Professional Documents
Culture Documents
research-article2015
EDQXXX10.1177/0891242415617239Economic Development QuarterlyAntonelli and Fassio
Abstract
The study implements the Schumpeterian notion of creative reaction to articulate and test the hypothesis that the shift to
the knowledge economy in advanced economies is the result of the creative reaction of firms, caught in out-of-equilibrium
conditions by the fast globalization of product and factor markets since the last decade of the 20th century. Advanced
countries specialized in the generation and exploitation of knowledge because of its relative abundance stemming from their
sophisticated knowledge governance mechanisms and the large stock of knowledge. In turn, this had strong positive effects
on total factor productivity. The empirical analysis confirms that in advanced economies the specialization in knowledge-
based activities substituted for the previous specialization in mass manufacturing activities supporting the increase of total
factor productivity. The new specialization in knowledge-intensive activities has been stronger in cases where patent activities
have been more intense and the exposure to international trade and the revenue per capita have been larger.
Keywords
globalization, creative response, technological change, structural change, knowledge economy, productivity growth
Since the last decade of the twentieth century, advanced The Creative Response
economies have been experiencing radical structural change,
with a marked decline of mass production in manufacturing The 1947 contribution by Joseph Alois Schumpeter, “The
industries and an increasing specialization in knowledge- Creative Response in Economic History,” has received rela-
intensive business services (KIBSs) and knowledge- tively little attention in the economic literature and yet can
intensive manufacturing (KIM). This has been accompanied and should be considered the synthesis of a long research
by a rapid pace of introduction of new technologies, coupled process. In this groundbreaking article, Schumpeter provides
with total factor productivity (TFP) growth. A large literature an articulated and coherent framework of analysis of techno-
has analyzed the characteristics, the effects, and some of the logical change as an endogenous process that much deserves
causes of these twin processes of structural and technological a wider readership and systematic use. This study uses
change (Bonatti & Felice, 2008; Buera & Kaboski, 2012; Schumpeter’s framework to explore the intertwined dynam-
Nickell, Redding, & Swaffield, 2008; Schettkat & Yocarini, ics of technological and structural change that have been tak-
2006). ing place in advanced economies in the decades of the
This study applies the Schumpeterian notion of creative twentieth and twenty-first centuries. Let us recall briefly the
response—enriched by the new economics of knowledge founding arguments of the Schumpeterian notion of techno-
and Kuznets’s view that technological innovation and struc- logical change as a form of creative reaction.
tural change are endogenous and strictly intertwined—to Schumpeterian firms are exposed to continual mismatches
develop and empirically test an articulated explanatory between their own expectations and the actual developments
framework. According to this framework, the shift to the of product and factor markets. Unexpected changes in product
knowledge economy was driven by the search for a new and factor markets expose them to imbalances and to operat-
source of competitive advantage, stirred by globalization, ing in out-of-equilibrium conditions that cannot be fixed in the
and based on the strong and unique knowledge base of short term. In textbook accounts of the working of equilibrium
advanced economies with positive effects in terms of TFP
growth (Kuznets, 1965; Schumpeter, 1947). 1
Università degli Studi di Torino, Collegio Carlo Alberto, Moncalieri, Italy
The rest of the study is organized as follows: The next 2
CIRCLE, Lunds Universitet, Lund, Sweden
section articulates the Schumpeterian approach and extracts
Corresponding Author:
the main hypotheses. The section following that presents the Cristiano Antonelli, Collegio Carlo Alberto, Via Real Collegio 30, 10024
empirical evidence. The conclusion summarizes the main Moncalieri, Italy.
results and explores their policy implications. Email: cristiano.antonelli@unito.it
4 Economic Development Quarterly 30(1)
markets, firms, unable to elaborate on correct expectations, are Such a competitive pressure was enhanced by the liberal-
forced to exit by their competitors. In the Schumpeterian ization of international financial markets that followed the
framework, instead, firms are credited with the capability of globalization of international product markets: This deprived
trying to change their technologies, so as to adapt their produc- advanced economies of the traditional competitive advantage
tion conditions to the unexpected changes in the markets. of industrializing countries, based on the lower costs of capi-
Firms can react. Not all reactions, however, are deemed capa- tal and better access to financial markets. The liberalization of
ble of successfully changing their technologies. Here, the cru- international financial markets in fact provided industrializ-
cial distinction between adaptive and creative reactions is ing countries with the opportunity to access the sophisticated
introduced. Firms that do not have access to the necessary lev- financial markets of the advanced economies and to borrow
els of external knowledge will fail: Their reaction will be financial resources at costs that were close to those of the
adaptive. Firms that have access to high-quality knowledge original domestic competitors. Advanced countries had to
externalities can successfully change their technologies. Their search for other sources of competitive advantage.
reaction, supported by the availability of external competences Firms caught in out-of-equilibrium conditions could
and skills, will be “creative.” Technological change and hence implement a creative (as opposed to adaptive) response
TFP growth take place when firms, caught in out-of-equilib- because of the availability of knowledge as an input at costs
rium conditions because of the mismatch between their expec- far below equilibrium levels. The strong knowledge base
tations and the actual conditions of product and factor markets, (already in place because of the progressive increase of
react by taking advantage of an economic context that pro- knowledge intensity that occurred within manufacturing
vides access to external knowledge available at costs that are industries) was made possible through the Schumpeterian
below equilibrium levels. creative reaction and became the main source of a new spe-
This framework has an extraordinary analytical power cialization (Abramovitz & David, 1996).
