One or Many Kuznets Curves? Short and Long Run Effects of the Impact of Skill-Biased Technological Change on Income Inequality

Abstract

We draw on a dynamical two-sector model and on a calibration exercise to study the impact of a skill-biased technological shock on the growth path and income distribution of a developing economy. The model builds on the theoretical framework developed by Silverberg and Verspagen (1995) and on the idea of localised technological change (Atkinson and Stiglitz, 1969) with sector-level increasing returns to scale. We find that a scenario of catching-up to the high-growth steady state is predictable for those economies starting off with a high enough endowment of skilled workforce. During the transition phase, if the skill upgrade process for the workforce is relatively slow, the typical inverse-U Kuznets pattern emerges for income inequality in the long run. Small scale Kuznets curves, driven by sectoral business cycles, may also be detected in the short run. Conversely, economies initially suffering from significant skill shortages remain trapped in a low-growth steady state. Although the long-term trend is one of decreasing inequality, small-scale Kuznets curves may be detected even in this case, which may cause problems of observational equivalence between the two scenarios for the policy-maker. The underlying factors of inequality, and the evolution of a more comprehensive measure of inequality than the one normally used, are also analysed.

Full text

IZA DP No. 1223
One or Many Kuznets Curves? Short and Long
Run Effects of the Impact of Skill-Biased
Technological Change on Income Inequality
Gianluca Grimalda
Marco Vivarelli
DISCUSSION PAPER SERIES
Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
July 2004

One or Many Kuznets Curves?
Short and Long Run Effects of the
Impact of Skill-Biased Technological
Change on Income Inequality
Gianluca Grimalda
CSGR, University of Warwick
Marco Vivarelli
Università Cattolica Piacenza,
Max Planck Institute Jena and IZA Bonn
Discussion Paper No. 1223
July 2004
IZA
P.O. Box 7240
53072 Bonn
Germany
Phone: +49-228-3894-0
Fax: +49-228-3894-180
Email: iza@iza.org
This paper can be downloaded without charge at:
http://ssrn.com/abstract=571707
An index to IZA Discussion Papers is located at:
http://www.iza.org/publications/dps/
Any opinions expressed here are those of the author(s) and not those of the institute. Research
disseminated by IZA may include views on policy, but the institute itself takes no institutional policy
positions.
The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
company supported by Deutsche Post World Net. The center is associated with the University of Bonn
and offers a stimulating research environment through its research networks, research support, and
visitors and doctoral programs. IZA engages in (i) original and internationally competitive research in
all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research
results and concepts to the interested public.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available on the IZA website (
www.iza.org
) or directly from the author.

IZA Discussion Paper No. 1223
July 2004
ABSTRACT
One or Many Kuznets Curves? Short and Long Run
Effects of the Impact of Skill-Biased Technological
Change on Income Inequality
∗
We draw on a dynamical two-sector model and on a calibration exercise to study the impact
of a skill-biased technological shock on the growth path and income distribution of a
developing economy. The model builds on the theoretical framework developed by Silverberg
and Verspagen (1995) and on the idea of localised technological change (Atkinson and
Stiglitz, 1969) with sector-level increasing returns to scale. We find that a scenario of
catching-up to the high-growth steady state is predictable for those economies starting off
with a high enough endowment of skilled workforce. During the transition phase, if the skill
upgrade process for the workforce is relatively slow, the typical inverse-U Kuznets pattern
emerges for income inequality in the long run. Small scale Kuznets curves, driven by sectoral
business cycles, may also be detected in the short run. Conversely, economies initially
suffering from significant skill shortages remain trapped in a low-growth steady state.
Although the long-term trend is one of decreasing inequality, small-scale Kuznets curves may
be detected even in this case, which may cause problems of observational equivalence
between the two scenarios for the policy-maker. The underlying factors of inequality, and the
evolution of a more comprehensive measure of inequality than the one normally used, are
also analysed.
JEL Classification: O33, O41
Keywords: skill-biased technological change, inequality, Kuznets curve, catching-up
Corresponding author:
Gianluca Grimalda
Centre for the Study of Globalisation and Regionalisation (CSGR)
University of Warwick
Coventry, CV4 7AL
United Kingdom
Email: g.f.grimalda@warwick.ac.uk
∗
This paper is part of a research project sponsored by the International Labour Office, International
Policy Group, Genève. We thank Alan Hamlin, Ahmad Naimzada, Maria Cristina Piva, Giorgio Rampa,
Michela Redoano-Coppede, Roberto Tamborini, Grahame Thompson, Akos Valentinyi, Jüuso
Valimaki, Vittorio Valli, Matthias Weiss, Fabrizio Zilibotti and all of participants XIV Conference of the
Italian Association for the Study of Comparative Economic Systems (Napoli, February, 27-26, 2004),
the international conference on Economic Growth and Distribution: The Nature and Causes of the
Wealth of Nations (Lucca, June 16-18, 2004) and seminars in Southampton, Trento and Warwick
universities for their comments. Usual disclaimers apply.

1. Introduction
‘…Is the pattern of the older developed countries likely to be repeated in the sense that in
the early phases of industrialization in the underdeveloped countries income inequalities
will tend to widen before the levelling forces become strong enough first to stabilize and
then reduce income inequalities?’ (Kuznets, 1955, p.24).
In the last two decades, within-country income inequality (WCII) has shown different
patterns around the world. Even though the ‘average’ country can be said to have
experienced an upward trend during this period (Sala-i-Martin, 2002, Fig. 11)
1
, examples of
increasing and decreasing trends can be found in both developed and developing countries
2
.
Since several countries have at the same time been affected by a process of increasing
globalisation, intended as increased international trade and foreign direct investments, it has
been natural for economists to ask whether a causal link between globalisation and income
inequality exists. The focus of this paper is in particular on developing countries (DCs).
On the theoretical side, standard trade theory, based on the Stolper-Samuelson corollary
of the Heckscher-Ohlin theorem, actually predicts that in developing countries, where
abundant unskilled labour is cheap, one should observe trade driving the demand for the
unskilled-labour-intensive goods, thus decreasing WCII
3
. The main counter-argument to the
Stolper Samuelson theorem is based on the skill-enhancing-trade hypothesis (Robbins, 1996,
2003) which points out that trade liberalisation in DCs implies importation of machinery
from the North, leading to capital-deepening and (given capital-skill complementarities) to
rising relative demand for skilled labour
4
. That such a process of imported skill-biased
1
Sala-i-Martin (2002) considers the population-weighted average of within-country income inequality in a
sample that includes 88% of the world population.
2
In the group of developed countries, a rise in income inequality has been particularly evident in the US, in the
UK, and in Sweden, whereas it has remained constant, if not decreased, in Germany, France and Italy. Among
DCs, China, India, and the majority of the former Soviet Union Republics are reported to have experienced
rising inequality, whereas countries such as Indonesia, Turkey and Mexico appear to have experienced a trend
in the opposite direction (see Sala-i-Martin, 2002: 3; which is based on a critical analysis of the 1999 issue of the
Human Development Report; see also Cornia and Kiiski, 2001; Deininger and Squire, 1996).
3
An updated version of this theory, applied to DCs exporting manufacturing goods, can be found in Wood,
1994.
4
On the empirical side, some authors conclude that the opening process has nothing to do with increasing
WCII (Edwards, 1997; Higgins and Williamson, 1999; Dollar and Kray, 2001), while others show a positive
correlation in contrast with the Stolper-Samuelson prediction (Lundberg and Squire, 2001; Cornia and Kiiski,

2
technological change (ISBTC) has recently taken place in middle-income DCs has been
convincingly proven by Berman and Machin (2000 and 2004). On the grounds of this
literature, Vivarelli (2004) shows a significant impact of increasing import on the WCII,
using a sample of 34 DCs who recently engaged in opening their economies to international
trade.
This evidence opens the way to a reconsideration of the so-called Kuznets curve.
Kuznets’s seminal analysis refers to the long-term process of industrialisation and
urbanisation that affects countries at their early stages of development
5
. Kuznets’s ‘story’ is
that the shift of labour from the agricultural sector (where both per-capita income and
within-sector inequality are low) toward the industrial/urban sector (which starts small, with
higher per-capita income and a relatively higher degree of within-sector inequality), results in
an inverted U-shaped curve relating economic growth to WCII (Kuznets, 1955: Table 1,
p.13; see also Kuznets, 1963)
6
. In what follows, we shall refer to this account as Kuznets I.
By focusing on developed countries, ‘new’ growth theorists have argued that a similar
type of non-linear dynamics should also occur as a consequence of skill-biased technological
change (SBTC) (see Galor and Tsiddon 1996 and 1997; Aghion et al., 1999; Galor and Moav,
2000). The argument runs as follows. The introduction of an SBTC triggers an increase in
skilled labour demand and of the skill premium, thus determining an increase in inequality
and originating the first segment of the Kuznets inverted-U curve. Then, widening wage-
gaps induce the unskilled to invest more in the formation of human capital through
education, learning and training. Hence, as workers upgrade their skill levels the skilled
labour supply increases, thus reducing the skill premium and inequality, and giving rise to the
second segment of the Kuznets curve.
Although different accounts of the technological transition are consistent with this
general idea
7
, a Kuznets curve originates as a result of wage evolution and changes in the
2001; Ravallion, 2001).
5
In fact, Kuznets (1955: 4) offers empirical evidence spanning the 50-75 years prior to the 1950s for a sample
of developed countries. However, he points out that during this period only a decreasing trend of inequality
can be observed. Consequently, the time-scale necessary to observe a complete inverted-U pattern of initially
inequality-increasing and then inequality-narrowing trends may seemingly require even longer than a century.
6
Updated versions of the original Kuznets’s model have been offered, for instance, by Robinson, 1976; Fields,
1980, Bourguignon, 1990 and Greenwood and Jovanovic, 1990.
7
In particular, Aghion et al. (1999, Section 3.3) discuss two types of technological change: disembodied ‘general
purpose technologies’ and technological change embodied in machinery of different vintages. In both cases,
WCII follows a Kuznets curve where the initial skill-biased effect – enhancing inequality – is counterbalanced

