• No results found

Sourcing Strategies and the Characteristics of Firms

N/A
N/A
Protected

Academic year: 2022

Share "Sourcing Strategies and the Characteristics of Firms"

Copied!
78
0
0

Laster.... (Se fulltekst nå)

Fulltekst

(1)

Sourcing Strategies and the Characteristics of Firms

An Empirical Analysis of European Firms

Jørgen Ingerød Steen

Master in Economics

Department of Economics

Faculty of Social Sciences

UNIVERSITY OF OSLO

(2)
(3)

Sourcing Strategies and the Characteristics of Firms

An Empirical Analysis of European Firms

Jørgen Ingerød Steen

(4)

c 2017 Jørgen Ingerød Steen

Sourcing Strategies and the Characteristics of Firms http://www.duo.uio.no/

Printed: Reprosentralen, University of Oslo

(5)

Preface

This thesis concludes five incredible years I have spent at the University of Oslo, where I have had the pleasure of learning from and collaborating with many smart, inspiring and in many other ways great people. A special thanks goes to my supervisor, Karen Helene Ulltveit-Moe, whose guidance has been invaluable in the completion of this paper.

A thanks also goes to my family, and especially my parents. Their continued support motivates me, and has allowed me to follow my passion.

All remaining errors are my own.

By Jørgen Ingerød Steen May, 2017

(6)

Abstract

This thesis seeks to add to the existing knowledge on the fragmentation of value chains in manufacturing firms. Using a matched set of firm-level data on key financial data and a survey on European firms I provide complementary evidence of firm performance, and go beyond earlier literature by adding information on the organization of firms. I identify five strategies for sourcing intermediate inputs in production. The first part of the analysis largely confirms existing evidence that firms engaged internationally are on average more productive than domestic firms. Using capital-, skill, and R&D-intensity as proxies, I also find that a relatively large share of firms engaged in headquarter-intensive production chains are using either domestic or foreign integration to source intermediate inputs. Controlling for characteristics of productivity and headquarter-intensity, I still find that there are significant di↵erences in the organization of firms that employ di↵erent sourcing strategies. Exporting parts of final output is positively related to the use of foreign sourcing, while decentralized decision structures are positively related with all strategies more complex than domestic outsourcing. In addition firms using integration are relatively more likely to be the controlling part of a group, and use external funding than firms engaged in outsourcing. I also find support that the selection into sourcing strategies is based on di↵erent characteristics in the North and South of Europe.

(7)

Contents

1 Introduction 1

2 Theoretical Foundation 4

2.1 Global Sourcing of Production Inputs . . . 4

2.1.1 Demand . . . 5

2.1.2 The Firm’s Optimization Problem . . . 6

2.1.3 Equilibrium . . . 8

2.1.4 Organizational Forms . . . 10

2.2 Testable Empirical Predictions . . . 14

3 Related Empirical Literature 16 4 Data sources 18 5 Descriptive Statistics 20 5.1 Categorizing Observed Sourcing Strategies . . . 20

5.2 Productivity . . . 21

5.3 Headquarter Intensity . . . 24

5.4 Other Firm Characteristics . . . 27

5.5 Pareto Distribution . . . 29

6 Empirical Analysis 30 6.1 Empirical specification . . . 31

6.2 Empirical Results . . . 32

6.3 Discussion of the Results . . . 39

6.4 Northern Versus Southern Europe . . . 43

7 Conclusion 48

(8)

References 51

Appendix A Adjustments to the Dataset 55

A.1 Irrelevant Observation . . . 55 A.2 Sectors . . . 55 Appendix B Full Sample Descriptive Statistics 57 Appendix C Regression Results - Characteristics of the Firm 58 C.1 Descriptive Statistics - North vs South . . . 67

(9)

1. Introduction

The liberalization of international trade, advances in communications technology and lower transportation costs has given firms across the world a wide range of opportunities in the organization of their production processes. There has been a large increase in fragmentation of production processes over the last several decades (see Amiti and Wei, 2005; Hummelset al., 2001), and there is evidence that firms are increasingly specialized in specific parts of the production chain (Johnson and Noguera, 2016). While traditional theories and empirics focused on the opportunity to lower costs by outsourcing low-skilled manufacturing jobs, Blinder (2009) shows that a multitude of jobs for di↵erent levels of skill can potentially be outsourced. One example is outsourcing of tasks related to research and development (R&D), which traditionally has been linked with high-skilled jobs, and is now identified as a possible driver to increase firm growth (see Bøler et al., 2015; Quinn, 2000; Linder et al., 2003).

Motivated by higher profitability, firms seek to gain competitive advantage by optimizing the organization of their production processes. By allowing

specialized entities to produce inputs or carry out specific tasks, a firm can be more efficient by utilizing comparative advantage and economies of scope of its partners. Some firms choose to acquire their inputs via outsourcing, purchasing services or intermediate goods from producers outside the firm boundaries.

Others choose to integrate the production technology into their own facilities, either by replicating the technology or acquiring the technology-owning firm as an affiliate. Both of these strategies can be conducted both at home and abroad, and in the data we often observe a mix of strategies utilized to maximize profits.

Theories of heterogeneous firms in international trade (see Melitz, 2003) inspired new theories to explain the sourcing strategies for intermediate inputs chosen by individual firms. These advances in the theory and increased availability of

(10)

firm-level data, has motivated much of the recent empirical research in

international trade. Most of the studies are focused on whether a firm supplies it’s goods to a foreign market, and the productivity levels associated with

di↵erent methods of doing so (see Girmaet al., 2005; Bernard and Jensen, 1999;

Clerideset al., 1998). A smaller, but growing, subset of studies focus on the global sourcing of production inputs or trade in tasks, and these studies

concentrate on explaining the strategic decisions associated with the production process (Navarettiet al., 2010; Corcos et al., 2013). Common to most of the studies is that they are solely focused on quantitative characteristics, like productivity and size, determining the strategic choices. Another limitation is that domestic firms are usually treated as a residual group of firms that are not active abroad, which implies that the full set of available sourcing strategies in the home country is not reflected.

