KATHOLIEKE

UNIVERSITEIT
LEUVEN

DEPARTEMENT TOEGEPASTE
ECONOMISCHE WETENSCHAPPEN

RESEARCH REPORT 0146

MULTINATIONAL KNOWLEDGE SPILLOVERS

WITH CENTRALIZED VERSUS DECENTRALIZED
R&D: A GAME THEORETIC APPROACH
by
F. SANNA-RANDACCIO
R. VEUGELERS

0/2001/2376/46
 Multinational Knowledge Spillovers with Centralized versus
Decentralized R&D: a game theoretic approach

By

FRANCESCA SANNA-RANDACCIO
University of Rome "La Sapienza"
Via Buonarroti 12
00185 Rome, Italy
E-mail: fsr@dis.uniromal.it
Tel: 39-06-48299228
Fax: 39-06-48299218

And

REINlllLDE VEUGELERS
KULeuven and CEPR-Fellow
Naamsestraat 69
3000 Leuven, Belgium
E-mail: Reinhilde.Veugelers@econ.kuleuven.ac.be
Tel: 32-16-32 69 08
Fax: 32-16-326732

December 2001

Abstract
This paper provides a theoretical model on the trade-offs a MNE faces when assigning subsidiaries an
active role in innovation and organizing its R&D decentralized versus centralized. R&D
decentralization avoids having to adapt centrally developed innovations to local markets, being able to
use the specific know-how of the subsidiary. In addition R&D subsidiaries can be used to source
locally available external know-how. But the MNE has to organize the transfer of local know-how
internally so as to be able to benefit from this location specific know-how throughout the organization.
At the same time, decentralization of R&D to the subsidiary level intensifies the challenge of
effectively appropriating core technology know-how, preventing the spilling over of valuable know-
how to competitors, located in the foreign markets. While R&D decentralization has repercussions on
both intra-company technology transfers as well as inter company technology spillovers, it emerges as
a possible equilibrium outcome from the resulting strategic interaction between the foreign subsidiary
and local competition. The proposed model treats both internal and external spillovers in a game-
theoretic context explicitly recognizing that absorptive capacity is required to be able to use external
spillovers. The analysis suggests that a strong local know-how base is not a univocally positive factor
for locating R&D abroad and indicates the critical complementary role of managing internal and
external spillovers to capitalize on the benefits from R&D decentralization. It also shows that the
intensity of product market competition in the host country is important, especially in determining the
outgoing spillover costs.

Veugelers acknowledges support from FWO (G.OI31.98), PBO (98/KUU5) and CNRS (Enjeux economiques de
l'innovation). Sanna-Randaccio acknowledges travel grant support from the EU Commission (EU MESIAS
network) and from CNR. The comments from R. Belderbos, as well as the hospitality of Universiteit Maastricht,
where the second author was visiting, are gratefully acknowledged.
 Multinational knowledge spillovers with centralized versus
decentralized R&D : a game theoretic approach

1. Introduction

In the traditional literature on multinationals, following the seminal work of Dunning

(1988), multinational activities originate out of the R&D activities of the firm. But rather than
seeing the geographic dispersion of MNEs as a result of knowledge creation, the emphasis in
the literature has shifted towards seeing the geographic dispersion of MNEs as a source for
knowledge creation (see a.o. Cantwell (1995), Niosi (1999), Kuemrnerle (1997)). In the
current international environment innovation strategies require increasingly more global
sourcing, sensing new market and technology trends worldwide. All this implies a different
role for subsidiaries in the innovative strategy of the multinational enterprise, who become
important to access (local) external sources. The subsidiary is being viewed as a vehicle to
continually reassess and upgrade know-how on core products and technologies, to provide a
basis for new generations of innovative products, which can be used throughout the
multinational organisation A major challenge identified for the MNE is to find an
organisational system that is capable of transferring know-how across units and locations,
allowing locally generated know-how to be used globally.
This paper provides a theoretical model on the trade-offs a MNE faces when
assigning subsidiaries an active role in innovation and organising its R&D decentralized
versus centralized. The model considers R&D decentralization as a choice which allows to
use the specific know-how of the subsidiary and avoids having to adapt centrally developed
innovations to local markets. In addition R&D subsidiaries can be used to source locally
available external know-how. But the MNE has to organize the transfer of local know-how
internally so as to be able to benefit from this location specific know-how throughout the
organisation. At the same time, decentralization of R&D to the subsidiary level intensifies
the challenge of effectively appropriating core technology know-how, preventing the spilling
over of valuable know-how to competitors, located in the foreign markets. While R&D
decentralization has repercussions on both intra-company as well as inter company
technology transfers, it emerges as a possible equilibrium outcome from the resulting strategic
interaction between the foreign subsidiary and local competition. The proposed model
focuses on how the interplay of internal and external knowledge flows interacts with the
nature of host market competition to influence the choice of MNEs to effectively disperse
internationally its R&D. It treats both internal and external spillovers in a game-theoretic

2
context explictly recognizing that absorptive capacity is required to be able to use external
spillovers.
Abstracting from internal technology transfers, the impact of external know-how
spillovers on the incentives of firms to innovate has been widely studied in Industrial
Organization (see De Bondt (1996) for an overview). This literature focuses on the
importance of the strategic effects of spillovers, stressing the interaction with product market
competition. High spillovers can thus lead to underinvestment in R&D, when firms are
marketing substitute products. Recently, some LO. models have taken into account that firms
can manage these spillovers through organizational decisions. Cohen and Levinthal (1989)
pioneered the idea that firms can try to increase incoming spillovers by investing in
"absorptive capacity", i.e. spillovers are more efficient in reducing own costs when the firm
is engaged in own R&D. This notion of absorptive capacity has been integrated in the LO.
models on R&D cooperation by Kamien & Zang (2000). The influence of external R&D
spillovers on the incentive to engage in FDI has also been analyzed (Petit & Sanna-
Randaccio (2000».
Another related line of research is the geographical localisation of innovative
activities. Innovative activities are found to be highly clustered (Jaffe et al (1993), Audretsch
& Feldman (1996». The principle explanatory factor for clustering revolves around the
existence of knowledge spillovers. Since distance hinders the exchange of especially tacit
knowledge, proximity matters for being able to absorb external spillovers. Hence firms
agglomerate their R&D activities to be able to capitalize on external knowledge spillovers.
Gersbach & Schmutzler (1999) present a game-theoretic model of geographic clustering of
activities. A duopoly decides on the location of their R&D and production, while competing
in final output markets a la Bertrand. An agglomeration equilibrium where both firms would
choose their R&D site in the same location, requires simultaneously internal and external
spillovers: not only must firms learn something from each other in agglomeration, they must
also be able to transport this know-how internally. J

