GLOBAL VALUE CHAINS: The productivity contribution of ......(WIOD), while intangible valuation is obtained from the Intan-Invest database. Our unbalanced panel consists of 17 countries

GLOBALINTO Capturing the value of intangible assets in micro data to promote the EU’s Growth and Competitiveness

1

Deliverable 2.2

GLOBAL VALUE CHAINS:

The productivity contribution of intangible assets and participation in global value chains

Due date of deliverable: 2019/10/31

Actual submission date: 2019/10/31

Globalinto Identifier: WP2.

Author(s) and company: Ahmed Bounfour, Alberto Nonnis, Altay Ozaygen, Keoungoui Kim, Tatiana Bellieva, Université Paris-Sud, Université Paris-Saclay

Work package/task: WP2 / Task 2.2

Document status: final

Confidentiality: confidential / restricted / public

Document history

Version Date Reason for change

1

2

3

4

This project has received funding from the European Union’s Horizon 2020 The

mechanisms to promote smart, sustainable and inclusive growth under grant

agreement No 822259.


2

TABLE OF CONTENTS

1 Introduction 4

2 Global Value Chains and intangible capital: a literature review 5

2.1 Intangible capital as a driver of productivity 5

2.2 Measuring Global Value Chain participation 7

3 Data 12

4 Methodology 13

5 Results 17

5.1 The production function 17

5.2 Intangibles and GVCs 18

5.3 The mediating effect of GVCs 23

6 Conclusions 25

7 References 27


3

Summary

This deliverable studies the contribution to productivity of five types of intangibles

(R&D, computer software, branding, design and organizational capital) and global

value chain (GVC) participation.

The analysis is conducted at industry level. GVC participation is analysed using

network centrality measures calculated from the World Input-Output Database

(WIOD), while intangible valuation is obtained from the Intan-Invest database. Our

unbalanced panel consists of 17 countries and 16 sectors observed for the period

2000–2014. The analysis is based on a two-stage procedure. First, a proxy for total

factor productivity (TFP) is obtained by estimating a production function with

labour and capital as inputs. Endogeneity is accounted for by specific production

function estimation methods. Secondly, estimated TFP is used to evaluate how

several types of intangibles and GVC measures affect productivity. A mediation

analysis is performed to estimate jointly the impact of the two factors and evaluate

the role of GVCs in the transmission mechanism between intangibles and

productivity.

Both GVCs and intangibles were found to be significant drivers of productivity.

Moreover, our results revealed that intangibles have an impact via two channels: one

direct and other indirect (via GVC mediation). The main contribution of this study

is that our results show not only the contribution of intangibles and GVC integration

to productivity, but also the combined effect of the two. Intangibles are found to have

a secondary effect, mediated by GVC, which can be as large as the direct effect in

terms of magnitude.


4

GLOBAL VALUE CHAINS:

The productivity contribution of intangible assets and participation in global value chains

1 Introduction

Recent decades have seen drastic change in the organization of production chains; in

particular, production is becoming more fragmented across countries. The term, global

value chains (GVCs) refers to all of the activities that contribute to the end product – from

its initial conception to final distribution. These huge networks enable the exchange of

materials, intermediate inputs and services, and connect industries and firms located in

different countries. Porter (1990) notes that countries tend to specialize in the production

of specific products. This phenomenon is also seen in GVCs – countries that specialize in

particular sections of the chain import less specialized intermediate inputs from other

countries.

In this context, intangible capital is playing an increasing role, and it has been shown

to affect both participation in GVCs and productivity. Recent studies of intangibles have

pointed out that significant added value is created in the first and last stages of the

production chain, before and after manufacturing. The early stages concern product

design (R&D and design) and the later stages the sale of the product (branding, retail and

other post-sale services). In recent years, the added value of these stages has increased

considerably, accentuating the shape of the so-called smiling curve that depicts

contributions to added value in the various stages of the production chain.

Although few studies have examined intangible capital in the context of GVCs, recent

work has shown that participation in GVCs improves productivity (Baldwin et al., 2014)

and that investing in intangibles favours participation in GVCs (Jona-Lasinio et al., 2016).

What is missing is an evaluation of the combined effects of the two factors. This paper

aims to fill in this gap, by analysing the effect of investment in intangibles and GVC


5

participation on productivity, and the linkage between these two factors. In other words,

it explores two different channels through which intangibles affect productivity: one

direct and one indirect (via GVCs).

A key topic in GVC analyses concerns the way participation is defined. In this paper,

GVCs are seen as huge networks in which firms (or industries) are nodes that exchange

goods, materials and information with each other. Sectoral data on the import and export

of intermediate goods, and the level of participation of each unit of analysis is proxied

using measures of network centrality developed in the literature.

The analysis is divided into two steps. In the first, a Cobb-Douglas production function

is estimated with tangible capital and labour as inputs, while residuals are computed as

the measure of total factor productivity (TFP). This approach addresses a number of

problems associated with productivity analysis, the most important of which concerns the

endogeneity of capital in the estimation of production functions. We therefore estimate a

production function with the Olley and Pakes (1992) method, using capital investment as

a proxy to account for productivity shocks.

In the second step, the contribution of GVCs and intangibles, and their relationship to

TFP are analysed using the estimated TFP proxy. GVC participation is evaluated by

various network centrality measures. The hypothesis that GVCs mediate the effects of

intangibles is tested, consistent with the literature that argues that intangibles support

both productivity and GVC integration. In sum, this paper tests whether intangibles affect

productivity both directly and indirectly (via GVCs). The analysis is conducted on annual,

industry-level data. The panel consists of 17 countries and 16 sectors observed for 15 years,

from 2000 to 2014.