3
composition of the labour supply. Hence, these theories account for the recent rise of WCII
in developed countries in terms of the upward part of the Kuznets curve, and predict an
inequality-decreasing trend for the next years. The reason is that a period of 15-20 years
from the original SBTC is seemingly sufficient for the inequality-decreasing forces to
counteract the initial inequality-enhancing effect (Aghion et al., 1999, p. 1655). Given the
supposedly shorter time scale of the latter account with respect to Kuznets’s original, and
given the different unit of analysis – rich or middle-income countries vis-à-vis DCs – we
shall refer to this latter account as Kuznets II.
On the empirical side, the Kuznets curve was commonly accepted in the 70s (see
Ahluwalia, 1976), while more controversial results were found in the following years (see
Papanek and Kyn, 1986; Anand and Kanbur, 1993; Li, Squire and Zou, 1998). However,
more recent studies have given further support to the law (Barro, 2000)
8
. Similarly, Reuveny
and Li (2003) have found a 5% significant support for the existence of a Kuznets curve
using a sample of non-OECD countries over the period 1960-96.
The purpose of this paper is to analyse the impact of an ISBTC on WCII from a
theoretical viewpoint. More precisely, we want to investigate the extent to which the transfer
of skill-biased technology toward middle-income DCs can trigger a Kuznets II dynamics.
This is achieved by means of a ‘calibration’ exercise, in which a dynamical two-sector
macroeconomic model is applied to the case of DCs through calibrating values for its
parameters and initial conditions on data relative to a sample of middle-income DCs. In
particular, depending on the amount of skilled productive forces that the economy is
endowed with at the time of the ISBTC, and on the initial productivity of the skilled
intensive technology, different scenarios can be generated in terms of the effects of the
technological diffusion on WCII and the growth rate of the country. The theoretical
framework also enables us to take into account a number of factors affecting WCII in
addition to those highlighted in the Kuznets I and II accounts, such as (a) the evolution of
unemployment in both the skilled-intensive and the unskilled-intensive sector and (b) the
dynamics of income distribution between capital and labour.
by the diffusion of the new technology – following a logistic curve – combined with the adjustment of the
labour force trough learning, training and education.
8
In particular, a Kuznets curve emerges with clear and statistically significant regularity; the relationship
between the Gini coefficient and a quadratic in log GDP turns out to be statistically significant in a SUR panel
estimation based on a sample of 100 countries over the period 1965-95 (Barro, 2000: Table 6, p.23)

4
The main result of the analysis is that an evolution a là Kuznets of WCII appears indeed
possible in the long run, but this only happens in those countries in which the supply of
skilled labour is sufficiently high when the ISBTC takes place, so that the skill-intensive
technology successfully diffuses within the economy. Moreover, even in this case a
sufficiently slow process of upgrade for the workforce is necessary in order for this result to
obtain. On the other hand, the investigation also emphasises the possibility of failure in the
diffusion of the advanced technology within the economy, in particular when skilled labour
is initially in short supply. In fact, this can easily be the typical situation of those DCs
characterised by institutional constraints in their educational and training systems (including
firms’ inability to provide on-the-job training and to develop an adequate path for human
capital upgrade). Here, a vicious cycle sets in, of low investments in the high-tech sector and
persisting skill shortages due to the lack of incentives for the workforce to upgrade their
skills. This result is consistent with the technology-gap approach in emphasising the
possibility of multiple steady states in a country’s development process (see Fagerberg, 1994
for a review, and Fagerberg and Verspagen, 2002). As a result, the economy gets trapped in
a low-growth development path due to technological lock-in. In this case, depending on the
initial relative productivity of the skill-intensive technology, either a path of relatively low
inequality occurs, which leads to a scenario of substantial equality coupled with poverty, or
income inequality displays increases in the short run and is later reabsorbed.
This latter scenario is particularly noteworthy, as it engenders a pattern resembling a
Kuznets curve on a small scale. Diagrammatic and statistical analyses of the computer-
generated data help show that such a short-run pattern is associated with sectoral business
cycles, tensions in the labour market, and the dynamics of income distribution, all of which
are triggered by the ISBTC, rather than the underlying forces of the Kuznets II account. As
a result, the initial inequality-enhancing effect caused by the increase in the skill differential is
here compensated by a decrease in skilled labour demand rather than through adjustments in
skilled labour supply. The fact that such short-term Kuznets curves driven by the business
cycle also occur in the scenario of technological catching-up alongside the long-term one,
may be a cause of concern for the policy-maker. The reason is that, since these two
scenarios are observationally equivalent in the short run, it would be wrong to infer from the
observation of rising inequality that an advanced technology is diffusing among the
economy, as a superficial reliance on the Kuznets II account may suggest. In fact, the rising
pattern of inequality may be due to a short-term effect of the business cycle in the presence

5
of relevant skill shortages, even when the skill-intensive technology fails to take off in the
economy in the long run.
Overall, the four scenarios that are generated by this investigation are seen as possible
explanatory models of the different patterns of income inequality that are being observed in
DCs. In particular, the latter scenario may provide a plausible interpretative account for the
recent WCII dynamics in those middle-income globalizing DCs which have opened to
international trade but whose process of technological catching-up is stagnating (examples
are most Latin-American countries, some Middle-East and North-African countries and
previous Soviet Republics).
The theoretical underpinnings of the model and the analysis of its steady states are presented
in Section 2. The theoretical framework is based on Silverberg and Verspagen (1995) and it
consists of a dynamical two-sector model characterised by increasing returns to scale at the
sectoral level, which generates unbalanced growth and multiple steady states. In section 3
the initial conditions of the perturbed system are calibrated on real data from middle-income
DCs starting with a relatively high percentage of skilled agents. In this section we show that
the Kuznets II account – originally put forward for developed countries (see above) - can be
replicated with regard to middle-income countries engaged in a globalisation process.
Section 4 analyses the WCII dynamics in the case of substantial skill shortages leading to a
‘regressive’ dynamics of failure in technological catching-up. The two possible patterns of
inequality illustrated above – one with an overall decreasing trend and another with a short-
term spurt in inequality – are analysed. Section 5 concludes.
2. The model
2.1 General features of the model
There are three key assumptions underlying the model
9
. First, there exist a variety of
sectors in the economy - two in its simplest version - that are associated with technologies
having different degrees of skilled labour intensity. Their pattern of technical change is
localised (Atkinson and Stiglitz, 1969; Antonelli, 1995) and it is assumed that productivity
growth rates are positively related with the share of economic activity taking place within
each sector. This implies that there are increasing returns to scale at the sectoral level. If we
9
For an extensive presentation and discussion of the present model, see Grimalda (2002).

6
abstract away from the between-sector linkages, which are illustrated below, then the
relevant variables for each sector, that is, unit labour cost and labour demand, follow a
Lotka-Volterra, or predator-prey, model (Hirsch and Smale, 1974; Goodwin, 1967). This
generates continuing cyclical behaviour in these two variables, which is a consequence of the
dynamics of income distribution between capital-owners and workers. In fact, on this
account, if the system finds itself in a phase of high investments, the consequent excess of
labour demand will drive wages up, thus reducing the rate of profit and investment. In turn,
this will decrease the level of production and employment, so that wages drop and this
triggers a new phase of increase in investments.
The second basic assumption is that agents are boundedly rational (Simon, 1955; Nelson
and Winter, 1982; Hogarth and Reder, 1986), so that the aggregate behaviour of individual
choices follows a replicator type of dynamics (Weibull, 1995).
Third, labour markets do not clear instantaneously; rather, wages evolve in accordance
with the imbalances between demand and supply. In contrast, since the country is presumed
to sell its product on the world market, the demand for its output is assumed to be perfectly
elastic, so that any amount of output that is produced can be absorbed by the world market
at the given price. Hence, commodity prices will be assumed constant throughout the
analysis.
Given the presence of increasing returns to scale at the sectoral level, the model is
characterised by multiple steady states, which differ in relation to the sectoral specialisation
the economy undertakes and, consequently, to their growth rates, as convergence to the
skilled-intensive technology guarantees higher growth rates. Convergence is determined by
the structural conditions of the economy, such as the size of the adjustment costs sustained
by workers and entrepreneurs in order to ‘migrate’ to the alternative sector of the economy,
and by the dimension of skilled productive forces at the time of the ISBTC shock. In
particular, both these aspects highlight the relevance of an economy’s absorptive capacity of
advanced technologies as a key factor for catching-up (see Lall, 2004), and the scenarios
studied in our investigation show that such capacities are not necessarily created through
market mechanisms, at least in the presence of particularly adverse initial conditions.
Given the nature of the problem at hand, i.e. the impact of an ISBTC on WCII and
the adjustment that this induces, we analyse both the initial transition phase occurring in the
short run as well as the phase of convergence toward a steady state taking place in the long
run.

7
2.2 A formal analysis
The basic assumption of the model is that each of the two sectors of the economy is
associated with a particular technology, which differs from the other in its labour skill-
intensity. In particular, the ‘modern’ (in contrast with the ‘traditional’) sector of the economy
is associated with a skilled-labour (unskilled-labour) intensive technology, which, for
simplicity, exclusively requires skilled (unskilled) labour. Moreover, we assume that each
technology is uniquely associated with a technique of production, so that labour and capital
are used in fixed proportions. This enables us to take on a Leontief representation for each
of the two sectoral production functions:
1,2,min =






= i
c
K
LaQ
i
iii
(1)
L
1
and K
1
(L
2
and K
2
) represent the employment of skilled (unskilled) labour and capital in
the skill-intensive (unskilled-intensive) technology. c is the fixed coefficient of the content of
capital for one unit of output, assumed to be equal for the two technologies, whereas a
i
is
labour productivity. As illustrated in section 2.4, we shall characterise the two sectors in
terms of the high-tech and the low-tech sectors within manufacturing in middle-income
DCs. In this way, the model describes the transition of an economy catching up from a
relatively backward sectoral specialisation to a relatively advanced one, and could thereby be
applied to the study of the Kuznets II hypothesis to middle-income DCs (see Section 1).
The model’s dynamics is driven by the following basic equations.
ii
i
i
g
a
a
κ
=
•
(2)
()
()
()
()
{}





=−+−
<−+−
=
•
1 1,0min
1 1
iiii
S
ii
iiii
S
ii
i
i
if ygLx
if ygLx
y
y
κηγ
κηγ
(3)
()
()( )( )( )
[]
()( )()
()
()( )( )( )
[]
()( )()
()
()( )( )
[]