A more closely related line of previous research, linking internal and external
knowledge flows and subsidiaries of MNEs, can be found in Das (1987), who examines
whether parent firms will transfer technology to subsidiaries given that local rivals may
learn. This model is specifically set up for developing countries: the subsidiary is non R&D
active, but receives a transfer of technology from its parents, while it is competing as a

J Gersbach & Schmutzler (1999) consider two types of external spillovers: external spillovers when rival
production units are co-localized with own R&D sites, and knowledge complementarities among co-
localized R&D sites. In addition, the firms also need to consider that internal spillovers are required when
R&D is located separately from production. When interpreting their results in an FDI setting (see
Gersbach & Schmutzler (2000», their restriction to the use of only one R&D site per firm precludes a

3
leading firm facing a competitive fringe of local firms which are non R&D active but can
costIessly learn from subsidiaries. The paper finds that despite learning by local firms, it is
still worthwhile for the parent to transfer the better technology. A similar LDC setting is
used in Wang & Blomstrom (1992) who take into account that MNEs face a cost of
transferring internally technology, which will be higher for state-of-the-art technologies, and
that locals face a cost of learning. When the subsidiary competes in a differentiated duopoly
with the local firm who faces a technology gap, they find that technology transfers via FDI
are positively related to the level of host country's firms learning investment. The focus of
this literature on LDCs implies that only the internal transfers from headquarters to
subsidiaries and the external transfers from subsidiaries to local firms are considered, while
the competitive strncture the subsidiary is facing is one of a weaker local rival. The issue of
R&D decentralization is not at stake here.
With the majority of FDI located in developed countries, subsidiaries active in R&D
activities and technology sourcing becomes an important issue resulting in flows from local
external sources to subsidiaries, while the MNE has to ensure that the locally generated
know-how flows from the subsidiary towards the central level. In addition the issue of
appropriating know-how becomes more critical when local rivals are no technology
laggards. Our analysis of R&D decentralization by the MNE focuses on the interaction
between host product market competition and both inter- and intra-firm know-how transfers.
We consider both the two-way internal transfer of know-how between headquarters and
subsidiaries as well as the two-way external transfer of know-how between subsidiaries and
local competitors.
Before presenting the model set-up and results, the paper starts with an overview of
recent literature and stylized evidence on internationalization of R&D, which is used when
constructing the model.

2. Evidence on R&D decentralization by MNEs

Centrifugal demand and supply related forces for decentralization need to be traded
against centripetal forces (see Grandstrand et al (1992)). A decentralized R&D allows for
responsiveness to local differences. These are the demand oriented motives for
decentralization of R&D, where it is important to be close to "lead users" and adapt products
and processes to local conditions, often related to host market regulations. Supply oriented
motives for R&D decentralization relate to acquiring access to a wider range of scientific and
technological skills when technology sourcing (Kogut and Chang (1991)). A centralized R&D
function allows to capitalize on economies of scale when pooling R&D resources. The result

treatment of the issue of decentralizing R&D to the subsidiary level with consequent internal know-how

4
is internal transfers of know-how from central R&D labs to subsidiaries (Teece (1976)). In
addition centralization allows to better control R&D, minimizing leaking of information to
(potential) competitors.
Statistical evidence and survey results on R&D internationalization suggest that most
research still remains at corporate headquarters (e.g. Patel & Pavitt (1992)). But the
percentage of R&D carried out abroad is increasing rapidly (Grandstrand et al (1992), Caves
(1996), Serapio & Dalton (1999), Reger (2001)).
If technology sourcing is an (increasingly more) important motive for decentralizing
R&D this should be reflected in technology transfers from local sources to foreign
subsidiaries. Most of the empirical literature on technology spillovers uses patent
information to trace know-how flows. Almeida (1996) using US patent citations counts on a
sample of foreign subsidiaries in the US semiconductor industry, finds foreign subsidiaries to
cite regionally located firms significantly more. Frost (1998) finds proximity matters since
foreign subsidiaries cited other entities located in the same state more frequently. Also
Branstetter (2000) found Japanese firms investing in the US to have a significantly higher
probability of citing other US firms' patents.
But foreign subsidiaries not only acquire local know-how, they are also sources of
knowledge spillovers to the local economy. The empirical evidence on spillover benefits to
the local economy from FDI at the aggregate level, have generally failed to find robust
evidence of positive knowledge spillovers from multinational investment (see Blomstrom &
Kokko (1998), Mohnen (2001) for a review). Turning to firm level evidence for spillovers
from foreign subsidiaries to the local economy, Almeida (1996) finds that patents belonging
to foreign firms investing in the US are cited more by local US firms than other foreign firms.
Also Branstetter (2000) finds a higher probability of US firms citing Japanese firms when
they invest in the US.
The choice between centralizing and decentralizing R&D also has implications on the
internal know-how flows between parents and subsidiaries. While for centrally developed
innovations, know-how flows from parents to subsidiaries, when subsidiaries are assigned a
role in accessing and developing local specific know-how, this know-how needs to flow to
corporate level. Recent studies can more easily provide evidence for the transfers of know-
how from parents to affiliates, but find less conclusive support for the reverse direction, from
subsidiaries to headquarters. Frost (1998), using USPTO data for 1980-1990, found evidence
for the importance of headquarter patents for the innovations of subsidiaries, while patent data
provided only limited evidence for the transfer of know-how from subsidiaries to
headquarters.

transfers.

5
 Typically most empirical studies who trace know-how flows rely on patent citations.
But a vast amount of information is transferred without writing it down in patent application
or even in formal contracts, certainly in case of internal transfers. Using survey based
evidence to directly assess the occurrence of internal technology transfers between parents
and subsidiaries as well as between subsidiaries and other external local partners, Mansfield
& Romeo (1980» found that two third of UK firms indicated that their technological
capabilities were raised by technology transfers from US firms to their overseas subsidiaries.
But only 20% felt this effect was of importance. Veugelers & Cassiman (2002) using survey
data from a sample of Belgian innovation active manufacturing firms similarly found
evidence for not only technology transfers from parents to subsidiaries and from subsidiary to
external local partners, but also for the reciprocal flows of know-how.