In this thesis I reflect the full set of available strategies by distinguishing firms that are engaged in outsourcing and integration, at home or abroad. In addition, I do not limit my study to the usual quantitative measures of productivity, size and headquarter intensity, but use data from the 2010 EFIGE survey to identify qualitative information on other firm characteristics a↵ecting the strategic decisions related to sourcing strategies. Earlier research largely agree on the ranking of firm performance across sourcing strategies, but omit information on the organization of individual firms. This thesis contributes by identifying key di↵erences in the organizational structure of firms using di↵erent strategies, which could be the foundation for further research on causality between sourcing strategy, firm performance and other characteristics. In addition, the results are of interest to policy-makers who seek to facilitate conditions for running

profitable and productive businesses. By identifying di↵erences in characteristics between good and bad performers, this study could aid in firm-level policy-design to promote growth.

The empirical analysis builds on the theoretical work done by Antr`as and Helpman (2004), and all calculations are done in Stata version 14.2. I use a matched dataset of firm-level data from six countries in Europe, containing answers from the EFIGE-survey, conducted in 2010, and key financial data from

(11)

the Amadeus database. This dataset allows me to analyze both the quantitative and qualitative firm characteristics that are likely to a↵ect strategic decisions. I focus only on the sourcing of inputs, and di↵erentiate outsourcing and integration both at home and abroad. I sort firms into one of five mutually exclusive

sourcing strategies, based on assumptions on the complexity of using each one.

Firms are identified as using domestic outsourcing, domestic integration, foreign outsourcing, a mix of domestic integration and foreign outsourcing, or foreign integration1. I rely on a set of multinomial logit regressions to identify the relation between several firm-level characteristics and sourcing strategies.

The results from the quantitative part of the analysis is largely in line with earlier empirical evidence, confirming that higher productivity is associated with more complex sourcing strategies. This is in line with the theory, where only increasingly productive firms are able to overcome the increasing size of fixed costs associated with more complex strategies. I also find that firms sourcing inputs by integration are significantly more likely to be engaged in production intensive in headquarter-services.

The lack of a productivity gap between two sourcing strategies and the idea that the organizational structure of a firm has a significant impact on strategic

decisions motivates a qualitative study of additional firm characteristics across sourcing strategies. I find that domestic outsourcing firms di↵er from firms using complex strategies in a number of organizational ways. Exporting some of

produced output is positively associated with using foreign sourcing, and firms engaged in complex strategies are more likely to have decentralized decision structures. Being the controlling part of a group and using external finance is also more common for firms engaged in integration versus outsourcing. On the other hand, I find no preference of one strategy over another for firms controlled by a family owner.

The remainder of this thesis is organized as follows. Section 2 presents the theoretical foundation for firms’ chosen sourcing strategy. Section 3 presents related literature and section 4 describes the data sources. Section 5 presents

1I often refer to strategies in two groups, domestic outsourcing firms are the reference group, while all the others are part of the group defined as ”complex strategies”.

(12)

descriptive statistics for firms across sourcing strategies. Section 6 presents the empirical strategy and results of the empirical analysis. Section 7 concludes.

2. Theoretical Foundation

The latest advances in trade theory has been inspired by increased availability of firm-level data, showing large intra-sectoral di↵erences in the activities of

individual firms. This within-firm heterogeneity created a need for models that could explain choices made by individual firms2.

Melitz (2003) created a framework of within-industry firm heterogeneity in an environment of monopolistic competition to explain why some firms choose to export parts of their output while others only serve the domestic market. This has been the foundation for several extensions, among others Helpman et al.

(2004), who allow a second option for supplying the foreign market through foreign direct investment. However, firms do not only have a series of options available related to their output, but also to the organization of production lines.

In a globalized world a firm can choose to produce within or outside the boundaries of the firm and where to locate di↵erent parts of its production.

2.1 Global Sourcing of Production Inputs

Antr`as and Helpman (2004) mixes the setup from Melitz (2003) with a setting of incomplete contracts to explain the sourcing decisions of firms. The model

predicts where and how firms source their inputs, as a function of their

productivity. A firm can choose to produce everything within the boundaries of the firm, throughintegration, or acquire intermediate inputs through

arm’s-length contracts using outsourcing. Both integration and outsourcing can also either be carried out at home or abroad, creating a complex set of

2In contrast to traditional trade models that focused on the comparative advantage of nations or sectors, for example the Heckscher-Ohlin framework.

(13)

production strategies the firm can utilise. The model is based on the Melitz (2003) framework and uses the same demand function, while the production side has been adapted to create a bigger set of strategies in firms sourcing decision.

Antr`as and Helpman (2004) use a model of incomplete contracts by Antr`as (2003) to describe the relationship between the intermediate input producers and the final good producers.

2.1.1 Demand

The setting is a two-country world, where the countries are the North and the South. There is only one factor of production, labour, that is used to produce two di↵erent goods. All the consumers in the world have the same preferences, and the number of consumers is normalised to 1. There are two types of goods in the world. x0 is a homogeneous good and there are J sectors supplying a mass of di↵erentiated goods. The utility function is given by:

U =x0+ 1 µ

XJ

j=1

Xjµ, 0< µ <1

wherex0 is the homogeneous good andXj is the total consumption of goods from sector j and µis a parameter. Following Melitz (2003) the consumption within an industry j is given by a constant elasticity of substitution (CES) function

Xj =

Z

xj(i)di

1

of the consumption of di↵erent varieties xj(i). The range of iis determined endogenously in the model. It can be shown that the elasticity of substitution between two varietiesxj(i) and xj(k) is a constant = 1/(1 ↵). Furthermore it is assumed that ↵> µ so that varieties within a sector j are closer substitutes than the homogeneous good or varieties from a di↵erent sector. Also notice that µand ↵ does not have a subscript because it is assumed that these are the same for all sectors. With these assumption it is possible to derive the inverse demand

(14)

function for variety iin sector j:

pj(i) = Xjµ xj(i) 1 (2.1) The interpretation of the demand function is easier in its usual form:

xj(i) = X

µ

1

j pj(i)11. Because ↵<1, we can see that the demand fori in sector j is positively related to the total consumption X in sector j and negatively related to the price of i.

2.1.2 The Firm’s Optimization Problem

The producers face a perfectly elastic supply of labour in each country. The wage rate is denoted wl, l=N, S for the North and the South respectively. Because of the perfectly elastic labour supply these wages are fixed, and it is assumed that wN > wS.

The environmental setting for producing firms are the same as in Melitz (2003).

When creating a firm, one has to pay an entry fee. After this fee is paid, the firm draws a productivity level✓ from a known distributionG(✓). In order to start producing the firm has to incuradditional fixed costs of production. These additional fixed costs are a function of the sourcing strategy chosen, integration oroutsourcing, and the location of the production, at home or abroad.

