Professional Documents
Culture Documents
Antonelli. Income Inequality in The Knowledge Economy
Antonelli. Income Inequality in The Knowledge Economy
Antonelli. Income Inequality in The Knowledge Economy
PII: S0954-349X(20)30374-X
DOI: https://doi.org/10.1016/j.strueco.2020.07.003
Reference: STRECO 948
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
1
INCOME INEQUALITY IN THE KNOWLEDGE ECONOMY1
Cristiano Antonelli a, b,
Matteo Tubiana a, c
a
Dipartimento di Economia e Statistica Cognetti de Martiis, Università di Torino,
Lungo Dora Siena 100/A, 10153, Torino, Italy.
b
Collegio Carlo Alberto. Piazza Vincenzo Arbarello, 8, 10122 Torino, Italy.
c
Dipartimento di Scienze Aziendali, Economiche e Metodi Quantitativi, Università di
Bergamo. Via dei Caniana, 2, 24127 Bergamo, Italy.
Emails: cristiano.antonelli@unito.it; matteo.tubiana@unibg.it.
1
The funding of the research project PRIN 20177J2LS9 is acknowledged.
2
1. INTRODUCTION
The historical analysis of the distribution of income shows a clear
discontinuity in the long-term evolution at the end of the XX century. The
long-term reduction of income inequality levelled off in most advanced
countries in the last decades of the XX century. Since the beginning of the
XXI century, there is abundant evidence of the beginning of a new trend
towards increasing levels of inequality (Atkinson, Piketty, Saez, 2011).
The new trend towards increasing levels of income inequality parallels the
introduction and diffusion of new information and communication
technologies (ICT) and the globalisation of both product and financial
markets. This trend shapes the shift of advanced economies away from the
manufacturing industry as the pillar of their economic structure and the
emergence of knowledge-intensive business services (KIBS) as the new
key sectors (Atkinson and Piketty, 2007 and 2009; Piketty, 2014; Franzini
and Pianta, 2016).
This paper provides a theory of the relations among these three dynamics –
namely inequality, globalisation and knowledge economy transition – and
lays down a first empirical attempt to measure it. It articulates the
hypothesis that the new tradability of knowledge as an economic good that
can be exchanged as a service, capitalised in financial assets and intangible
property rights, rather than embodied in other tangible goods, together
with the radical structural change of the economic system towards a
knowledge economy are the cause of the increasing levels of income
inequality experienced by advanced countries.
3
The shift to the knowledge economy can be regarded as a radical change in
the economic structure of advanced countries. Following the way paved by
Kuznets (1955), radical changes in the economic structure of economic
systems are likely to affect income inequality with an inverted-U shape.
Income inequality raises when the change in the structure of the system is
radical and slows down eventually (Antonelli, 2019).
4
the transaction takes place. The risk, however, shifts from the customer to
the vendor as soon as the details of the piece of knowledge are shared. The
customer, who was – ex-ante – exposed to the opportunistic conduct of the
prospective vendor, may – ex-post – implement an opportunistic
behaviour, leave the knowledge marketplace and take advantage of the
information disclosure without any payment. Now the risk has fully shifted
to the vendor. The Chandlerian corporation, based upon the separation
between ownership and control and the vertical integration of knowledge
generation intra-muros enabled to handle the problems of the limited
appropriability and tradability of knowledge (Lazonick, 2010).
The use of the wide range of ICT and the strengthening of IPRs have
changed in depth the mechanisms of generation, appropriation,
exploitation and market exchange of knowledge as a property right in the
new markets for patents, as service in internal quasi-markets as well as in
arm‟s length transactions, and capitalised as an asset traded in financial
markets. Let us analyse them in turn.
The new IPR regime. The enforcement in March 1994 of the Agreement of
Trade-related Aspects of Intellectual Property Rights (TRIPs Agreement)
and the sequence of patent reforms in the USA led to the strengthening of
the IPRs regime and their globalisation (Pagano and Rossi, 2009). The
enhanced privatisation of knowledge with the strengthening of the IPR
regime can be regarded as one of the main institutional changes that
characterise and favour the emergence of the new knowledge economy.