3. The Model Set-up

The empirical evidence seems to suggest that the internationalization of R&D and
subsidiaries as technology sources have become an important trend. The model follows the
empirical evidence on MNE innovative strategies, with subsidiaries located in countries with
an own local know-how base and reciprocal intra- and inter-firm knowledge transfers. We
consider two countries (country I and 11). Country I is the home base of a MNE (firm 1)
which is a monopolist in the home market and controls a production subsidiary in country IT,
where also a local producer (firm 2) operates. The MNE has to decide whether or not to
decentralize its R&D activities to the subsidiary.
Both internal and external flows depend on whether the MNE decides to decentralize
or not. When the MNE's foreign production plant is active in R&D, it faces reciprocal
external know-how flows with the local competition, i.e. the decentralized R&D unit will be
able to absorb know-how from the local market. But at the same time the local competition
can absorb know-how from the decentralized R&D unit. As Figure 1 illustrates, in case of
centralization there are no external flows, i.e. there is no spilling over of MNE know-how to
local firms, but likewise no know-how can be sourced by the MNE. With respect to internal
flows, these remain unidirectional from parent to subsidiary in case of centralization, while
decentralization implies a bi-directional internal flow.
Insert Fig 1 here
The decentralization decision of multinational R&D is studied in a two stage game. In
the first stage firm 1 (the MNE) undertakes its R&D location choice taking into account how
this decision affects its own and its rival output decision. In the second stage the subsidiary
and the local producer -competing a la Cournot-, decide simultaneously how much to

6
produce and selI in country II, while the parent chooses as a monopolist the output to be sold
in country I.

3.1. Own R&D resources and intemal & extemal R&D spillovers

Both the MNE and the local firm are engaged in product innovation. Taking a short run
perspective, we assume that the total amount of resources devoted to R&D by each firm is
fixed. Thus with xm and xl we indicate the given level of own R&D resources respectively
from the MNE and the local producer. While the local firm produces and innovates only in
its home market, the MNE must decide where or not to decentralize its R&D activities.
At: Ad -
xm =X m =X m (I)

(2)

where superscript c represents R&D centralization and d R&D decentralization. Note that not
only R&D resources are fixed for each firm, but also that the total R&D resources are
assumed to be the same in case of centralization or decentralization.
Although the MNE's R&D resources are fixed at the corporate level, the R&D
resources individually available to the parent and the subsidiary varies according to the

MNE's R&D location decision. The MNE can locate alI of its R&D resources xm in country

I. This is the case of centralization. Alternatively, it can locate a share a of its total R&D
resources in country II assigning an innovative task to the subsidiary, which is the case of
R&D decentralization. Thus the own R&D resources of the parent and the subsidiary in the
case of R&D centralization are given by:

x~ =xm (3c)

x~' =0 (4c)

while in the case of R&D decentralization:

x% = (1-a)xm (3d)

x~ = dim (4d)

The total effective know-how, which each plant can use for product innovation, is not
only composed of own R&D resources, but also includes R&D resources of other plants
within the same firm or from other firms, at least to the extent that these resources spill over
across firm and country boundaries. As to internal knowledge transfer between subsidiary
and parent, the know-how generated by the MNE in each market is transferred to the other
unit. The parameter f3IP indicates the share of know-how produced by the parent which is

transferred to the subsidiary. Reciprocally, the parameter f3 1.\. indicates the share of know-

7
how produced by the subsidiary which is transferred to the parent (see Fig. 1). These internal
transfers are imperfect, not only because of the costs associated with transferring know-how
but also because of the need to adapt transferred know-how. We have

13 1 ~ 1 (Si)

In the case of f3/p, a parameter value below 1 reflects the classic cost of adapting the
centrally developed knowledge in the home lab to host market conditions. These costs arise
from the fact that the products and processes developed by the parent need to be modified to
satisfy requirements in the host country. The more dissimilar the home and the foreign
market, the larger the need for adaptation, i.e. the smaller will be f3 II' •
As to external knowledge transfer between the MNE and the local competitor, we
assume that there is knowledge dissemination only if there is R&D proximity (cf
agglomeration literature). This implies that only when the MNE decides to decentralize its
R&D, there will be external spillovers with the local competition. These spillovers are two-
way. On the one hand will decentralization create the possibility to source local know-how.

These are the incoming spillovers f3 Xl. On the other hand, locating R&D resources to the

local market open up these resources for spillovers to the local competitors. These are the

outgoing spillovers 13 Xs .

iff a>O. (Sii)

The assumption of localized spillovers furthermore implies that even if a > 0, there is no
involuntary transmission to the local firm of the knowledge generated by the parent in

country I, that is for x~ = (1- a)xm • Only the decentralized R&D resources are spillover-
prone. This implies that the MNE can influence the flow to external local producers through
its decentralization decision.
In addition we account for the fact that the extent to which external spillovers are
integrated in the own knowledge base depends on the absorption capacity of the receiver. 2

The own R&D resources serve to develop the absorptive capacity of the firm. Thus we have
that the external spillovers received from the local firm by the subsidiary is given by

(13 Xl dim )x/. This implies that the MNE can influence the importance of incoming external

spillovers through the amount of R&D resources which are decentralized. Similarly, the

external spillovers received by the local firm is given by (13 XSXt )dim . This implies that the
stronger the R&D base of the local competitor, the more important the outgoing external

2 Note that we only consider absorptive capacity for external spillovers, but ignore it for internal spillovers.

8
spillover effects will be for the local firm. We impose the restriction that fJ XI dim ~I and

fJ x'xi ~ 1.

Given the previous assumptions, we can now characterize the effective know-how
base for each company. This effective know-how base, being a combination of own R&D
and of the know-how obtained through internal and external spillovers, represents the amount
the firm would have had to invest in research in the absence of spillovers to obtain the same
research output. The effective know-how bases of the multinational parent, of the subsidiary
and of the local firm in case of R&D centralization are:

x~ =xm (6c)

X,~ =f3 IP xm (7c)

(8c)

In case of R&D decentralization, the effective know-how bases are :

Xd ,,= (I - a )-Xm + fJIs ax- m + fJIs (fJ Xl mm

- )-
xl (6d)
X~ = dim + f3Ip (l-a)xm + (f3 Xldim)Xl (7d)

xf =Xl + (f3 Xsxl)dim (8d)

3.2. Market competition

We now turn to characterizing the second stage, which is the market competition stage.
Linear demand functions are considered in both markets, which are perfectly segmented.
While the parent firm is a monopolist in its home market, in the local market its subsidiary
competes a la Cournot with the local competitor. Since product innovation is examined, the
position of the respective demand curve for each producer depends on its effective know how
base. We thus have:

(9)

(10)

(11)

with k=c, d. The parameter b i (b n) is inversely related to market size in country I (country II).
The parameter <p captures product differentiation. The higher <p the less differentiated the
goods produced by the subsidiary and the local firm and thus the more intense is product
market competition in country II.