Production of final goods require two inputs hj(i) and mj(i), where hj(i) is associated with headquarter services and mj(i) is associated with component manufacturing. All final goods production must take place in the North, because it is assumed hj(i) cannot be outsourced and it can only be produced in the North. Output of final goods is given by a Cobb-Douglas function of the two inputs

xj(i) =✓

hj(i)

j

j mj(i) 1 ⌘j

1 j

, 0<⌘j <1 (2.2) where ✓ is the realized firm-level productivity level and ⌘j is a measure of the headquarter intensity in sector j. A higher ⌘j then implies that headquarter services are relatively more important in the production of i. Production of both

(15)

hj(i) and mj(i) uses one unit of labour per unit of output, and while hj(i) can only be produced in the North, mj(i) can be produced either in the North or in the South. This setup implies that there are two types of active firms (i)

final-good producers providing headquarter services (H) and (ii) component manufacturers supplying intermediate inputs (M). All final-good producers H must contract with component manufacturers M for the supply ofmj(i). In addition to the choosing whether to obtain intermediate inputs from within or outside the boundaries of the firm, H has to choose whether to contract with a M located in home or abroad.

To enter the market, a final-good producer must pay a sunk fixed cost wNfE, before observing the productivity level ✓. When a firm decides to start

production, it must search for a supplier ofmj(i) either in the North or in the South, and simultaneously decide whether tointegrate (V) or outsource (O) the production of the intermediate input. LetwNfkl where k={V, O} denotes the sourcing method andl ={S, N}is the location of M. Antr`as and Helpman define the term organizational form as the combination of k and l, and wNfkl is then referred to as the fixed organizational cost.

In line with Melitz (2003) the relationship between productivity and fixed costs drive di↵erences in the organizational structures of firms. It is assumed that the fixed costs are higher when M is located abroad and that the costs associated with integration are higher than with outsourcing. This implies a ordering of the size of the organizational fixed cost:

fVS > fOS > fVN > fON. (2.3) The relationship between the final-good producers and the component

manufacturers is characterised by a setting of incomplete contracts. It is not possible for H and M to sign an enforceable contract ex ante because the parties are not able to commit to not renegotiate the initial contractex post. The

negotiation of a contract is therefore heldafter M has produced the intermediate input. This negotiation follows a two-party Nash bargaining game. This implies that the parties negotiate over the surplus created by the contractual

(16)

relationship. Based on their relative bargaining power the final-good producer gets a share 2(0,1) and the component producer gets a share (1 ). The bargaining power of the final-good producer is a function of the organizational form H have chosen. M is assumed to have no outside option independent of the organizational choice of H because it specialies mj(i) to the specific H. In the case of outsourcing H also have zero outside option, because it does not have ownership rights to the produced quantities of mj(i) if the contract breaks down.

A breakdown of negotiation then implies xj(i) = 0. If H chooses integration it has ownership rights to mj(i), so it has implicitly bought a right to fire M and seize the produced quantities. It is necessary to assume a cost of firing M, because if firing M is costless, H will always have an incentive to do soex post.

Realizing this M would producemj(i) = 0 and an outsourcing dominate

integration in all cases. As a result, by firing M, H loses a share (1 l) of final good production because it is not as e↵ective in utilising mj(i) individually as in cooperation with M. Furthermore, assume that N > S because of the added complexity of retrieving the intermediate inputs across borders.

Location and ownership structure is chosen by a profit maximizing final-good producer. The component manufacturer has to pay a fixed fee to participate in a contract with H, to ensure that M participates at minimum cost for H. There is an infinite supply of M in both countries, so the return for M is equal to the participation fee plus the outside option, which is assumed to be zero.

2.1.3 Equilibrium

The potential revenue from final good sales if a contract is signed is R(i) = R

p(i)x(i)di, where the subscript j has been dropped because a firm operates within a given sector. Inserting for (2.1) and (2.2) yields

R(i) = Xµ

h(i)

↵⌘ m(i)

1 ⌘

↵(1 ⌘)

. (2.4)

If, on the other hand, the contract breaks down, the outside option of M is always zero while the outside option of H depends on the ownership structure

(17)

and the location of M.

With outsourcing the outside option of H is also zero. Therefore, H will get R(i) and M gets (1 )R(i).

Vertical integration gives more bargaining power to the final-good producer, because H now has a positive outside option. If the contract breaks down H can still produce lx(i) after firing M, so this is the outside option for H. In line with the Nash bargaining game, H’s return will be its outside opportunity plus a share

of the surplus [1 ( l)]R(i) generated by the contract. This implies that H will get ( l)R(i) + [1 ( )]R(i) while M gets (1 )[1 ( )]R(i). Both H’s and M’s payo↵ is defined as a share of revenues. Let klR(i) denote the payo↵ for H with the organizational form defined by the combination of k and l.

Remembering that N > S it is possible to order the size of kl:

N

V = ( N)+ [1 ( N)]> VS

= ( S)+ [1 ( S)]> ON = OS = (2.5) where ON = OS because there are no outside option for any of the firms in these cases. So integration leads to a higher fraction of revenues to H, implying that H has more bargaining power under integration.

Because H and M are not able to sign an enforceable contract ex ante, both maximize their own payo↵s individually. In particular H maximizes

klR(i) wNh(i) and M maximizes (1 kl)R(i) wlm(i). Using the first-order conditions from these maximization problems, they obtain a function for the total profits

kl(✓, X,⌘) = X ↵)/(1 ↵)↵/(1 ↵) kl(⌘) wnfkl (2.6) where

l

k(⌘) = 1 ↵[ kl⌘+ (1 kl)(1 ⌘)]

{(1/↵)(wN/ kl)[wl/(1 kl)]1 }↵/(1 ↵). (2.7) In the profit function X and ⌘ are taken as given by the individual firm, because these are industry-level measures, while ✓ is the realized firm-level productivity.

Because H can choose the fee, t, that M has to pay up front to participate in the contract and H faces an infinite supply of M, the optimal choice for t will imply

(18)

M kl = 0. This again implies that ⇡lHk =⇡kl(✓, X,⌘), so H will maximize the total profits.

Using (2.6) it is possible to see that there must be a cuto↵ level of ✓ at which profits are zero.

kl(✓, X,⌘) = max

k2{V,O},l2{N,S}kl(✓, X,⌘) = 0 (2.8) and this will implicitly define a cuto↵ value ✓ below which H is not able to make positive profits and will exit the market.