Many have compared the current trends towards the reinforcement and
extension of IPR to the enclosure of common land that preceded and
actually enabled the Industrial Revolution. The enhanced privatisation of
knowledge is clearly necessary to support the new tradability of
knowledge as a service. The strengthening of the IPR, in fact, reduces the
risks of opportunistic behaviour associated with the information
asymmetry in knowledge exchange and supports the specialisation of firms
in the generation and market exchange of knowledge as a service, rather
than an input embodied in other goods (Gallini, 2002; Pagano, 2014;
Aghion, Howitt, Prantl, 2015).
The new IPR regime supports the growth of the new markets for
knowledge. Next to the trade of knowledge as a service, knowledge is
traded as a patent and a license (De Rassenfosse, Palangkaraya, Webster,
5
2016). The trade as IPRs is actually complementary to the trade of
knowledge as a service, since the exchange of property rights is
implemented by the provision of dedicated services that enable customers
to take advantage of the proprietary knowledge that is acquired. In
traditional IPRs markets, transactions took place mainly across borders
between corporations active on the supply side, and foreign manufacturing
firms on the demand side (Arora, Fosfuri and Gambardella, 2001;
Branstetter, Fisman, and Foley 2006). In the new larger and denser
markets for IPRs, transactions take place to a more significant extent
within national borders between scientific entrepreneurship and small
knowledge-intensive firms active on the supply side and corporations on
the demand side (De Marco, Scellato, Ughetto, Caviggioli, 2017;
Caviggioli and Ughetto, 2013; Monk, 2009).
6
while avoiding their respective shortcomings. Venture capitalists, like the
innovative banker, perform the screening of new projects well beyond the
limits of the internal competence of the incumbents, accessing the broad
spectrum of professional and scientific competencies available in the
marketplace. Like the corporation, it is able to mobilise competent
managers to assist in the development of the start-up. Like the corporation,
it can raise equity to fund innovations and actually participate not only in
failures – as it is the case of the banker – but also to successes. The take-
over of successful start-ups after their IPO in the stock markets becomes a
significant source of new knowledge. After the take-over, the new small
high-tech companies are delisted and become part of the corporation that,
in so doing, acquires advanced technological competence and effective
prototypes, well screened and tested with respect to both the engineering
and the marketing side. The take-over of small high-tech firms substitutes
intra-muros R&D activities within corporations. Most importantly, for the
focus of this paper, it enables venture capitalists, including scientific
entrepreneurs, to appropriate the economic value of knowledge, capitalised
as an asset. The exploitation of knowledge capitalised as equity takes
place in the stock markets that perform the new function of markets for
knowledge embodied in knowledge intensive assets (Kortum and Lerner,
2000; Gompers and Lerner, 2004).
7
of employment retained within headquarters that became knowledge
intensive service providers.
8
extensibility that, in turn, increase the levels of appropriability. The new
dedicated and idiosyncratic knowledge, generated on purpose for a specific
customer, enters and remains in the stock of internal knowledge of
suppliers and becomes an additional input for its eventual recombinant
generation and sale embodied in new knowledge services to other
customers. Customers have not access to the stock of knowledge, whereas
the providers retain the full command of the knowledge outputs and of the
procedures and methodologies that enabled them to elaborate specific
applications and solutions to the problems of the customers. Customers
can purchase the applications of the “algorithms”, not the “algorithms”
themselves. Appropriability is endogenous to both the generation and the
trade of knowledge as a service (Gans and Stern, 2017). Vendors can
appropriate large knowledge rents because they can take advantage of the
dynamic increasing returns triggered by the limited exhaustibility of
knowledge. The profitability of KIBS increases as they can keep selling
specific knowledge and yet increasing their stocks of generic knowledge.
Corporations rely more and more on external sources for their knowledge
inputs. Research and development activities are more and more outsourced
to specialised suppliers that act as providers of specialised knowledge
inputs. On the one side, the radical change in knowledge tradability has
positive effects on the efficiency of knowledge generation as it enables
higher levels of division of labour and specialisation among a variety of
specialised suppliers. The reduction of the adverse effects of the not-
invented-here syndrome, limiting the capability of corporations to
overcome the enclosures generated by the limited scope of their own
competences, comes as a consequence. The increased tradability of
9
knowledge also has positive effects in terms of larger opportunities to use
the market as a screening and selection device, better able than internal
hierarchies to sort projects that have more opportunities to succeed.
