1. Exam assignment research design
Below you find the introduction to a scientific article that reports on a study of the use of external scientists to solve R&D problems in firms. Your task is to analyze the way the authors formulate their research problem, using Van de Ven (2007). In particular, you can draw upon chapter 3 in your analysis. Also state an argument whether you approve to the way the authors formulated their research problem or not. Your discussion should be no longer than 500 words.
3
4
2. Exam assignment qualitative methods
Based on the methods literature and your own experience with working with qualitative data and analysis, discuss the benefits and drawbacks of qualitative data analysis and coding according to the proposed Gioia method (as for instance discussed in the Gioia, Corley and Hamilton, 2013, paper). Discuss the following (Maximum 750 words total):
a) What are the main opportunities provided by the Gioia method for coding and analyzing qualitative data in the social sciences? What are the main benefits of this approach compared to other means of analyzing and coding qualitative data? Draw on the Gioia paper but also other methods literature to argue for this approach.
b) What are the major drawbacks of this approach to analyzing and coding qualitative data? Reference methods literature to argue for your answer.
5
3. Exam assignment quantitative methods
For the quantitative part of the exam, you are tasked with evaluating the methods section found below. Please remember to take into account the different themes covered in the three quantitative sessions when evaluating the data and methods, and answering the questions.
1. Comment on the values related to exploring the data, reliability and construct validity, and the regression analysis. Are they OK and if so why, and are there some values that are more questionable, and if so explain why? (25 credits) (maximum 500 words)
2. Have we carried out all the necessary tests for a rigorous analysis? Do you miss some tests that you would add, and if so why and what would they add? (10 credits) (maximum 250 words)
3. Interpret the results vis–vis our hypotheses. (5 credits) (maximum 250 words)
HYPOTHESES
H1: Greater headquarters involvement in the development of an innovation will positively affect transfer performance effectiveness. H2: Greater use of formal hierarchical governance tools by headquarters in the innovation transfer process will negatively affect transfer performance effectiveness. H3: Greater use of expatriates from the sending subsidiary to the receiving subsidiary during the transfer will positively affect transfer performance effectiveness. H4: An established relationship between the sending and receiving subsidiaries will positively affect transfer performance effectiveness. H5: Relationship building between the sending and receiving subsidiary will positively affect transfer performance effectiveness.
DATA AND METHODS
The data used in this research was collected between 2002 and 2005 and covers 169 intra
MNE innovation transfer projects in great detail. Innovations in subsidiaries were identified
through snowball sampling, which is appropriate when the population is difficult to define
and no comprehensive listing exists (Hair et al., 2006). The data can be traced back to 72
6
innovation development projects hosted by 63 subsidiaries belonging to 23 different MNEs
headquartered in the US and Europe. The sending subsidiaries span 14 countries and the
receivers 31 countries.1 Different industries are represented in the sample, for example,
manufacturing, telecommunications, transportation, and the steel industry. The innovation
selection criterion was based on the novelty and specific value to the organization. This
follows the 2005 OECD definition of innovations, that is, the implementation of a new or
significantly improved product (good or service), or process, a new marketing method, or a
new organizational method in business practices, workplace organization or external
relations (p. 47). This selection was done by the innovating/developing subsidiary.
Moreover, the innovations had to have the potential of being transferred and they also had to
have been completed one to 10 years prior to the interview. Sampling innovations that have
transfer potential means that the dataset contains some innovations that have not been subject
to transfer. These innovations are excluded in the present analysis.
One potential sample bias is that it only contains successful innovations, in terms of
having been developed. However, given the question at hand, this bias is almost intrinsic
since the transfer of unsuccessful innovations is highly unlikely and would not add anything
to the MNEs competitive advantage. Successful in this sense does not imply subsequent
market success.
The data was collected through face-to-face interviews on site at the subsidiaries where
the respondent answered a structured questionnaire – an approach similar to surveys with the
advantage of being able to target the respondent in person and knowing exactly who answers
the questionnaire. The respondents had been involved in the innovation development and
were usually R&D managers, project managers, or subsidiary CEOs. In relation the transfer
projects, even if the data was collected at the sending subsidiary, the innovations had been
1 More specifically, the senders are located in: Sweden, Taiwan, Italy, France, UK, US, Germany, Belgium, Finland, Austria, Czech Republic, Denmark, the Netherlands, and Switzerland.