that can be successfully applied to the analysis of the radical The large supply of knowledge externalities not only
structural and technological change toward the knowledge favored the creative reaction of firms, but also made possible
economy that has characterized the advanced economies in the transition away from mass production in manufacturing
the past decades. We consider such a phenomenon to be the industries toward the new specialization in KIBS and KIM.
result of the creative reaction of advanced economies to the Firms, and advanced countries at-large, learned how to take
threats of globalization, with the aim of making the most advantage of the relative abundance of knowledge as an
effective use of their strong knowledge base as a key input input, the advantages of increasing returns in the generation
and as a source of competitive advantage (Antonelli, 2008, of knowledge, and the sophisticated mechanism of knowl-
2011). edge governance already at work. This allowed them to
The starting point can be found in the opening of interna- direct technological change and to develop a new specializa-
tional product (and, eventually, factor) markets that took tion based on the generation and exploitation of technologi-
place in the second part of the twentieth century. The increas- cal knowledge, not only as an input but also as an output
ing levels of wages and the reduced flexibility of labor expe- (Antonelli & Fassio, 2011, 2014).
rienced in many advanced economies pushed firms to Advanced economies were discovered to have a strong and
globalize by delocalizing the most labor-intensive phases of distinctive competitive advantage in the generation of techno-
their production processes to reimport them later to serve logical knowledge, and they learned how to exploit it with the
their product markets in advanced economies (Caves, 2007; systematic introduction of KIBS. The new competitive pres-
Dunning & Lundan, 2008). sures forced them to introduce radical technological and struc-
The increasing weight of labor-intensive imports from tural changes directed at making the best and most effective
labor-abundant countries was the result of the merging of the use of their strong knowledge base as the main output and
interests of global corporations—as sellers in advanced input on which they could build a long lasting and exclusive
countries but producers in labor abundant ones—with those competitive advantage. They did so by exploiting its relative
interests in labor-abundant countries eager to take advantage abundance and low relative cost in their domestic economies,
of the large domestic product markets of advanced econo- as well as its low international mobility. The strength of their
mies. The progressive liberalization of the domestic markets local knowledge base has provided advanced economies with
of advanced economies and the strong reduction of all tariff the opportunity to gain a new competitive advantage by spe-
and nontariff obstacles to international trade undermined the cializing in the generation of KIBS (Boden & Miles, 1999;
competitive advantage of the manufacturing industries of Doloreux & Shearmur, 2012; Muller & Doloreux, 2009).
advanced economies. As a consequence, advanced econo- Knowledge-intensive activities are in fact strongly local-
mies were pushed to react to the increasing competitive pres- ized. The generation of knowledge is a highly localized pro-
sure of the exports of labor-intensive industrializing cess rooted in the web of interactions and complementarities
countries, both in their domestic markets and in the rest of of a variety of agents and institutions. These agents and insti-
the world economy (Kang & Lee, 2011; MacDonald, 1996). tutions have a strong impact on the actual levels of knowledge
Antonelli and Fassio 5
Figure 1. Flowchart of the system dynamics of technological and structural change stemming from the creative response to
globalization.
cumulability. For these reasons, technological knowledge is a The crucial role of user–producer interactions to generate
scarce and idiosyncratic resource that only a few countries new technological knowledge becomes a powerful mecha-
can command—those with high levels of stocks of techno- nism in retaining and implementing a small but highly KIM
logical knowledge embodied in skills, human capital, and a industry in advanced countries (Von Hippel, 1988). The mass
dedicated institutional setup that make possible the flow of offshoring by manufacturing would clearly weaken the quality
the necessary knowledge interactions (Antonelli, 2011; of knowledge user–producer interactions in advanced coun-
David, 1993). tries. The mass offshoring of manufacturing would quickly
Knowledge tacitness and its limited appropriability erode the knowledge base of advanced countries and under-
require the occurrence of sophisticated mechanisms of gov- mine the source of their own competitive advantage. The
ernance of knowledge transactions-cum-interactions, both advantages of proximity in knowledge user–producer interac-
among firms and between the business sector and the public tions would migrate to the countries of destination. The new
research system. This makes possible the generation and complementarity between a large array of KIBS industries and
exploitation of knowledge as an economic activity based on a smaller but highly KIM industry in advanced countries is a
the necessary levels of the division of labor, specialization, key component of the new knowledge economy.
and exchanges among research units with different incentive The mechanism at work was characterized by the typical
mechanisms. Such sophisticated systems of governance are self-reinforcing dynamics of feedback, for which the strong
the result of long-term historical processes that include pub- knowledge base of advanced economies is at the same time a
lic research activities along with business firms and foster cause and a consequence of the process. The specialization in
reciprocal relations between scientific and technological knowledge-based activities of advanced economies was the
knowledge that other countries cannot easily imitate or result of the relative and comparative abundance of knowledge-
reproduce (Chesbrough, 2003; Geuna, 1999). generating activities in their economies, yet it further increased
The intrinsic tacitness of technological knowledge has one the division of labor and enhanced the mechanisms of knowl-
additional very important consequence: the complementarity edge generation at work in advanced economies.