2 Global Value Chains and intangible capital: a literature

review

2.1 Intangible capital as a driver of productivity

This paper relates to two strands of literature, one on intangible capital and the other on

GVCs. While there is growing interest in the investigation of each of the topics, there are

few works that cover both subjects.


6

With respect to intangibles, the literature recognises that they play an important role

in growth (Corrado et al., 2005; Gu and Lev, 2011; Ståhle et al., 2015). From a company

perspective, intangibles are considered to be the most valuable asset that companies have

(Bounfour, 2003). The growth in TFP observed in recent decades has not been explained

by growth in the use of traditional inputs – a puzzle that intangibles help to solve.

However, their classification and measurement remains challenging, given their

ambiguous nature. At firm level, there is no consensus regarding whether they should be

included in balance sheets as expenditure or capital investment. Historically, the first

approach has been followed. Intangibles were considered as expenditure related to

production inputs; this meant that they were not included in national accounts as capital

formation, potentially leading to a mismeasurement of output growth. New approaches

that treat intangibles as investments have recently been introduced (e.g. Corrado et al.,

2005) and some types (such as R&D expenditure), have recently started to be included in

national accounts.

Notwithstanding these limitations and difficulties in their identification and

measurement, their role as a driver of productivity has been studied extensively. For

example, Corrado et al. (2005) find that intangibles made a major contribution to TFP in

the 1990s. Marrocu et al. (2012) studied a panel of European firms for the period 2002–

2006, and observed a considerable increase in investment in intangible compared to

tangible capital. At the same time, they found that the three categories of intangibles they

defined (human, technological and public capital) made a significant contribution to

productivity. Similar results at firm level are found by Oliner et al. (2007) and O’Mahony

and Vecchi (2009), who also find a spillover effect at industry level for R&D-intensive

sectors. Arrighetti et al. (2014) investigated the factors that lead Italian firms to invest in

intangibles, and their results show that specific characteristics, such as firm size, human

capital and historical intangible asset base are most important. Hall et al. (2005) focused

on patents, and found a positive correlation between patent citations and market value.

Sandner and Block (2011) examined trademarks and found a positive impact of different

trademark indicators on firm value. Sectorial level analyses are provided by Corrado et al.

(2016). The study found, for a panel of 10 European countries, a positive contribution of

five types of intangibles (R&D, branding, design, training and organizational capital) to


7

productivity growth. Similarly, Greenhalgh (2006) found that R&D and European

patenting (but not UK patenting) tended to increase market value in the United Kingdom.

The literature that combines intangible capital with GVCs is less extensive. A GVC

approach to measuring intangibles is provided by Chen et al. (2017) – the study considers

the difference between final product value and the cost of tangible factors at each stage of

production. Dedrick (2017) provides an analysis of the smartphone sector, and compares

four major brands to evaluate which captures the highest share of profits from intangibles

in GVCs. Jona-Lasinio et al. (2016) uses country-level data, and found that intangibles

were among the factors that led to participation in GVCs. They also distinguished between

intangibles that favour forward participation (in which a country provides inputs to other

countries further along the chain) and backward participation (where a country receives

inputs from other countries in the chain).

2.2 Measuring Global Value Chain participation

Some of the GVC literature has focused on empirical analyses and case studies (Dedrick,

2010; Sturgeon and Kawamaki, 2010; OECD, 2013; WIPO, 2017; UNCTAD, 2013; World

Bank Group et al., 2017). In particular, Roos (2017) emphasises the role of GVCs as a

driver for economic activity.

More studies have attempted to evaluate GVC participation through social network

analyses. As Jackson (2014) points out, the network perspective gives an insight into

economic behaviours that are difficult to catch through other means, and makes it

possible to explain the links created by the interaction of people or institutions. GVC flows

are often presented in the form of Input-Output (I-O) tables. These tables make it possible

to analyse macro-level interactions, such as international trade between countries, each

of which is represented as a node in the network. At a more granular level, each node

represents a national economic sector, which helps to distinguish between flows of

services and products. Network scores are used to evaluate the position of countries

and/or industries within these flows. Moreover, these analyses make it possible to

investigate the input-output relationship between countries and shed light on

interdependencies among them. In this context, GVCs are weighted networks, where the

weight of the link is measured by monetary transactions between national industries.


8

Different research teams have provided various ways to process I-O table data. Most

are free and can be downloaded. For example, Koopman et al. (2014) proposed an

accounting framework that integrates official trade statistics and national accounts 1 .

Yamano (2016) described the development of OECD inter-country I-O tables 2

highlighting that new indicators are needed for evidence-based policymaking. Lenzen et

al. (2012, 2013) developed an environmentally-extended, multi-region input-output table

(MRIO) model that take into account carbon, water, ecological footprinting, and a life-

cycle assessment3. Their tables include (among other data) labour inputs, air pollution,

energy use, water requirements, land occupation, various emissions and primary inputs

to agriculture. There are two versions of the database; the open version consists of 26

harmonized sectors, while the closed version is more granular. Another I-O table that

takes into account environmental indicators is EXIOPOL 4 , also known as the Multi-

Regional Environmentally Extended Supply-Use Table (Tukker et al. 2013). Finally,

Timmer et al. (2015) developed the World Input-Output Database5 (WIOD) within the

framework of the International System of National Accounts. The 2016 edition of the

WIOD comprises 43 countries and 56 sectors, and a model for the rest of the world – the

latter includes all other trades outside these countries and sectors. The present study

draws upon WIOD data (see section 3). Table 1 provides an overview of the various I-O

tables developed by different research projects.

