3. Research Method
3.4. Model Specification
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Finally, the study includes some control variables that will help the research to increase its consistency. The name of those variables is defined as follows:
3.4. Model specification
As the previous paragraphs showed, the representation of innovation would take in almost all the analysis (except for the operationalization through the share of innovating sales) the form of a dummy variable. Due to this, the research will consider the conventional practice of using a logistic model to evaluate the propensity to innovate.
This propensity to innovate will be modeled as follow:
𝑝𝑟𝑜𝑏(𝑦𝑖 = 1) = 𝑒𝑥𝑝𝑋𝑖𝛽 1 + 𝑒𝑥𝑝𝑋𝑖𝛽 Table N° 07
Name Description Type Range
Year Years of operation Continuous[1,151]
Industry Technology intensity Ordinal [1,3]
Size Size of the firm Ordinal [1,4]
Control Variables Table N° 06
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
X is the set of independent variables that the research will include in the analysis. The suffix "i" represents the firm. And 𝛽 represent the set of parameters to estimate and each of them will show us a change in the logit of the probability associated with a unit change in the independent variable holding all other predictors constant.
3.4.1. Model 1: Networks
In our first model, the research will build the simplest model assuming that the propensity to innovate is influenced by the Knowledge Network, the Production Network and Customer Networks. Using the simple version of the network variables (the binomial version), the model will be us following:
log ( 𝑝
1 − 𝑝) = 𝛽0+ 𝛽1∗ 𝑁𝑒𝑡𝐾𝑛𝑜𝑤 + 𝛽2∗ 𝑁𝑒𝑡𝑃𝑟𝑜𝑑 + 𝛽3∗ 𝑁𝑒𝑡𝐶𝑢𝑠𝑡 + 𝜇
Using the strongest version of the network variables (built through the nature of the relation); due to each network variable having four categories (to the less strong relation to the strongest relation), our model will need three categories of each variable, so the model will be as follow:
log ( 𝑝
1 − 𝑝) = 𝛽0 + 𝛽1∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤1+ 𝛽2∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤2+ 𝛽3∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤3+ 𝛽4
∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤4+ 𝛽5∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑1+ 𝛽6∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑2+ 𝛽7
∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑3+ 𝛽8∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑4+ 𝛽9∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡1+ 𝛽10
∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡2+ 𝛽11∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡3+ 𝛽12∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡4+ 𝜇
3.4.2. Model 2: Networks + R&D investment + Absorptive capacity
The second model will include the other two more important independent variables according to our literature review: R&D investment and absorptive capacity. For RD
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
investment, it will be used the measure related with the presence of RD investment. For the absorptive capacity, the presence of a formal RD laboratory will be used. Using the simple version of the network variables (the binomial version), the model will be us following:
log ( 𝑝
1 − 𝑝) = 𝛽0+ 𝛽1∗ 𝑁𝑒𝑡𝐾𝑛𝑜𝑤 + 𝛽2∗ 𝑁𝑒𝑡𝑃𝑟𝑜𝑑 + 𝛽3∗ 𝑁𝑒𝑡𝐶𝑢𝑠𝑡 + 𝛽4∗ 𝑅𝐷_𝐼 + 𝛽5
∗ 𝐴𝑏𝑠𝐶𝑎𝑝 + 𝜇
Using the strongest version of the network variables, the model will need three categories of each variable, so the model will be as follow:
log ( 𝑝
1 − 𝑝) = 𝛽0 + 𝛽1∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤1+ 𝛽2∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤2+ 𝛽3∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤3+ 𝛽4
∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤4+ 𝛽5∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑1+ 𝛽6∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑2+ 𝛽7
∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑3+ 𝛽8∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑4+ 𝛽9∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡1+ 𝛽10
∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡2+ 𝛽11∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡3+ 𝛽12∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡4+ 𝛽13∗ 𝑅𝐷𝐼 + 𝛽14∗ 𝐴𝑏𝑠𝐶𝑎𝑝 + 𝜇
3.4.3. Model 3: Networks + R&D investment + Absorptive capacity + Control Variables
The third model will include the independent variables and also the three control variables that are used more in the literature: age of the firm, size, and technology intensity. Using the simple version of the network variables (the binomial version), the model will be us following:
log ( 𝑝
1 − 𝑝) = 𝛽0+ 𝛽1∗ 𝑁𝑒𝑡𝐾𝑛𝑜𝑤 + 𝛽2∗ 𝑁𝑒𝑡𝑃𝑟𝑜𝑑 + 𝛽3∗ 𝑁𝑒𝑡𝐶𝑢𝑠𝑡 + 𝛽4∗ 𝑅𝐷𝐼+ 𝛽5∗ 𝐴𝑏𝑠𝐶𝑎𝑝 + 𝛽6𝑌𝑒𝑎𝑟 + 𝛽7∗ 𝑆𝑖𝑧𝑒2 + 𝛽8∗ 𝑆𝑖𝑧𝑒3+ +𝛽9∗ 𝑆𝑖𝑧𝑒4+ 𝛽10∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡2
+ 𝛽11∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡3+ 𝜇