10
In these figures, the share of employment accounted by manufacture
declines for every advanced country, with the exception of the Czech
Republic. Symmetrically, in the historical period analysed, the relevance
of KIBS is increasing sharply in European regions: it climbs in Sweden
from around 5% to almost 9%, in the UK from 7% to 10.6%, the
Netherlands from 8% to almost 10%.
These changes, in turn, affect in-depth the income distribution with two
different processes: i) the polarisation of labour markets, caused by the
decline of demand and increased unemployment of standard labour and
augmented demand of creative; ii) the increase of profits associated to the
generation, appropriation and exploitation of knowledge, mostly unshared
with low-skilled workers. Respectively, both wage and rent (wealth)
inequalities increase. Let us analyse each process in turn.
11
Figure 1. Employment share2 of KIBS and Manufacture all OECD
countries. Source: OECD.
2
Drawing from OECD National Accounts tables, we exploit indexes of the share of employment (in persons) accounted
by KIBS and Manufacture as a proxy for their economic centrality.
12
Wage Inequality. The demise of the Chandlerian corporation engenders the
reduction of average wages at the system level: in large corporations, blue-
collar workers did benefit of high unit wages, far above baseline wage
levels in the economy at large, stemming from their high levels of
unionisation. In the corporation, in fact, unit wages did exhibit a strong
positive relationship with firms‟ size and profitability (Mueller, Ouimet,
Simintzi, 2017; Card, Heining Kline, 2017; Farber et al., 2018). Within
manufacture, instead of decreasing wages, we observe a contraction of
employment. Indeed, the bargaining power of unionised labour, even if
amid a declining trend across OECD countries (see Figure 3), defends
sticky wages in standard labour (Figure 4). Overall, the share of workers
protected by unions and benefitting of above-average wages declines.
13
Figure 3. Employment and Average Wages on base 100 in 1990 (or at first
later date available). KIBS and Manufacture comparison. Source: OECD.
14
Standard labour employed in the production of manufacturing goods is
exposed to the stiff competition raised by the entry of new large, low-wage
and labour abundant economies. Employment contraction is associated
with sticky wages. The markets for creative labour, able to participate
actively into the generation and exploitation of technological knowledge,
on the opposite, are protected by international competition by the unique
conditions of accessing and using the large stock of technological
knowledge, cumulated through time and implemented by high-quality
knowledge governance mechanisms. Such markets are characterised, on
the demand side, by the positive effects exerted by the knowledge-
intensive direction of structural change. The increased levels of knowledge
derived demand have, indeed, positive effects on both job opportunities
and wage levels. The increasing levels of wage inequality are reinforced
by the upstream shift of profits in the global value chains, their
concentration in the KIBS sectors and the participation of professional and
scientific entrepreneurship in their appropriation.
15
knowledge rents directly as profits (Audretsch and Link, 2018). In the
knowledge economy, the share of income paid to capital is larger than in
the corporate economy. As a consequence, income inequality increases
through higher rent (wealth) inequality.
The new tradability of knowledge, hence, has direct effects on the levels of
knowledge appropriability, increasing knowledge rents and the
polarisation of labour markets, differently remunerating creative and
standard workers, with a consequential increase in the levels of wage
inequality. The increase in both wage and rent inequalities reinforces the
increase in income inequality.
3. EMPIRICAL ANALYSIS
3.1 Data
We build a panel dataset of countries borrowing from two sources.
Namely, i) the Standardized World Income Inequality Database (SWIID
8.1, Solt, 2016) which is a valuable and extensive source for income
inequality data (in the forms of Gini indexes3) normalised across countries
and sources; and ii) various tables from the OECD repository.