1 https://www.aeaweb.org/articles?id=10.1257/aer.104.2.459 2 https://www.oecd.org/sti/ind/inter-country-input-output-tables.htm 3 https://worldmrio.com/ 4 https://exiobase.eu/ 5 http://www.wiod.org/home

https://www.aeaweb.org/articles?id=10.1257/aer.104.2.459https://www.oecd.org/sti/ind/inter-country-input-output-tables.htmhttps://worldmrio.com/https://exiobase.eu/http://www.wiod.org/home


9

Table 1: I-O databases developed by different research teams.

Author Database Name

Geographical coverage (country)

Sectoral coverage

Period Specificity

Yamano (2016)

OECD ICIO 67 36 2005–2013

Data from OECD member and non-member states

Timmer et al. (2015)

WIOD 43 and 1 RoW 56 2000–2014

Data from national accounts

Koopman et al. (2014)

Based on GTAP and UN COMTRADE

26 41 2004 A framework that aims to breaks up a country’s gross exports

Lenzen et al. (2012, 2013)

Eora MRIO 187 15909 (products and sectors)

1990–2009

Environmental indicators

Tukker et al. (2013)

EXIOPOL v.3 43 and 5 RoW 200 products and 164 industries

1995–2011

Environmental indicators, 40 emissions

The network analysis of I-O tables provides new insights into exchanges between

industries and countries (Amador et al., 2018; Korniyenko et al., 2017; Criscuolo and

Timmis, 2017; Carvalho, 2014; Kagawa et al., 2013), the structure of these transactions

(McNerney et al., 2013) and their evolution (Suh et al., 2013; Xing et al., 2016; Cerina et

al., 2015; Grazzini and Alessandro, 2015). Measurement is another topic that has been

studied (Blochl et al., 2011; Cerina et al., 2015)

Carvalho (2014) argues that the network perspective supports the development of

testable hypotheses, and that the use of I-O data provides an environment for empirical

explorations. For example, Criscuolo and Timmis (2017) examine the world economy and

identify central hubs, and peripheral countries and sectors using network centrality

metrics. The latter authors find that there is a shift in manufacturing value chains from

traditional production centres, and that different service sectors are gaining importance

in various countries. Furthermore, the change in network structure is helping some firms

to catch up. McNerney et al. (2013) show that, globally, manufacturing, food, chemicals,

services, and extraction industries are the most important. Kagawa et al. (2013) found


10

clusters of environmentally-important industries in the Japanese automobile supply

chain and, in particular, four industries that are most important for CO2 emissions’

reductions. Suh et al. (2013) observe that service and manufacturing industries are

separate in Korea, and explore the changing structure of the Korean mobile phone

industry. Xing et al. (2016) study economic development through the interaction of

different industries. The authors show that industries with higher flow betweenness or

random walk centrality foster intensive spreading effects in their industrial chain.

The interconnectedness and dependency of industries and countries brings several

risks that are analysed within the network resilience literature (Contreras and Fagiolo,

2014; Ando, 2014; Acemoglu et al., 2015; Korniyenko et al., 2017). Korniyenko et al.

(2017) use GVCs to evaluate the supply fragility of traded goods globally. The authors

assess the potential consequences of a supply shock resulting from importing goods from

specific countries. This assessment can be used to forecast the outcome of conflicts,

natural disasters and political instability. Various parameters, such as the size of the

country and the type of the shock play an important role in the economic outcome of a

supply outage (Contreras and Fagiolo, 2014). The latter authors show that shocks

affecting final demand have an impact on both the economy and production.

However, the various models that have been developed to measure diffusion in the

network show that these economic shocks do not change input and output relations

between different economic sectors and technology. Ando (2014) focuses on the United

States, and studies how a shock in the country could affect aggregated output. At the same

time, it has been shown quantitatively that the direct effects of shocks are limited,

compared to network-based propagation and that the geographic network has an

important effect on the propagation of a shock to other industries, due to collocation

(Acemoglu et al., 2015).

Various network scores are used to evaluate the position of an economic sector within

GVCs. Table 2 provides a summary of the literature. Eigenvector centrality indicates the

importance of a node by measuring the importance of neighboring nodes (Bonacich, 1972;

Katz, 1953). The distinctive aspect of eigenvector centrality is that it is measured with

relation to the centrality value of other nodes, while most other centrality measures focus

on the node itself.


11

Table 2: Network analyses based on I-O tables found in the literature.

Network parameters

Countries analysed Database

Amador et al. (2018) Strength, hub and authority

Germany, US, China, Japan

WIOD

Criscuolo and Timmis (2017)

Eigenvector Japan OECD ICIO

Carvallo (2014) Eigenvector US US BEA

Xing et al. (2016) Flow betweenness China WIOD and China I-O table

Cerina et al. (2015) Strength degree, PageRank, community coreness

-- WIOD

McNerney et al. (2013)

Weibull distribution US OECD I-O table

Suh et al. (2013) Degree, closeness, betweenness

Korea Korean I-O table

Contreras and Fagiolo (2014)

Random walk betweenness

European countries Eurostat I-O tables

Blochl et al. (2011) Random walk, modified betweenness

-- OECD STAN

For instance, a manufacturing industry connected to other important industries is more

likely to have a bigger influence or play a more important role in the network, thus its

eigenvector score is high. Heo and Lee (2019) argue that, compared to 14 other centrality

indices, eigenvector centrality can clearly highlight inter-industry convergence. This

observation is underlined by its use as an indicator of the importance or influence of an

industry in earlier studies (Rapach et al., 2015; Heo and Lee, 2019; Qian et al., 2017).