9
3.3. Firm profits

The MNE profits are given by the sum of profits gained in the two markets. Thus in
the case of R&D centralization:

fIem = n ep + n se (12c)

fI e[ =n[e (13c)

(14c)

(15c)

(16c)

if R&D decentralization:

(12d)

(13d)

(14d)

(15d)

(16d)

The negative _I.CaXm )2 term in the MNE profitability in case of R&D

2
decentralization (12d) is introduced to capture the presence of economies of scale in R&D,
which according to the literature represent a major centripetal force. If R&D is decentralized
these economies of scale are not fully exploited, implying an increase in R&D expenditure for
the same amount of resources devoted to R&D.

4. Main results

4.1 Location factors influencing where to establish a foreign R&D lab

In this section we will characterise what drives the profitability of the MNE in case of R&D
decentralization. Identifying the subsidiary specific profitability drivers in case of
decentralization, will allow us to highlight which are the main location factors influencing the

10
decision of the MNE to undertake R&D activities in a specific foreign market. We focus the
discussion on local market size or cost conditions and the local know-how base as location
factors.
By solving for optimal output in the case of decentralization, respectively for the
parent in country I and the subsidiary and the local firm in country II, we obtain:

-d M p (1- ar~'m 13 Is ciX

m 13 Is (13 Xl ciX
m )"Xl
q = -- + +- - - + -'----"---"'-'-'- (17)
P 2b l 2b l 2b l 2b l

Ad M, 2f3 IP{I)-
-a Xm (2f3XI ax- m -cP )-
Xl
(7--cPf3XS-)-
Xl ax m
q., = (4-cp ~ )bl/ + 2
(4-cp )bl/
+
(4-cp )b ll
2 + 2
(4-cp )b ll
(18)

and Ms =(2-cp)A ll -2c, +cp c, represent the initial (no

innovation) demand-cost margin. When discussing plant level profits, we can focus our
attention on these equations since the profitability of each unit (parent, subsidiary) is
increasing in its own equilibrium output 3.

The main focus is the impact on subsidiary profitability. Expression (18) shows that,
as expected, the subsidiary output, and hence profitability, increases with host country market
size (captured by Ms and by the demand slope 11b1/)' This represents a first important
location factor. The second term represents the extent to which central R&D resources
(1- a)xm can still be deployed by the subsidiary, even in case of decentralization, but only

imperfectly given imperfect internal transfers f3 11' :0; I. The better able the subsidiary is in
directly using the central R&D resources, the higher the subsidiary profitability. Hence,
closer matched locations requiring less adaptation of central know-how are more interesting
location sites, ceteris paribus.
Less evident as location factor is the role of local know-how (xl) and incoming and

outgoing external spillovers to which we now turn. The interesting aspect brought about by
expression (18) is the way in which these knowledge transfers interact with product market
competition.
From a technology sourcing perspective, the size of the local know-how base (Xl)

serves as an important location factor. The model shows that the local rival's own R&D
(X,) affects subsidiary profitability in three ways, two of which are captured by the third

term in Eq. (18). The first effect is due to the incoming external technological spillovers

enjoyed by the subsidiary when undertaking innovative activity in loco «13 Xl ciXm )Xl ). This

11
effect is positive. But the extent to which q.~ is increasing in xl due to this effect depends on
the subsidiary absorption capacity, which depends on the own R&D resources decentralized
to the subsidiary.
The second effect of local know-how Xl on subsidiary profitability comes via product

market competition and is negative. A higher level of local producer R&D resources, xl will
make the local competitor stronger in the product market which will have a negative impact
on the subsidiary output level. This impact is stronger the more intense the competition is in
the product market between the two producers (i.e. the higher is q».
Thirdly, xl increase the local firm ability to capture the outgoing external spillovers
generated by the subsidiary. This is captured by the last term in Eq. (18). Via this route, the

local know-how base has a negative impact on q1 and hence on affiliate profitability. A local
firm with a stronger R&D base will be better able to absorb know-how from the subsidiary
thus becoming a stronger competitor. This effect depends on product market competition (as

shown by the fact that the outgoing external spillover /3 xSxlciim is multiplied by q». For

instance, if q> = 0, /3 Xs has no effect on q1 since in this case the subsidiary is not affected

negatively if its know-how is disseminated to the local economy.

The necessary and sufficient condition for q.~ to be increasing in the local firm's

R&D (Xt ) is thus :

(2/3 Xl - q>/3 Xs )dXm > q> . (19)

which shows that the probability that the condition holds, increases with product
differentiation (lower q», and with the difference in intensity between incoming and outgoing
external spillovers. Hence we compare the net effect via external spillovers with the
competition effect. Eq.(19) is the condition that needs to hold for the R&D investment by the
local firm (Xt ) to lead to an increase in the subsidiary profitability.

The size of the local know-how base Xl affects also the profitability of the parent

company. Expression (17) allow us to discuss the impact of Xl on the parent plant when

decentralizing R&D to the subsidiary. The last term in Eq. (17) shows that, in the case of
decentralization, the parent plant profits from the incoming external spillovers when

allocating R&D resources in country II. In fact fi% (and thus parent's profitability) is

increasing in x, (the amount of R&D undertaken by the local producer) at least to the extent

12
that f3 Xl (external spillovers of know-how from the local producer to the subsidiary) exists.
Furthermore, since the absorption capacity of each unit is firm-specific, the amount of
potential spillovers appropriated by the subsidiary and then passed to the parent is increasing

in the subsidiary'S own R&D (aX m ). Hence the more R&D is decentralized to the subsidiary

the larger the positive effect from sourcing local spillovers on the parent plant profits.
However the benefits for the parent from the learning which is associated to locating R&D
activities abroad depends also on the MNE's ability to transfer this knowledge internally

from the subsidiary to other units of the MNE (i.e. on f3ls). Thus the internal and external
transfer mechanisms interact in determining the final effect for the parent.