3
From the word of the SWIID creator, Frederick Solt: «I think the clearest explanation of the Gini index is that it is half
the average difference in income between all pairs of units—say, households—as a percentage of the mean income of
those units. Okay, I said ―clearest,‖ not necessarily ―clear.‖ Anyway, it has a theoretical range of 0 (all households have
the same income) to 100 (one household has all the income and the rest have none), but Ginis below 20 or above 60 are
rare in the real world. There are good reasons to prefer other measures of inequality, and there are many options, but the
Gini is by far the most widely available» (https://fsolt.org/blog/2017/07/28/the-swiid-source-data.html)
16
SWIID is the output of a strenuous effort to generate freely available data
on inequality covering the maximum range of countries and years, through
harmonisation and imputation of various data sources4. We exploit it to
define our dependent variable: income inequality. As Florida and
Mellander (2016) report, income and wage inequality describe different
geographical distributions since they reflect different but interlinked
phenomena. Income inequality, indeed, is determined both by wage and
rent inequalities. Income inequality is, therefore, a proper measure to
investigate our research questions.
The Gini index is just one of the many measures used to grasp inequality.
We opted for it in order to maximise the length and breadth of our panel,
given that we want to explore the transitions of economies from one
regime to another. Figure 6 in the Appendix shows the correlations
between some indexes of inequality: the Gini on disposable and market
income from SWIID and other three indexes from OECD – the Palma
Ratio, the Percentile Ratio and the Gini, all of them computed on
disposable income. Different measures of inequality computed on the
same type of income are highly and positively correlated, whereas the
correlation between Ginis on market and disposable income is average
intensity. In the econometric analysis that will follow, we go for the Gini
on market income as the main dependent variable. Indeed, the difference
between market and disposable income is the mediation of the welfare
state redistributing income within the population. At this stage, we are
interested in the unmediated, direct effect of the knowledge economy
transition in income, duly accounting for state-level idiosyncrasies.
Figure 5 describes the market and disposable income inequality trends for
a panel of OECD countries. It portraits a generalised condition of rising
inequality but qualified by a certain degree of heterogeneity from country
to country. Most of the plotted countries experience rising inequality from
4
Jenkins (2015) examines two major sources of inequality data, namely SWIID and WIID. He expresses a conditional
favor for the use of the latter for comparative studies. SWIID, indeed, is the output of a not completely clear multiple
imputation methodologies on WIID raw data, pointing to maximize the countries time-series and making them
comparable, properties that are not provided by WIID. WIID, instead, leaves to the researcher the duty to match and
harmonize the series, with a large amount of discretion. Even though we are thankful for the insightful exercise, we do
not agree with Jenkins conclusions in two regards. First, as he brings evidence, empirical analysis from WIID and
SWIID for developed countries, whose data availability is larger, is consistent. Second, we think that relying on the
research efforts of established and trustable academics and institutions is necessary and advisable for new,
parsimonious, frontline research insights to be provided.
17
the „80s onwards, on average, as documented on the regional scale by
Rosés and Wolf (2018). France and Italy exhibit the renown U-shaped
trend (Piketty and Saez, 2003). As announced, market and disposable
income inequality follow very similar trajectories for most countries, but
figures in the lower panel exhibit more pronounced shapes. This indicates
that redistribution policies are at work.
18
Figure 5. Trends in income inequality and redistribution. Source: SWIID.
19
Another available measure of a country‟s knowledge intensity is R&D
investments as a portion of GDP. The amount of resources devoted to
research is undoubtedly correlated with the emergence of knowledge-
based firms, but the two are not measuring the same phenomenon. Indeed,
R&D investments, both from the business and the public sectors, may take
place – actually took place – even without incurring in a knowledge-
dedicated sector. KIBS firms now perform an increasing share of business
related R&D investments.
We are aware that the dynamics of knowledge production and use can
account for only a fraction of a vast, structural phenomenon as inequality,
which is mostly feeding itself. In the framework we proposed, the
globalisation of product and financial markets covers a pivotal role as a
trigger of the knowledge transition of industrialised countries. We measure
it as exports plus imports relative to GDP, extracted from the World
Bank‟s World Development Indicators. To comply with the argument put
forward above about the distributional struggles arising from de-
unionisation, we add the coverage of unions on employment, or union
density, to the list, as well as the unemployment rate. Moreover, we use
the log of GDP per capita at PPP to control for countries‟ economic
development. All these variables come from the OECD datawerahouse.
20
on 15-64 persons) and age-adjusted mortality rate (as an indicator of
societal life quality), once again from the OECD repositories.