A more intuitive indication of importance is production received from, and given to, other

industries. Strength is a network index that is designed to capture this simple property.

It is divided into three indices depending on the target: strength-all, strength-in and

strength-out. Strength-all is the summation of what an industry receives and provides,

while strength-out and strength-in are the summation of outputs and inputs,

respectively. However, in industry networks these measures are often biased due to loops

formed when an industry consumes its own production. While eigenvector centrality and


12

strength focus on the weight of adjacent edges, betweenness centrality highlights the role

of the node as a bridge that links different clusters of nodes. Nodes that have high

betweenness scores are also called brokers (Brandes, 2001; Freeman, 1978). Blochl et al.

(2011) used betweenness centrality to show the key sectors of national industries using

global I-O tables.

3 Data

Our analysis draws upon a panel covering the period 2000–2014 consisting of annual

industry-level data for 17 countries6 and 16 sectors, following the Industrial Classification

of All Economic Activities (ISIC Rev. 4) classification of industries. 7 Data on GVCs,

intangibles, value added and production factors were also included. Data sources included

the World Input-Output Database (WIOD) for GVC data, Intan-Invest for data on

intangibles and value added, and EU Klems3 for production factors (Table 3).

In particular, the WIOD dataset developed by Timmer et al. (2015) includes bilateral trade

flows between sectors in different countries. Each sector produces outputs that are either

used as inputs by other sectors or are end products. At the same time, each sector receives

outputs from other sectors and uses them as inputs.

Table 3: Data sources

Database Variables

Global Value Chain World Input-Output

Database (WIOD) Intermediate inputs, trade flows

Intangibles Intan-Invest Investment (at 2010 prices) in R&D, computers

and software, branding, design and organizational capital

Production factors

EU Klems Labour (hours worked by sector) and capital stock

(real fixed capital stock at 2010 prices)

All of these exchanges can be seen as part of a huge network in which industries are

nodes and flows of goods or services (measured in monetary terms) are edges. The

6 Austria, the Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Italy, Luxembourg, Portugal,

Slovakia, Slovenia, Spain, Sweden, the United Kingdom and the United States. 7 In order to make the WIOD database compatible with the other databases employed in the study, several industry

subcategories were aggregated (R and S into R-S, D and E as D-E and M and N as M-N). 3 The 2017 EU Klems release is used, see Jäger (2016) for details.


13

centrality of each sector indicates its participation in GVCs.8 Investment data is used to

proxy intangibles, following the classification presented in the Intan-Invest database

(Corrado et al., 2016). Specifically, the focus is on R&D, computer software, branding,

design, and organizational capital.

GVC participation is measured by network centrality measures. These measures

indicate the different roles of nodes within network and help to identify influential nodes

(Bloch et al., 2016). These important nodes are, in this study, industrial sectors within

countries, and can be found and measured by widely-accepted social network algorithms

(Bloch et al., 2016). Different centrality measures reflect the different roles that nodes can

have in a network.

This paper focuses on three measures. The first, and most intuitive, is the centrality

score, which is calculated as the sum of all transactions involving a node. It is also possible

to distinguish between transactions flowing in and flowing out of a sector. In other words,

it distinguishes between the amount of inputs a sector receives from others (backward

GVC participation) and the amount of inputs it supplies to others (forward GVC

participation). The second measure is the eigenvalue centrality score, which is calculated

by counting the number of ties that a node has with others. This measure is used for

robustness, as its interpretation is similar to node strength centrality. Finally,

betweenness centrality is calculated. This measures each node’s importance in linking

other nodes (Brandes, 2001; Freeman, 1978).

4 Methodology

This study investigates the relationship between productivity, GVCs and intangible

capital. The analysis is performed in two stages. In the first stage, a TFP ex-intangible

measure is computed by estimating a Cobb-Douglas production function with only labour

and tangible capital as inputs. The formula is as follows:

8 In order to simplify the network for analysis, all transactions between industries below $10 million USD were omitted.


14

𝑌𝑐,𝑖,𝑡 = 𝐴𝑐,𝑖,𝑡𝐾𝑐,𝑖,𝑡𝛽1 𝐿𝑐,𝑖,𝑡

𝛽2 (1)

where Y is value added, A is TFP, K is capital stock and L is labour measured as total hours

worked by employees, while the subscripts c, i and t denote country, industrial sector and

time.4 To reflect endogeneity in inputs, the above production function is estimated with

the methodology proposed by Olley and Pakes (1992), which uses capital investment to

control for productivity shocks.9 Equation (1) can be written as the logarithm:

(2)

In this specification, inputs are known to be correlated with the error term. Estimates

with standard methods yield biased and inconsistent coefficients.

Following Olley and Pakes (1992), the error term in equation (2) can be divided into two

components, ω and η:

(3)

of which only ω is correlated with inputs and therefore causes simultaneity. When a

shock ω occurs, it affects firms’ decision rules. In other words, even though a shock to ω

is not observed by the econometrician, it is observed by the firm, affecting their

investment choices. Therefore, firms’ investment decisions are a function of both capital

and ω.