The last term in Eq. (18) accounts for the "own R&D effect", which is certainly

positive since f3 Xs Xl ::; 1. This term shows that the subsidiary output and hence profitability is

increasing in the amount of R&D undertaken in country II (aX l11 ), even when allowing for the
fact that part of this knowledge may leak to the local firm. As discussed supra, the size of the
outgoing external spillover is influenced by the absorption capacity of the local firm, i.e.

f3 xSXI' and the intensity of product market competition.

The positive effect of own R&D on q.~ is reinforced by the fact that the subsidiary
absorption capacity, and thus the amount of learning from the local producer, i.e. the external
incoming spillover, depends on aXm , as shown in the third term in Eq. (18). Hence own

R&D resources in the subsidiary plant also serve to enhance the learning effect from the local
economy.

The necessary and sufficient condition for q~ to be increasing in the subsidiary's

own R&D (aX m ) is

f3ls (f3 Xl xl + 1) > 1 (20)

4.2 The MNE decision to decentralize its R&D activities

4.2.1. Conditions for R&D decentralization

The solution of the first stage game allows to identify under which condition the
MNE will decide to decentralize its R&D activities, i.e. locate a share of its R&D activities in
country II assigning a role in its overall research effort to the subsidiary operating there.
The MNE will choose decentralization iff:

13
 (21)

Hence discussing the decentralization decision involves comparing both subsidiary and parent
profitability in case of decentralization versus centralization.
Let us recall that:

(22)
We consider first the effect of R&D decentralization on parent's and subsidiary's variable
profits separately, before analyzing the overall effect. This way of proceeding will allow us
to highlight how R&D decentralization affects the profitability of the different units of the
MNE.4

From Eq. (22), since output levels are nonnegative, qJ ~ 0 (with k=c,d and j=p,s,l),

we have that:

(23)

(24)

Thus the necessary and sufficient conditions for (q~ -q~»O and (q.~ -q.~»O represent

sufficient conditions for the MNE variable profits to increase when R&D is decentralized.

4.2.2. The impact of R&D decentralization on the parent's plant profits

We first discuss the effect of R&D decentralization on the parent plant's profitability.
As to the effect of R&D decentralization on the parent equilibrium output level (and thus on
parent's profitability), we have that:
.d.c (l-f3Is)aX m f3Isf3Xl aXmXt
q,,-q,,=- 2b + 2b (25)
I I
The first term captures the negative impact that R&D decentralization has on the
parent equilibrium output level. This effect arises from the short run nature of the problem
examined. Being in the short run, the total amount of resources devoted to R&D by the MNE

4 Disentangling parent and affiliate profits from the decentralization choice is interesting when considering
the bargaining process internally within the MNE. For instance if the decision to decentralize R&D will
only be taken if at least each party benefits, this implies that (if~ - if~;) >0 and (if.~ - if.;') >0 need to
hold in addition to (21).

14
is given and consequently the choice of allocating R&D abroad implies lower R&D resources
at home. This has a negative repercussion on the parent equilibrium output and hence
profitability. This effect is at least partly compensated by the fact that the subsidiary transfers

the know-how it creates back to the parent. This transfer is however imperfect with I3i.1" :::; 1 .

The negative effect is thus mitigated by internal transfers f3 ls , and hence depends on the
ability of the subsidiary to transfer know-how to the central level. Only in the extreme case of

perfect internal transfers f3ls = 1 this negative term disappears.

The second term of (25) captures the beneficial effect of R&D decentralization on the
parent's output level from being able to gain access to the foreign pool of potential spillovers
generated by local producers, cf supra. The benefit for the parent however again depends on

the value of f3ls. Thus the ability of the subsidiary to transfer back knowledge to the parent

acts as a filter. Whether the parent benefits from the incoming external spillovers also
depends on the absorption capacity of the subsidiary. Since the subsidiary absorption
capacity depends on its own R&D (given by dXm ), both the positive and the negative effects
of R&D decentralization on parent's profitability are rising in the amount of R&D resources
allocated to the subsidiary.

The necessary and sufficient condition for q~ - q~ > 0 and hence for parent profits

to be larger in case of decentralization, i.e. if;~ - if~ > 0, is condition (20):

f3!'>' (13 Xl Xl + 1) > 1. This condition clearly indicates that the ability of the subsidiary to

channel back know-how is a crucial determinant of the effect of R&D decentralization on

parent's profitability. It stresses the crucial role of internal knowledge management within the
MNE. The condition also brings to attention the importance of the interaction between
internal and external knowledge transfer mechanisms. The larger the incoming external

spillovers, the more likely that (20) will hold. Although the sign of (q~ - q~), does not

depend on the amount of R&D resources decentralized, the magnitude of the positive
(negative) effect is increasing in aXm and the size of country 1.

4.2.3. The impact of R&D decentralization on the subsidiary's profits

The effect of R&D decentralization on the subsidiary output (and thus the impact on
subsidiary profitability) depends upon:

(26)

15
 The first term in Eq. (26) is connected to the adaptation motive for R&D
decentralization, i.e. with the demand related motives for establishing a foreign R&D lab.
When the MNE allocates R&D resources to the host country (dim) instead of devoting them

to the parent lab, the foreign lab's innovative effort is tailored to satisfy local needs and
benefits from proximity with local production. Thus the subsidiary can avoid the adaptation
costs that it would have to incur if the MNE had chosen to centralize all R&D in the home

country. The lower f3 Ip , the smaller the share of the knowledge generated by the parent
which is of use in the host country, and thus the greater the benefits of localizing R&D
where there is production. That is why the benefits of undertaking dim R&D in loco is

weighted by the term (1 - f3 Ip) which can be considered as representing the unit cost of
adapting to local conditions the know-how transferred by the parent to the subsidiary. If

f3 Ip = 1 the first term in Eq. (26) vanishes, showing that in the extreme case of perfect
internal knowledge transfer from the parent to the subsidiary (and thus in the absence of
adaptation costs) there is no incentive to decentralize R&D due to this motive.
The second term in Eq. (26) reflects the supply related motives for R&D
decentralization, connected with the learning motive. It captures the effect of incoming
external spillovers which arise because of the proximity between the subsidiary lab and the
local producer lab. By decentralizing R&D, the MNE becomes able to absorb from the local
firm, benefiting from incoming external spillovers. Eq. (26) shows that the positive effect of
incoming external spillovers on the subsidiary profitability is not affected by product market
competition, i.e. connected with <p and thus does not work via product market competition.
However there are also dangers associated to localizing R&D resources close to local
competitors. These costs are represented by the third term in Eq. (26) capturing the effect of
the outgoing external spillovers. Due to lab proximity, at least part of the know-how created
by the subsidiary will leak to the local producer. The dissemination of the subsidiary's own
R&D to the local firm has a negative impact on the subsidiary profitability since it increases
the local firm competitiveness in the product market. Thus the outgoing external spillovers
sort their effect via product market competition. That is why the intensity of the negative
impact of the outgoing external spillovers depends on cp (the product differentiation
parameter). The higher product differentiation and thus the less intense product market
competition (the lower cp), the lower the costs of R&D proximity. The extent to which the
local producer can benefit from these spillovers depends on its absorptive capacity which in
tum is determined by its own R&D resources XI' Thus the stronger the know-how base of the

local competitor, the larger the negative impact on subsidiary profits of the outgoing external
spillovers.