7
transferred to more than one receiver (on average, the innovations in our sample were
transferred to 2.35 receivers). This allows respondents to compare, for instance, transfer
effectiveness across projects. The questionnaire had been pre-tested in two pilot interviews
and minor changes were made in order to eliminate ambiguous questions and phrasings as
well as to exclude erroneous indicators. By having access to managers with detailed
knowledge of the specific innovations, a deeper understanding could be gained (Denrell,
Arvidsson and Zander, 2004), as well as the possibility to discuss the questions with the
respondents. This approach allows targeting the appropriate respondent and detecting
inconsistencies in the answers during the interview, hence increasing reliability and face
validity of the data.
Measures
The advice of Boyd, Gove and Hitt (2005) was followed and single measure indicators were
avoided. Multiple indicators were used in both the dependent and independent variables. This
approach minimizes measurement error, is parsimonious, and offers a multifaceted
representation of the underlying construct (Hair et al., 2006). Additionally, as recommended
by Cox (1980), seven-point Likert-type scales were used to obtain the data on innovation
transfer in MNEs. Besides the subjective estimations by the respondents, distance measures
using secondary data, patenting, and size were included as control variables. The constructs
were identified in an iterative process where coefficient alphas as well as theoretical issues
were considered (Churchill, 1979; Nunnally, 1978). The constructs were theoretically valid
and empirically verified. Subsequently, factor analysis was used in order to confirm the
constructs discriminant validity.
8
Dependent variable
The dependent variable – knowledge transfer effectiveness – reflects the adoption and use of
the transferred knowledge within the receiving unit. The responses focused on circumstances
related to completeness, ease, and timeliness of the adoption and use and follows previous
recommendations and discussions in the literature (Kostova and Roth, 2002; Leonard-Barton
and Sinha, 1993; Repenning, 2002; Szulanski 1996). Compared to earlier studies,
concentrating on the extent of knowledge flows between firm subsidiaries our method of
depicting transfer performance is the degree of transfer effectiveness in terms of investigating
the actual adoption and use of the innovation at the receiving subsidiary (Ciabuschi et al.,
2011a). This reflects key ideas in the knowledge based view (Grant, 1996) and reflects
utilization of transferred knowledge.
Transfer performance effectiveness is measured as a four-item construct where the
respondents were asked to indicate on a scale from 1 (totally disagree) to 7 (totally agree)
whether: ,
counterpart adopted the innovation very quickly>, and
adopt by this counterpart>. One final item was included in this construct and was measured
on a similar scale from 1 (not at all) to 7 (very high):
has been completed>. The internal construct reliability (cronbachs alpha) was 0.817. These
four items were summed and averaged to form the dependent variable in the following
statistical analysis. The dependent variable is distinct from other variables in the analysis, as
Table 1 shows.
9
Independent variables
The first dimension of headquarters subsidiary-level influence is whether or not they have
been involved in the innovation development of the innovation subject to transfer and build
on and extend the attention-based view (Bouquet, Morrison and Birkinshaw, 2009; Ocasio,
1997). Headquarters involvement in innovation development is captured in a four-item
construct where the respondents were asked to indicate, on a scale from 1 (totally disagree) to
7 (totally agree) whether:
innovation>,
innovation>, , and
MNE HQ has taken important initiatives for developing the innovation>. The four indicators
were summed and averaged in order to form the construct used in the regression analysis.
Internal construct reliability was 0.908.