Where i and t stand for country and year subscripts, KE is a matrix filled
by the knowledge economy transition measure – the centrality of KIBS
and BERD expenditures on GDP – and matrix X contains a set of control
variables progressively plugged into the model. To protect our estimates
from omitted variables, we proceed with a two-ways FE estimation,
therefore controlling for both the unobserved time-invariant heterogeneity
at the country level and the endogenous time effect, usually referred as
business cycle effect. All covariates enter the regression at the same point
in time of the dependent variable. Table 1 reports the first set of results.
5
The longest sample of countries involved in the main empirical exercises include: Austria, Australia, Belgium,
Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Norway, New Zealand, Poland,
Portugal, Spain, Sweden, United Kingdom, United States. Depending on the covariates, the list of countries shrinks.
21
The first column introduces the knowledge economy variables only, which
are positive and significant. In the second round, variables accounting for
the distributional issues related to employment comes in: the estimated
parameters are significant and with expected signs. In the third column,
variables related to the structure of the economy step in but are not much
significant. Finally, the fourth column introduces the demographic
variables: the strongest, and significant, (negative) effect comes from the
elderly dependence ratio6. This first set of regressions, already introducing
a wide range of controls, does confirm the expected positive correlation
between the knowledge economy transition and income inequality within
advanced countries.
6
The VIF of the relevant variables (KIBS % and BERD expenditures) for these regressions is satisfactory, well below
10. Indeed, the highest value is that of KIBS (~5) followed by GERD (~2.3). However, the VIF for KIBS is pushed up
by the introduction of population density rather than by GERD, even though the two are correlated at 0.43.
22
Table 1. Two-ways Fixed Effect estimator7.
Dependent variable:
(Log of) Gini on Market Income
(1) (2) (3) (4)
BERD on GDP 0.027*** 0.026*** 0.030*** 0.024**
(0.007) (0.006) (0.007) (0.008)
KIBS % (Persons) 0.009* 0.012* 0.016*** 0.015***
(0.004) (0.004) (0.004) (0.004)
Openness -0.0001 -0.0001 -0.0002
(0.0001) (0.0001) (0.0002)
Union density -0.001** -0.001** -0.001*
(0.0004) (0.0005) (0.0005)
Unempl rate 0.003*** 0.004*** 0.005***
(0.0005) (0.001) (0.001)
HC 0.062. 0.063*
(0.033) (0.030)
GDP pc 0.037 0.066*
(0.026) (0.030)
EPL 0.020 0.035*
(0.016) (0.015)
Pop density -0.001*
(0.0003)
Elderly dep. ratio 0.003**
(0.001)
Mortality 0.001
(0.001)
7
An R2 of 0.32 may appear low respect to a model with such a wide range of variables. However, one has to remind
this is a within R2, that is, it finds the total sum of squares on the demeaned outcome variable. The overall R2, obtained
through a LSDV with the very same model, scores as high as 0.93. Considering that inequality is a inherent, path-
dependent phenomenon, we think a 0.32 within-R2 is a satisfactory score.
23
Table 2 and Table 3 report a few robustness checks, with two different
estimators. The first column in Table 2 reproduces the last one in Table 1
as a reference to the main results. The second column tests the same model
on all available OECD observations, including considerably more
countries8. The most relevant differences are that, in Table 2, the
coefficient of BERD increases in size, whereas Openness and Mortality
get statistical significance.
24
(Bell, Jones, 2015). The rationale is to consider the data as hierarchical –
two levels: country and time – and estimate a random intercept model to
account for the multilevel structure. Variables enter the regression as
yearly deviations from the country mean (the shape of one-way FE
model). Year dummies are used too in order to control for common
shocks. Other than deviations from the mean, addressing the within
variance component, the researcher can insert variables country averages,
grasping the between component, or any other level-invariant
characteristic. Estimation happens via ML optimisation.
The first column in Table 3 reproduces, once again, the last one in Table 1
as a reference to the main results. Then, the second column introduces the
labour regulation indexes. Even though KIBS and BERD coefficients‟
shrinks in size and significance, they resist the new fine-grained controls.
Regarding column two, Social security laws is the only statistically
significant variable, displaying an inequality-protecting effect. The
remaining three columns repropose the robustnesses illustrated for Table
2, with the alternative Within-Between RE estimator and the labour
regulation indexes. No relevant issues are spotted. These results confirm
the hypotheses, but it seems appropriate to take into account the
limitations of our investigation.