−−−−
+
−>−−−−−−−
+
−
−>−−−−−−−
+
=
•
otherwiseyuyu
c
yuyuifyuyu
c
yuyuifyuyu
c
111
1
1)(11 1)(111
1
1)(11 1)(111
1
2211
1122211222
2211122111
κ
α
κνκτκ
α
κνκτκ
α
κ
κ
(4)

8
() () ()
() () ()

















>−






−












−−






−
−−






−
>−


















−
−−






−
=
•
otherwise
tw
s
L
stw
s
L
ifw
s
L
sw
s
L
ss
w
s
L
sw
s
L
ifw
s
L
sw
s
L
ss
s
0
)()(1)(
1
)(1
1
1
1
)(1
1
)(11
1
1
22
2
1
1
22
2
2
2
11
1
2
2
11
1
µµβ
µµβ
(5)
Equation 2 describes the evolution of labour productivity in a generic sector i. It is based on
the idea of localised technical change, which makes technical knowledge a public good at the
sectoral level but not at the economy-wide level. In particular, technical change is path-
dependent and triggered by a learning-by-doing process, which links productivity increases
with the density of economic activity in a sector; hence, productivity growth rates are
proportional to the share of capital invested in a sector. k denotes the capital share of
investment in the skilled intensive technology. g
i
are parameters that characterise the
productivity gains in the various sectors of the economy. A realistic assumption is that the
skill-intensive technology is, ceteris paribus, able to guarantee higher productivity growth rates.
Thereby, we assume that g
1
>g
2
.
y
i
is the unit cost of labour for sector i: That is,
i
i
i
a
w
y ≡
, where w
i
and a
i
are sectoral
wages and productivity levels respectively. The growth rate of y
i
, as represented in equation
(3), is made up of two components. The first is given by the excess of labour demand -
denoted by x
i
- over supply - denoted by L
i
S
. In particular, sectoral labour demand is defined
as
ca
K
x
i
i
i
=
. In other words, the wage growth rates depend on the excess of labour demand
over supply. The speed at which labour market imbalances impinge upon wages is measured
by the parameter
γ
, which will be assigned a value that implies – in the basic one-sector
version of the model – cycles of expansion and recession of a 10-year length. The second
component is associated with a redistributive mechanism independent of market forces,
which assigns a ‘bonus’ to wages equal to a portion
η
i
of sectoral productivity gains. Such a
component can best be seen as an institutional arrangement that accrues a fixed amount of
productivity gain to wages, and which is affected by the relative strength of capitalists and
workers in the bargaining process over income distribution. We allow for the two
redistributive parameters
η
i
to differ across sectors, so that bargaining may take place at the
sectoral level rather than at the economy-wide level. Given the ‘Harrodian’ flavour of the

9
model, caused by the sectoral Leontief-type technologies, a condition of structural
unemployment for the workforce (firms) obtains if
η
i
is strictly greater (lower) than one
10
.
Equation (4) expresses the rule of motion for capital share invested in the skill-intensive
sector, which is constructed in accordance with the replicator dynamics (Turner and Soete,
1984; Silverberg and Verspagen, 1995). The basic idea is that firms are boundedly rational
and – due to cognitive and informational limitations – strive to maximise their profits by
imitating more successful agents. Accordingly, only a fraction of them select the more
profitable action at each instant of time. In particular, some firms will migrate from the less
profitable to the more profitable sector at each instant of time, where such a portion
depends on the size of the difference in the profit rates – the bigger the profit rate in a
sector, the more likely the news will spread and/or firms will execute the ‘right’ action - and
on the exogenous parameter
α
- an index of both the speed with which information is
diffused among firms and the velocity at which intersectoral switches can occur. This flow
of firms adds to the ‘normal’ accumulation of profits in each sector, which follows the
behavioural rule typical of Kaldorian models that capital-owners reinvest all of their profits
in either sector, whereas workers consume all of their income (Kaldor, 1957)
11
. The
possibility of firms being rationed because of labour shortages is also taken into account by
means of the variable u
i
, which represents the degree of capacity utilisation of capital in
sector i
12
.
An additional aspect is taken into account in sectoral capital accumulation, that is, a
firm’s switch to the currently more profitable sector is conditional on the payment of an
adjustment cost, which is expressed in (4) by the functions
ν
1
(k) for the upgrade and
ν
2
(k)
for the downgrade costs respectively
13
. We assume that such costs vary depending on a
10
Due to the lack of data for sectoral unemployment rates, in the specification of the model in the following
sections, the two coefficients
η
i
will be assigned a value such that the steady state sectoral unemployment rates
coincide with the aggregate one for the economy, for which data are available.
11
Nothing substantial would change in the model if workers’ propensity to consume and entrepreneurs’
propensity to invest was constant, but less than one.
12
Formally, u
i
is defined as follows:







≤
>
=
S
ii
S
ii
i
S
ii
i
Lxwhen
Lxwhen
K
Lca
u
1
where K
i
is the absolute level of capital present in each sector.
13
Given the characterisation of technique 1 as skilled-labour intensive, we shall define upgrading the migration

10
firm’s degree of specialisation in a particular technique, so that the higher the specialisation,
the lower the cost of taking up the related technology. Such a degree of specialisation is
thought of as an immutable characteristic of the firm, acquired prior to the undertaking of
economic activities, and it affects solely the adjustment costs, not productivity. Moreover,
specialisation is technique-specific, so the higher the specialisation in a specific technique,
the lower the specialisation in the alternative one. This enables an ordering of firms on the
[0 , 1] interval, depending on their higher or lower degree of specialisation in technique 1 vis-
à-vis technique 2. In particular, the higher a firm’s specialisation in technique 1, and the lower
its specialisation in technique 2, the closer it will lie to the left hand-side of the interval, and
vice versa. Note that when we refer to an agent as ‘skilled’ we do not refer to the ease with
which s/he can upgrade, but only to whether s/he is currently employed in the skilled-
intensive sector. Finally, the choice of the parameters related to these functions makes the
upgrade costs generally higher than the downgrade costs
14
.
Equation (5) describes the rule of motion for skilled labour, which is denoted by s. It is
analogous to equation (4) in that workers’ movements across sectors are triggered by the
comparison of the expected wage earned in the two alternative sectors, net of the payment
of an adjustment cost that decreases in their level of sector-specific specialisation. Costs are
represented by the functions
µ
1
(s) and
µ
2
(s), which have the same interpretation as the
functions
ν
1
(k) and
ν
2
(k) illustrated above. Similarly,
β
, like α, measures the information
diffusion rate among workers.
from unskilled-intensive technology to skilled-intensive, and downgrading the movement in the opposite
direction.
14
The functional form that has been used in the simulations is as follows:
(
)
1
1
τ
κκν
=
and
(
)
(
)
2
1
2
τ
κκν
−=
.
τ
1
and
τ
2
are parameters determining the magnitude of the upgrade costs: the higher the parameter, the higher the
cost for each member of the population to improve their skill. The assumption
τ
1
>
τ
2
.implies that upgrade
costs are ceteris paribus greater than downgrade costs. Note that the entrepreneur associated with point 0 on the
interval [0,1], will have at the same time the highest possible specialisation in terms of the high-tech technology,
and thus the adjustment cost for moving from the low-tech to the high-tech sector is 0, and the least capacity in
mastering the low-tech technology, so that the adjustment cost for moving from the low-tech to the high-tech
sector is the highest possible, i.e. s/he has to spend her/his whole yearly profit. As
κ
increases, so does the
cost for upgrading, whereas the cost for downgrading decreases. Despite the choice of the adjustment costs
functions seeming to be based on a rather stringent assumption, the results of the model prove to be robust to
many possible specifications.

11
2.3 The Steady States of the Model
The steady states of the system can be divided into three categories: convergence toward
a high-growth equilibrium, convergence toward a slow-growth equilibrium, and a balanced
growth path in which both sectors of the economy coexist. By convergence we mean the
process that leads asymptotically to the complete allocation of capital and labour to one of
the two sectors. That is, if the country operates on the international scene, as is the case in
this model, convergence is equivalent to specialisation in the production of one of the two
commodities. The balanced growth path solution, instead, depicts a situation in which the
two sectors grow at the same rate.
The local stability of the first two types of steady state cannot be assessed on purely
analytical terms
15
. Still, the extensive simulation analysis that has been conducted shows that
these are stable attractors of the system for a feasible constellation of parameters. In
contrast, the solution associated with the balanced growth path can be ruled out immediately
as unstable. In what follows the three types of steady state will be presented in more detail.
2.3.1 High-growth steady state
A)
()






===
−
−=−== 1 0 edundetermin
1
1 1 1
221
1
1111
sxygxcgy
γ
η
κ
This solution is characterised by convergence to skilled-intensive technology. It holds
under the condition that
η
1
be greater than 1
16
, thus implying a positive level of
unemployment for skilled labour. One can also note that a greater speed of adjustment in
the labour market, as measured by coefficient
γ
, helps reduce the level of unemployment,
which at the limit for
γ
converging to infinity is equal to zero. Hence, the introduction of
non-instantaneous market clearing within the model brings about structural unemployment.
Instead
α
does not play a role within this specification
17
. Although the value for y
2
turns out
15
This is due to the presence of some purely imaginary eigenvalues making the system locally non-hyperbolic
(Guckhenheimer and Holmes, 1990). For an extensive discussion of the dynamical properties of the system,
see Grimalda (2002).
16
A substantially similar steady state also holds for the case
η
1
<1, though it is now capital rather than labour to
be rationed in equilibrium.
17
The case investigated in Grimalda (2002), where labour supply is fixed in each sector and unable to migrate,

12
to be undetermined, the subsequent numerical analysis clearly shows that such a variable
tends to the value of 1, i.e. to the situation of zero profits in the sector that remains residual
in the economy.
2.3.2 Low-growth steady state
We also find a steady state symmetric to (A), which is characterised by convergence
toward the unskilled-intensive sector. Thus, it brings about a lower growth rate in
equilibrium:
(B)
()






===
−
−=−== 0s 0 edundetermin
1
1 1 0
112
2
2221
xygxcgy
γ
η
κ
Solution (B) is an equilibrium with ‘structural unemployment’ in the leading sector of the
economy, i.e. sector 2, and, again, extinction of the residual one; this solution holds under
the restriction that
η
2
is greater than 1. Note that unemployment amounts to
γ
η
1
2
−
in the
steady state. The properties of stability of these steady states are the same as those found for
the case of convergence towards the first sector.
2.3.3 Balanced growth path
This is the only steady state in which both technologies coexist:
(C)
(
)
(
)
() ()














−
−−=
+
−+
=
−
−=
+
−+
=
+
=
21
2
2
21
212
2
21
1
1
21
212
1
21
2
1
1
1
1
1
1
ggsx
gg
cggg
y
ggsx
gg
cggg
y
gg
g
γ
η
γ
η
κ
would be different. In that setting,
α
enters the expressions for y
1
and y
2
, and as it tends to infinity, which
corresponds to the case of perfect information and rationality of the agents (see section 2.3), then the sectoral
profit rates are equal, which makes firms indifferent in choosing between the two sectors. Hence, the
traditional neoclassical condition of full employment and cross-sector equality in profit rates may be viewed as
a limit case of the present model.