between KIBS and a new KIM industry. The foundations of Summing up the results of the analysis carried out so far,
this new complementarity are found in three factors: we can articulate the sequential system dynamics represented
in Figure 1, where we identify three nodes:
1. The transfer and repeated use of knowledge are pos-
sible only if and where dedicated interactions 1. The increased competitive pressure of the exports
between users and producers take place. from newly industrializing countries undermined the
2. Geographic proximity plays a crucial role in support- traditional specialization in mass manufacturing
ing dedicated interactions, which are characterized industries of advanced countries and stirred the reac-
by the advantages of proximity both in geographic tion of firms, caught in out-of-equilibrium conditions
and cultural terms. Agents operating at a distance by the rapid change of both product and factor mar-
across countries with different languages and cultural kets. Their creative reaction—as opposed to “adap-
traditions have much larger problems participating in tive” reaction—was made possible by the strong
the dedicated interactions that are necessary to access knowledge externalities stemming from the large
existing knowledge and to learn how to use it again stock of technological knowledge and the size of the
as an input. current research and development (R&D) activities,
3. Close personal relations are necessary to convert sci- and this led to a substantial increase of the pace of
entific knowledge generated by the public research technological change.
system and especially by universities into technologi- 2. The quality of the knowledge base itself became a
cal knowledge. supporting factor that helped direct the strong winds
6 Economic Development Quarterly 30(1)
of technological change toward the introduction of procedure should allow us to avoid the typical endogeneity
new innovative routines specializing in the genera- problems related to the estimation of causal effects.
tion of technological knowledge. The effect of
increasing returns at the system level in the genera- The Patent Equation. The first equation concerns the cre-
tion of technological knowledge, together with the ative reaction of firms to unexpected shocks in factor and
endowment of sophisticated knowledge governance product markets induced by globalization. Firms are
systems, helped identify knowledge as the strategic induced to react creatively to the mismatch between their
resource—both as an input and as an output—on expectations and the actual market conditions by means of
which to build a new competitive advantage and a technological change. The possibility of a creative reaction
new specialization. The rapid increase in the extent by firms will depend on the availability of external knowl-
of the market for knowledge aided the fast growth of edge in the system. We proxy this effort by means of the
KIBS industries. The creative reaction of firms led to number of patent applications.1 The main feature of a pat-
a new specialization, characterized by a much smaller ent is the novelty of its content with respect to the previous
KIM industry, complemented and supported by an technology. For this reason, we believe patents are a good
array of KIBS industries. This complementary com- proxy of the ability of firms caught in out-of-equilibrium
bination became the new core of the economic sys- conditions to react creatively to the mismatch between
tem and had the final effect of accelerating a major expectations and actual market conditions through the
structural change and the transition from the indus- introduction of actual innovations.
trial economy to the new knowledge economy. In our setting, three factors affect the ability of firms to
3. When and where the reaction to globalization was react creatively: (a) the knowledge externalities stemming
“creative,” it favored the faster introduction of from the existing stock of technological knowledge; (b) the
technological innovations and the creation of a new country-level endowment of knowledge embodied in the
specialization based on a small but highly produc- researchers employed in R&D activities; and (c) the level of
tive manufacturing industry coupled with a strong openness to trade of a country, which induces firms to react
array of KIBS. In such cases, countries could foster creatively. In the first equation, the pace of technological
technological change and hence obtain relevant change, proxied by the number of patent applications, is
increases of TFP. The creation of KIBS industries explained by the overall stock of patents, by the amount of
enabled the sustainable TFP growth of the entire R&D activities, and by the level of openness to trade:
economic system.
PAT int it = a0 + a1STOCKpatit −1 +
a2 RESit −1 + a3 ( EXPit −1 IMPit −1 ) + (1)
The Empirical Evidence
ηi + λt + ε it
The Model where i denotes country and t denotes the year. We choose to
To empirically test our hypotheses, we adopt a country-level adopt a specification in which variables are expressed in
perspective. We believe that this is a suitable way to investi- intensities over total employment, to avoid any possible
gate the dynamics of structural change within advanced country-size effect in our estimates. PATint is (the log of) the
economies. Indeed, most of the factors that induced the number of patent applications per person employed,
change in the productive structure of countries can only be STOCKpat is (the log of) the cumulated stock of patents per
analyzed at the country level. Tariff regimes, which influ- person employed, RES is the number of researchers per 1,000
ence the openness to trade of national economies, are stipu- of people employed, and EXP/IMP measures the level of
lated between countries. In turn, tariff regimes influence the openness to trade of a country as a ratio between exports and
dynamics of product markets and the degree of penetration imports: The higher the level of imports (and hence the lower
of foreign competition, which is the main inducement effect the ratio), the greater the mismatch between national firms’
for the creative response of firms. expectations and actual market conditions because of com-
As Figure 1 shows, our theoretical framework can be petition from foreign countries. All the independent variables
described as a sequence of causal relationships whereby the are 1-year lagged, as a first attempt to reduce the possible
increasing international integration of factor and product problems of endogeneity due to reverse causality. We control
markets because of globalization leads three things: (a) to for country effects (ηi) and common time trends (λt) through
creative reactions by firms, (b) to the increasing specializa- the use of country and time dummies. The idiosyncratic error
tion in KIBS, and ultimately, (c) to the growth of TFP. To test term in Equation (1) is denoted by εit.