The use of formal hierarchical governance tools and sanctions by headquarters is
captured by four items and is similar to measures employed by Gates and Egelhoff (1986) and
Tsai (2002). The respondents were asked to indicate, on a scale from 1 (totally disagree) to 7
(totally agree), to what extent:
innovation with the counterpart>, and
any sanctions by HQ with the counterpart (Reversed)>. Moreover, the respondents were
asked to indicate on a scale from 1 (not at all) to 7 (very much) whether the transfer of the
innovation was driven by: and . These
four items were summed and averaged to form the construct. The cronbachs alpha of this
construct is 0.632. The use of subsidiary expatriates is reflected in a two-item construct and
builds on Gupta and Govindarajans (2000) measure, and Galbraiths (1973) integrative
mechanisms. The respondents were asked to indicate, on a scale ranging from 1 (not at all) to
7 (very high):
10
managers was used>. The respondents were also asked to indicate, on a scale from 1 (not at
all) to 7 (very much):
personnel between the developer and the receiver>. The indicators were added and averaged
to form the scale. The cronbachs alpha returned with a value of 0.743.
Established relationships, that is, dyadic transfer experience in the sender-receiver
relationship, is a two-item construct where the respondents indicated to what extent, (besides
the focal innovation discussed during the data collection) on a scale from 1 (not at all) to 7
(very much): and
shared knowledge>. The indicators were summed and averaged in order to form the
construct, which had a coefficient alpha of 0.738. This construct builds on literature
highlighting experiences role in knowledge transfer (Ingram and Baum, 1997).
Finally, relationship building between the sending and receiving subsidiaries during the
innovation transfer was captured using a three-item construct drawing on Ghoshal and
Bartletts (1988) framework concerning socialization mechanisms as well as the indicators
used by Persson (2006). The respondents were asked to indicate on a scale from 1 (not at all)
to 7 (very high) the level of use of: ,
teams, project groups etc.> and
regarding the innovation transfer>. The indicators were summed and averaged. Cronbach
alpha was 0.732.
Control variables
In order to more fully specify the model, a number of control variables were introduced. Age
was included since older subsidiaries are more established in their business networks and
have a tendency to be more autonomous (Forsgren, 1990); they can also exhibit a higher
innovative capability (Cohen and Levinthal, 1990; Foss and Pedersen, 2002). To control for
11
age, the logarithm of the number of years the subsidiary had been operating on the market
was included in the regression equation.
Size measured as the natural logarithm of the number of developing subsidiary
employees is used as a proxy for many subsidiary-related characteristics. Research has shown
that large subsidiaries have greater intra-firm bargaining power (Mudambi and Navarra, 2004)
and size can also affect knowledge transfer even if the knowledge has a low relevance (Yang
et al., 2008). Research has also used size as one indicator for valuable knowledge stock,
which can be of greater overall value for the MNE (Gupta and Govindarajan, 2000).
Basic research is captured with the help of a dummy variable. If the subsidiary
conducted research considered to be core, the variable was coded 1; if the subsidiary did not
conduct any basic research, the observation was coded 0. Knowledge developed by a
subsidiary performing core activities is likely to be more easily adopted, building on
absorptive capacity logic. In order to control whether knowledge sharing activities are
stimulated in the MNE, this was included as a single-item variable. The respondents were
asked to indicate, on a scale of 1 to 7, how important knowledge sharing was in the
performance evaluation made of them. This has been shown to have a positive impact on
knowledge transfer flows in previous studies (Bjrkman et al., 2004). To control for the target
subsidiarys knowledge-receiving ability, we employed a measure capturing unit similarity of
the innovation transfer partners. This is a two-item construct capturing how similar the sender
and receiver are regarding technological and organizational features. The respondents were
asked to indicate, with regard to the receiver, on a scale of 1 (totally disagree) to 4 (neither)
up to 7 (totally agree) whether: and
.2 The indicators were summed
and averaged, thus forming the construct. Internal reliability was 0.738. A dummy variable
2 These items were reverse coded in order to capture the similarities between the subsidiaries involved in the knowledge transfer.