Ic,i,t = ft(ωc,i,t,Kc,i,t) (4)

Under the assumption that ω follows a first order Markov process, and inverting (4),

it is possible to develop the following expression for productivity shock:

ωc,i,t = ft−1(Ic,i,t,Kc,i,t) = θt(Ic,i,t,Kc,i,t) (5)

9 Levinsohn and Petrin (2003) argue that, in some cases, intermediate inputs are a better proxy than investment for

productivity shocks. This paper uses investment, for two reasons. First, intermediate inputs are highly correlated with the intangible measures used in the second step; and second, using sectoral data rules out the possibility of zero investment values that are often present in firm-level data, and therefore the possibility of violating the hypothesis of strict monotonicity of the proxy.


15

provided that it is possible to find a control variable Ic,i,t that can proxy investment

meaning that productivity η is the only unknown variable in the model. Therefore, using

(5) and (3), equation (2) can be rewritten as:

logYc,i,t = β0 + β1 logKc,i,t + β2 logLc,i,t + θt(Ic,i,t,Kc,i,t) + ηc,i,t (6)

Olley and Pakes (1992) and Levinsohn and Petrin (2003) estimate the model in two

steps. The first is to regress Y on L and on a non-parametric approximation of θ in order

to retrieve a consistent estimate of the labour coefficient. This procedure also makes it

possible to estimate the term β0 + β1 logKc,i,t + θt(Ic,i,t,Kc,i,t) that is used in the second step

to estimate the capital coefficient by bootstrapping. If the above term is denoted as γc,i,t,

for any value of βk, θt can be computed as the difference:

θt = γc,i,t − β1 logKc,i,t (7)

In order to see which βk is the ‘true’ beta, we assume that θt(Ic,i,t,Kc,i,t) follows a Markov

process:

θt = g(θt−1) + ζt (8)

where E(ζt|It−1) = 0

Estimating the latter non-parametrically, and regressing θt estimated in the previous

step on its lagged values, estimated values of ζt will converge asymptotically to the true ζt

only if the tested βk is the true beta.

Estimated TFP measures computed in the previous step are used in the second stage of

the analysis in order to evaluate the effects of investment in intangibles and participation

in GVCs. In particular, the following model is estimated:

TFPc,i,t = α + β GVCc,i,t + γ intc,i,t + µc,i,t (9)

where GVCc,i,t is a measure of network centrality intended to capture the level of

integration of sector i in GVCs, while intc,i,t is a vector of intangible capital components

including R&D, software and databases, design, branding and organizational capital.

Since variables in the intangible vector have a high degree of multicollinearity, each

component is evaluated in separate regressions; specifically, each component is evaluated

with a term that represents all other intangibles except the one considered. This approach


16

is used in order to avoid the problem of omitting relevant variables in each specification.

The following network centrality measures are used: node strength centrality, eigenvalue

centrality and betweenness centrality; each is considered in a separate specification.

At this point, we link our analysis to the extant literature on intangibles, productivity

and GVCs by evaluating the mediating effect of GVCs in the transmission mechanism

between intangibles and productivity. The literature highlights that intangibles have a

considerable direct effect on productivity and encourage participation in GVCs. We

therefore test whether they also have an indirect effect on productivity through the GVC

itself. The path diagram shown in Figure 1 shows the hypothesized relations among

variables.

Figure 1: Potential links between GVCs, intangibles and productivity

We perform a mediation analysis that estimates the direct effects of both intangibles

and GVCs on TFP, and uses the relationship between GVCs and intangibles to evaluate

the indirect effect of intangibles on productivity (via GVCs). This method exploits the

relationship between GVCs and intangibles in the analysis and evaluates the joint

contribution of both to productivity. The model is tested using the multilevel data method

proposed by Krull and MacKinnon (2001), in order to estimate all coefficients

simultaneously (rather than estimating every path in the diagram in separate equations)

and to account for panel effects.

We estimate the following three equations and combine coefficients to find direct,

indirect and total effects. First, we evaluate the impact of intangibles on TFP without

mediation:


17

TFPc,i,t = α0 + α1intc,i,t + ec,i,t

Second, we estimate the effect of intangibles on GVC:

(10)

(11)

Third, we estimate the impact of both intangibles and GVC on TFP:

TFPc,i,t = γ0 + γ1 GVCc,i,t + γ2 intc,i,t + ξc,i,t (12)

Finally, we combine the coefficients of the above equations and evaluate their

significance with bootstrapping simulations.

5 Results

5.1 The production function

Here we present the results of the first stage of the analysis, namely the estimation of the

production function, in Table 4. As noted in the previous section, we obtain our estimates

using the Olley and Pakes (1992) method, with capital investment as a proxy for

productivity shocks.

Table 4: The production function

The results at this stage are consistent with the literature at industry level. Next, we use

the residuals from the above regression as a proxy for TFP, which we use in the

remaining part of the analysis.


18

5.2 Intangibles and GVCs

In the second stage, we evaluate the impact of GVCs and intangibles on TFP. We proxy

GVC participation with three network centrality measures: node strength centrality,

eigenvalue centrality and betweenness centrality. Five types of intangibles are included

in the model: R&D, software and databases, design, branding and organizational capital.

Given the high degree of collinearity between the different intangible components, we

evaluate each intangible type in separate regressions taking care to avoid omitted

variable bias. The model is expressed as follows:

𝑇𝐹𝑃𝑡 = 𝛽0 + 𝛽1𝐺𝑉𝐶𝑡𝑗+ 𝛽2𝑖𝑛𝑡𝑡

𝑚 + 𝛽3(𝑖𝑛𝑡𝑡𝑡𝑜𝑡 − 𝑖𝑛𝑡𝑡

𝑚) + 𝜇𝑡 (13)

where the index j denotes the three network scores mentioned above, while m is the five

types of intangibles.10 All variables are considered in logarithms.