this specific type of process, we have specified a model artic-
ulated in a sequential system of three equations: (a) a patent The KIBS Equation. Our hypothesis is that the technological
equation, (b) a KIBS equation, and (c) a TFP equation. This effort put forward by firms exposed to a mismatch between
Antonelli and Fassio 7
their expectations and the actual market conditions led to the R&D in purchasing power parity (PPP) dollars over total
increasing specialization of advanced economies in KIBS. employment, and PATINT is (the log of) the number of patent
This was again because of the increasing level of integration applications over total employment. Besides the usual coun-
of markets induced by the globalization. To measure this try and time effects, eit is the idiosyncratic error term.
effect, we introduce our second equation, in which the weight
of KIBS in each national economy is explained by two fac-
Data
tors: (a) the level of technological effort put forward by the
firms (as proxied by patents) and (b) the degree of openness We built a data set that includes almost all Organisation for
of a country to foreign trade. Moreover, our hypothesis is Economic Co-operation and Development (OECD)–member
that manufacturing activities and KIBS are only partial sub- countries; that is, the most advanced economies that experi-
stitutes and that for small shares of KIM activities, the two enced the shift to knowledge-intensive activities described
activities are instead complementary: Therefore, we intro- earlier. The time period chosen for the analysis goes from
duce the share of manufacturing activities and its squared 1990 to 2007, during the years in which the information and
term to control for this effect. To have a proxy for the level of communication technologies (ICT) revolution showed its
labor productivity, we also include the level of income per effect on the real economy and before the worldwide finan-
capita (INCOPC) to control for the proximity to the frontier cial crisis started in 2008. Even if the processes of outsourc-
of economic efficiency.2 ing of production toward emerging economics started long
before 1990, the specialization in KIBS activities was only
KIBSit = b0 + b1PAT int it −1 + possible after the ICT revolution of the early 1990s.
b2OPEN it −1 + b3 MANU it −1 + b4 MANUsqit −1 + (2) Therefore, we claim that focusing on this limited time period
b5 INCOPCit −1 + ηi + λt + uit allows us to identify a specific feature of the economic spe-
cialization of advanced capitalistic countries that was only
In Equation (2), KIBS denotes the share of employment in possible after the introduction of ICT and that has been radi-
KIBS, PATint is (the log of) the number of patent applica- cally different from the outsourcing processes that occurred
tions per person employed, OPEN indicates the share of in the previous decades.
trade (import and export) over total gross domestic product The number of patent applications per person employed at
(GDP), MANU denotes the share of employees in manufac- the country level (PATint) is obtained from the World
turing sectors over total employment, and MANUsq indicates Intellectual Property Organization (WIPO) database, which
its squared term. INCOPC indicates income per capita. provides the number of patent cooperation treaty (PCT)3 pat-
Again, ηi and λt are the usual country and time effects, while ent applications, attributing each patent to the country in
uit is the idiosyncratic error term. which the applicant is a resident. The WIPO database is very
well suited for international comparison because it provides
The TFP Equation. The creative reaction of firms faced with data that are not affected by the choice of the specific national
globalization, which led to the increasing weight of KIBS, or regional patent office (as in the case with U.S. Patent and
allowed for the actual increase of TFP in advanced econo- Trademark Office or European Patent Office). From this
mies. The technological change induced by KIBS had a viewpoint, the WIPO database seems less exposed to the bias
direct impact on the increase of output that could not be of the country of origin of the applicant. To check for the role
explained by labor and capital. Therefore, in our last equa- of previously existing knowledge, we also compute the
tion, we test the impact of KIBS on the level of TFP. More- cumulated stock of patents per person employed, following
over, we control for the other factors that are usually the procedure suggested by Hall, Jaffe, and Trajtenberg
associated with technological change—that is, the endow- (2005).4
ment of human capital, the level of expenditures in R&D, We mainly identify knowledge-based economic activities
and the technological effort of firms, as proxied by the num- with KIBS. An extensive literature has dealt with the problem
ber of patent applications. of the identification of KIBS among the usual sectoral classi-
fications (Boden & Miles, 1999; Di Maria, Bettiol, De Marchi,
TFPit = c0 + c1KIBSit −1 + & Grandinetti, 2012; Muller & Doloreux, 2009). According to
c2 HKit −1 + c3 R & D int it −1 + (3) this literature, the economic activities considered as KIBS rely
c4 PAT int it −1 + ηi + λt + eit heavily on professional knowledge, are themselves primary
sources of information and knowledge, and use knowledge to
In Equation (3), the TFP indicates the level of total factor produce intermediate services for their clients’ production pro-
productivity, KIBS denotes the share of employment in cesses. Following Freel (2006), who identifies KIBS sectors
knowledge-intensive business services, HK indicates the with the two-digit sectors 72, 73, and 74 of the International
share of enrolled students in tertiary education over the total Standard Industrial Classification of All Economic Activities
population, R&Dint is the log of the ratio of expenditures in (ISIC) Rev. 3 classification, we use the OECD STructural
8 Economic Development Quarterly 30(1)
ANalysis (STAN) database and classify as KIBS the following Table 1. Description of Variables.