Antr`as and Helpman (2004) also assume free entry of H firms. This implies that the expected profits of entering the markets must be equal to the fixed cost of

entry. Z 1

kl(✓, X,⌘)dG(✓) =wNfE. (2.9) To see this imagine that there is a positive expected value of entering the market.

Then an infinite number of firms would like to enter the market, driving the expected profits down to zero.

2.1.4 Organizational Forms

A final-good producer faces di↵erent tradeo↵s when choosing its organizational form. Considering the location of M is that there are lower variable costs in the South, however, at the same time the fixed costs are higher there. Deciding between integration or outsourcing H will get a higher share of the revenue under integration, but this also removes some of the incentives for M because he knows he can be firedex post. To describe the di↵erences between the organizational forms, Antr`as and Helpman examine two types of sectors. One with relatively low headquarter intensity, defined as the component-intensive sector, and a relatively high headquarter intensity, defined as the headquarter-intensive sector.

The component-intensive sector is characterized by a low ⌘ the headquarter-intensive sector by a high ⌘.

As seen in figure 2.1 the selection into di↵erent organizational forms are related

(19)

Figure 2.1: Within-sector selection into organizational form. From Antr`as and Helpman (2004)

to the productivity level ✓. Notice that in both sectors the least productive firms exit, but there are also organizational di↵erences in the sectors. First consider the component-intensive sector. Antr`as and Helpman show that profits are decreasing in kl. Based on the ordering in (2.5), this implies that outsourcing dominates integration. However, OS = ON, so with respect to the share of revenue obtained, H is indi↵erent between outsourcing in the North and in the South. However, there are di↵erences in both the variable and the fixed costs depending on the location of M. M located in the South has lower variable costs but higher fixed organizational costs compared to the North. For a setting where the di↵erences in the variable costs are relatively small compared to the

di↵erence in fixed costs, figure 2.2 depicts the profits earned in the two di↵erent organizational forms.

The figure shows that firms with productivity level below (✓M) make negative profits and exit the market. Firms with productivity ✓M <✓<✓M ON have highest profits choosing outsourcing in the North and those with ✓>✓M ON outsource in the South. Notice that ⇡SO has a lower intercept and a steeper slope because of the larger fixed costs and lower variable costs associated with outsourcing in the South. Based on the free entry condition, the cuto↵ levels for the organizational forms can also be calculated

M =X ↵)/↵

wNfON

NO(⌘)

(1 ↵)/↵

(2.10)

NM O =X ↵)/↵

 wN(fOS fON)

OS(⌘) ON(⌘)

(1 ↵)/↵

. (2.11)

(20)

Figure 2.2: Equilibrium in a component-intensive sector. From Antr`as and Helpman (2004)

From figure 2.2 the intercept denoted (✓M ON )↵/(1 ↵) depends on the relative size of the wage di↵erences and the di↵erences in fixed organizational costs. Noticeably if wN/wS >(fOS/fON)(1 ↵)/↵(1 ⌘) the intercept between ⇡NO and ⇡OS will be below the x-axis, and only firms that outsource in the South will the active type of firms.

From figure 2.1 it is evident that the equilibrium in the headquarter-intensive sector has a richer set of organizational forms. Again total profits depend on , but when ⌘ is large, profits are positively correlated with . This implies that H want to follow the ordering from 2.5 to obtain the highest profits. In the same way as in the component-intensive sector, the relative sizes of variable costs, fixed costs and the productivity level create a selection into di↵erent

organizational forms. Now the combination of wN > wS, (2.3), (2.5) and ✓ create the profit-curves depicted in figure 2.3.

(21)

Figure 2.3: Equilibrium in a headquarter-intensive sector. From Antr`as and Helpman (2004)

Firms with productivity below ✓H exit the market. Firms with productivity level

H <✓ <✓NHV outsource in the North, ✓HVN <✓ <✓NHO integrate in the North,

HON <✓<✓HOS outsource in the South and the most productive firms ✓>✓SHO obtain highest profits by integrating in the South (vertical FDI). In this figure it is assumed that OS > VN, and this depends on an assumption on the relative importance of the tradeo↵s that the firm faces. By assuming that the wage di↵erence is relatively large compared to the di↵erence in VN and the following ordering holds for the steepness of the slopes in figure 2.3

S

V(⌘)> SO(⌘)> VN(⌘)> ON(⌘). (2.12) As in figure 2.2 the existence of di↵erent organizational forms are dependent on our assumptions, and by changing these we can have situations where the

(22)

intersections between any of the forms are placed so that some of them may not exist. Based on the assumptions above, the cuto↵s in figure 2.3 can be calculated:

H =X(↵ µ)/↵

wNfON

ON(⌘)

(1 ↵)/↵

NHO =X(↵ µ)/↵

 wN(fVN fON)

N

V (⌘) ON(⌘)

(1 ↵)/↵

NHV =X(↵ µ)/↵

 wN(fOS fVN)

OS(⌘) VN(⌘)

(1 ↵)/↵

SHO =X(↵ µ)/↵

 wN(fVS fOS)

VS(⌘) OS(⌘)

(1 ↵)/↵

(2.13)

As long asG(✓) has support on a high enough✓, the only thing that is certain is that the organizational form of vertical FDI exists.

2.2 Testable Empirical Predictions

The theory presented in this section inspires the empirical analysis in section 6.

The baseline analysis concentrates on the aspects of these predictions, before I extend the model to incorporate other aspects of firm heterogeneity. Specifically three main predictions lay the foundation for the analysis is section 6.

1. Firms engaged in production intensive in headquarter-related services are more likely engage in integration, while firms in manufacturing-intensive production lines will prefer outsourcing.

The intuition lies in the contractual environment of the model. Since the output is non-contractible ex-ante, both parties are subject to a possible holdup problem in the ex-post bargaining over produced outputs. Since both parties know this, there is a problem of underinvestment ex-ante, which can be alleviated by giving property rights of the output to the party that is most important in the production. In the headquarter-intensive production, integration gives the headquarter property rights of the output

(23)

produced by the manufacturer, hence the headquarter firm is incentivized to avoid underinvestment. In manufacturing-intensive production lines, the manufacturer is relatively more important than the headquarter, and the argument follows analogously to explain why outsourcing is preferred in manufacturing-intensive production.