13
Its main characteristic is that productivity is the same in the two sectors, and there is
rationing of either capital or labour depending on whether the coefficient
η
i
is less or greater
than 1. Since both sectors evolve according to the same growth rate, the economy can be
said to follow a balanced growth path. An analysis of the local properties of stability of this
steady state shows that such an outcome is in fact unstable. The economic reason is to be
found in the property of cumulativeness of sector-specific technology. If this state is
perturbed, then sectoral productivities will differ, thus attracting some firms to move to the
more profitable technology. As a consequence, the sector that ‘by accident’ happens to be
more profitable will experience positive sectoral economies of scale that will suffice to break
the balance between the two profit rates, triggering a snowball effect of convergence
towards one of the steady states illustrated above.
2.4 Modelling the impact of an ISBTC on a low-growth steady
state
As discussed in the introduction, we model globalisation as a way to implement SBTC in
a previously technologically backward country. SBTC is introduced directly through FDI,
multinational plants and import of more advanced capital goods, and indirectly through
exposure to international competitiveness, so that more commodities become tradeable and
domestic firms are induced to update their own technologies.
Despite the basic setting of the model being devised for a closed economy, we can
investigate the impact of globalisation by means of a theoretical exercise, which consists in
studying the evolution of the system after a low-growth steady state – supposedly a good
representation for a DC lagging behind in the technological ladder - is perturbed as an effect
of an ISBTC. In other words, we suppose that the economy shifts from the low-growth
steady state to a position corresponding to the introduction of an SBTC into the economy.
The extent of this shift is derived from real data, so as to reflect the actual weight of
advanced technologies in a sample of middle-income countries during the 80s and 90s. The
evolution of the system from the new starting position is then analysed, and in particular we
focus on whether the country can successfully catch up and converge toward the high-
growth steady state, and on whether a Kuznets type of dynamics can be triggered along the
adjustment path.

14
As for the ‘calibration’ exercise of determining the magnitude of the ISBTC shock and
the structural parameters of the economic system, we focus on the manufacturing sector and
draw on the classification offered by the OECD Structural Analysis (STAN) database that
divides the whole manufacturing sector into one group of high-tech and one of low-tech
industries
18
. We then collect population-weighted averages during the 80s and 90s for a
group of middle-high income and one of middle-low income countries for the relevant
variables of the model (see the Appendix).
Relying on this calibration, the evolution of WCII is studied by applying the Gini index
to some relevant categories of income. A first measure is built in accordance with the
Kuznets I and II accounts, which only consider the dynamics internal to labour income
distribution. Since in our model there are two such categories, that is, skilled and unskilled
labour, and a third of unemployed workers, the relevant cumulative population distribution
and their related income is the following:
()





+−
0 1
22
11
21
w L
w L
LL
We call the resulting inequality measure the restricted Gini index (RGI). An important
caveat, though, is that our index only takes into account between-group inequality, whereas it
neglects within-group inequality, as all of the agents belonging to each group are assumed to
earn the same income. This obviously leads to a substantial under-estimation of inequality in
absolute terms in our model. Nevertheless, we still believe that the main results of our
analysis are not affected by this aspect, especially because it is not a-priori clear whether there
exist significant differences in within-group inequality across the two groups.
A second index of inequality can be computed by considering capital income as well as
labour income. We shall refer to this as the comprehensive Gini index (CGI). The categories of
income that are considered are now as follows:
18
Mainly high-tech sectors are those having higher than average R&D expenditure as a measure of either value
added or output. See the Appendix for further details.

15
()
(
)
[]
()









−
−+++−+
1
0 11
22
11
22
11
2121
Kr κnu
Kr κ nu
w L
w L
uunLLn
κκ
n is here the ratio between the capital-owners population and that of employees, so that
the total population has a size of 1+n
19
. The first category is now given by the sum of
workers and entrepreneurs who are unemployed; the second and the third categories are
occupied skilled and unskilled workers as in the RGI. The fourth and fifth categories are the
profit earned by entrepreneurs active in the high-tech and low-tech sectors respectively,
which is given by the relative interest rate multiplied by the aggregate level of capital. Since
there is an additional factor of dispersion in CGI with respect to RGI, the income inequality
measured by the former will be higher than the latter.
3. Evolution of income distribution as a result of a ‘progressive’
technological catching-up with skill-upgrading
3.1 A Kuznets curve scenario
We first conduct a simulation where data are drawn from the sample of middle-high
income countries. Parameters have been assigned the following values on the basis of
theoretical considerations and real data
20
(see Appendix: Table 1):
19
Note that a characteristic of the model is that movement between the two populations of workers and
capital-owners is not allowed. Observations of the relative size of employers vis-à-vis employees for developing
countries (see e.g. KILM 2001 database, International Labour Office, Geneva) appear to imply a value for n as
being below 5%, so we set n=4% in the simulations.
20
In particular, values for the sectoral productivity growth rates g
1
and g
2
are drawn directly from the data. c, i.e.
the inverse of capital productivity, has been assigned a value such that the implied capital income share is one
third of total income in the high growth steady state. This is, in fact, the value generally used in growth
accounting exercises to estimate capital income share (see e.g. Mankiw et al., 1992: 410). This implies a capital
income share of roughly 17% for the low-growth steady state, which accords with the idea that DCs have a
lower capital income share than developed ones. γ has been assigned a value of 2.5, so that the business cycle
has a length of 10 years in the basic single-sector version of the model (see section 2.1). The values of η
i
have
been determined in such a way that the level of average unemployment is equal to 7.499% in both sectors,
which is the average value found in the data. The value of
α
is taken from Soete and Turner (1984); given that
β
plays the same role as
α
as an index of agents’ degree of bounded rationality, it has been assigned the same

16
{g
1
= 3.955% ; g
2
= 1.503%; U=7.499%; c=8.428 ; γ=2.5 ; η
1
= 5.470 ; η
2
= 13.473 ; α=1 ;
β = 1 ; λ
1
= 0.5 ; λ
2
= 10 ; τ
1
= 3 ; τ
2
= 10 }
It is worth noting that the high-tech sector productivity growth outstrips low-tech
productivity by 2.45%. Moreover, the implied value for structural unemployment is roughly
7.5%, and a complete cycle of recession and expansion in the basic single-sector component
of the model is of 10 years; consequently, a different duration of the cycle should be
attributed to the inter-sectoral dynamics.
As for the choice of the system’s initial conditions, as argued in the previous section, we
suppose the economic system is located in the slow-growth steady state, that is, case (B) in
section 2.3.2, before the ISBTC shock. Hence, we take the associated steady state value for
y
2
as the initial condition for the simulation. We then determine the value for y
1
in
accordance with the productivity differential and skill premium of high-tech with respect to
low-tech sectors that result from the data relative to middle-high income countries (see
Appendix: Table 1). These imply in particular a productivity advantage and a skill ratio for
the high-tech sector of comparable size: 46% for the former and 47% for the latter. The
starting value for skilled labour demand x
1
and supply s has been drawn from the average
employment in the high-tech sector found in the sub-sample considered, which is
approximately 31% of the workforce. We assume that both labour markets start off from a
situation of unemployment of the same magnitude as that in the steady state. These
considerations provide the following initial conditions for the endogenous variables of the
system:
value. Admittedly, the pairs of
λ
and
τ
- which determine the mobility costs for workers and entrepreneurs
respectively - are parameters for which finding an empirical counterpart appears problematic. To have a rough
idea of their interpretation, one should bear in mind that when
λ
1
and
τ
1
equals 1, then the median agent, i.e.
the agent located in the centre of the [0,1] interval, will have to spend half of her/his yearly income in order to
upgrade. Agents laying to the left (right) of the median agent will have to spend less (more) than her/him, with
a portion of their yearly income equal to zero (1) for the agents at the left (right) extreme of the interval.
Besides, the cost for the median agent increases as the parameters decrease. A perfectly symmetrical
interpretation holds for the downgrade cost parameters
λ
2
and
τ
2
. The particular values chosen imply that
upgrading costs for workers are relatively high in comparison with those for firms - for instance because some
of these firms are multinational companies with a higher level of expertise in adapting to new technical
paradigms than the local workforce - whereas downgrading costs are relatively lower for both.