25
Table 2. Two-ways FE estimator. Robustness checks.
Dependent variable:
(Log of) Gini on Market Income (Log of) Gini on Disposable Income
(1) (2) (3) (4)
BERD on GDP 0.024** 0.034*** 0.035*** 0.051***
(0.008) (0.007) (0.010) (0.010)
KIBS % (Persons) 0.015*** 0.015*** 0.013**
(0.004) (0.003) (0.005)
KIBS % (Persons, 2nd) 0.008**
(0.003)
Openness -0.0002 -0.001*** -0.0002 -0.0001
(0.0002) (0.0001) (0.0002) (0.0002)
Union density -0.001* -0.001 -0.001* -0.002*
(0.0005) (0.0004) (0.001) (0.001)
Unempl rate 0.005*** 0.004*** 0.004*** 0.003**
(0.001) (0.001) (0.001) (0.001)
HC 0.063* -0.001 0.065* -0.032
(0.030) (0.025) (0.031) (0.036)
GDP pc 0.066* 0.025 0.058. 0.050
(0.030) (0.026) (0.031) (0.038)
EPL 0.035* 0.030* 0.030. 0.036*
(0.015) (0.012) (0.016) (0.014)
Pop density -0.001* -0.0003 -0.001. -0.001*
(0.0003) (0.0003) (0.0003) (0.0004)
Elderly dep, ratio 0.003** 0.003* 0.003. 0.004*
(0.001) (0.001) (0.001) (0.002)
Mortality 0.001 0.003** 0.002 0.006***
(0.001) (0.001) (0.001) (0.002)
26
Table 3. Within-Between Random Effect Estimator. Robustness checks.
Dependent variable:
(Log of) Gini on Disposale
(Log of) Gini on Market Income
Income
(1) (2) (3) (4) (5)
BERD on GDP 0.024*** 0.017** 0.029*** 0.027*** 0.041***
(0.007) (0.006) (0.005) (0.007) (0.008)
KIBS % (Persons) 0.013*** 0.012** 0.012*** 0.010*
(0.003) (0.004) (0.003) (0.005)
KIBS % (Persons, 2nd definition) 0.007**
(0.003)
Openness -0.0001 -0.0001 -0.001*** -0.0001 -0.0001
(0.0002) (0.0002) (0.0001) (0.0002) (0.0002)
Union density -0.001** -0.001* -0.001* -0.001** -0.002**
(0.0005) (0.0005) (0.0003) (0.001) (0.001)
Unempl rate 0.005*** 0.004*** 0.003*** 0.003*** 0.002**
(0.001) (0.001) (0.0005) (0.001) (0.001)
HC 0.078** 0.041 -0.005 0.053. -0.058
(0.029) (0.028) (0.022) (0.029) (0.038)
GDP pc 0.063** 0.050* -0.007 0.044. 0.038
(0.022) (0.023) (0.020) (0.025) (0.030)
Pop density -0.001* -0.001. -0.0002 -0.0004 -0.001*
(0.0003) (0.0003) (0.0003) (0.0003) (0.0004)
Elderly dep, ratio 0.003*** 0.003** 0.003** 0.002. 0.004**
(0.001) (0.001) (0.001) (0.001) (0.001)
Mortality 0.001 0.001 0.004** 0.002 0.006**
(0.002) (0.002) (0.001) (0.002) (0.002)
Employment laws -0.070 -0.180* -0.090 -0.559**
(0.074) (0.078) (0.077) (0.190)
Social security laws -0.345* -0.333* -0.390** -0.308
(0.144) (0.141) (0.150) (0.369)
Civil rights 0.012 -0.141 0.026 0.268
(0.084) (0.088) (0.083) (0.216)
Collective relations laws 0.080 0.212* 0.094 0.456.
(0.094) (0.100) (0.093) (0.243)
27
We might conjecture that there are chances for some common trends to
exist, which might yield a correlation where there is none, instead.
Therefore, we test for the presence of a unit root in the measures for
inequality, KIBS and GERD quota, but the variables look stationary10.