16
 The necessary and sufficient condition for q.~ - q.~ > 0 (and hence for subsidiary

profits to be higher in case of R&D decentralization, i.e. for it,~ - it.~ > 0) becomes:

(27)
from which we have

(28)

If the external spillover parameter is symmetric (13 Xl = f3 x,), condition (28) always

holds since <p <2 and f3lp :51 . Thus if the intensity of external technological spillovers (i.e.

13 x) is sector/technology specific, the subsidiary eqUilibrium output (and profits) will

increase when R&D is decentralized for any value of a, 13 11', and <po This is the case since
the positive effect of the incoming external spillovers is direct, with incoming external
spillovers a pure externality, while the negative effect of the outgoing external spillovers is

mediated via competition in the product market. On the other hand, if (f3 Xl = f3 XS),

condition (20) required for parent's profitability to rise, does not necessarily holdS. Thus the
model suggests that R&D decentralization is more likely to result in higher profitability for
the subsidiary than for the parent unit.

If the external spillover parameter is asymmetric (f3 XI '* f3 x"), we obtain from Eq.

(28) that the sufficient condition for q~ - q.~ > 0 (and for subsidiary profits to be higher in

case of R&D decentralization, i.e. for it: - it.~ > 0) becomes:

(29)

Note that condition (29) is riot overly restrictive and allows for the outgoing spillover
parameter to be larger than the incoming spillover parameter.
Another special case is when the local firms are no direct competitors, for instance
when they would be research institutes or firms with related technologies but which are
unrelated in the product market. If <p=O, the negative effect of the outgoing external spillovers
vanishes. This means we would have in this case the subsidiary to profit from decentralization

( it: - it; > 0 ) for all values of the other parameters.

Overall, it is likely that the decision to assign an innovative role to the subsidiary will
result in an increase of the latter profitability, which explains the subsidiaries' call on part of
the R&D resources of the MNE.

5 If f3 Xl = f3 Xs this implies that since f3 Xl Xl :51 , condition (20) does not hold for f3 J•• ~0.5.

17
 4.2.4. The impact on the MNE's total profits

To assess the overall effect on the MNE profitability however requires that due
consideration should be given not only to the impact on the subsidiary profitability but also to
the effects on parent profitability and to the role of the additional R&D costs due to foregone
economies of scale. In total, we have six effects that form the benefits and the costs from
R&D decentralization:

• (B.I) avoidance of adaptation of central innovations by the subsidiary (first term in (26»
• (B.2) benefits from the incoming external spillovers to the subsidiary (second term in (26»
• (B.3) benefits from the incoming external spillovers to the parent (second term in (25»

• (C.I) the foregone economies of scale in R&D (last term in (22»

• (C.2) the impact on the subsidiary from outgoing external spillovers (third term in (26»
• (C.3) the loss of R&D resources for the parent plant which are decentralized to the subsidiary but
not fully recoverable due to imperfect internal transfers (first term in (25»

In order to fully solve the first stage game (i.e. the R&D decentralization choice) we
need to compare the total MNE's profits corresponding to each of the two potential states
(decentralization/centralization) (Eq. (22)). The complexity of the expressions for equilibrium
profits makes it difficult to perform analytical comparisons to fully characterize the
equilibrium choices in general. Nevertheless, we can discuss the factors driving costs and
benefits of decentralization and characterize the outcome for some special cases. Hence,

rather than evaluating the conditions required for fI~ - fI~, > 0, i.e. for decentralization of
R&D to be the solution of the game, we will concentrate mostly on discussing factors that can
promote R&D decentralization by analysing the sign of the partial derivative of exogeneous
factors on the overall net profits from R&D decentralization.

4.3 Factors affecting the impact of R&D decentralization on total MNE's

profits

Of particular interest to consider as factors influencing the R&D decentralization

decision by the MNE are the local know-how base and the mechanisms to transfer know-how
internally and externally.

4.3.1. Local know-how base

18
 A first important factor affecting fI~, - fI~, and hence the R&D decentralization

decision, is the local know-how base XI' From supra we know that a strong local know-how

base increases the benefits from incoming external spillovers both for the subsidiary and the
parent (B.2 and B.3). But at the same time it enlarges the cost of outgoing external spillovers
(C.2) since the local rival will have a stronger absorptive capacity. Already at the subsidiary
level, the net effect of a local know-how base can be positive (i.e. C.2 < B.2). For this Eq.
(19) was the necessary and sufficient condition, cf supra. If we consider the overall effect of
the local know-how base on the incentives to decentralize, Eq.(19) is not only a necessary and
sufficient condition to increase subsidiary profitability, cf supra, it is also a sufficient
condition for the local R&D base XI to act as a driver for R&D decentralization (Le. for

o(II d _ II c )
m m >O),as shown in expression (31):
ax[

Eq. (31) is not always positive, which would imply that a local know-how base is not
a univocally positive factor for R&D decentralization This would happen when the cost of
outgoing external spillovers becomes very important. This cost would start to dominate when
competition is strong, i.e. <p large, and the learning is asymmetric, i.e. 13 x, is high, while 13 XI
is low. 6

4.3.2. Internal transfer of know-how

Another important factor determining the size of both benefits and costs to
decentralization is the process of internal transfer of know-how within the MNE. We have
that

(32)

Eq. (32) shows that the incentive to decentralize R&D is increasing in 13 1." • A better
internal transfer of know-how from subsidiary to parent results in lower costs of
decentralization (lower C.3) and in larger benefits from incoming external spillovers to the

6 For instance. it can easily be checked that for the extreme case when ( 13 x,' XI )=1. 13 XI =0. and <p>0.
(31) will indeed be negative if 13 11' <0.5. If f3lp is large. we require for (31) to be negative that qd, is large
enough. Eg P
f3l =1 requires qd, >113.