17
{ }
0.311s ,2877.0 , 8805.0 ,6373.0 ,8733..0 ,3982.0;1;466.1
1122.21
=
=
=
=
==== xyxyaa
κ
This point is characterised by a position of ‘advantage’ for the skill-biased technology, in the
sense that the labour productivity for skill-intensive technology is greater than for the other
technology, but it also has a higher ‘potential’ for growth, as g
1
is larger than g
2
.
However,
skilled labour wages are also higher by an amount that slightly exceeds the productivity
advantage, so that firms are initially almost indifferent between the two technologies in
terms of profit rates, as y
1
is almost equal to y
2
.
The long-run outcome of this scenario is the specialisation of the economy in the high-tech
sector (Figure 1). During the transition, a pattern similar to a Kuznets dynamics originates
for both RGI and CGI (Figure 2 and 3). They reach a peak after 100 years, and then
converge to their new steady state level, which is associated with steady state (A) (see section
2.3.1). Since (A) implies a higher capital income share than (B), income inequality measured
by CGI shifts to a greater value in the new steady state
21
.
Diagrammatic and statistical analyses confirm that the usual mechanism underlying the
Kuznets II account is at work here. In fact, there exist a number of explanatory factors for
income inequality in the model. As far as the RGI index is concerned, inequality can be
affected by (a) the amount of the skill differential; (b) skilled labour unemployment and (c)
unskilled labour unemployment; (d) the proportion of skilled labour in the total. An
additional factor is relevant in the determination of CGI, that is, (e) the distribution of
income between labour and capital
22
. Figures 1, and 4 to 7 portray the evolution of each of
these factors over the 0-250 years span
23
.
21
More precisely, the RGI is equal to 0.072 in both steady states, as the only source of inequality is here given
by the ratio of unemployed workers to the total workforce, and this is by assumption the same in the two
steady states. Instead, the CGI increases from 0.151 in the low-growth steady state to 0.343 in the high-growth
one due to the higher capital income share associated with the high-growth steady state. Note that the initial
values for both the RGI and the CGI are actually higher than the values associated with the initial steady state.
This is of course due to the fact that the initial conditions for the simulation exercise differs from the initial
steady state by the amount of disturbance triggered by the ISBTC.
22
Capital-owners’ unemployment may also be a relevant factor of inequality; however, given the ‘Harrodian’
flavour of the model and the choice of parameters, this is equal to zero in the steady state, and is negligible
during the transition phase.
23
Figure 4, portraying the evolution of the skill differential, ends in period 50 because the variable follows an
exponential trend afterwards.

18
We split the analysis into three sub-periods. In the first 66 years, the only factors that
may cause a rise in inequality are the skill differential, skilled labour unemployment, and, as
far as the CGI is concerned, labour income share. Indeed, the proportion of skilled workers
remains flat in this period, as adjustment costs are too high and the wage differential still too
low to make migration profitable for workers. Moreover, unskilled labour unemployment
fluctuates around a rather flat trend. Statistical analysis reveals that wage differential and
labour income share are the most important factors in explaining RGI and CGI respectively,
whereas unskilled labour unemployment has some influence on RGI. Moreover, skilled
labour unemployment turns out to be insignificant (see Appendix: Table 3 and 4, column
(a)). The latter result is probably due to the small size of skilled labour unemployment in this
phase. In fact, in the successive period, from year 66 to years 106-120 (the RGI reaches its
peak earlier than the CGI), two additional factors cause the upward trend of income
inequality to be more pronounced than before - and to lose its cyclical pattern: firstly, the
start of workforce migration from the unskilled to the skilled sector (see Figure 1) - due to
the enlargement of the ‘rich’ side of the population - has a positive effect on inequality. This is
the case at least when the richer proportion of the population is relatively small
24
. Moreover,
such a movement has the effect of making the (still rising) skilled labour unemployment
quantitatively more significant than before; and secondly, the substantial fall in the labour
income share – which is clearly converging toward its new steady state level - increases the
inequality measured by the CGI even further. In fact, the percentages of skilled labour force,
labour income share and skilled labour unemployment all appear statistically significant and
24
The migration of the workforce toward the skilled sectors has, in fact, two contrasting effects on the Gini
index. On the one hand, there is a scale effect, whereby the proportion of poor individuals in the population
decreases. On the other hand, there is a relative poverty effect, which implies that the labour share of the poor
shrinks. The scale effect has a negative impact on inequality, whereas the relative poverty effect increases
inequality. This can be shown clearly if we concentrate on the Gini index and assess the impact of a change in
the fraction of the poor on the distribution of labour income, leaving the skill wage differential constant. It can
be shown that the following formula holds:








−=
22
1
2
1
dL
d
dL
dRGI
σ
where
σ
is the labour share of unskilled labour,
and its derivative with respect to L
2
is always positive. Hence, the two terms within brackets represent the scale
and the relative poverty effects respectively. If we take unemployment to be constant, so that a decrease in L
2
implies a one-to-one increase in L
1
, which of the two effects prevails depends on the magnitude of L
2
. More
precisely, the derivative of RGI is positive for values of L
2
less than some threshold level. Hence, when the
proportion of unskilled labour is large, as is the case initially, the relative poverty effect dominates the scale
effect, and thus inequality tends to grow. The opposite occurs as L
2
, exceeds such a threshold value, which
occurs after period 100 in this scenario.

19
add to the skill differential as explanatory factors for WCII in this period (see Appendix:
Table 3, column (b) and (c); Table 4, column (b), (c) and (d)).
As approximately half of the workforce has migrated to the skilled labour sector, the
Kuznets curve starts its reversal. Since the wage differential is still rising, and this has an
unambiguous positive effect on inequality, the inequality-decreasing effect of the other
factors must offset the impact of the skill differential. First, a sharp reduction in the
unemployment rates in both labour markets can be observed (Figures 5 and 6). Second,
labour income share rises after year 122. Finally, the scale effect due to the continuous shift
of workers to the ‘rich’ side of the income distribution now has the result of mitigating
inequality (see footnote 25). Skilled labour unemployment, the supply of skilled workers, and
labour income share, all turn out to be statistically significant, whereas the wage differential
is uncorrelated with the inequality indexes (see Appendix: Table 3, column (d) and (e); Table
4, column (e), (f) and (g)). In the final part of the period, after nearly all the populations of
workers and capital-owners have migrated to the high-tech sector, the decreasing trend in
inequality tends to smooth, and the two indexes converge toward their steady state values.
The whole cycle takes as long as 150-200 years to complete, which seems to be in
accordance with the secular long-term trend envisaged by Kuznets’ original account (see
footnote 5). However, it has to be said that the length of the cycle crucially hinges upon the
value of the adjustment costs
λ
1
,
λ
2
and
τ
1
,
τ
2
(see next section). According to the
simulations conducted, the shortest time it can take to reach a peak in the Kuznets curve in
this model– which obviously occurs in the complete absence of any adjustment cost – is 27
years. Another characteristic of the model is that during the initial period in which
adjustments in the workforce have yet to take place, several ‘short-term-Kuznets cycles
appear to occur, each of the approximate duration of 10 to 13 years (Figure 2 and 3). Such
short-term Kuznets curves are driven by the business cycle and by the inter-sectoral
dynamics of capital allocation, and they will be investigated in more detail in section 4.

Figure 1: Evolution of skill-intensive capital
share and skilled labour supply share
Figure 2: Evolution of RGI
Figure 3: Evolution of CGI
Figure 4: Evolution of the wage differential
Figure 5: Evolution of skilled labour
unemployment
Figure 6: Evolution of unskilled labour
unemployment
κ
s

21
Figure 7Evolution of Labour Income Share
3.2 A Scenario with Decreasing Inequality
In order to better appreciate the relevance of the magnitude of the adjustment costs for
the outcome of the simulation, we have run some simulations with low adjustment costs for
both workers and capital-owners
25
. The main difference with respect to the previous
scenario is that the workforce starts migrating toward the skilled-intensive sector from the
very outset, and the transfer of capital toward this sector is faster (Figure 8). Figure 9 depicts
the long-run evolution of RGI. It is apparent that the evolution of income inequality is
entirely different from before, following a decreasing trend that progressively converges
toward its steady state value. What causes the steep drop in RGI in the first couple of years
is the fact that the migration of the workforce towards the skill-intensive sector is initially so
rapid that the skilled wage differential actually decreases in the early stages of this simulation
(Figure 10). After this, the wage differential starts to increase, which is nevertheless
counterbalanced by fast migration toward the skill-intensive sector
26
. Therefore, the inverse-
U shaped pattern observed in Figures 2 and 3 is by no means a necessary feature of income
25
In particular, this scenario has been obtained for values of the adjustment cost parameters equal to
λ
1
=
λ
2
=
τ
1
=
τ
2
=10. That is, parameters for the downgrade are left unchanged with respect to the previous case,
whereas those relative to the upgrade are modified so as to imply lower adjustment costs. See also the previous
note. More precisely, taking as a reference the parameter λ, a value of 1, which denotes the situation in which
the median worker has to spend half of her/his yearly wage to upgrade, implies a period of slightly less than a
hundred years to reach the peak of the Kuznets. With λ=3, 60 years are needed, etc.
26
See also footnote 25 as to the interaction of a scale effect and a relative poverty effect in the dynamic of
income inequality.

22
inequality along the transition path toward the high-growth steady state, but crucially hinges
upon the workforce’s rapidity in skill upgrading, which in turn depends on the magnitude of
the adjustment costs.
Figure 8: Evolution of skill-intensive capital
share and skilled labour supply share
Figure 9: Evolution of RGI
Figure 10 Evolution of Skill Differential
4. Evolution of income distribution as a result of a ‘regressive’
technological lock-in without skill upgrading
4.1 A Scenario with Decreasing Inequality
We now turn to the analysis of a different scenario, where the initial conditions and the
relevant parameters have been derived from data relative to the sub-sample of middle-low-
income countries (See Appendix: Table 2):
{g
1
= 4.362% ; g
2
= 2.145%; U= 7.658%; α=1 ; c=7.642 ; γ=2.5 ; β = 1 ; η
1
= 5.389 ; η
2
=
9.925 λ
1
= 0.5 ; λ
2
= 10 ; τ
1
= 3 ; τ
2
= 10 }
The differences with respect to the previous scenario in productivity growth rates and
unemployment rates appear to be rather marginal. What instead proves to be significantly