The outputs from fixed effects estimations are presented in Tables 4, 5, 6 and 7. The

results suggest that all of the types of intangibles considered are strong and significant

drivers of productivity in every specification. GVC participation is positive linked with

both strength and eigenvalue centrality, but not betweenness. This suggests that having

many connections matters for productivity, while being a bridge for other industries does

not. Tables 6 and 7 distinguish between two components of node strength centrality11, and

capture the distinction between forward and backward integration.12 Forward integration

measures the goods that flow out from a sector, while backward integration indicates the

goods flow into that sector. Forward integration happens when industry i provides inputs

to sector j (exports along the GVC), while backward integration occurs when industry i

receives inputs from sector j (imports along the GVC).13 Our results suggest that both

forward and backward integration matter for productivity.

10 Country and industry subscripts are removed to simplify the notation. 11 The network literature often refers to these measures as in-strength and out-strength centrality. 12 The distinction between the two components is recurrent in the GVCs literature. See, for example, Jona-Lasinio et al.

(2016) and Koopman et al. (2010). 13 For the purpose of this study, we use the terms importer and exporter in a more broad way beyond their

conventional understanding, as, in our analysis, transactions include also exchanges between sectors within the same

country.


19

Table 5: Node strength centrality


20

Table 6: Node strength centrality. Forward integration.


21

Table 7: Node strength centrality. Backward integration

To improve robustness, Table 8 uses a second measure of centrality, namely

eigenvalue centrality. The interpretation of this measure confirms a positive effect on TFP.

Table 9 uses a third measure, betweenness, to capture the extent to which a node is

important in linking other nodes in the network. Here, betweenness is not significant,

meaning that acting as a bridge for other industries does not have any direct effect on

productivity.


22

Table 8: Eigenvalue centrality


23

Table 9: Betweenness centrality

5.3 The mediating effect of GVCs

In this subsection, and following Baron and Kenny (1986), we evaluate the role of GVC as

a mediator between intangibles and productivity. The literature suggests that intangibles

both support GVC participation and enhance productivity (Jona-Lasinio et al., 2016).

Using the multilevel data technique proposed by Krull and MacKinnon (2001), we


24

estimate a system of simultaneous equations and evaluate the joint effect of intangibles

on TFP, both directly and indirectly via GVC.

Table 10 presents the results of estimates for all five types of intangibles considered.

Indirect, direct and total effects are reported in the first three rows, and the proportion of

the mediated effect in the fourth row. The direct effect is equal to the intangible coefficient

in equation (12). The indirect effect is computed as the product of the intangible effect on

GVC – the coefficient in equation (11), and the GVC effect on TFP – the first coefficient in

equation (12). The total effect is the sum of direct and indirect effects, while the total effect

mediated is the ratio between the indirect and total effect.14

Table 10: Mediation analysis results

In four of the five cases, the results provide evidence of the presence of a mediating

effect of GVC integration. Organizational capital is the exception. Brand and design have

an effect that accounts for around 60 percent of the total; R&D accounts for 45 percent;

while software and computers have a smaller but still significant indirect effect (about 16

percent of the total).

The indirect effect of intangibles (via GVC) is bigger than their direct effect in two

cases; it is almost as big as the direct effect in the case of R&D; and it is small or not

14 Standard errors are obtained via bootstrap simulations with a number of replications set to 500.


25

significant in the remaining two cases. Therefore, GVC integration has an important role

to play as a mediator in the majority of the categories of intangibles considered.

6 Conclusions

This sectoral analysis empirically investigated the impact of intangibles and GVC

participation on productivity for the period 2000–2014. The results revealed positive and

significant effects of five types of intangibles, confirming their importance for

productivity. Furthermore, the results demonstrated the mediating role of GVCs in the

relationship between intangibles and productivity, highlighting that intangibles

contribute to productivity both directly and indirectly (via GVC participation).

The paper makes several contributions to the literature on intangible capital and GVCs.

First, our study builds upon and extends prior research that investigates relationships

between intangibles, GVCs, and productivity. Although earlier work provides evidence

that intangibles and GVCs are drivers of productivity, to the best of our knowledge, the

joint effect of these two factors has not been examined. This paper addresses this gap, and

we provide an integrative examination of intangibles, GVCs and productivity. We found a

positive impact of both intangibles and GVC participation on productivity, measured with

network scores. Furthermore, our analysis identified two channels through which

intangibles affect productivity: one direct and one indirect, via GVC mediation. This result

provides new insights, as prior studies have pointed out that intangibles support GVC

participation but the magnitude of the effect on productivity is smaller than direct effects.

Second, we examine a broad range of intangible assets, including R&D, software and

databases, design, branding, and organizational capital. Furthermore, we adopt multiple

approaches to network measurements: node strength centrality (including forward and

backward integration), eigenvalue centrality, and betweenness centrality, which all help

to investigate relationships in more detail, provide finer-grained comparisons and

improve robustness.