two-digit ISIC Rev. 3 sectors: renting of machinery and equip-
PATint Log of the number of PCT patent applications
ment (71), computer and related activities (72), research and (WIPO) per employed person
development (73), and other business activities (74).5 We then STOCKpat Log of the cumulated stock of PCT patents
calculate the share of persons engaged in these sectors against (WIPO) per employed person
the total employment of each country, as provided by STAN, RES Number of researchers per 1,000 of employment
and compute the share of workers in KIBS sectors (KIBS). (OECD)
Also, the number of employees in the manufacturing sector EXP/IMP Ratio of exports over imports (OECD)
comes from the STAN database and includes all sectors classi- KIBS The share of employment in KIBS for each
fied as “manufacturing” in the ISIC Rev. 3. We then calculated country (OECD)
the share of persons engaged in manufacturing sectors over OPEN Share of trade on GDP (OECD)
total employment (MANU). INCOPC Level of income per capita (PENN World Tables)
Trade data proceed from the OECD database: Specifically, MANU Share of employment in the manufacturing sectors
for each country (OECD)
we collected data on the share of trade on GDP (OPEN) and
MANUsq Square of the share of employment in the
the ratio of exports to imports (EXP/IMP). These measures manufacturing sectors for each country (OECD)
indicate the degree of openness to trade of a country and its TFP Total factor productivity (OECD)
capability to be competitive in the international markets. The HK Share of enrolled students in tertiary education
number of researchers per 1,000 people employed (RES) was on the total population (UNESCO)
also taken from the OECD Research and Development R&Dint Log of the ratio of expenditures in R&D (in PPP
Statistics, together with the data on the expenditures in R&D dollars) over total employment (OECD)
(in PPP dollars) over total employment (R&Dint). The share
Note. KIBS = knowledge-intensive business service; R&D = research and
of enrolled students in tertiary education on the total popula- development; PPP = purchasing power parity; PCT = patent cooperation
tion, used as a proxy for human capital (HK), instead comes treaty; WIPO = World Intellectual Property Organization; OECD =
from United Nations Educational, Scientific and Cultural Organisation for Economic Co-operation and Development; UNESCO =
Organization data. The database also includes the level of United Nations Educational, Scientific and Cultural Organization.
income per capita of each country (INCOPC), as provided by
the Penn World Tables, using the ratio of PPP dollars (in
Table 2. Descriptive Statistics, 1990 to 2007.
2005 constant prices) per equivalent adult.
The data used to calculate the levels of TFP were taken Variable Observations M SD Minimum Maximum
from the OECD–STAN database for the whole economy.
PATINT 313 4.482 1.524 −2.950 6.721
The (log of) the level of TFP was calculated, as in Bernard STOCKpat 317 6.123 1.760 −2.111 8.819
and Jones (1996), in the following way: RES 303 6.967 2.719 2.340 17.773
EXP_IMP 311 1.036 0.227 0.522 1.902
Yit
ln(TFPit ) = βit ln KIBS 300 0.096 0.029 0.040 0.176
Lit OPEN 321 0.730 0.371 0.160 1.728
(4)
Yit INCOPC 324 0.032 0.008 0.008 0.056
+(1 − βit ) ln
Kit MANU 317 0.174 0.042 0.099 0.287
MANUsq 317 0.032 0.016 0.010 0.082
Following the Euler theorem and assuming constant TFP 301 11.627 0.172 11.190 12.013
returns to scale, we calculate β as the share of labor compen- HK 306 0.037 0.011 0.011 0.068
sation over total GDP. Y denotes real GDP, L is the level of R&Dint 241 6.365 0.700 3.133 7.426
total employment, and K is the net stock of capital: All mea-
sures are taken from the OECD–STAN database.6 Real GDP
and capital stock series have then been deflated by the PPPs in Equation (4). Indeed, we suspect that even if we adopted a
deflator to obtain comparable measures in PPP dollars for all 1-year lag to avoid reverse causality issues in the estimation,
the countries in the database. and even if we included country dummies to control for unob-
The final database is a fairly balanced panel of 18 OECD served heterogeneity, some of the independent variables of our
countries and 18 years (1990-2007).7 In Tables 1 and 2, the equations might still suffer from endogeneity problems. In
descriptive statistics of the variables are displayed. panel data, the main limitation of the usual ordinary least
squares (OLS) fixed effects estimators is the assumption of
strict exogeneity, according to which the level of an indepen-
The Methodology dent variable today must be uncorrelated with past, present,
Our estimation strategy aims to address the endogeneity prob- and future exogenous shocks (the idiosyncratic error term).
