansson, P., Lundberg, L. and Melchior, A.
(eds.), Technology and International Trade, Aldershot, Edward Elgar

Page 20 of 23 L. Boggio and L.Barbieri

EXPshareit =

EXPORTit
i EXPORTit

(41)

with i = 1,..., 33 and t = 1993,..., 2007 , and so

EXPGRi =

log( EXPsharei ,2007 ) log( EXPsharei ,1993 )

14

(42)

ULCAVi =

ULCGRi =

t ULCit
15

(43)
ULCi

i ULCi EXPshareav

1, (44)

where ULCi is the average growth of the ULC of country i over the period considered, i.e.
and

ULCi =

log(ULCi ,2007 ) log(ULCi ,1993 )

14

(45)

EXPORTit
15
EXPshareav =
.
(46)
i
EXPORTit
t
15
i
33

An additional independent variable computed on the wages and introduced in Section
5.2 is
WAGESit
WAGESAVi = t
, (47)
15
i.e. the average country wage computed on the disposable time period.
t

B. Panel data analysis

In the panel data analysis, the variables are computed also considering their time-series
dimension: thus, the dependent variable EXPGR is the annual growth rate defined
modifying equation (42) as follows:

EXPGRit = log( EXPsharei ,t ) log( EXPsharei ,t 1 ), (48)

with i=1,,33 and t=1993,,2007.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

with 14 being the number of annual periods divided into the 19932007 interval.
The independent variables are the average unit labour costs (ULCAV) and their
average growth rate (ULCGR) over the 15years computed as replicator equations

International competitiveness:controversies and empirical evidence Page 21 of 23

The independent variables are the unit labour costs (ULCit) and their annual growth
rate (ULCGR) computed as replicator equations:

ULCGRit =

ULCit
i [ ULCit ( t EXPshareit )]

1, (49)

where ULCit is the annual growth of the ULC of country i from time t 1 to t,i.e.

ULCit = log(ULCit ) log(ULCi ,t 1 ) (50)

C. Technology variable selection

In order to introduce into our equations a variable expressing the technical progress, we
shall consider the variables studied in the work of Archibugi and Coco (2005) to capture the technological capabilities of a country. In their paper, alternative measures are
considered of national technological capabilities that generate country rankingsthese
are precisely the measures proposed by the World Economic Forum (WEF Technology
index), the UN Development Program (UNDP Technology Achievement index),
the UN Industrial Development Organization (Industrial Development Scoreboard
UNIDO), the RAND Corporation (Science and Technology Capacity Index STCI)
and a measure proposed by the authors (Technological Capabilities Index ArCo).20
The variables considered in Archibugi and Cocos paper, from our viewpoint, suffer
an important shortcoming: they cover only a small part of our equations time interval.
To obviate it, we shall consider the use of two variables that do not suffer that shortcoming, namely per capita GDP and per capita GERD (gross expenditure on R&D),21
and test their ability to fulfill the role of indicators of national technological capabilities
by comparing them with Archibugi and Coco variables.
20
Similarities and differences between the various methodological results at the aggregate level are analyzed without considering the nature of technological capabilities available in each country. Specifically, the
authors show that the results frequently diverge because of the different social indicators taken into account.
Nevertheless, a strong similarity in the rank correlations is interpreted as evidence of the reliability of each
indicator. This supports the idea that it is impossible to generate a unique measure of technological capabilities, but methodological variety helps create a better understanding of social phenomena. Moreover, all
ingredients of technological capability are as relevant as the final measure, and in order to assess the impact
of technological capabilities on economic and social indicators, individual indicators and sub-indexes should
be taken into account.
21
See Appendix Afor more details.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

We considered the descriptive statistics of the variables concerned; see Table 5.

As far as the export variable is concerned, we can note that the within variability overcomes the between variability, meaning that the variability over time in each
country in the period under consideration prevails over the variability of exports
between countries.
On the contrary, as far as the ULC variable is concerned, the between variability
overcomes the within variability, pointing at wider differences among countries than
the individual dynamics.
As regards the ULCGR, the results seem to highlight a predominance of the individual dynamic variability over the variability across countries.

Page 22 of 23 L. Boggio and L.Barbieri

Table5. Descriptive statistics of the panel data variables

Variable
EXPGR
ULC
ULCGR

Mean
overall
between
within
overall
between
within
overall
between
within

0.014
0.590
-0.490

1.Data refer to 33 OECD countries over the period 19932007.

Std.Dev.

Observations

0.077
0.037
0.69
0.132
0.119
0.057
11.267
3.338
10.758

N=440
n=33
N=440
n=33
N=440
n=33

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

For the sake of homogeneity with the latter indexes (almost all of which are computed over the 19972000 or 19952000 period), averages of per capita GDP and per
capita GERD of the 19952000 period are calculated (namely GDPAV and GERDAV,
respectively). The sample of countries selected is common to the Archibugi and Coco
set, i.e. 28 countries.
The rankings of our indicators, reported in Table6, are very similar to those of the
indicators considered by Archibugi and Coco (2005, p.186, Table2), as confirmed by
Table7, which shows for the 28 countries the co-graduation matrix among rankings
by means of Spearman correlations.
There is a high correlation between our indicators and the other indexes confirming
our hypothesis that the two variables could be valid indicators of national technological capabilities.
The highest correlation coefficients, however, are found for GERDAV. GDPAV shows
significant correlation, in one case equal to GERDAV, but lower in the othercases.
In view of these results, we choose GERDAV as the main variable that should capture the innovation capability of a country.

International competitiveness:controversies and empirical evidence Page 23 of 23

Table6. Technology indicators
GDPAV

14
9
10
12
28
21
5
7
6
24
25
16
4
17
3
15
27
8
19
11
26
22
23
18
20
1
13
2

4
6
8
5
28
23
15
12
7
20
24
10
17
14
9
22
26
3
18
1
27
19
25
21
16
13
11
2

Table7. Spearman correlations

GERDAV
GDPAV
WEF
UNDP
ARCO
RAND

GERDAV

GDPAV

WEF

UNDP

ARCO

RAND

1
()
0.75
(5.74)
0.67
(4.59)
0.83
(7.73)
0.90
(10.55)
0.93
(12.62)

1
()
0.67
(4.61)
0.72
(5.28)
0.74
(5.69)
0.78
(6.39)

1
()
0.85
(8.06)
0.78
(6.73)
0.79
(6.60)

1
()
0.88
(9.43)
0.89
(9.75)

1
()
0.96
(16.89)

1
()

1.t-statistics are in parenthesis, showing that all correlations are significant at the 1% level.

Downloaded from http://cje.oxfordjournals.org/ at New York University on February 5, 2016

Australia
Austria
Belgium
Canada
China
Czech Republic
Finland
France
Germany
Greece
Hungary
Ireland
Israel
Italy
Japan
Korea
Mexico
Netherlands
New Zealand
Norway
Poland
Portugal
Slovak Republic
Slovenia
Spain
Sweden
United Kingdom
United States

GERDAV