Advances in Spatial Science - Editorial Board Manfred M. Fischer Geoffrey J.D. Hewings Phần 8 pptx

On the other hand, all other variables that indicate technology stemming from

either the MNE group or the subsidiary itself show evidence of the transformation

and exploitation of acquired knowledge into particular needs of the MNE and the

subsidiary.3

The model employed for RQ1 is the following:

RDLi ¼ b0 þ bjRAC þ bkPAC þ blROLE þ bmCV þ ei (12.1)

where RDL is the existence of a R&D laboratory, RAC stands for variables

measuring realized absorptive capacity, PAC for those measuring potential absorp￾tive capacity, ROLE identifies various subsidiary roles assigned by the MNE group

and CV for all control variables taken into consideration. In line with the cited

literature, we use industry’s technology intensity, mode of entry (new company or

joint venture), years of operation and region of origin (whether the MNE originates

from the EU, the USA or the Pacific Rim), as control variables.

For RQ2, the dependent variable is the ordered answer (from 4 to 1) of question

7c (R&D carried out by own laboratory), as the source of technology based on the

formulation discussed above. In particular, this RQ considers the second stage in

the developmental process of a subsidiary’s AC, (once it already runs an own R&D

laboratory), to check for factors affecting the intensity of its RAC. In this model we

also use measures of potential and realized AC that we used in RQ1. However, the

firm has now another element of RAC, namely, the scientific personnel hired to

equip the laboratory, thus we also include here the number of scientific personnel as

an extra variable of RAC.

The equation used for RQ2 is the following:

OWNRDi ¼ b0 þ bjRAC þ bkPAC þ blROLE þ bmSROLE þ bnei (12.2)

where the dependent variable is OWNRAD (the importance of sourcing the R&D

from own R&D lab as indicated in questionnaire response 7c). Once again, RAC

stands for variables measuring realized absorptive capacity, PAC for those measur￾ing potential absorptive capacity, ROLE identifies various subsidiary roles assigned

by the MNE group and CV for all control variables taken into consideration. In this

RQ we also include as explanatory variables the roles assigned to the existing R&D

labs. As control variables, we use industry’s technology intensity, the age of the

R&D lab (years of operation)4 and the region of origin.

The dependent variable employed for investigating the impact of PAC and RAC of

the subsidiary is the total turnover.5 In this stage, the R&D laboratory is in operation,

3

For a description of variables falling into either of the two categories, see Appendix 1.

4

As we examine the intensity of own RAC (own R&D lab), and unlike RQ1, the years of operation

of the subsidiary is not relevant, while the age of the R&D lab is.

5

A number of performance variables are plausible. Our focus on turnover from sales is in line with

the focus of the resource-based view (RBV), in particular Penrose’s view (see Pitelis 2002, for an

extensive discussion).

12 Multinational Enterprise and Subsidiaries’ Absorptive Capacity 267

thus, besides RAC belonging primarily to the MNE group, the subsidiary has further

enhanced its AC by developing its own research unit hence in addition to variables of

RAC and PAC used above, we hereby include the presence of an R&D laboratory.6

The equation used for RQ3 is the following:

PERFi ¼ b0 þ bjRAC þ bkPAC þ blROLE þ bnei (10.3)

where PERF stands for performance (the subsidiary’s total turnover) and the other

variables are previously explained.

Results

Each one of the three RQs was estimated by using three independent regression

models. The definition of the variables used in the tables below as well as selected

sample correlation matrices showing the strength of association between groups of

variables may be found in Appendix A. The results of conditional X2 tests that

examine the lack of independence among pairs of variables of interest are also

available on request.

RQ1:

Model 1: The impact of AC on the likelihood of establishing an R&D lab –

Table 12.1.

Our results show that the likelihood of establishing an R&D lab depends on prior

PAC of the subsidiary: the higher the dependence of the subsidiary is on R&D

carried out for it by local scientific institutions, thus the higher is its PAC the higher

the likelihood is of establishing an R&D lab (note that other measures of either PAC

or RAC do no enter significantly in the equation although it appears that the higher

the dependence of the subsidiary is on existing AC, the lower the likelihood of

establishing an R&D lab). It follows that PAC measured as the subsidiary’s

exposure to external knowledge, seems to enhance AC by inducing subsidiaries

to develop their own R&D lab in order to be able to transform acquired knowledge

to their own procedures and technologies adopted to their own needs, in line with

the fourth dimension of Zahra and George (2002).

Our results indicate that subsidiaries aiming at developing and producing new

products (WPM) and subsidiaries aiming at producing and exporting already exist￾ing products (SMR) are more likely to develop an R&D laboratory, as compared to

subsidiaries that target the internal (UK) market only (TMR).

As regards to the control variables, we find that the longer a subsidiary operates

in a particular location the more likely it is to create its own R&D unit. We also note

6

We do not include the number of scientific personnel here, because this belongs to the R&D lab,

so by including the existence of the laboratory by definition we account for the scientific personnel

engaged in the lab.

268 C. Kottaridi et al.

that new companies and joint ventures decrease the likelihood of establishing a lab

(if the method of establishing the subsidiary is by taking over an existing company

then the corresponding coefficient is positive, thus implying an increase in the

likelihood of establishing an R&D lab).