2. Firms sort into more complex strategies depending on their productivity.

From low to high the model predicts: Exit, domestic outsourcing, domestic integration, foreign outsourcing, foreign integration.

This ranking is based on the assumptions on the relative sizes of fixed and variable costs related to each sourcing strategy. To overcome higher levels of fixed costs, the firm must be more productive. It is assumed that using integration incurs a larger fixed cost than outsourcing, and that foreign sourcing is more costly than domestic sourcing. The ranking above is based on the assumption that the fixed costs of foreign sourcing dominates fixed costs of integration, hence firms must be more productive to engage in foreign outsourcing than in domestic integration. As is evident, by

adjusting the assumptions on the relative sizes of fixed costs, a multitude of rankings are possible.

3. The prevalence of complex sourcing strategies is higher in sectors with more heterogeneity in the productivity distribution.

This prediction relies on the assumed distribution of the productivity draw, G(✓). Assuming that ✓ is drawn from a Pareto-distribution implies that a larger variance in the distribution increases the presence of complex strategies. When the variance increases, the Pareto-distribution has a heavier right tail, which increases the probability of drawing a high

productivity. This implies that a larger share of firms are able to overcome the fixed costs associated with complex strategies.

(24)

3. Related Empirical Literature

Although there has been a large increase in empirical studies on individual firms after Melitz (2003), the relatively poor availability of data still limits the scope of these studies. However, as firm-level data has become more available, the

knowledge of causes and e↵ects of individual firms’ strategic decisions have also increased and there is a growing body of evidence on the relation between productivity and sourcing strategies. On the other hand, there are still areas related to the structure of the firm and finances that need further research (see Jabbour, 2012; Bernard et al., 2012).

Hummels et al.(2001) show why measuring the value of fragmented sourcing strategies is not as straightforward as measuring the value of exports. The

increasingly fragmented production chains, especially internationally, implies that it is necessary to measure the value added by each part of the chain, not merely the value of the final product. They measure vertical specialization as the value added by production, by subtracting the value of imported inputs from the value of exported goods. Johnson and Noguera (2016) build on this framework, and find that trade in intermediate inputs make up 60-70% of world trade. They also find that the ratio of value added over value of exports is decreasing, indicating that firms are more specialized in their production, exploiting comparative advantages across several producers in a production chain. With trade in

intermediate inputs being such a large part of world trade, Bernard et al. (2009) provide evidence that the strategic decisions of individual firms to enter and/or exit trade has impacts on the aggregate trade flows between countries.

Several studies report evidence on firm characteristics of firms that use foreign sourcing versus domestic sourcing firms. Nunn and Trefler (2013) specifically tests the sector-level hypotheses from the theory above, using transaction-level data from the US. Di↵erentiating between foreign outsourcing and foreign

(25)

integration, they find that intra-firm trade increases with the level of headquarter intensity and productivity. They also find a cuto↵-level for headquarter intensity, below which there is no significantly positive levels of intra-firm trade. In

addition they provide evidence that a more dispersed productivity distribution in a sector is associated with a higher share of intra-firm trade. Tomiura (2007) provide further evidence on the international sourcing of production inputs.

Concentrating on firms that use foreign outsourcing, foreign integration and firms that are exporting, he finds that there are significant di↵erences in the productivity distributions across sourcing strategies in Japanese manufacturing industries. The productivity distribution for foreign integration firms is most heavily right-skewed, indicating that these are more productive than foreign outsourcing firms.

Common to most literature is that it either focuses on the international or the domestic fragmentation of production lines. However, there is some research that comprehensively study the full set of available strategies. Federico (2010) reports evidence on the productivity ranking of firms across the sourcing strategies for Italian manufacturing firms. He finds that firms using domestic outsourcing are least productive, while foreign integration firms are most productive. He also finds that foreign outsourcing firms are less productive than firms using domestic integration. In a comparable study, Kohler and Smolka (2009) di↵erentiate real headquarters and other buyers of intermediate inputs to match more closely to the contractual environment in the theory . They find the same ranking for firms using foreign integration and domestic outsourcing, however, in contrast to Federico (2010) they find that foreign outsourcing firms are more productive than domestic integration firms.

Studies of the causal link between productivity and sourcing strategies are relatively rare. Analogous to the discussion on causality between firm

performance and internationalization of sales (through exports or foreign direct investment), causality can run in di↵erent directions. The two main theories are of self-selection into complex strategies based on pre-existing productivity

di↵erences, the other is a learning e↵ect where being engaged in complex sourcing strategies increases productivity growth over time. In his paper, Federico (2010),

(26)

only has cross-sectional data on the sourcing strategy employed by firms, however, about two thirds of the sample has longitudinal data on productivity.

His preliminary discussion supports the theory of self-selection, which is the one presented in the theory by Antr`as and Helpman (2004). Kohler and Smolka (2014) has longitudinal data for both productivity and sourcing strategies of manufacturing firms in Spain, and they too find support that the self-selection e↵ect is stronger than the learning e↵ect of being engaged in complex strategies.

4. Data sources

The dataset used in this paper has data on a matched set of individual firms from two distinct databases. One part is from a project coordinated by Bruegel, to better understand the features of European firms in a global economy

(EFIGE). The other is from a database on key financial data, Amadeus, maintained by the Bureau of van Dijk.

The EFIGE-dataset is the result of a survey conducted in 2010, and the questions cover the years from 2007-2009. It has answers from 14,758 firms 3 from the manufacturing sector in seven European countries.

Table 4.1: Number of firms in the EFIGE survey, by country

Country Number of firms

Austria 443

France 2,973

Germany 2,935

Hungary 488

Italy 3,021

Spain 2,832

UK 2,066

Total 14,758

Source: EFIGE-dataset

3See appendix for some adjustments made to the dataset.

(27)

The survey concentrates on topics related to the internationalization decisions within firms, with questions about sales abroad, production facilities, and global sourcing of inputs. Along with these it also contains questions on other topics related to the structure of the firm, workforce characteristics, financial structure, markets and pricing, and innovation. The survey was only conducted on firms with more than 10 employees, which results in an oversampling of large firms.

Weights have been constructed to correct for this bias, and adjust the sample to ensure it is representative4

While the EFIGE-dataset is more concentrated on the decisions of the firm, it is restricted in its coverage of quantitative financial data. To cover both qualitative and quantitative aspects of individual firms, the EFIGE questionnaire has been paired with data from Amadeus. Amadeus contains key financial data on about 21 million European firms, which has been standardized to make comparisons across borders possible. Along with standard measures of operating turnover it also contains information on the number of employees, costs of production, and value of assets. The data from Amadeus has also been used to create measures of firm-level productivity that I will utilize when taking the theory to the data.