Finally, from a methodological perspective, we draw upon a comprehensive, combined

dataset of intangibles and GVC participation measures at the industry level for the fifteen-

year period. The analytical approach is based on a two-stage procedure. In the first stage,


26

a TFP proxy was computed by estimating a production function with labour and capital

as inputs. In the second stage, the contribution of intangibles and GVC integration to

estimated TFP was assessed. At this stage, the two factors were tested both as separate

regressors and as combined predictors. The aim was to evaluate their joint effect via a

mediation analysis, which revealed that GVC was a mediator between intangibles and

productivity.


27

7 References

Acemoglu, D., Akcigit, U., and Kerr, W. (2015). Networks and the Macroeconomy: An

Empirical Exploration. NBER Working Paper Series, (21344):97.

Amador, J. a., Cabral, S., Mastrandrea, R., and Ruzzenenti, F. (2018). Who’s Who in

Global Value Chains? A Weighted Network Approach. Open Economies Review,

29(5):1039–1059.

Ando, S. (2014). Measuring US sectoral shocks in the world input–output network.

Economics Letters, 125(2): 204–207. https://doi.org/10/gfzjp7

Arrighetti, Alessandro, Fabio Landini, and Andrea Lasagni. (2014). Intangible assets

and firm heterogeneity: Evidence from Italy. Research Policy, 43(1): 202-213.

Baldwin, J. R., Yan, B., et al. (2014). Global value chains and the productivity of

Canadian manufacturing firms. Citeseer.

Baron, R. M. and Kenny, D. A. (1986). The moderator–mediator variable distinction in

social psychological research: Conceptual, strategic, and statistical considerations.

Journal of personality and social psychology, 51(6):1173.

Bloch, F., Jackson, M. O., and Tebaldi, P. (2016). Centrality Measures in Networks.

arXiv:1608.05845 [physics].

Blochl, F., Theis, F. J., Vega-Redondo, F., and Fisher, E. O. (2011). Vertex centralities

in input-output networks reveal the structure of modern economies. Physical Review

E, 83(4).

Bonacich, P. (1972). Factoring and weighting approaches to status scores and clique

identification. The Journal of Mathematical Sociology, 2(1): 113-120.

Bounfour, A. (2003). The Management of Intangibles: The Organisation’s Most

Valuable Assets. Routledge Advances in Management and Business Studies. Routledge.

Brandes, U. (2001). A faster algorithm for betweenness centrality. Journal of

Mathematical Sociology, 25(2): 163–177.


28

Carvalho, V. M. (2014). From Micro to Macro via Production Networks. Journal of

Economic Perspectives, 28(4): 23–48.

Cerina, F., Zhu, Z., Chessa, A., and Riccaboni, M. (2015). World input-output network.

PLoS ONE, 10(7): 1–21.

Chen, W., Gouma, R., Los, B., Timmer, M., et al. (2017). Measuring the income to

intangibles in goods production: a global value chain approach. Technical report, World

Intellectual Property Organization-Economics and Statistics Division.

Contreras, M. G. A. and Fagiolo, G. (2014). Propagation of Economic Shocks in Input-

Output Networks: A Cross-Country Analysis. arXiv:1401.4704.

Corrado, C., Haskel, J., and Jona-Lasinio, C. (2016). Intangibles, ICT and industry

productivity growth: Evidence from the EU. In Jorgenson, D. W., Fukao, K., and

Timmer, M. P., editors, The World Economy, pages 319–346. Cambridge University

Press, Cambridge.

Corrado, C., Hulten, C., and Sichel, D. (2005). Measuring capital and technology: an

expanded framework. In Measuring capital in the new economy, pages 11–46.

University of Chicago Press.

Criscuolo, C. and Timmis, H. (2017). The changing structure of GVCS: Are central hubs

key of productivity? In 2017 Conference of the Global Forum of Productivity, Budapest.

Dedrick, J. (2010). Who profits from innovation in global value chains?: A study of the

iPod and notebook PCs. Industrial and Corporate Change, 19(1):81–116.

Freeman, L. C. (1978). Centrality in social networks conceptual clarification. Social

networks, 1(3): 215–239.

Grazzini, J. and Alessandro, S. (2015). An empirical analysis of the global input-output

network and its evolution. UNIVERSITA CATTOLICA DEL SACRO CUORE Dipar-`

timento di Economia e Finanza, (31).

Greenhalgh, Christine, and Mark Rogers. (2006). The value of innovation: The

interaction of competition, R&D and IP. Research Policy, 35(4): 562-580.


29

Gu, F. and Lev, B. (2011). Intangible Assets: Measurement, Drivers, and Usefulness. In

Managing Knowledge Assets and Business Value Creation in Organizations:

Measures and Dynamics, pages 110–124. IGI Global.

Hall, Bronwyn H., Adam Jaffe, and Manuel Trajtenberg. (2005). Market value and

patent citations. RAND Journal of economics, 36(1): 16-38.

Heo, P. S. and Lee, D. H. (2019). Evolution patterns and network structural

characteristics of industry convergence. Structural Change and Economic Dynamics.

Jackson, M. O. (2014). The Past and Future of Network Analysis in Economics. SSRN

Scholarly Paper ID 2520284, Social Science Research Network, Rochester, NY.

Jäger, K. (2016). Eu klems growth and productivity accounts 2017 release, statistical

module1. Description of methodology and country notes for Spain.

Jona-Lasinio, C. S., Manzocchi, S., and Meliciani, V. (2016). Intangible assets and

participation in global value chains: An analysis on a sample of european countries.

Rivista di Politica Economica, 7(9): 65-95.

Kagawa, S., Okamoto, S., Suh, S., Kondo, Y., and Nansai, K. (2013). Finding

environmentally important industry clusters: Multiway cut approach using

nonnegative matrix factorization. Social Networks, 35(3):423–438.