Nonetheless, we experiment with an Error Correction Model, estimated
through OLS, taking into account the presence of a co-integration between
the variables at stake. However, results (Table 6 in the Appendix) are not
robust: the autoregressive component of the model absorbs most of the
effects and BERD expenditures on GDP are barely significant in the long
run. We tested a dynamic model, augmented by the first lag of our
dependent variable, with the two-steps, difference-GMM estimator.
Similarly to the ECM case, results are not robust (Table 6 in the
Appendix): the KIBS centrality index is only slightly significant.
Interestingly, the autoregressive component is not.
4. CONCLUSIONS
This paper dives into the interplay between the radical structural change
occurring in the transition to the knowledge economy and the increase of
income inequality. The analysis explores the effects of the changes in the
organisation of the generation, exploitation and appropriation of
technological knowledge on the rising levels of income inequality
experienced by advanced countries since the end of last century. We
10
We perform the Im et al. (2003) (IPS) and Maddala and Wu (1999) (MADWU) panel unit root tests, available for
unbalanced panel data in the plm R package, version 2.2-3. The IPS test on BERD investments on GDP does not reject
the null of a unit root, whereas the MADWU test does. We then perform the same tests on the residuals of the Gini
index regressed on BERD: both reject the null.
28
introduced a theoretical framework where the increasing levels of income
inequality are a direct consequence of the new specialisation of advanced
countries in the knowledge generation and exploitation based on the KIBS
industries and the consequent polarisation of labour markets.
Exploiting available data mostly from the OECD data warehouse and the
Standardized World Income Inequality Database, we brought evidence of a
novel correlation that confirms the validity of the hypotheses. Even though
no causality is claimed in the paper, quite an effort has been made to stress
the main correlation with a heterogeneous set of controls and various
model specifications. Still, results were not robust in the models dealing
with the autoregressive component of inequality. This is a limitation of the
present analysis, which we aim at addressing specifically in future
research.
The evidence confirms the emerging limits in the ability of the knowledge
economy to share and distribute the returns from the generation of
technological knowledge that are now better appropriated but only by a
smaller portion of stakeholders.
29
ACKNOWLEDGMENTS
The authors acknowledge the comments of the referees and the director of this
journal, many attendants to the international conference ―Innovation and industrial
economics‖ held at the Nanjing University, June 2018 and the 20 th AISSEC
Conference ―Rise and Decline of Economies: A Comparative Perspective‖, Collegio
Carlo Alberto, Torino, October 2018 and Guido Pialli. The funding of the research
project PRIN 20177J2LS9 is also acknowledged.
30
REFERENCES
Aghion, P., Howitt, P., Prantl, S. (2015), Patent rights product market
reforms and innovation, Journal of Economic Growth 20(3), 223-262.
Atkinson A.B., Piketty, T. (eds.) (2007), Top incomes over the Twentieth
century: A contrast between continental European and English speaking
countries. Oxford University Press, Oxford.
Atkinson A.B., Piketty, T. (eds.) (2009). Top Incomes over the Twentieth
century: A global perspective. Oxford University Press, Oxford.
Atkinson, A.B., Piketty, T., Saez, E. (2011), Income inequality in the long
run history, Journal of Economic Literature 49(1), 3-71.
31
Audretsch, D.B., Link, A. (2018), Sources of knowledge and
entrepreneurial behavior, Toronto University Press, Toronto.
Autor, D., Dorn, D., Hanson, G.H., Pisano, G., Shu, P. (2020), Foreign
competition and domestic innovation: Evidence from US patents,
American Economic Review, forthcoming.
Bloom, N., Garicano, L., Sadun, R., Van Reenen, J. (2014), The distinct
effects of information technology and communication technology on firm
organisation, Management Science 60, 2859-2885.
Card, D., Heining, J., Kline, P. (2013), Workplace heterogeneity and the
rise of West German wage inequality, Quarterly Journal of Economics
128(3), 967-1015.
Caviggioli, F., Ughetto, E. (2013), The drivers of patent transactions:
Corporate views on the market for patents, R&D Management 43(4) 318-
332.
Choi, C. (2010), The effect of Internet on service trade, Economics Letters
109, 102-104.
32
Crouch, C. (2019), Inequality in post-industrial societies, Structural
Change and Economic Dynamics 51, 11-23.
De Marco, A., Scellato, G., Ughetto, E., Caviggioli, F. (2017), Global
markets for technology: Evidence from patent transactions, Research
Policy 46(9), 1644-1654.