lems related to the estimation of our equations, as summarized Although we believe that present and future shocks will most
Antonelli and Fassio 9
likely be uncorrelated with the independent variables of our of the parameters of interest, it only has to verify that the
model, in some cases past shocks could indeed have an effect orthogonality conditions hold (Bogliacino & Pianta, 2013;
on some regressors. Wooldridge, 2002). This makes it the preferred estimator for
The final specification of the model is structured in three our system of equation, since we can be sure that if our
sequential equations where the two key independent vari- choice of instruments is legitimate, we will be able to consis-
ables (PATint) and (KIBS) are dependent variables of the pre- tently estimate the relevant coefficients.
vious equations:
OLS IV IV
PATint PATint
Note. OLS = ordinary least squares; IV = independent variable. All models include country and time dummies. In columns (2) and (3), excluded
instruments are the number of researchers per 1,000 of employment in time t − 2 and t − 3. Robust standard errors in parentheses.
*p < .1. **p < .05. ***p < .01.
respect to exports, the higher the level of exposure of a coun- MANU and MANUsq of 0.3 and −1.3, respectively, we can
try to foreign competition. The results in column (3), in calculate the level at which the effect of manufacturing share
which we still control for the endogeneity of researchers, are on KIBS starts to be negative; this amounts to approximately
in line with these expectations: The lower the ratio of exports 0.11 (11%). This result is important, as it confirms that KIBS
to imports (and hence the greater the imports), the higher the and KIM are complementary when their shares are close to
creative reaction of firms, as proxied by the number of pat- parity. The increase of KIBS substitutes the decline of manu-
ents per employed person. facturing only when the share of the former is too small and
Table 4 presents the results of the estimation of Equation the share of the latter exceeds approximately the 10% level.
(2)—the KIBS equation, which explains the level of KIBS in Finally, to control for the proximity to the frontier of economic
each OECD country. According to our theoretical frame- advances, we also control for the level of wages and labor pro-
work, we expect that both the openness to trade of a country ductivity, proxied by the income per capita (INCOPC); again,
(which we proxy with the share of trade over total GDP) and we find that the coefficient is positive and significant.
the creative reaction of firms proxied by the number of pat- In column (2), we control for the possible endogeneity of
ents should increase the specialization of countries in KIBS. PATint, and we exploit the sequential pattern of our system of
Moreover, we also expect manufacturing activities to exert a equations: We use as instruments for PATint the independent
substitution effect on KIBS only after a certain threshold; a variables used in Equation (1)—that is, the stock of patents
small KIM industry should instead be complementary with per employed person and the level of researchers per 1,000 of
the existence of KIBS. employment at time t − 2. The 2SLS estimates do not change
Column (1) presents the normal OLS estimation of the much with respect to the previous estimates: The significance
model, including country and time dummies. The results show of PATint is slightly reduced but remains significant at the 5%
that the ratio of trade over total GDP has a positive and signifi- level. Moreover, the Kleibergen–Paap F statistics show that
cant effect on the share of employment in KIBS. The same the instruments are not weak, and the Hansen test on overi-
positive and significant effect is found for the number of pat- dentification indicates that the instruments are truly exoge-
ents per employed person. The share of manufacturing over nous. In this case, however, we find that the endogeneity
total employment (MANU) is also found to be positive, problem is not relevant because the endogeneity test over the
although its squared term (MANUsq) is negative. This indi- PATint variable cannot reject the null hypothesis of exogene-
cates the existence of an inverted-U relationship between the ity of the variable. In column (3), instead, we instrument only
share of manufacturing and the share of KIBS in total employ- the variable that measures the openness to trade (OPEN), as
ment: After reaching a certain threshold, the effect of manu- we suspect that this variable might also suffer from endogene-
facturing becomes negative. Looking at the two coefficients of ity problems. As instruments, we use the lagged levels of
Antonelli and Fassio 11
OLS IV IV IV
Note. OLS = ordinary least squares; IV = independent variable; KIBS = knowledge-intensive business service. All models include country and time
dummies. In column (2), excluded instruments are the number of researchers per 1,000 people employed in time t − 2 and the stock of patents per
employed person in time t − 2. In column (3), excluded instruments are the openness to trade in time t − 2 and income per capita in time t − 2. In column
(4), excluded instruments are the number of researchers per 1,000 people employed in time t − 2, the stock of patents per employed person in time t −
2, the openness to trade in time t − 2, and income per capita in time t − 2. Robust standard errors in parentheses.