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Using the strongest version of the networks variables, the model will need three categories of each variable, so the model will be as follow:
log ( 𝑝
1 − 𝑝) = 𝛽0 + 𝛽1∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤1+ 𝛽2∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤2+ 𝛽3∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤3+ 𝛽4
∗ 𝐴𝑑𝑁𝑒𝑡𝐾𝑛𝑜𝑤4+ 𝛽5∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑1+ 𝛽6∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑2+ 𝛽7
∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑3+ 𝛽8∗ 𝐴𝑑𝑁𝑒𝑡𝑃𝑟𝑜𝑑4+ 𝛽9∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡1+ 𝛽10
∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡2+ 𝛽11∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡3+ 𝛽12∗ 𝐴𝑑𝑁𝑒𝑡𝐶𝑢𝑠𝑡4+ 𝛽13∗ 𝑅𝐷𝐼 + 𝛽14∗ 𝐴𝑏𝑠𝐶𝑎𝑝 + 𝛽15𝑌𝑒𝑎𝑟 + 𝛽16∗ 𝑆𝑖𝑧𝑒2 + 𝛽17∗ 𝑆𝑖𝑧𝑒3 + 𝛽18∗ 𝑆𝑖𝑧𝑒4 + 𝛽19∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡2+ 𝛽20∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡3 + 𝛽21∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡4+ 𝜇
3.4.4. Model 4: Origin of network + R&D investment + Absorptive capacity + Control Variables
Finally, using the variable origin of the network, the research will attempt to give an answer to the second research question. Because the variable "Origin" has five categories, the analysis will create four dummy variables that allow us to analyse the impact of the links with the Asia Pacific. It will take the lack of connections as a basis to compare.
log ( 𝑝
1 − 𝑝) = 𝛽0 + 𝛽1∗ 𝑂𝑟𝑖𝑔𝑖𝑛2+ 𝛽2∗ 𝑂𝑟𝑖𝑔𝑖𝑛3+ 𝛽3∗ 𝑂𝑟𝑖𝑔𝑖𝑛4+ 𝛽4∗ 𝑂𝑟𝑖𝑔𝑖𝑛5+ 𝛽5
∗ 𝑅𝐷𝐼+ 𝛽6∗ 𝐴𝑏𝑠𝐶𝑎𝑝 + 𝛽7𝑌𝑒𝑎𝑟 + 𝛽8∗ 𝑆𝑖𝑧𝑒2 + 𝛽9∗ 𝑆𝑖𝑧𝑒3 + 𝛽10∗ 𝑆𝑖𝑧𝑒4 + 𝛽11∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡2+ 𝛽12∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡3 + 𝛽13∗ 𝑇𝑒𝑐ℎ𝐼𝑛𝑡4+ 𝜇
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
CHAPTER 4
RESULTS AND MAIN FINDINGS
4.1. Characteristic of the sample
The survey interviewed 1,452 enterprises. The main proportion is composed of big companies and it represents around 74.31% (1079 enterprises) of the sample. The second big group is the small enterprise with 18.11% of the sample (263 enterprises).
Then, the sample is composed of a few proportions of microenterprises (58 enterprises) and medium enterprises (52 enterprises).
In terms of economy activity, there are six industries that have at least five percent of participation in the sample: food products (21.07), Rubber and plastics products (9.71), Metal products except weapons and ammunition (8.26), Chemicals and chemical products (8.13), Wearing apparel (6.27) and textiles (6.2). These six industries represent almost 60% of the sample. The less represented industries (with less than one percent of the sample) are two: Coke and refined petroleum products (0.83) and Computer, electronic and optical products (0.41). Using the codification related to technology intensity, it is possible to realize that more than 50% of the sample is composed of the
Table N° 08
Size Frequency Percent
Microenterprise 58 3.99
Small Enterprise 263 18.11
Medium Enterprise 52 3.58
Big company 1,079 74.31
Total 1,452 100
Sample by Size
‧
low technology industries that in theory are less innovative than the other industries. It reflects the necessity to control our results considering this fact.
Also, another important characteristic of the sample is that the year of operation is on average 22 years old but that happens for the value of some really old firms (the oldestis151 years old). In fact, the median show us that less than 50% of the companies are less than 18 years old.
Table N° 10
Leather and related products 15 32 2.2
Wood and products of wood and cork 16 39 2.69
Paper and paper products 17 36 2.48
Printing and reproduction of recorded media 18 39 2.69 Coke and refined petroleum products 19 12 0.83
Chemicals and chemical products 20 118 8.13
Pharmaceuticals 21 48 3.31
Rubber and plastics products 22 141 9.71
Other non-metallic mineral products 23 71 4.89
Basic metals 24 33 2.27
Fabricated metal products except weapons and ammunition25 120 8.26 Computer, electronic and optical products 26 6 0.41
Electrical equipment 27 48 3.31
Machinery and equipment n.e.c. 28 47 3.24
Motor vehicles, trailers and semi-trailers 29 38 2.62 Other transport equipment except ships and boats 30 20 1.38
Furniture 31 32 2.2
Other manufacturing except medical and dental instruments32 24 1.65 Repair and installation of machinery and equipment33 30 2.07
Total 1,452 100
Sample by Sector
Freq. Percent
Low Technology 828 57.02
Medium technology 299 20.59
Medium and high technology 325 22.38
Total 1,452 100
Sample by Technology Intensity
‧
Analysing the sample, the research finds that the innovation widely practiced by the firms is the innovation in Product with 46.76%. The second innovation that is more widely practiced by the Peruvian manufacture enterprises is innovation in Commercialization with 43.04%. Innovation in Organization is third more practiced innovation with 40.15%. And finally, the Innovation in Process is the least practiced with 39.88 % of the sample.