The dataset also contains information on the main sector each firm is active in.

To control for sector fixed e↵ects, each firm is assigned to a NACE Rev. 1.1 sector on the two-digit level. This implies that each firm is categorized as active in one of 19 sectors5.

The advantage of this dataset is the availability of qualitative data on the individual firms. While evidence from quantitative data on input-output tables, size of firms and productivity has been extensively researched, most of them are missing qualitative data on the structure of firms that is not observable through publicly available financial accounts.

The EFIGE survey has only been conducted once, resulting in a cross-sectional dataset. This implies that I am not able to make causal conclusions on the results. However, the results will describe characteristics of di↵erent firms and

4See Altomonte and Aquilante (2012) for methodology on constructing weights and ensuring a representative sample.

5See appendix A.2 for more information on the sectoral categorization.

(28)

sourcing decisions that are still valuable in policy construction and further research.

5. Descriptive Statistics

5.1 Categorizing Observed Sourcing Strategies

The EFIGE survey allows me to categorize firms based on answers to key questions about the production and ownership of the firm. Because EFIGE is concentrated on the internationalization of firms, it is especially well suited to identify international sourcing of production. However, taking the theory to the data is not straightforward because the observed sourcing strategies are generally more complex than the ones described in the theory. About 45% of firms in the sample utilize two or more of the strategies defined above, therefore it is not trivial to define strategy categories. To account for these observed complexities, I have created 5 mutually exclusive categories. Each firm is categorized into what is assumed to be the most complex sourcing strategy applied in it’s production chain.

I employ a broad definition of outsourcing, as all firms that have purchased either services or intermediate goods are defined as outsourcing firms. I do not di↵erentiate whether these intermediate inputs are bought through arm’s-length contracts with partner firms or through a market for these goods or services.

Domestic outsourcing (DO) firms are then defined as firms that buy the intermediate inputs in the domestic economy and do not engage in any other sourcing strategy. On the other hand, if a firm purchases intermediate inputs abroad it is defined as a foreign outsourcing (FO) firm. Firms that combine DO and FO are also defined as foreign outsourcing firms.

Domestic integration (DI) firms are firms that report ownership of domestic affiliates. A firm is categorized as DI if this is the only strategy, or if it uses a mix of DO and DI. Firms engaged in foreign integration (FI) has answered that

(29)

they run part of their production abroad through direct investment. This is assumed to be the most complex strategy, so all firms that use FI either as only strategy or in combination with any other strategy is categorized here.

To recognize that sourcing strategies are more complex in the data than in the theory, I have added an intermediate group between FO and FI. This category includes firms that use both domestic integration and foreign outsourcing, or DO, DI and FO simultaneously. This group will be referred to as DI & FO.

Table 5.1 displays the number of firms active in each sourcing category. The most prevalent strategy is outsourcing, as about 78% of the sample use either foreign or domestic outsourcing. There are about the same number of firms using domestic integration and the mixed strategy of foreign outsourcing and domestic integration. The least prevalent category is foreign integration, with only 5.6% of firms in the sample. These results are similar to earlier empirical results (see Kohler and Smolka, 2009; Federico, 2010)

Later results will depend on the use of productivity measures on individual firms.

Data on total factor productivity (TFP) is available for about half the sample, and the same categorization is displayed for this sub-sample in panel (2) of table 5.1. The availability of TFP data for Austrian firms is very low, with only 20 firms left when restricting the sample. This sample is too small for the empirical analysis in section 6, so I drop Austrian firms from the core sample.

There is no evidence of bias towards any category that would make the sample less representative. Since all estimations will be based on the core sample, I will only present descriptive statistics of the reduced sample in the following sections.

Corresponding tables for the full sample are placed in the appendix.

5.2 Productivity

The theory assumes that fixed costs of each sourcing strategy increases with the strategy’s complexity. To overcome these additional fixed costs and self select into complex strategies, firms must be increasingly productive.

(30)

Table 5.1: Firms’ sourcing strategies

(1) Full Sample (2) Core Sample

Sourcing strategy N. of

firms Percent N. of

firms Percent

Domestic Outsourcing 5,632 43.8% 2,822 42.8%

Domestic integration 1,134 8.8% 541 8.2%

Foreign outsourcing 4,423 34.4% 2,345 35.5%

Domestic integration & Foreign

outsourcing 941 7.3% 519 7.9%

Foreign integration 719 5.6% 370 5.6%

Total 12,849 100.0% 6,597 100.0%

Source: Author’s elaborations on the EFIGE-dataset.

In this study, productivity is measured by total factor productivity (TFP), and is included in the dataset (see Altomonteet al., 2012). Based on a standard

Cobb-Douglas production function, TFP measures how much a firm produces, for a given amount of inputs. A firm that can produce more with the same amount of inputs as a competitor, has a higher TFP and is more productive. To avoid possible simultaneity bias between unobserved productivity shocks and observed inputs of labor and materials, TFP has been calculated using the Petrin and Levinsohn (2003) production function6.

Based on prediction 2 in section 2.2 it should be possible to rank sourcing strategies based on productivity. The assumptions on relative sizes of fixed costs across strategies a↵ect the predicted ranking. With my additional strategy (DI&FO) and the baseline assumptions the model predicts that low productivity firms choose DO, DI firms have low-medium productivity, FO firms have medium productivity, DI&FO firms have high-medium productivity and high productivity firms choose FI. The equilibrium solution also predicts that revenues will follow the same rankings.

Table 5.2 presents some descriptive statistics on revenues, number of employees, and productivity. In line with the predictions, DO firms have the lowest

6For more information on the exact method for estimating TFP, see Altomonteet al.(2012, Appendix 2).

(31)

Table 5.2: Descriptive Statistics for Productivity, Turnover, and Employees

N. of firms

Avg.

turnover (in 1,000 EUR)

Avg. n. of employees

Total Factor Productivity

DO 2,822 6,805.73 37.27 0.882

DI 541 17,729.71 68.90 1.061

FO 2,345 19,471.00 76.71 1.043

DI&FO 519 26,616.98 88.83 1.215

FI 370 100,432.64 364.45 1.308

Total 6,597 17,278.37 71.38 0.994

Source: Author’s elaborations on the EFIGE-dataset.

productivity and the smallest revenues, while FI firms are largest in revenues and have the highest productivity. Also DI & FO firms have the expected

medium-high levels of both revenue and productivity. On the other hand, DI and FO firms are not clearly ranked. While there is evidence that DI firms are

marginally more productive, FO firms have larger revenues, a relationship that is not reflected in the model. The ranking in terms of the number of employees follows the same pattern, where the smallest are DO firms, DI&FO firms are medium-large and FI firms are the largest. Again, the DI and FO firms are very similar in number of employees.