23
different is the proportion of skilled workers: this is now substantially smaller than the
middle-high income country case, as it only amounts to 21% as opposed to the 31% of the
previous case. This clearly reflects the fact that high-tech sectors are relatively undersized in
middle-low-income countries in comparison with middle-high income ones. The initial
conditions have been computed using the same method as in the previous section, and the
following values obtain:
() () ()
(
)
(
)
() () ()
}198.00 , 879.00 ,726.00
,836..00 0.214,0s ,278.00;10;415.10{
112
221
===
=
=
===
xyx
yaa
κ
The wage premium, being roughly 48% vis-à-vis a productivity premium of 41%, causes the
profit rate in the low-tech sector to be initially higher than the high-tech sector. Figure 11
represents the behaviour of the system in the long run, and shows that the economy
converges to the low-growth steady state associated with specialisation in low-tech
technology. This outcome is caused by the evolution of sectoral productivity and the inter-
sectoral dynamics of capital allocation. Although the high-tech sector starts off with higher
productivity, the relative abundance of unskilled labour makes the low-tech sector overall
more profitable. Hence, high-tech capital share follows a decreasing trend over time, the
fluctuations being due to the cyclical pressures on wages in the unskilled labour market.
Given the presence of increasing returns to scale at the sectoral level, the rising
concentration of firms in the low-tech sector brings about higher productivity growth rates,
so that the economy specialises in the low-tech sector. On the other hand, the presence of
relatively high adjustment costs initially prevents workers from transferring to the high-tech
sector. Such an incentive does not improve over time, because the low concentration of
firms in the high-tech sectors causes the wage differential to decrease over time (Figure 12).
Consequently, the proportion of workers employed in the skilled sector initially remains
constant and then decreases when the skill differential has actually turned in favour of the
unskilled wage (Figure 11).
To sum up, technological lock-in is ultimately determined by the structural conditions of
the economy, and in particular by both the initial shortage of skilled labour and
entrepreneurs and the presence of relevant adjustment costs in skill upgrading, that is,
institutional constraints in the education and training systems. Since the rate of technical
innovation is driven by the level of concentration of capital in each sector, the lack of a
sufficient critical mass of skills, both in the workforce and in firms, gives rise to a vicious

24
circle of decreasing investments in the high-tech sector, decreasing demand for skilled
labour and decreasing productivity growth rates. This produces the characteristic
snowballing effect leading to technological lock-in towards the backward technique.
Figures 13 and 14 describe the behaviour of RGI and CGI respectively in the very long
run,. Given the lack of convergence toward the high-tech sector, no secular Kuznets curve
takes place here. The initial impact of the ISBTC on inequality is, in fact, negative
27
.
Subsequently, both indexes follow a decreasing trend, which suffers an abrupt reversal in
period 65, which is due to the ‘switch’ from unskilled to skilled labour by the poorest
recipients of labour income (in Figure 12, the switch occurs when the curve crosses the level
of 1). Statistical analysis (see Appendix: Table 5) enables us to discern that skill differential
and labour-capital income shares are the major determinants in the short run of RGI and
CGI respectively. In the long run, although serious problems of collinearity prevent the use
of statistical analysis, it is apparent that the adjustments in the labour supply also affect the
RGI.
27
Note, though, that the steady state value for RGI and CGI is different from the initial value observed in this
scenario. In particular, RGI measures 0.074 and CGI is equal to 0.187 in the steady state. See also footnote
23.If one took this value as the reference point, then it would be inappropriate to talk about a decreasing
impact of the shock on inequality even in the short run.
Figure 11: Evolution of skill-intensive capital
share and skilled labour supply share
Figure 12: Evolution of the wage differential
κ
s

25
Figure 13: Evolution of RGI
Figure 14: Evolution of CGI
4.2 A Scenario with a Short-Term Rise in Inequality
So far we have based our choice of parameters and starting condition values on the
average data in our possession. Let us now depart only slightly from this approach by
investigating the impact of a further increase in the initial conditions of the productivity
premium on skill-intensive technology. The reason lies in the fact that in both the previous
scenarios the skill wage premium is – albeit only marginally – higher than the productivity
premium, thus causing the skill-intensive technology to be initially less profitable than its
alternative. This translates into an initial decrease in capital invested in the high-tech sector
(see Figures 1 and 11). In the present section, we instead suggest investigating the case in
which skill-intensive technology starts off from a robust enough position of advantage to
ensure that the immediate impact on the high-tech capital portion of total capital is positive.
Hence, we alter the initial conditions of the previous scenario by increasing the productivity
premium for high-tech technology by a further 20% than was assumed in the previous
section. This leads to the productivity premium at time 0 being equal to 62% rather than
41%
28
. All of the other parameters are left unchanged as regards the previous scenario; thus
the following initial conditions obtain:
0.214})0(s ,198.0)0( , 733.0)0(
,726.0)0( ,836..0)0( ,316.0)0(;1)0(;698.1)0({
11
22.21
===
=
=
=
=
=
xy
xyaa
κ
28
A productivity differential at least as wide as that assumed in this scenario is the case for Indonesia, Morocco
and Guatemala in our sample of middle-low income countries (See Appendix: Table 2).

26
As Figure 15 shows, the long-run outcome is not different from that of the previous
scenario: the economy fails to catch up and the shares of capital and labour employed in the
high-tech sector eventually fade away. However, high-tech technology now shows a much
stronger ‘resilience’ than in the previous case, as after more than a century nearly a quarter
of capital is still invested in the high-tech sector.
Such a high persistence of demand for skilled labour in the presence of a severe skill
shortage determines a significant change in the evolution of WCII with respect to the
previous scenario. Not only is inequality on average higher in this case (compare Figures 16
and 17 with 13 and 14) for both indexes, but also the impact of the SBTC on RGI is now
positive, and the inequality rise in this index is initially quite steep. Consequently, in the first
30 years (Figure 19), RGI is characterised by an inverted-U pattern a là Kuznets, whereas no
significant trend can be recognised in the CGI, the initial impact of the SBTC being in fact
largely negative (Figure 20). Cycles of an average duration of 23 years – thus longer than the
previous scenarios (see Figures 2 and 13) - then occur repeatedly for RGI, following a long-
term trend of decreasing inequality. The CGI dynamics is are made more erratic by the
interplay between wages and profits and the frequent reshuffles in their relative rankings (see
Figure 18). As the workforce shifts to the low-tech sector, the inequality indexes settle on
their steady state levels.
Even in this case the dynamics in the skill differential (Figure 21) appears to be the
major driving force for RGI, as in fact both curves share a similar double-peak pattern. In
addition to this, unemployment rates initially decrease in both sectors (Figure 22 and 23);
when they start rising then the upward trend in RGI becomes steeper. Statistical analysis
confirms the major role of the skill differential as a determinant of the RGI in the long run,
although unskilled labour unemployment plays some part as well (Appendix: Table 6,
column (a) and (b)). Labour income share (Figure 24) instead has the biggest part in
affecting the fluctuations in the CGI, although the skill differential is also relevant in the
short run (see Table 6, column (c) and (d)).
This scenario, which is obtained for what appears as not too large a deviation from the
data collected, clearly highlights the possibility of fluctuations in income inequality that
present the inverted-U shape typical of the Kuznets I and II accounts. However, analysis
reveals that the underlying economic mechanism is indeed different from that on which the
traditional Kuznets accounts are grounded. No shift in workforce takes place here, as the
adjustment costs prove too high to induce workers to abandon the low-tech sector. Instead,
they are the components of the sectoral business cycle, the intersectoral allocation of capital,

27
and the tensions triggered in the labour market as a consequence of the ISBTC shock, which
play a major role. By way of illustration, what causes the fluctuations in RGI in the short run
is the following mechanism. The productivity advantage of the high-tech sector keeps skilled
labour demand constantly high, which triggers a rise in the wage premium. This is what
causes labour income inequality to increase in the short run. The wage differential rise is
such as to offset the inequality-decreasing effect due to the reduction of unemployment.
Hence, as the economy enters a phase of recession, inequality rises even more quickly.
After this, the economy enters a phase of expansion in the low-tech sector, whereas the
high-tech sector is approaching a phase of recession, so the wage differential reduces. Since
the low-tech sector weighs more than the high-tech, the overall effect is to reduce labour
income inequality. Similar mechanisms, which are led by the sectoral business cycle in the
process of adjustment triggered by the ISBTC shock, also hold for the short-term
fluctuations in inequality observed in the previous two scenarios. The analysis of CGI shows
a different initial pattern, but even in this case inequality is subject to fluctuations that are
due to the business cycles of the two sectors comprising the economy.
Figure 15: Evolution of skill-intensive capital
share and skilled labour supply share
Figure 16: Evolution of RGI
Figure 17: Evolution of CGI
Figure 18 Evolution of income per group
π
1
s
π
2
w
2
w
1
κ

28
Figure 19: Evolution of RGI (short run)
Figure 20 Evolution of CGI (short run)
Figure 21: Evolution of skill differential (short
run)
Figure 22: Unskilled labour unemployment
(short run)
Figure 23 Evolution of skilled labour
unemployment (short run)
Figure 24 Evolution of labour income share
(short run)

29
5. Conclusions
On the basis of the analyses carried out in the previous sections, the following
conclusions can be put forward.
1) The modern interpretation of Kuznets’ Law (Kuznets II) – originally devised for the
developed countries – can also be applied to globalised middle-income DCs
characterised by ISBTC and catching-up (that is, the capacity to converge to high-tech
sectors in the long run).
2) However, in order for this to be the case, it is necessary that the process of migration
towards the skill-intensive sector of the workforce be slowed down by significant
adjustment costs. Only in this way can a poverty effect overcome a scale effect in the
dynamic of income inequality. If this condition does not hold, RGI follows an entirely
different pattern from that conjectured by Kuznets.
3) Although we cannot provide more than a sketchy character as regards this aspect, the
timing of the model seems to imply a much longer time scale for the Kuznets reversal to
happen than what is advocated by the Kuznets II account (see introduction). This is due
to the presence of relevant adjustment costs in the skill upgrade process. Therefore, the
idea that income inequality may be a temporary phenomenon reabsorbed automatically
through the working of market forces and labour supply adjustment should be put under
serious scrutiny by the policy-maker, as its persistence may in fact prove to be socially
too costly not to require intervention. This conjecture should be matter of further
investigation.
4) Alongside this long-run Kuznets curve, short-term ‘micro’ Kuznets curves also emerge.
These are essentially associated with the business cycle of the economy; in particular the
determinants of such short-term cycles are: the inter-sectoral dynamics of capital
allocation, which determine labour demand; the evolution of the labour income share
and of sectoral unemployment; and the wage differential, all of which are triggered by
the ISBTC.
5) In the case of lock-in globalised DCs (those unable to converge to the high-tech pattern
of growth), no long-run Kuznets curve emerges, as workers do not migrate to the high-
tech sector of the economy due to adjustment costs which are too high.
6) In particular, RGI may exhibit a decreasing trend right from the start, which may be seen
as a scenario of equality coupled with poverty. However, if the productivity differential
of the ISBTC is sufficiently high, then RGI initially increases, so that a Kuznets curve