De Rassenfosse, G., Palangkaraya, A., Webster, E. (2016), Why do patents
facilitate trade in technology? Testing the disclosure and appropriation
effects, Research Policy 45(7), 1326-1336.
Dorn, F. (2016), On data and trends in income inequality around the
world, CESifo DICE Report, 4/2016.
Farber, H.S., Herbst, D., Kuziemko, I., Suresh, N. (2018), Unions and
inequality over the twentieth century: New evidence from survey data,
NBER Working Paper 24587
Feenstra, R.C., Inklaar, R., Timmer, M.P. (2015), The next generation of
the Penn World Table, American Economic Review 105, 3150–3182.
33
Garicano, L., Rossi-Hansberg, E. (2006), Organization and inequality in a
knowledge economy, Quarterly Journal of Economics 121 (4), 1383-1435.
Goldfarb, A., Greenstein, S., Tucker, C. (2015), Economic analysis of the
digital economy, University of Chicago Press, Chicago.
Gompers, P.A., Lerner, J. (2004), The venture capital cycle, MIT Press,
Cambridge.
Hartmann, D., Guevara, M. R., Jara-Figueroa, C., Aristarán, M., &
Hidalgo, C. A. (2017), Linking economic complexity, institutions, and
income inequality, World Development, 93, 75-93.
Im, K. S., Pesaran, M. H., Shin, Y. (2003), Testing for unit roots in
heterogeneous panels, Journal of Econometrics 115(1), 53-74.
34
Mueller, H.M., Ouimet, P.P., Simintzi, E. (2017), Wage inequality and
firm growth, American Economic Review 107(5) 379-383.
Pagano, U., Rossi, M.A. (2009), The crash of the knowledge economy,
Cambridge Journal of Economics 33(4), 665-683.
Piketty, T., Saez, E. (2003), Income inequality in the United States, 1913–
1998, Quarterly Journal of Economics, CXVIII, 1-39
35
Stockhammer, E. (2017), Determinants of the wage share: A panel
analysis of advanced and developing economies, British Journal of
Industrial Relations 55, 3–33.
36
APPENDIX
37
Table 4. Variables description. Source: own elaboration.
Variable Source Type Description
Gini Mkt SWIID Dependent Gini on market income
Share of Persons
OECD National
KIBS % Main explanatory accounted by KIBS on
Accounts
total economy
Business-related
BERD on GDP OECD Main explanatory
expenditure in R&D
Export plus imports
Openness World Bank WDI Main control
relative to GDP
Unemployment rate of
Unempl rate OECD Main control
the population
Coverage of unions on
Union density OECD Main control
employment
combination of
workers’ years of
HC Penn World Tables Main control
schooling and assumed
return from education
GDP pc OECD Main control GDP at PPP per capita
Index of stringency of
EPL OECD Robustness control
the labour market
Pop density OECD Robustness control Population density
Share of over 65 on 15-
Elderly dep ratio OECD Robustness control
64 persons
Age-adjusted mortality
Mortality rate OECD Robustness control
rate
Aggregate index on
Employment laws Botero et al. (2004) Robustness control individual employment
relations
Aggregate index on
Collective relations laws Botero et al. (2004) Robustness control collective employment
relations
Aggregate index on
Social security laws Botero et al. (2004) Robustness control social security and
health expenditures
Aggregate index on
Civil rights Botero et al. (2004) Robustness control
civil rights
38
Table 5. Summary statistics of models variables. Source: own elaboration.
39
Figure 7. Proportion of missing by variable. Source: own elaboration.
40
Table 6. GMM and ECM estimations. Only main regressors displayed.
Dependent variable:
(Log of) Gini on
(Log of) ∆Gini on Market Income
Market Income
difference GMM ECM
(1) (2)
(Log of) Gini on Market Income (L1) 0.298 -4.164**
(0.329) (1.390)
KIBS % (Persons) 0.026*
(0.013)
BERD on GDP -0.011
(0.047)
∆BERD on GDP 0.463
(0.290)
BERD on GDP (L1) 0.355*
(0.181)
∆ KIBS % (Persons) 0.088
(0.142)
KIBS % (Persons) (L1) -0.066
(0.078)
41