19
parent (higher B.3). For instance, perfect internal transfers from the subsidiary to the
parent, 13 1., =1, will eliminate the parent plant's adaption cost (C.3).
A better internal transfer from the parent to the subsidiary does not always discourage

decentralization? In fact f3ll' has an ambigous effect on the incentive to invest abroad in
R&D. We have that

a (fr~ - fr~,) (33)

af3 ll'

from which

. a(rr~af3-rr~,)
sIgn lP
sign{(l-a);i.~
~
_q.~) (34)

which implies that, if the subsidiary's output in case of decentralization is large enough,
a more efficient internal transfers from the parent may actually act as incentive for R&D
decentralization

4.3.3. External transfer of know-how

Since both parent and subsidiary enjoy higher profits in case of decentralization when
there are more spillovers from the local source (see (B.3) & (B.2», we have

8 (fI~ -fr~)
(35)
813 XI

This means that higher spillovers from the local source to the MNE univocally acts as
an incentivator for R&D decentralization. But of course: since subsidiary profits will
decrease with higher spillovers to the local source (see C.2), we have at the same time:

Cl (fr~ - fr~) _ xr
[2aXm l d <0
(36)
Clf3xS qJ (4_qJ2) q.I

Expression (36) implies that being able to prevent spillovers to the local source
univocally improves the case for R&D decentralization. However since the negative role of
13 X.,. depends on cp, Eq.(36) suggests that investing in knowledge protection measures for

lowering 13 x., (such as employee conduct rules or measures to reduce knowledge-worker

exits) is especially important for firms facing close competitors in production.

? Perfect internal transfers from the parent, f3lP=1, eliminate the subsidiary's adaption motive for
decentralization (B.1), leaving only the incoming external spillovers as drivers for R&D decentralization.

20
 If the external spillover parameter is symmetric (f3 XI = f3 x., ), the total effect will be
positive since already at the subsidiary level B.2. > C.2, cf supra.

4.3.4. Cross effects

Given the importance of both internal and external knowledge transfers, it is
interesting to discuss cross effects. It can easily be shown that 8:

a2(TI~, - TI~,) > 0 (37)

a/Flaw"'
Expression (37) implies that there is a complementarity between internal and external
know-how transfers: a better internal know-how transfer process within the MNE, increases
the efficiency of mechanisms used to acquire external know-how and vice versa. In other
words, the role of the spillovers generated by local innovators as a centrifugal factor to induce
a MNE to locate R&D in that country rests also on the ability of the MNE's subsidiary to
transfer internally knowledge across geographic boundaries and vice versa.
The cross effects with respect to XI similarly allow to discuss which complementary

forces the MNE can use to maximize the positive impact from local know-how sourcing on
overall MNE profitability. It is quite easy to show that
a\TI d _TIt")
m m >0. (38)
axlaf3 I'. ,
and when condition (19) holds
a 2 (TId _TIt")
01 m >0. (39)
axl af3 xl '

a 2 (TId _TIt")
m 01 <0. (40)
axl af3 x., '

When the MNE locates R&D abroad, the MNE is much more likely to benefit from
the technology sourcing advantage from decentralization when it has an efficient internal

know-how transfer process (high f31.v ); when it has an efficient process to absorb externally
available know-how (high f3 XI); and when it can prevent know-how from spilling over to

competitors all to easily (low f3 x.. ). These results illustrate again the complementarity

between an efficient knowledge management system and the technology sourcing motive for
R&D decentralization.
A final important factor to consider as complementary force in technology sourcing is
the amount of R&D resources allocated to the subsidiary dim' serving as absorption capacity

21
for external know-how acquisition, but at the same time opening up the possibility of
appropriation by local competition.
While condition (27) is sufficient for

(41)

conditions (20) and (27) are sufficient for

a (TIdno -TI")
2
m >0' (42)
af3 xl aaxno '

a2(TId m
-TI")
m >0; (43)
af3 I.' aaxm
This implies that having more R&D resources located at the subsidiary level
increases the extent to which the MNE at the corporate level can benefit from local learning
as well as internal know-how transfers from the subsidiary when decentralizing R&D.
Equivalently, a higher external spillover level from local sources (high 13 XI) and a better

ability to use subsidiary know-how at corporate level (high f3ls ) will push the MNE, when
deciding on the optimal amount of decentralizing R&D resources, to allocate more resources
to the subsidiary level. On the contrary, a high level of outgoing external spillovers, i.e. low
level of appropriability, (high 13 x.,), will lead the MNE to allocate less R&D resources to the
subsidiary .
Finally, condition (19) together with (20) and (27) are sufficient for
a2 (TIdm _TIC)
m >0' (44)
aXI aaxm '

Since conditions (19), (20) & (27) are more likely to hold with a high 13 XI and a low
<p and/or low 13 x." this implies that a sufficiently large incoming external spillover level

while having weak competition or high enough appropriability of subsidiary know-how are
sufficient conditions for the size of the local know-how base to act as a stimulus for allocating
more R&D resources to the subsidiary level.

5. Conclusions
This paper provides a theoretical model on the trade-offs which a MNE faces when
assigning subsidiaries an active role in innovation and organizing its R&D decentralized.
R&D decentralization avoids having to adapt centrally developed innovations to local