30
analogous to those occurring in the alternative scenario (section 3.1) can be detected in
the short run. The impact on CGI is instead always negative in this case, as the major
determinant of this index is the labour income share rather than the skill differential.
On the whole, this study shows that the observed increase in WCII in many globalising
middle-income DCs in the ‘90s can be interpreted as the first segment of a short-term
Kuznets inverted-U curve. However, this dynamics can be originated either by a skill-biased
technological transition similar to the one detected in developed countries (catching-up; see
Section 3), or by a ‘regressive’ dynamics without any catching-up and skill-upgrading in the
long run (lock in; see Section 4).
Summing up, the scenarios illustrated in section 3.1 and section 4.2 lead to opposite
outcomes in the long run in terms of patterns of technology adoption, sectoral specialisation
on the international markets, and growth. What turns out to be problematic is that since a
cyclical pattern of WCII coexists in the short term with the long-term pattern of structural
change described by the Kuznets II account, these two scenarios turn out to be
observationally equivalent in the short run.
Although putting forward economic policy prescriptions is outside the purpose of this
paper, some tentative implications may be drawn from the analysis developed here. It is in
fact apparent that the policy-maker’s agenda should be very different in these two scenarios.
A mere reliance on redistributive policies may suffice in scenario 3.1 in order to alleviate the
social costs of inequality, especially when the time necessary in order to reach the ‘peak’ of
the Kuznets curve – and thus to start the redistribution of the benefits of growth to those
social groups that have initially been ‘left behind’ - is too long, due to the adjustment costs.
Structural reforms would instead be needed in scenario 4 to prevent the economy from
being locked in to a poverty trap. In particular, policies promoting the skill upgrading of the
workforce and of the local entrepreneurial forces, and in general those policies facilitating
inter-sectoral migration – which in fact may also imply geographical migration - appear as
necessary steps for breaking out of technological lock-in.

31
6. Appendix
6.1 Data Used in Calibration
Table 1: Middle-high income countries:
Average compound rate for relevant variables over the period 1980-2000 (or the closest period when not available)
Country
Productivity
ratio HT/LT
Wage ratio
HT/LT
Skilled
employment
Productivity
growth HT
Productivity
growth LT
Productivity
Difference
Unemployment
(%) Population Weight
Argentina 1.061 1.380 0.257 NA NA NA 8.084 25,051,010 5.627
Brazil 1.832 1.838 0.304 7.677 4.641 3.035 5.127 166,045,568 37.299
Chile 0.947 1.503 0.154 3.451 1.690 1.761 8.425 14,821,700 3.329
Croatia 1.280 1.163 0.272 NA NA NA 9.744 4,396,570 0.987
Czech Rep. NA NA 0.438 NA NA NA 4.790 10,294,900 2.312
Gabon 1.389 1.112 0.076 4.379 3.656 0.723 NA 1,167,290 0.262
Hungary 1.311 1.157 0.349 0.471 0.767 -0.296 8.427 10,114,000 2.271
Malaysia 1.381 1.193 0.355 -5.602 -8.145 2.543 4.733 22,180,000 4.982
Mexico 1.296 1.291 0.350 1.999 -0.185 2.185 3.429 95,225,432 21.390
Poland 0.962 1.171 0.345 4.371 -0.757 5.128 12.436 38,666,152 8.685
Puerto Rico 3.237 1.409 0.391 NA NA NA 16.438 3,860,000 0.867
Slovak Rep. 0.781 0.990 0.399 3.541 2.266 1.274 12.840 5,390,660 1.210
Slovenia 1.044 1.056 0.340 NA NA NA NA 1,982,600 0.445
South Africa 1.311 1.534 0.249 -0.589 0.165 -0.754 20.816 41,402,392 9.300
Trinidad &Tobago 2.277 1.112 0.128 10.251 -2.698 12.949 16.285 1,285,140 0.288
Uruguay 1.386 1.542 0.129 0.489 1.408 -0.918 9.333 3,289,000 0.738
Population-weighted
average 1.465 1.477 0.311 3.955 1.502 2.452
7.499

32
Table 2: Middle-low-income countries
Average compound rate for relevant variables over the period 1980-2000 (or the closest period when not available)
Country
Productivity
ratio HT/LT
Wage ratio
HT/LT
Skilled
employment
Productivity
growth HT
Productivity
growth LT
Productivity
Difference
Unemployment Population Weight
Bolivia 0.484 1.050 0.089 1.129 0.484 0.644 5.7 7,950,000 0.844
Colombia 1.236 1.303 0.198 3.305 1.906 1.398 11.005 40,804,000 4.334
Costa Rica 1.302 1.279 0.136 1.827 3.694 -1.866 NA 3,653,060 0.388
Dominican Rep. 1.636 1.947 0.044 2.071 5.292 -3.221 17.742 8,103,210 0.860
Ecuador 1.078 1.220 0.136 -1.191 2.332 -3.523 8.292 12,175,000 1.293
Egypt 1.237 1.322 0.205 1.836 0.356 1.479 8.057 61,580,000 6.541
El Salvador 1.530 1.273 0.128 -5.643 -7.338 1.695 9.228 6,035,000 0.641
Guatemala 1.621 2.019 0.129 1.208 0.523 0.685 NA 10,799,000 1.147
Indonesia 2.196 1.991 0.133 7.434 4.716 2.718 5.34 203,678,368 21.635
Iran 1.194 1.131 0.224 3.718 -0.619 4.338 NA 61,850,000 6.570
Iraq 1.270 1.038 0.181 NA NA NA NA 22,327,630 2.371
Jordan 1.108 1.337 0.138 -0.244 -2.974 2.730 14.4 4,597,350 0.488
Latvia 0.731 1.069 0.377 5.326 1.070 4.256 11.433 2,449,000 0.260
Lithuania NA 1.020 0.246 NA NA NA 11.08 3,703,000 0.393
Morocco 1.784 1.832 0.144 0.584 1.193 -0.608 17.4083 27,775,000 2.950
Panama 1.442 1.300 0.066 -0.825 -1.481 0.655 NA.. 2,764,000 0.293
Peru 0.953 1.248 0.194 NA NA NA 7.5 24,801,000 2.634
Philippines 1.410 1.563 0.185 3.474 4.917 -1.443 7.709 72,775,448 7.730
Russian Fed. 0.756 0.950 0.385 NA NA NA 8.84 146,899,008 15.604
Thailand 1.586 1.368 0.135 7.765 -1.546 9.312 2.347 59,793,500 6.351
Tunisia 1.307 1.268 0.136 NA NA NA NA.. 9,333,300 0.991

33
Table 2 (continued)
Turkey 1.209 1.402 0.224 3.248 2.132 1.115 8.488 63,391,000 6.733
Ukraine NA 0.727 0.342 NA NA NA 6.585 50,303,000 5.343
Venezuela 1.014 1.325 0.200 -0.025 -1.980 1.954 9.775 23,242,000 2.468
Yugoslavia 1.190 1.192 0.278 NA NA NA NA.. 10,616,000 1.127
Population-
weighted average
1.415 1.488 0.213 4.362 2.145 2.216
7.657
Sources: The classification of a country as middle-high or middle-low has been drawn from the UNU/WIDER-UNDP World Income Inequality
Database (WIID), Version 1.0, 2000.
Data about population refer to the year 1998 and have been drawn from the World Development Indicators (WDI) database, International Bank for
Reconstruction and Development – The World Bank, Washington, 2002.
In the table, HT denotes High-Tech and LT Low-Tech.
All of the other data are drawn from the UNIDO, Industrial Statistics Databases, accessed through ESDS International, University of Manchester
The reference period is 1980-2000, or the closest possible to this.
The sectors classified as high-tech are, in the 3-digit ISIC2 Revision, as follows:
351 (Industrial Chemicals); 352 (Other Chemicals); 382 (Machinery, except electric); 383 (Machinery electric); 384 (Transport Equipment); 385
(Professional & Scientific Equipment)
The sectors classified as low-tech are all of the remaining ones within manufacturing.
This classification follows that suggested by OECD, STAN Database (2001), Annex 3 of the accompanying documentation (available at:
http://www.oecd.org/dataoecd/60/28/21576665.pdf). More precisely, the STAN Database proposes a distinction between high-tech, medium high-
tech, medium low-tech and low tech. In our analysis, we have grouped together the first two categories, i.e. the high-tech and the medium-hi tech, as
this seemed more appropriate for countries at intermediate stages of development.

34
6.2 Results of the regressions on observations derived from simulations
In all of the following regressions, we have expressed the variables in logarithms in order to remove the effect due to the wage differential following
an exponential trend. Whenever problems of serious collinearity appear to emerge, we have omitted in turn each of the regressors being strictly
collinear with others. In particular, this hinders the possibility of running regressions for longer than the 0-30 span for the scenario of section 4.1.
Standard deviations are reported in parentheses. Levels of significance are indicated as follows: *** denotes 99% significance; ** denotes 95%
significance; * denotes 90% significance.
Table 3: Analysis of scenario section 3.1 (A Kuznets curve scenario as a result of a ‘progressive’ technological catching-up) - RGI
Dependent Variable: RGI
(a)
0-66
(b)
66-106
(c)
66-106
(d)
106-200
(e)
106-200
Constant -2.010 -1.795 -0.631 -0.051 -1.838
Wage premium 1.120***
(0.169)
0.491***
(0.140)
- -0.364
(1.120)
-0.329
(1.120)
Skilled labour
unemployment
0.137
(0.136)
0.381***
(0.050)
0.329***
(0.050)
0.601***
(0.170)
-
Unskilled labour
unemployment
0.201**
(0.120)
-0.050
(0.333)
0.002
(0.333)
-0.023
(1.334)
-0.030
(1.334)
Percentage of skilled
workers
- - 0.577***
(0.147)
- -0.776***
(0.187)
R² 0.99994 0.99992 0.99996 0.99966 0.99976
N 660 400 400 940 940