Katz, L. (1953). A new status index derived from sociometric analysis. Psychometrika,

18(1): 39-43.

Koopman, R., Powers, W., Wang, Z., and Wei, S.-J. (2010). Give credit where credit is

due: Tracing value added in global production chains. Technical report, National

Bureau of Economic Research.

Koopman, R., Wang, Z., and Wei, S.-J. (2014). Tracing Value-Added and Double

Counting in Gross Exports. American Economic Review, 104(2): 459–494.

Korniyenko, Y., Pinat, M., Dew, B. (2017). Assessing the Fragility of Global Trade: The

Impact of Localized Supply Shocks Using Network Analysis. IMF Working Papers,

17/30.


30

Krull, J. L. and MacKinnon, D. P. (2001). Multilevel modeling of individual and group

level mediated effects. Multivariate behavioral research, 36(2):249–277.

Lenzen, M., Kanemoto, K., Moran, D., and Geschke, A. (2012). Mapping the Structure

of the World Economy. Environmental Science & Technology, 46(15): 8374–8381.

Lenzen, M., Moran, D., Kanemoto, K., & Geschke, A. (2013). Building Eora: A Global

Multi-Region Input–Output Database at High Country and Sector Resolution.

Economic Systems Research, 25(1): 20–49.

Levinsohn, J. and Petrin, A. (2003). Estimating production functions using inputs to

control for unobservables. The review of economic studies, 70(2): 317–341.

Marrocu, Emanuela, Raffaele Paci, and Marco Pontis. (2011). Intangible capital and

firms’ productivity. Industrial and Corporate Change, 21(2): 377-402.

McNerney, J., Fath, B. D., and Silverberg, G. (2013). Network structure of inter-

industry flows. Physica A: Statistical Mechanics and its Applications, 392(24): 6427–

6441.

OECD (2013). Interconnected economies: benefiting from global value chains. OECD

Publishing.

Oliner, S. D., D. Sichel, and K. Stiroh. "Explaining productive decade, finance and

Economics Discussion Series." Federal Reserve Board (2007).

Olley, G. S. and Pakes, A. (1992). The dynamics of productivity in the

telecommunications equipment industry. Technical report, National Bureau of

Economic Research.

O’Mahony, Mary, and Michela Vecchi. (2009). R&D, knowledge spillovers and

company productivity performance. Research Policy, 38(1): 35-44.

Porter, M. E. (1990). New global strategies for competitive advantage. Planning

Review, 18(3): 4–14.

Qian, Y., Hardle, W. K., and Chen, C. (2017). Industry Interdependency Dynamics in a

Network Context. SSRN.

https://doi.org/10/gfwmxs


31

Rapach, D., Strauss, J., Tu, J., and Zhou, G. (2015). Industry Interdependencies and

Cross-Industry Return Predictability. SSRN Electronic Journal.

Roos, G. (2017). Knowledge management, intellectual capital, structural holes,

economic complexity and national prosperity. Journal of Intellectual Capital, 18(4):

745–770.

Sandner, Philipp G., and Joern Block. (2011). The market value of R&D, patents, and

trademarks. Research Policy, 7(40): 969-985.

Ståhle, P., Ståhle, S., and Lin, C. Y. (2015). Intangibles and national economic wealth –

a new perspective on how they are linked. Journal of Intellectual Capital, 16(1):20–57.

Sturgeon, T. J. and Kawamaki, M. (2010). Global value chains in the electronics

industry: Was the crisis a window of opportunity for developing countries ? World Bank

Policy Research Working Papers.

Suh, Y., Kim, C., and Kim, M. (2013). Exploring Structure of Mobile Ecosystem: Inter-

Industry Network Analysis Approach. International Journal of Industrial and

Manufacturing Engineering, 7(4): 617–620.

Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R., and De Vries, G. J. (2015). An

illustrated user guide to the world input–output database: The case of global

automotive production. Review of International Economics, 23(3):575–605.

Tukker, A., de Koning, A., Wood, R., Hawkins, T., Lutter, S., Acosta, J., … Kuenen, J.

(2013). Exiopol – Development and Illustrative Analyses of a Detailed Global Mr Ee

Sut/Iot. Economic Systems Research, 25(1): 50–70.

UNCTAD (2013). World investment report 2013–global value chains: Investment and

trade for development. world investment report 2013.

WIPO (2017). World Intellectual Property Report 2017: Intangible capital in global

value chains. Geneva: World Intellectual Property Organization. Technical report.

World Bank Group; IDE-JETRO; OECD; UIBE; World Trade Organization.

(2017). Global Value Chain Development Report 2017: Measuring and Analyzing the

Impact of GVCs on Economic Development. Washington, DC: World Bank. © World

https://doi.org/10/gcpb6t


32

Bank. https://openknowledge.worldbank.org/handle/10986/29593 License: CC BY

3.0 IGO.”

Xing, L., Ye, Q., and Guan, J. (2016). Spreading effect in industrial complex network

based on revised structural holes theory. PLoS ONE, 11(5):1–18.

Yamano, N. (2016). OECD Inter-Country Input–Output Model and Policy Implications.

In Uncovering Value Added in Trade: New Approaches to Analyzing Global Value

Chains, pages 47–59. World Scientific.

Documents

GLOBAL VALUE CHAINS: The productivity contribution of ......(WIOD), while intangible valuation is obtained from the Intan-Invest database. Our unbalanced panel consists of 17 countries