RQ2

Model 2: Assessing the impact of the type of an existing R&D lab on the

importance of the lab’s research as a source of technology for the subsidiary –

Table 12.2.

The importance of an established lab’s research as a source of technology for the

subsidiary significantly depends on the number of scientific personnel (RAC) while

the dependence of the subsidiary on internal to the MNE group technology lowers

the importance of the established R&D lab as a source of technology.

PAC as captured by the collaborations of the subsidiary with other firms

enhances the significance of an R&D lab as a source of technology.

With respect to the role of the subsidiary: the R&D lab appears to be of high

importance as a source of technology for subsidiaries that develop and produce new

products and the other way around for subsidiaries that produce and export inter￾mediate goods. Note that, as in Model 1, the impact from the role of the subsidiary

in developing and producing new products is higher than that of the other roles of

the firm (the coefficient of WPM is higher in absolute magnitude).

Table 12.1 Assessing the impact of AC on the likelihood of establishing an R&D lab

Dependent variable: LAB

Estimation method: ML – Binary logit

Observations used in estimation: 173

Robust std. errors from QML covariance

Variable Coefficient Std. error z-Statistic Prob.

C 5.6621*** 1.559341 3.631100 0.0003

EU 2.71805*** 0.925917 2.935529 0.0033

AM 2.24389** 0.950761 2.360101 0.01838

PAC 2.68776*** 0.968915 2.773986 0.0055

SDH 1.06039*** 0.393084 2.697620 0.0070

YO 0.02771*** 0.009201 3.012031 0.0026

NC 0.887073* 0.548129 1.618367 0.1056

JV 1.51331* 0.808497 1.871762 0.0612

TMR 0.49259** 0.225744 2.182062 0.0291

SMR 0.59033*** 0.231013 2.555379 0.0106

WPM 0.91869*** 0.240056 3.826997 0.0001

EXTT 0.83760** 0.416383 2.011615 0.0443

EXST 0.101017 0.292255 0.345646 0.7296

MNET 0.158813 0.226687 0.700584 0.4836

MNERD 0.023550 0.218030 0.108011 0.9140

COLRD 0.255565 0.351836 0.726375 0.4676

Log likelihood 85.52783 Hannan–Quinn criter. 1.292046

Restr. log likelihood 118.8690 Avg. log likelihood 0.494381

LR statistic (15 df) 66.68235 McFadden R-squared 0.280487

Probability(LR stat) 1.73E08

In models presented, the number of observations appears less than total replies – this is due to the

fact that there might be some non-responses in one or more of the questions

12 Multinational Enterprise and Subsidiaries’ Absorptive Capacity 269

Turning to the type of the R&D unit, if the lab was established to either develop

new products for the subsidiary’s market or to carry out basic research then it increases

the importance of its research as a source of technology for the subsidiary. The lab’s

importance as a source of technology is higher if it has been established for developing

and producing new products for the firm’s market than if it has been established to

carry out basic research (the coefficient of LIL is higher in absolute magnitude).

RQ3

Model 3: Assessing the impact of establishing an R&D lab on the perfor￾mance of the subsidiary (as measured by total turnover) – Table 12.3.

It appears that RAC plays an important role in the subsidiary’s performance. It is

noteworthy that among the various measures of RAC, operating a R&D laboratory

significantly increases the subsidiary’s sales. Also, prior RAC, i.e. the dependence

of the subsidiary on internal technology (from within its MNE group) enhances its

performance.

Regarding the roles of the subsidiaries, those established in order to produce and

export existing products turn out to have higher sales compared to subsidiaries that

were established in order to develop and produce new products.

Concluding Remarks and Policy Implications

The goal of our research is to make progress in terms of modeling AC, where the

focal unit of analysis is the MNE subsidiary, by bringing together different concep￾tual perspectives. Building on Zahra and George (2002) and Veugelers (1997) we

Table 12.2 Assessing the impact of the type of an existing R&D lab on the importance of the

lab’s research as a source of technology for the subsidiary

Dependent variable: OWNRD

Estimation method: ML –Ordered Logit

Observations used in estimation: 86 (if LAB ¼ 1)

Robust std. errors from QML covariance

Coefficient Std. error z-Statistic Prob.

EU 2.019458 1.368237 1.475956 0.1400

AM 2.480446* 1.471074 1.686146 0.0918

PAC 3.20297** 1.550129 2.066232 0.0388

SDH 0.188542 0.664942 0.283547 0.7768

AGE 0.009156 0.010890 0.840768 0.4005

NOPER 0.002468** 0.001102 2.239616 0.0251

RPS 1.00095** 0.470813 2.125999 0.0335

WPM 1.37954*** 0.390908 3.529072 0.0004

MNET 1.02546** 0.485460 2.112338 0.0347

COLRD 1.27781** 0.585120 2.183834 0.0290

IIL 1.00404*** 0.337238 2.977232 0.0029

LIL 1.58368*** 0.597474 2.650630 0.0080

Log likelihood 50.51169 Hannan–Quinn criter. 1.695812

Restr. log likelihood 73.99900 Avg. log likelihood 0.587345

LR statistic (12 df) 46.97463 LR index (Pseudo-R2) 0.317400

Probability(LR stat) 4.71E06

270 C. Kottaridi et al.