35
Table 4: Analysis of scenario section 3.1 (A Kuznets curve scenario as a result of a ‘progressive’ technological catching-up) - CGI
Dependent variable: CGI
(a)
0-66
(b)
66-120
(c)
66-120
(d)
66-120
(e)
120-200
(f)
120-200
(g)
120-200
Constant -2.289 -1.499 0.628 -1.111 0.614 -1.396 -2.046
Wage premium 0.799***
(0.169)
1.117***
(0.242)
- - -0.107
(0.968)
-0.072
(0.968)
-0.067
(0.968)
Skilled labour
unemployment
0.096
(0.136)
0.616***
(0.077)
0.562***
(0.077)
0.432***
(0.077)
0.713***
(0.089)
- -
Unskilled labour
unemployment
0.147
(0.196)
-0.027
(2.373)
0.011
(2.373)
0.016
(2.373)
- - -
Percentage of skilled
Workers
- - 1.097***
(0.217)
- -0.875***
(0.111)
Labour income
share
-2.830***
(0.025)
- -1.737***
(0.139)
- -1.555***
(0.066)
R ² 0.99991 0.99912 0.99955 0.99976 0.99964 0.99986 0.99997
N 660 540 540 540 800 800 800

36
Table 5: Analysis of scenario section 4.1 Table 6: Analysis of scenario section 4.2
Decreasing inequality with a ‘regressive’ technological lock-in Short-term rise in inequality with a ‘regressive’ technological lock-in
Dependent
variable: RGI
Dependent
variable: CGI
Dependent variable: RGI Dependent variable: CGI
(a)
0-30
(b)
0-30
(a)
0-30
(b)
0-200
(c)
0-30
(d)
0-200
Constant -1.183 -1.653 Constant -0.240 -1.067 -1.247 -1.723
Wage premium 1.439***
(0.063)
0.778***
(0.063)
Wage premium 1.183***
(0.081)
1.537***
(0.170)
0.743***
(0.081)
0.939***
(0.170)
Skilled labour
unemployment
0.103
(0.190)
0.067
(0.190)
Skilled labour
Unemployment
0.060
(0.700)
0.040
(1.664)
0.015
(0.700)
0.032
(1.664)
Unskilled labour
unemployment
0.407***
(0.133)
0.259**
(0.133)
Unskilled labour
Unemployment
0.250
(0.210)
0.418***
(0.163)
0.328*
(0.210)
0.268*
(0.163)
Percentage of
skilled workers
- - Percentage of
skilled workers
- 0.188*
(0.143)
- -0.074
(0.143)
Labour income
share
- -4.583***
(0.029)
Labour income
share
- - -2.371***
(0.051)
-3.451***
(0.043)
R ² 0.99999 0.99994 0.99995 0.99953 0.99993 0.99901
N 300 300 300 2000 300 2000

37
References:
- Aghion, P., Caroli, E., and C. GarcíaPeñalosa, (1999), Inequality and Economic Growth:
The Perspective of the New Growth Theories, Journal of Economic Literature, 37, 1615-60.
- Ahluwalia, M. (1976), Income Distribution and Development, American Economic Review, 66,
128-35.
- Anand S. and S.M. Kanbur (1993), The Kuznets Process and the Inequality-Development
Relationship, Journal of Development Economics, 40, 25-52.
- Antonelli, C. (1995), The Economics of Localized Technological Change and Industrial Dynamics.
Kluwer, Boston.
- Atkinson, A. B. and J. E. Stiglitz, (1969), A New View of Technological Change, Economic
Journal, 79, 573-8.
- Barro, R. J. (2000), Inequality and Growth in a Panel of Countries, Journal of Economic
Growth, 5, 5-32.
- Berman, E. and S. Machin (2000), Skill-Biased Technology Transfer around the World,
Oxford Review of Economic Policy, 16, 12-22.
- Berman, E. and S. Machin (2004), Globalization, Skill-Biased Technological Change and
Labour Demand, in Lee, E. and M. Vivarelli, (eds), Understanding Globalization, Employment
and Poverty Reduction, Palgrave Macmillan, New York, forthcoming.
- Bourguignon, F. (1990), Growth and Inequality in the Dual Model of Development: The
Role of Demand Factors, Review of Economic Studies, 57, 215-28.
- Cornia, G. A. and S. Kiiski (2001), Trends in Income Distribution in the Post-World War
II Period, WIDER Discussion Paper no. 2001/89, World Institute for Development
Economics Research, United Nations University, Helsinki.
- Deininger; K. and Squire, L. (1996). ‘A New Data Set Measuring Income Inequality’, The
World Bank Economic Review, 10, 565-91
- Dollar D. and A. Kraay (2001), Trade, Growth and Poverty, mimeo, Development Research
Group, The World Bank, Washington, June.
- Edwards, S. (1997), Trade Policy, Growth and Income Distribution, American Economic
Review Papers and Proceedings, 87, 205-10.
- Fagerberg, J. (1994). Technology and international differences in growth rates, Journal of
Economic Literature, 32, 1147-75
- Fagerberg, J. and Verspagen, B. (2002). Technology-gaps, innovation-diffusion and
transformation: an evolutionary interpretation, Research Policy, 31, 1291-1305
- Fields, G. (1980), Poverty, Inequality and Development, Cambridge University Press, Cambridge.

38
- Galor, O. and Moav, O. (2000). Ability-Biased Technological Transition, Wage Inequality
and Economic Growth, Quarterly Journal of Economics, 115, 469-97
- Galor, O. and Tsiddon, D. (1996). Income Distribution and Growth: The Kuznets
Hypothesis Revisited, Economica, 63, supplement, S103-S117.
- Galor, O. and Tsiddon, D. (1997). ‘Technological Progress, Mobility and Economic
Growth’, American Economic Review, 87, 363-82
- Goodwin R M (1967), ‘A Growth Cycle’, in: Feinstein, CH (ed): Socialism, Capitalism and
Economic Growth, London: Mac Millan
- Greenwood, J. and B. Jovanovic (1990), Financial Development, Growth and the
Distribution of Income, Journal of Political Economy, 98, 1076-1107
- Grimalda, G. (2002). Growth with Competing Technologies and Bounded Rationality, Southampton
University Discussion Papers N. 0205
- Guckhenheimer, J. and Holmes, P. (1990): Nonlinear Oscillations, Dynamical Systems and
Bifurcations of Vector Fields, Berlin: Springer Verlag
- Higgins, M. and J. G. Williamson (1999), Explaining Inequality the World Round: Cohort
Size, Kuznets Curves, and Openness, NBER Working Paper no. 7224, National Bureau of
Economic Research, Cambridge (Mass.).
- Hirsch M.W. and Smale S. (1974): Differential Equations, Dynamical Systems, and linear Algebra,
London: Academic Press
- Hogarth, R., and M. Reder (eds.), (1986). Rational Choice, Chicago: The University of
Chicago Press
- Kaldor, N. (1957). “A Model of Economic Growth, Economic Journal”, Economic Journal,
67, 591-624
- Kuznets, S. (1955), Economic Growth and Income Inequality, American Economic Review,
45, 1-28.
- Kuznets, S. (1963), Quantitative Aspects of the Economic Growth of Nations, Economic
Development and Cultural Change, 11, 1-80.
- Lall, S. (2004). The Employment Impact of Globalisation in Developing Countries, in Lee,
E. and M. Vivarelli, (eds.), Understanding Globalization, Employment and Poverty Reduction,
Palgrave Macmillan, New York, forthcoming.
- Li, H., L. Squire and H. Zou (1998), Explaining International and Intertemporal Variations
in Income Inequality, Economic Journal, 108, 26-43.
- Lundberg M. and L. Squire (2001), The Simultaneous Evolution of Growth and
Inequality, mimeo, The World Bank, Washington.

39
- Machin, S. and J. Van Reenen (1998), Technology and Changes in the Skill Structure:
Evidence from Seven OECD Countries, Quarterly Journal of Economics, 113, 1215-44
- Mankiw, N. G., Romer, D. And Weil, D. (1992) ‘A Contribution to the Empirics of
Economic Growth’, Quarterly Journal of Economics, 107, 407-37.
- Nelson R, Winter S. G. (1982), An Evolutionary Theory of Economic Change, Cambridge MA:
The Belknap Press of Harvard University
- Papanek, G. and O. Kyn (1986), The Effect on Income Distribution of Development, the
Growth Rate and Economic Strategy, Journal of Development Economics, 23, 55-65.
- Ravallion, M. (2001), Growth, Inequality and Poverty: Looking beyond Averages, mimeo,
Development Research Group, The World Bank, Washington, June.
- Reuveny, R. and Q. Li (2003), Democracy, Economic Openness, and Income Inequality:
An Empirical Analysis’, Comparative Political Studies, 36.
- Robbins, D. J. (1996), HOS Hits Facts: Facts Win; Evidence on Trade and Wages in the
Developing World, Development Discussion Paper no. 557, Harvard Institute for International
Development, Harvard University, Cambridge (Mass.).
- Robbins, D. J. (2003), The Impact of Trade Liberalization Upon Inequality in Developing
Countries: A Review of Theory and Evidence, working paper, International Policy Group,
International Labour Office, Geneva
- Robinson, S. (1976), A Note on the U Hypothesis Relating Income Inequality and
Economic Development, American Economic Review, 66, 437-40.
- Sala-i-Martin, X. (2002), The Disturbing ‘Rise’ of Global Income Inequality, NBER
Working Paper no. 8904, National Bureau of Economic Research, Cambridge (Mass.).
- Silverberg, G. and B. Verspagen (1995), “An evolutionary model of long term cyclical
variations of catching up and falling behind”, Journal of Evolutionary Economics, 5, 209-28
- Simon, H. (1955). ‘Behavioral Model of Rational Choice’, Quarterly Journal of Economics, 69,
99-118
- Soete, L. and Turner, R. (1984). “Technology Diffusion and the Rate of Technical
Change”, Economic Journal, 94, 612-24
- Vivarelli, M. (2004). Globalization, Skills and Within-Country Income Inequality in
Developing Countries, in Lee, E. and M. Vivarelli, (eds.), Understanding Globalization,
Employment and Poverty Reduction, Palgrave Macmillan, New York, forthcoming.
- Weibull, J. (1995), Evolutionary Game Theory, Mit Press, Cambridge: (Mass.).
- Wood, A.(1994), North-South Trade, Employment and Inequality. Changing Fortunes in a Skill-
Driven World, Clarendon Press, Oxford.