In figure 5.1 the distribution of TFP has been divided into deciles. Remember that the firms are categorized in what is assumed to be the most complex strategy it is active in. The figure signals that more productive firms are more likely to use complex strategies. It is evident that the share of firms that use domestic outsourcing as it’s most complex sourcing strategy is decreasing as productivity increases. For the firms with the lowest 10% of productivity, more than half choose DO as its most complex strategy, for the top 10% this is reduced to about 21%. The trend across TFP deciles is that the prevalence of DO decreases, while it increases for all other strategies. The selection e↵ect seems to be most pronounced for FI firms. For deciles 1-7 less than 5% of firms use FI, while in the 10th decile this has increased to 13%.

(32)

020406080100

Percent of firms

1 2 3 4 5 6 7 8 9 10

Deciles of TFP

FI DI&FO FO DI DO

Figure 5.1: Share of active firms in each sourcing category

Source: Author’s elaborations on the EFIGE-dataset. Notes: Deciles of TFP are created by sorting TFP in ascending order, and then splitting it into 10 equally large groups. Decile 1 is then the 10% of firms with lowest productivity. The vertical axis measures the prevalence of each sourcing strategy, by percent of firms active in each one.

5.3 Headquarter Intensity

In the theoretical model, sourcing of intermediate inputs is contracted on between a firm supplying headquarter services (H) and an intermediate inputs supplier (M). Based on the contractual environment between H and M, it is assumed that as the importance of investments carried out by H increases, the prevalence of integration should also increase. In line with prediction 1 in section 2.2 this implies that the prevalence of integration should increase with headquarter intensity. Because headquarter intensity is not directly observable in the data, I use three measures to proxy its level (see Antr`as and Yeaple, 2014).

Capital intensity, measured as total fixed assets per worker, reflects the importance of headquarter services in a firm. This is based on an assumption that the production of intermediate goods is more labor intensive than the production of headquarter services. There are possible caveats to using capital

(33)

intensity as a measure of headquarter intensity. While headquarter services are likely to be intensive in capital related to computers and other technical forms of assets, component production is likely to be intensive in machinery and

buildings. Nunn and Trefler (2013) disaggregate the capital stock into parts that are contractible and non-contractible between H and M. They find that

specialized capital, like machinery has a significant e↵ect on intra-firm trade, while non-specialized equipment has a zero or negative e↵ect. My data does not allow this level of disaggregation, so I rely on the amount of total fixed assets per worker.

Skill intensity is assumed to be positively correlated with the importance of headquarter services. I measure skill intensity by the share of white-collar workers in the workforce. White-collar workers are defined as ”Office, clerical, administrative, sales, professional, and technical employees” by Herman and Abraham (1998, p.74). These tasks are assumed to represent the work associated with headquarter services, as opposed to manufacturing of intermediate inputs.

Theoretically this measure seems to be a good proxy for headquarter intensity, however, it is plausible that with recent technological advancements the

manufacturing of intermediate inputs is increasingly handled by machines. If this is the case, it could increase the need for white-collar workers in manufacturing firms, to control and maintain the automated parts of production.

The last proxy for headquarter intensity is R&D intensity. This variable is assumed to capture the idea that in relationships between H and M firms, the H firm will do the innovation and design of products at their headquarters, and then use the comparative advantage of its partners to produce the final goods.

This variable su↵ers from the fact that there is di↵erent types of R&D, while it is reasonable to assume that the headquarters invest in the R&D related to the final product, manufacturers of intermediate goods also have an incentive to research and develop their production processes to increase productivity and reliability. The data does not di↵erentiate between product and process R&D. I measure R&D intensity as the share of R&D workers on total employment (see Castellani and Zanfei, 2007). A possible caveat with this measure is that it shows how much of it workforce is concentrated around R&D, however, it might not

(34)

Table 5.3: Descriptive Statistics for Headquarter Intensity

N. of firms

Avg. capital stock per employee (in

1,000 EUR)

Avg. share of white-collar

workers

Avg. share of R&D workers

DO 2,822 56.92 32.8% 6.3%

DI 541 101.54 39.1% 7.2%

FO 2,345 61.66 35.7% 7.9%

DI&FO 519 101.64 40.0% 6.7%

FI 370 81.73 43.9% 10.7%

Total 6,597 66.44 35.2% 7.1%

Source: Author’s elaborations on the EFIGE-dataset. Notes: All averages are weighted averages. Avg. capital stock per employee is measured by fixed assets over number of employees. Share of white-collar workers is the ratio of white-collar workers over total number of employees, share of R&D workers is the ratio of R&D workers over total number of employees.

reflect the total amount of resources spent on R&D if a firm chooses to outsource some or all of its R&D processes.

Table 5.3 presents some descriptive statistics on the measures of headquarter intensity across sourcing strategies. Both capital per employee and and share of white-collar workers increase when moving from outsourcing to integration. Since these are proxies for headquarter intensity, this supports prediction 1 from

section 2.2. However, the ranking changes for the locations of integration across the di↵erent measures of headquarter intensity. While DI (and DI&FO) firms have higher capital intensity than FI firms, the opposite is true for the share of white-collar workers. R&D intensity shows a slightly di↵erent ranking. While FI firms are clearly the largest and DO firms are the smallest, DI&FO firms have medium-low intensity, DI firms have medium intensity and FO firms display medium-high intensity of R&D workers. However, the di↵erences between the medium levels of R&D intensities are relatively small, hence it is difficult to clearly rank them. In light of the results on R&D intensity, the reversed rank between DI and FI firms in capital and skill intensity could be a result of

(35)

di↵erent types of headquarter intensities. It is possible that DI firms are more concentrated on the administrative tasks related to headquarter services, while FI firms more heavily invested in innovative research.