*p < .1. **p < .05. ***p < .01.
openness to trade and of income per capita in time t − 2. The results in column (2) show that the instruments are not weak
results show that the instruments are both relevant and exog- (the Kleibergen–Paap F statistics reject the null hypothesis of
enous, providing strong support for their inclusion. The coef- weak identification) and that they are not correlated with the
ficient of OPEN is still positive and significant. The equation error term, since the Hansen test of overidentifica-
endogeneity test cannot reject the null hypothesis of exogene- tion accepts the null hypothesis of exogeneity of the instru-
ity of the variable, but the p value is much lower than in the ments. The size and significance of the coefficient of the
case of PATint, showing that the openness to trade is probably KIBS variable, instead, is still large and strongly significant.
more likely to be affected by endogeneity. Finally, in column The endogeneity test on the KIBS variable, however, suggests
(4), we instrument both PATint and OPEN with the instru- that the variable is truly exogenous to the error term. Even if
ments used in the previous two specifications: In this case, our model includes country and time dummies, that should
too, we find that instruments are valid and that the coeffi- capture most of the unobserved heterogeneity present in the
cients of our variable of interest remain positive and strongly error term. In column (3), we also check whether including
significant. the R&D expenditures and the number of patent applications
In Table 5, we estimate Equation (3), the TFP equation. In per employed person (respectively, R&Dint and PATint)
column (1), we present OLS estimates, in which we include somehow affects the KIBS coefficient, since these are factors
both the share of KIBS and the level of human capital of a that are also likely to affect the levels of TFP. The results of
country (HK), as proxied by the number of enrolled students column (3) show that both PATint and R&Dint play an impor-
over the total population. As expected, both variables display tant role in explaining the level of TFP, and that the coeffi-
positive and significant coefficients. The positive coefficient cients of KIBS and HK (human capital) are not affected by the
of KIBS in particular confirms the hypothesis that knowledge- inclusion of these additional variables (also see Table 6).
intensive business activities have a fundamental role in
explaining the overall level of TFP of advanced economies.
Conclusion
To check whether the positive coefficient of KIBS represents
a true causal relation, we instrument it with its own past levels This study has analyzed the process of structural and techno-
to check whether an endogeneity problem is present. The logical change that led to the creation of a knowledge
12 Economic Development Quarterly 30(1)
OLS IV IV
TFP TFP
Note. OLS = ordinary least squares; IV = independent variable; KIBS = knowledge-intensive business service; TFP = total factor productivity. All models
include country and time dummies. In column (2), excluded instruments are the share of employment in KIBS in time t − 2 and t − 3. Robust standard
errors in parentheses.
*p < .1. **p < .05. ***p < .01.
Table 6. Countries Included in the Regressions. turn, became a powerful attractor, directing the introduction
Australia Germany of technological changes with a strong bias in favor of
Austria Italy knowledge-intensive activities. Increasing returns in the gen-
Belgium Japan eration of knowledge at the system level and sophisticated
Canada The Netherlands knowledge governance procedures favored the identification
The Czech Republic Norway of knowledge as the strategic input toward which it was more
Denmark Spain profitable to direct the introduction of new technologies.
Estonia Sweden The empirical evidence of the OECD countries from the
Finland The United Kingdom years 1990 to 2007 confirms that the set of variables so far
France The United States identified plays a strong and significant role in explaining the
growth of KIBS industries and their progressive substitution
for the manufacturing industry at the heart of the economic
economy based on the mix of a strong KIBS and a thin but system of advanced economies.
highly specialized KIM industry. In the period covering the The implications of the interpretative framework, elabo-
end of the twentieth and the early twenty-first centuries, this rated on in this research and supported by strong and reliable
mixture took the place of the mass manufacturing industry empirical evidence, are important both for economics and
that had been at the heart of the economic system of advanced economic policy. With respect to economics, we claim that
economies. This study has elaborated on an interpretative the interpretative framework (based on the Schumpeterian
framework based on the Schumpeterian notion of innovation notion of innovation as a form of creative reaction that takes
as a creative response that can take place only when firms place when favorable external conditions are available)
caught in out-of-equilibrium conditions can access relevant deserves a much broader use, as it provides important tools
knowledge externalities. They then implement these exter- for understanding how structural and technological change
nalities with the analysis of technological congruence; this are endogenous and strictly intertwined.
enables them to grasp the advantages of the direction of tech- Our analysis has two important policy implications: (a)
nological change toward the most intensive use of the inputs the central role of support to the generation and use of tech-
that are locally and comparatively most abundant. nological knowledge and (b) the new complementarity
The strong knowledge base that advanced economies between KIBS and KIM. Let us analyze them in detail.
have in place could support the reaction of firms, making that Economic policy should support as much as possible the
reaction a creative one. The increasing role of knowledge, in generation of knowledge, along three main axes: (a) by
Antonelli and Fassio 13
9. Because we are using the number of patent applications comparison. Journal of Analytical and Institutional Economics,
divided by the level of employment, our dependent variable is 39, 319-342.
not a positive integer, and therefore we cannot apply the usual Dunning, J. H., & Lundan, S. M. (2008). Multinational enterprise
count data models. and the global economy (2nd ed.). Cheltenham, England:
10. The joint null hypothesis of the Hansen J statistics is that the Edward Elgar.
instruments are valid—that is, uncorrelated—with the error Freel, M. (2006). Patterns of technological innovation in knowl-
term and that the excluded instruments are correctly excluded edge-intensive business services. Industry and Innovation, 13,
from the second stage. 335-358.
Furman, J., Porter, M. E., & Stern, S. (2002). The determinants of
national innovative capacity. Research Policy, 31, 899-933.
References Geuna, A. (1999). The economics of knowledge production.