8 The full expression is not reported for sake of space.

22
markets. In addition R&D subsidiaries can be used to source locally available external know-
how. But the MNE has to organize the transfer of local know-how internally so as to be able
to benefit from this location specific know-how throughout the organization. At the same
time, decentralization of R&D to the subsidiary level opens up the possible spilling over of
valuable know-how to competitors located in the foreign markets.
The proposed model focuses on how the interplay of internal and external
knowledge flows interacts with the nature of host market competition to influence the choice
of MNEs to effectively disperse internationally its R&D. The intensity of competition in the
local market emerges as important in determining the size of both benefits and costs to R&D
decentralization. It is especially significant in determining the outgoing spillover costs. In the
absence of local competitors in production, the subsidiary will always profit from R&D
decentralization in our model. But even if there is local competition to worry about, the cost
from outgoing external spillovers is outweigh ted by the benefit from incoming external
spillovers at the subsidiary level, at least when external spillovers are symmetric.
In addition, the model indicates that a strong local know-how base is not necessarily
a motive for R&D decentralization. While it increases the benefits from incoming external
spillovers both for the subsidiary and the parent, at the same time it enlarges the cost of
outgoing external spillovers since the local rival will have a stronger absorptive capacity.
This cost could start to dominate when competition is strong and the external spillovers are
asymmetric, sufficiently in disfavor of the MNE.
We also find that a more efficient internal know-how transfer process within the
MNE from the subsidiary to the parent univocally acts to promote R&D decentralization. It
will increase the benefits from incoming external spillovers to the parent. But on the other
hand, a more efficient transfer of know-how from the parent to the subsidiary makes the
motive for avoiding adaptation by the subsidiary less prevailing. Nevertheless it does not
always discourage the MNE from investing abroad in R&D. A better internal know-how
transfer process within the multinational increases the efficiency of mechanisms used to
source external know-how and vice versa. Hence, the results clearly illustrate the
complementarity between an efficient internal and external knowledge management system
and the technology sourcing motive for R&D decentralization. Another important
complementary force increasing the efficiency of technology sourcing is the decentralized
know-how base, which serves to absorb local know-how.
While the model allows to discuss the forces driving costs and benefits of R&D
decentralization within multinational firms, a full characterization of the decentralization
choice of the multinational requires numerical simulations, given the complexity of the
setting. Our future research will move towards obtaining predictions from the model which
can be tested against data on R&D decentralization. Extending the model, such as to allow

23
for endogeneous R&D resources, or local competitors reciprocally locating R&D abroad,
would allow for model results which are closer to most empirical settings.

24
 Parent
Firm
x dp=(1-a)x m

INTERNAL
TRANS
FERS
i
- ~I'

~XI

Local
Subsidiary EXTERNAL
Firm
xd~=axm TRANSFERS
XI

W'

Fig IA: Internal and External Knowledge spillovers with Decentralized R&D

25
 i ..
i •
i -.
i lUI
Parent
i_
i c»
Firm i.,.
xCp=x m i ....
i~
i~
i -.
INTERNAL
TRANS .-.-.-.-.-.-.-.-.-l-~.-._._.~~~_~~~~~_
FERS

Local
Subsidiary Firm
XI

Figure IB: Knowledge Transfers with Centralized R&D

26
References

Almeida, P., 1996, Knowledge Sourcing by Foreign MNEs: Patent Citation Analysis in the
US Semiconductor Industry, Strategic Management Journal, 155-165.
Audretsch, D. and M. Feldman, 1996, R&D spillovers and the geography of innovation and
production, American Economic Review, 86, 630-640.
Blomstrom, M. and Kokko, A., 1998, Multinational Corporations and Spillovers, Journal of
Economic Surveys, 12,3,247-277.
Brandstetter, L., 2000, Is FDI a channel of Knowledge Spillovers: Evidence from Japanese
FDI in US, NBER WP 8015.
Cantwell, J .A., 1995, The globalization of technology: what remains of the product cycle
model? Cambridge Journal of Economics, 19, 155-174.
Caves, R., 1996, Multinational Enterprise and Economic Analysis, Cambridge University
Press, Cambridge.
Cohen W. and D. Levinthal, 1989, Innovation and learning: the two faces of R&D, The
Economic Journal, 99, 569-596.
Das, S., 1987; Externalities, and technology transfer through multinational corporations,
Journal of International Economics, 22, 171-182.
De Bondt, R., 1996, Spillovers and innovative activities, International Journal of Industrial
Organisation, 15, 1-28.
Dunning, J., 1988, The Eclectic Paradigm of International Production: a restatement and some
possible extensions, Journal ofInternational Business Studies, 19, 1-31.
Frost, A., 1998, The geographic sources of innovation in the multinational enterprise: US
subsidiaries and host country spillovers, 1980-1990, PhD Sloan School of Management,
MIT.
Gersbach, H. and A. Schmutzler, 1999, External spillovers, internal spillovers and the
geography of production and innovation, Regional Science and Urban Economics, 29,
679-696.
Gersbach, H. and A. Schmutzler, 2000, Foreign Direct Investment, R&D locations and
Welfare, mimeo University of Zurich.
Grandstrand, 0., L. Hakanson, S. Sjolander, 1992, (Eds.) Technology Management and
International Business, Wiley & Sons.
Jaffe, A., M. Trajtenberg and R. Henderson, 1993, Geographic localization of knowledge
spillovers as evidenced by patent citations, Quarterly Journal of Economics, 577-598.
Kamien,M. and Zang, I. "Meet me halfway: research joint ventures and absorptive capacity."
International Journal of Industrial Organization, October 2000, 18(7), pp.995-1012.
Kogut, B. and S. Chang, 1991, Technological capabilities and Japanese foreign direct
investment in the US, Review of economics and Statistics, 401-413.
Kuemmerle, W., 1997, Building effective R&D capabilities abroad, Harvard Business
Review, 61-70.
Mansfield, E. and A. Romeo, 1980, Technology transfer to overseas subsidiaries by US based
firms, Quarterly Journal of Economics, 737-750.

27
Mohnen, P., 2001, International R&D Spillovers and economic growth, in M.Pohjola (Ed.)
Information technology, Productivity and Economic Growth, Oxford University Press,
2001.
Niosi, J., 1999, The internationalization of industrial R&D: from technology transfer to the
learning organization, Research Policy, 28,107-117.
Patel, P. and K. Pavitt, 1992, Large firms in the production of the world's technology: an
important case of non-globalization, in O. Grandstrand, L. Hakanson, S. Sjolander (Eds.)
Technology Management and International Business, Wiley & Sons, 53-74.
Petit, M.L. and F. Sanna-Randaccio, 2000, Endogenous R&D and foreign direct investment in
international oligopolies, International Journal of Industrial Organization, 18,339-367.
Reger, G., 2001, Differences in the Internationalization of Research and Technology between
Western European, Japanese and North American Companies, mimeo, Univ of
Brandenburg.
Serapio, M. and D. Dalton, 1999, Globalization of industrial R&D: an examination ofFDI in
R&D in the US, Research Policy, 28, 303-316.
Teece, D., 1976, The Multinational Corporation cost of International Technology Transfer,
Cambridge, MA.
Veugelers, R. & B. Cassiman, 2002, Innovative strategies and know-how flows in
international companies: some evidence from Belgian manufacturing, R. Lipsey &
Muchielli, J.L.(eds.), Multinational Firms and Impact on Employment, Trade &
Technology, London, Routledge.
Wang, J. and M. Blomstrom, 1992, Foreign investment and technology transfer: a simple
model, Journal of International Economics, 137-155.

28