12
indicating whether the innovation subject to transfer was patented or not was included in the
model. Patenting proxy codification of knowledge and connects to the potential ease with
which the knowledge might be transferred (Tallman and Chacar, 2011). Additionally,
distances and differences between countries in which subsidiaries are located may influence
knowledge transfer. Therefore, we controlled for distances in a number of dimensions. The
geographic distance between the locations was calculated for each transfer project. The
number of kilometers was calculated using MapCrow. This measure was transformed into the
natural log of the distance measure and is consistent with the approach of other studies using
geographic distance (e.g., Hansen and Lovs, 2004). Cultural distance was controlled for by
using Kogut and Singhs 1988 index, expressed as:
4 CDj= {(Iij IiN)2/Vi}/4 , i=1
where CD is the cultural distance between the subsidiary host countries, Iij is the score
of the receiving subsidiarys country on the ith dimension, and IiN is the score of the sending
subsidiarys country in this dimension. Vi represents the score variance in the specific
dimension. Institutional distance was measured building on the approach of Gaur et al. (2007)
and Xu, Pan and Beamish (2004). The institutional dimensions found in the Executive
Opinion Survey of the Global Competitiveness Report (2005) were explored and a factor
analysis was conducted (principal component with varimax rotation and Kaiser
normalization). The institutional environment is captured by a seven-item construct that
loaded on a single factor having a coefficient alpha of 0.961. This data was matched to our
data calculating the institutional distance between the host countries of the sending and
receiving subsidiaries. The relative economic differences were captured by estimating
differences in GDP per capita between the host countries of the subsidiaries (Tsang and Yip,
13
2007). Data was obtained through the Total Economy Database (2006). Following Tsang and
Yip (2007), we created a measure for relatively more developed countries in relation to the
other part of the dyad. This measure can be expressed as:
ln(GDPsender)-ln(GDPreciever) if GDPsender GDPreciever and = 0 if GDPsender < GDPreceiver
Further assessment of the data
The use of perceptual measurements can be problematic because of social desirability and
self-assessment bias. This is mitigated by the face-to-face interviews.
All relevant indicators were included in a principal component factor analysis
(principal component with varimax rotation and Kaiser normalization, see Table 1). The
KMO-value returned at 0.638. The Bartletts test of sphericity was at a 0.001 significance
level. In the principal component analysis, six factors were extracted with an eigen value
above 1. The seventh factor returned with an eigen value of 0.837.
Some cross-loadings were observed: the first occurred for the item of headquarters
instruction to share the innovation with the counterpart on the construct of headquarters
participation during the development with a value of 0.436. The second cross-loading relates
to the respondents reporting whether the innovation transfer occurred without any sanctions
from headquarters (reversed) with a value of 0.354 on the construct of headquarters
participation during the development.
To investigate whether there is a correlation between two or more predictor variables
augmenting the estimated R2 of the model, the Variance Inflation Factor (VIF) was calculated,
see Table 2.
14
RESULTS
The paper examines how different organizational mechanisms affect knowledge transfer
effectiveness. The mean values, standard deviations, and correlation matrix for all the
variables are presented in Table 2. In order to estimate the models, Ordinary Least Squares
regressions were used. In Table 3, the results from the multiple regression analysis are
reported.
15
Table 1 Factor analysis with Varimax rotation
Variable