Abramovitz, M., & David, P. A. (1996). Technological change and Cheltenham, England: Edward Elgar.
the rise of intangible investments: The U.S. economy growth Hall, B. H., Jaffe, A., & Trajtenberg, M. (2005). Market value and
path in the twentieth century. In Employment and growth in the patent citations. RAND Journal of Economics, 36, 16-38.
knowledge-based economy (pp. 35-60). Paris, France: OECD. Kang, S. J., & Lee, H. (2011). Foreign direct investments and de-
Antonelli, C. (2008). Localized technological change: Towards the industrialization. World Economy, 34, 313-329.
economics of complexity. London, England: Routledge. Kuznets, S. (1965). Economic growth and structure. London,
Antonelli, C. (Ed.). (2011). Handbook on the economic complexity England: Heineman.
of technological change. Cheltenham, England: Edward Elgar. MacDonald, J. M. (1996). Does import competition force effi-
Antonelli, C., & Fassio, C. (2011). Globalization and innovation cient production? Review of Economics and Statistics, 76,
in advanced economies. In G. Libecap (Ed.), Advances in the 721-727.
study of entrepreneurship, innovation and economic growth Muller, E., & Doloreux, D. (2009). What we should know about
(Vol. 22, pp. 21-46). Cambridge, England: Emerald. knowledge-intensive business services. Technology in Society,
Antonelli, C., & Fassio, C. (2014). The economics of the light 31, 64-72.
economy: Globalization, skill-biased technological change, Nickell, S., Redding, S., & Swaffield, J. (2008). The uneven pace
and slow growth. Technological Forecasting & Social Change, of deindustrialization in the OECD. World Economy, 31, 1154-
87, 89-107. 1184.
Bernard, A. B., & Jones, C. I. (1996). Comparing apples to oranges: O’Mahony, M., & Timmer, M. P. (2009). Output, input, and pro-
Productivity convergence and measurement across industries. ductivity measures at the industry level: The EU KLEMS data-
American Economic Review, 86, 1216-1238. base. Economic Journal, 119, 374-403.
Boden, M., & Miles, I. (Eds.). (1999). Innovation and the service- Ostrom, E., & Hess, C. (Eds.). (2006). Understanding knowledge
based economy. London, England: Cassell Academic. as a commons: From theory to practice. Cambridge: MIT
Bogliacino, F., & Pianta, M. (2013). Profits, R&D, and innova- Press.
tion—A model and a test. Industrial and Corporate Change, Patel, P., & Pavitt, K. (1994). The nature and economic importance
22, 649-678. of national innovation systems. STI Review, 14, 9-32.
Bonatti, L., & Felice, G. (2008). Endogenous growth and chang- Schettkat, R., & Yocarini, L. (2006). The shift to service employ-
ing sectoral composition in advanced economies. Structural ment: A review of the literature. Structural Change and
Change and Economic Dynamics, 19, 109-131. Economic Dynamics, 17, 127-147.
Buera, F. J., & Kaboski, J. P. (2012). The rise of the service econ- Schumpeter, J. A. (1947). The creative response in economic his-
omy. American Economic Review, 102, 2540-2569. tory. Journal of Economic History, 7, 149-159.
Caves, R. E. (2007). Multinational enterprise and economic analysis Von Hippel, E. (1988). The sources of innovation. Oxford, England:
(3rd ed.). Cambridge, England: Cambridge University Press. Oxford University Press.
Chesbrough, H. (2003). Open innovation: The new imperative for White, H. (1980). A heteroskedasticity-consistent covariance
creating and profiting from technology. Boston, MA: Harvard matrix estimator and a direct test for heteroskedasticity.
Business School Press. Econometrica, 48, 817-838.
Cohen, W. M., Nelson, R. R., & Walsh, J. P. (2000). Protecting Wooldridge, J. M. (2002). Econometric analysis of cross section
their intellectual assets: Appropriability conditions and why and panel data. Cambridge: MIT Press.
U.S. manufacturing firms patent (or not) (NBER Working
Paper No. 7552). Cambridge, MA: National Bureau of Author Biographies
Economic Research.
Cristiano Antonelli (1951) holds the chair of political economy of
David, P. A. (1993). Knowledge property and the system dynam-
the University of Torino. He is fellow of the Collegio Carlo Alberto,
ics of technological change. In L. Summers & S. Shah
where he guides the Bureau of Research on Innovation, Complexity,
(Eds.), Proceedings of the World Bank Annual Conference
and Knowledge (BRICK). He is the managing editor of Economics
on Development Economics (pp. 215-248). Washington, DC:
of Innovation and New Technology.
World Bank.
Di Maria, E., Bettiol, M., De Marchi, V., Grandinetti, R. (2012). Claudio Fassio is a post-doc fellow at the Centre for Innovation,
Developing and managing distant markets: The case of KIBS. Research, and Competence in the Learning Economy (CIRCLE) at
Economia Politica, 29, 361-379. the University of Lund. His research interests focus on the economics
Doloreux, D., & Shearmur, R. (2012). How much does KIBS con- of innovation, science and technology; on these topics he has pub-
tribute to R&D activities of manufacturing firms? A sectoral lished several articles in international peer-reviewed journals.