Since capital intensity, skill intensity and R&D intensity are all supposed to proxy headquarter intensity, there could be problems of collinearity between them. Table 5.4 shows the correlation between these measures. The correlations between capital intensity, skill intensity and R&D intensity are low, and there is no evidence of a linear relationship between the variables in the first column. As expected the second column reveals some correlation between skill and R&D intensity, since white-collar workers are more likely to participate in designing and innovating products. Unreported graphical studies of the measures supports the evidence from table 5.4, suggesting that I do not need to worry about

multicollinearity between these regressors.

Table 5.4: Correlation matrix for measures of headquarter intensity

Variables (K/L) (WH/L) (R&D/L)

(K/L) 1

(WH/L) 0.0399 1

(R&D/L) 0.0251 0.1542 1

Source: Author’s elaborations on the EFIGE-dataset. Notes: (K/L) is the capital intensity measured as total fixed assets per worker, (WH/L) is skill intensity measured as share of white-collar workers in the workforce, and (RD/L) is R&D intensity measured as share of workers engaged in R&D activities over total employment.

5.4 Other Firm Characteristics

While the previous sections have broadly supported the quantitative predictions from the theory and largely reflects the body of evidence presented in earlier literature, the EFIGE dataset is not limited to these measures. I have access to a number of key characteristics concerning the ownership and managerial structure of firms, in addition to some information on the access to external funds.

(36)

Table 5.5: Descriptive Statistics for Firm Characteristics

DO DI FO DI&FO FI

Exporting 59.4% 63.0% 83.6% 84.2% 94.1%

Family-

controlled 66.2% 56.4% 55.9% 55.5% 44.1%

Centralized

decisions 75.7% 64.3% 69.1% 64.0% 53.2%

Controlling 2.6% 51.8% 5.1% 50.5% 41.1%

External

funding 54.1% 57.5% 50.5% 59.7% 46.5%

Source: Author’s elaborations on the EFIGE-dataset. Notes: Exporting, family controlled, centralized decisions, controlling, and external funding are all measured as the ratio of firms that sort into each category over the total number of firms in the sample.

Table 5.5 shows some descriptive statistics for these measures, across all sourcing strategies.

Considerable heterogeneity in these variables across sourcing strategies suggests that these are significantly di↵erent firms sorting into each category.

A firm is defined as an exporter if it sell at least some of it’s production abroad.

The descriptive statistics suggests a positive correlation between being an exporter and complex sourcing strategies. While only 59% of DO firms are exporters, more than 94% of firms engaged in FI are also exporting a share of their production.

Family-owned firms often represent a large part of the family’s total wealth, inducing risk aversion in adapting new technologies and managerial structures.

The family-controlled firm is defined as a firm where the CEO or the head of the firm is the individual that owns it, or a part of the family of the individual that own it. The data suggests a negative relation between complex sourcing

strategies and a family member being the CEO of a firm.

While family ownership refers to the ultimate owners of a firm, the everyday conduct of business will in many cases be heavily dependent on middle

(37)

management. Still, firms di↵er in their managerial structure in the rights to make strategic decisions. A firm is defined as having centralized decision rights if most of the strategic decisions has to go through the CEO, leaving little autonomy to the individual middle manager. More than two thirds of DO firms have

centralized decision rights, while this figure is reduced to about half for FI firms.

About 26% of the sample is part of either a foreign or domestic group of firms.

The firm is defined as controlling if it answered in the EFIGE survey that it is head of the group or if it own affiliates either in the home country or abroad.

There is a large discrepancy between firms engaged in outsourcing and integration for this category. While only 5% of firms engaged in FO controls other firms in a group, this raises to about 50% of DI firms.

Firms engage in external financing when they acquire funds that have not been generated within the firm boundaries. This can be either through credit,

securities, external equity or other financial instruments. External finance can be essential to obtain enough funds to undertake profitable and productivity

enhancing investments. Firms that have answered that they used external

funding in 2008-2009 fall into this category. About half the sample used external funding in this period, and there is relatively little heterogeneity between the sourcing strategies.

5.5 Pareto Distribution

Most of the predictions in section 2.2 are a result of the assumption that firms draw their productivity, ✓, from a Pareto distribution. Most of the results

derived above, at least partially, confirm the predictions from the theory. Still, it is informative to check whether this is a result of appropriate assumptions or not.

Figure 5.2 shows both the observed values of TFP and a theoretical Pareto distribution. Based on the figure, the assumption that the productivity draw is taken from a Pareto distribution is justifiable.

(38)

0.2.4.6.81Density

0 5 10 15 20

Total Factor Productivity Observed values of TFP

Theoretical Pareto probability density function

Figure 5.2: Observed and theoretical distribution of TFP (2008)

6. Empirical Analysis

As we have seen in sections 2, 3, and 5 there has been substantial research done on the relationship between firm performance and internationalization modes, both in terms of sales and sourcing of production inputs. This thesis does not limit the analysis to the usual quantitative measures of productivity, size, headquarter intensity, and innovation, but adds information on other firm

characteristics in terms of ownership, managerial structure, and financing. In the following DO is the reference strategy, while all other strategies are part of the group of ”complex strategies”.

To identify di↵erences in the organizational structure of firms engaged in di↵erent sourcing strategies, I keep the observed strategy as the outcome

variable. Then I use the firm characteristics described in section 5 as explanatory variables to identify any of these are associated with using one strategy over another. This allows me to identify both quantitative and qualitative di↵erences

Referanser

RELATERTE DOKUMENTER

73 This included managers and teachers at madrassas and schools, leaders and officials of local government, alumni of madrassas and notable donors from the community,

However, at this point it is important to take note of King’s (2015) findings that sometimes women can be denigrated pre- cisely because they are highly able

The dense gas atmospheric dispersion model SLAB predicts a higher initial chlorine concentration using the instantaneous or short duration pool option, compared to evaporation from

Based on the above-mentioned tensions, a recommendation for further research is to examine whether young people who have participated in the TP influence their parents and peers in

Figure 5.3 Measured time series of the pressure for HK 416 N at two different directions from the shooting direction, with and without flash suppressor, at 84 cm from the muzzle..

Azzam’s own involvement in the Afghan cause illustrates the role of the in- ternational Muslim Brotherhood and the Muslim World League in the early mobilization. Azzam was a West

Along the R&amp;D dimension, we estimate the model on the distribution of R&amp;D effort (intensity and level), correlations between R&amp;D intensity and firm size and

There had been an innovative report prepared by Lord Dawson in 1920 for the Minister of Health’s Consultative Council on Medical and Allied Services, in which he used his