Factor loading Communality Factor 1: HEADQUARTERS INVOLVEMENT IN DEVELOPMENT The MNE HQ has participated closely in developing this innovation 0.913 0.872 The MNE HQ has brought competence of use for the development of this innovation 0.864 0.814 The MNE HQ has been important through specifying requests 0.899 0.832 The MNE HQ has taken important initiatives for developing the innovation 0.785 0.678 Eigenvalue 3.790 % Variance 19.950 Factor 2: TRANSFER PERFORMANCE EFFECTIVENESS The counterpart adopted the innovation very quickly 0.649 0.512 The innovation has been very easy to adopt by this counterpart 0.836 0.720 The performance of the innovation transfer was very satisfactory 0.811 0.692 To what extent the innovation transfer has been completed 0.490 0.240 Eigenvalue 3.154 % Variance 16.600 Factor 3: HEADQUARTERS HIERARCICHAL GOVERNANCE TOOLS The MNE HQ has formally instructed you to share this innovation with the counterpart 0.526 0.519 The transfer of the innovation has occurred without any sanctions by HQ with the counterpart (Reversed) 0.561 0.610 Requirement from HQ 0.486 0.236 HQ evaluation system 0.686 0.575 Eigenvalue 2.350 % Variance 12.367 Factor 4: RELATIONSHIP BUILDING Temporary training at partner sites 0.790 0.729 Cross-unit teams, project groups etc. 0.806 0.764 Face to face meetings were used in the communication regarding the innovation transfer 0.761 0.679 Eigenvalue 1.847 % Variance 9.720 Factor 5: ESTABLISHED RELATIONSHIP They previously cooperated with the receiver 0.867 0.811 They previously had shared knowledge 0.823 0.801 Eigenvalue 1.284 % Variance 6.757 Factor 6: SUBSIDIARY EXPATRIATES To what extent, with regard to the transfer of the innovation exchange of managers, was used 0.881 0.837 To what extent the transfer of the innovation was driven by moving personnel between the developer and the receiver 0.817 0.802 Eigenvalue 1.225 % Variance 6.446 Total variance explained 71.840
16
Table 2 Correlation and descriptive statistics
MEAN S.D. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 1. Transfer performance effectiveness 5.211 1.312 1 2. Age 3.528 0.903 0.036 1 3. Size 5.414 1.590 0.041 0.112 1 4. Basic research 0.555 0.498 0.132 0.402** 0.224** 1 5. Knowledge sharing 4.148 1.774 0.147 0.081 0.272** 0.173* 1 6. Unit similarity 5.557 1.55271 0.465** -0.213** -0.158 -0.060 0.002 1 7. Patent 0.569 0.496 -0.111 -0.102 -0.212** 0.019 -0.240** 0.093 1 8. Physical distance 5.735 3.578 -0.001 -0.008 0.024 -0.035 -0.062 0.016 0.051 1 9. Cultural distance 0.618 0.809 0.015 0.049 -0.148 -0.012 -0.023 0.052 0.106 0.635** 1 10. Institutional distance 0.491 0.572 0.041 -0.114 0.059 -0.025 -0.147 -0.057 -0.070 0.464** 0.379** 1 11. Economic differences 0.052 0.124 0.067 -0.067 0.060 -0.008 0.209** 0.052 -0.007 0.391** 0.270** -0.042 1 12. Headquarters involvement in dev. 2.110 1.580 -0.147 -0.350** -0.149* -0.182* 0.005 0.099 0.024 0.073 0.080 -0.057 0.150 1 13. Headquarters hierarchical tools 2.601 1.569 -0.039 0.056 0.324** 0.228** 0.169 -0.054 -0.180* 0.048 0.123 -0.020 0.150 0.348** 1 14. Subsidiary expatriates 1.893 1.472 -0.390** -0.042 -0.064 -0.137 0.032 -0.264** 0.034 -0.049 -0.037 -0.026 -0.150 0.081 -0.098 1 15. Established relationship 4.777 1.660 0.289** 0.085 0.046 0.007 0.151 0.255** 0.074 -0.110 0.009 -0.202** 0.002 -0.039 0.108 -0.028 1 16. Relationship building 4.025 1.753 -0.004 0.000 -0.016 -0.020 -0.056 -0.037 0.012 -0.138 -0.167* -0.059 -0.185* 0.241** 0.093 0.175* 0.256** 1 VIF value 1.450 - - 1.597 1.629 1.433 1.277 1.310 1.271 1.963 1.585 1.409 1.478 5.528 1.528 1.185 1.336 1.246 Spearmans correlation ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
17
Table 3 Results from the Ordinary Least Squares regression analysis a
Regressor Model s.e. Age -0.006 0.145 Size 0.085 0.083 Basic research 0.125 0.250 Knowledge sharing 0.062 0.066 Patent -0.202* 0.236 Unit similarity 0.368*** 0.077 Physical distance -0.019 0.041 Cultural distance -0.010 0.162 Institutional distance 0.071 0.216 Economic differences 0.075 1.020 Headquarters involvement in development -0.002 0.081 Headquarters hierarchical governance tools -0.240** 0.082 Subsidiary expatriates -0.239** 0.077 Established relationship 0.211* 0.072 Relationship building 0.039 0.066 Diagnostics N 169 R2 0.368 Adj.R2 0.274 F-statistics 3.915*** a Values are standardized parameter estimates p<0.1, *p<0.05, **p<0.01, ***p<0.001