Despite this, the innovation tends to be informal. Only the 4.22% of the sample applications for obtaining a patent to protect their invention in the period 2012-2014.
In historical review, the percentage improves but it is still lacking in comparison with Table N° 11
‧
their innovation. This evidence is crucial when it is time to understand and select the correct measure of innovation. Remembering the work of Arocena & Sutz (2000), innovation in Latin America tends to be informal and for this reason, using patents is not a good approach to measure innovation as it is in developed countries.
Also, for the innovation in product, the degree of novelty shows us an interesting aspect of innovation. Thus, it is interesting that the innovation related with the improvement of a product or service tends to be higher than the innovation related with a new product or service. This aspect reminds us of the signalled by Heijs & Buesa (2016). The incremental innovation tends to be more frequent than radical innovation. Thus, the invention of new product in the strict sense is not the main source of innovation, but the constant improvement is the main driver of the innovation.
Table N° 13
Innovation in Product by Degree of Innovation
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.3. Networks
With regard to the networks; the Production network is the most important linkage for Peruvian companies with 59.89% of firms that have a relationship with a supplier, competitor or firms of the same group. Also, the customer network is another important linkage with 49.45%. The less important in term of linkages is the Knowledge network with 27.81% of the firms that have a relationship with universities, research institutes or laboratories.
4.4. Analyzing the regressions
Using the basic model described in the research method. The research finds that the presence of a production network and knowledge network are highly significant in the probability to innovate (in all its four dimensions). In the case of the customer network, it seems not to be important for innovation (at least using this measure). Analysing the results, it is possible to observe that the odds of innovation in product (the probability to innovate over the probability to not innovate) of the firms that have relationships with other firms (production network) is 2.02 bigger than the odds of the innovation in product of the firms that do not possess any kind of relationship with other firms. In a similar way, the research finds that the odds of innovation in process, innovation in the organization, and innovation in commerce tend to be bigger (19.6, 1.83 and 1.52 respectively) for the firms that possess a relationship than firms that do not have any
Table N° 15
Type of Networks Percentage Knowledge Network 27.81 Production Network 59.85
Customer Network 49.45
Networks
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
higher probability to innovate in the four types of innovation. For the case of the knowledge network, the analysis of the odd ratios reveals that the odds to innovate (in the four types of innovations) for the firms that own a relationship with knowledge components are bigger (1.61, 1.70, 1.85 and 1.78 respectively) than the firms that are not connected with knowledge components.
Using the improved measure of innovation, it is interesting to see a stronger degree of innovative relationships with the production networks and knowledge networks can be significant in the probability to innovate compared with the weaker relationships.
For instance, analysing the degree of relations with production components (competitors, suppliers, and firms of the same economic group), it is possible to observe that the odds of innovation in product for the firms that have relationship more related with Joint studies, Engineering and designs and Investigation, and development is 2.28 bigger than the odds of innovation in product for the firms that do not have relationships or only have commercial relationship with other production components. Also, observing the second strong type of innovative relationship; Training, Testing of
Table N° 16
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
products/processes, and Technical assistance; it is possible to observe that the odds of innovation in product for the firms that have an educational relationship with other firms is 1.95% bigger than the firms that only have a commercial relationship or do not have any relationship.
Despite having a result with less significance, the relationship related to the information required has a probability in product innovation that is 1.58 times larger than companies
Table N° 17
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
that do not have contact with other companies or are limited to only a commercial relationship.
In the case of process innovation, it is also observed that those companies that have a relationship with other companies focused on joint work, process engineering, research, and development present a probability to innovate in the process that is2.5 higher with regard to companies that do not have any kind of relationship with other companies or the relationship is purely commercial. Also, an important and unexpected result is the fact that the firms that have relationships with other firms focused on finances present a probability to innovate in process of 2.4 times larger than those companies that do not have a relationship or only have a commercial relationship. Similar results are found for the innovation in organizations and the innovation in commercialization where the relationship related to financial requests and joint studies with other firms present the largest odds in the logistic regression.
Analysing the degree of relationships of the companies that have links with the knowledge infrastructure (universities, public institutes, private institutes and non-university laboratories), the research finds that the relationship focused on joint studies, research and development and engineering is the most important for the four types of innovation according to significance and value of the odds ratio (2.07 for innovation in product, 2.133 for innovation in process, 2.099 for innovation in organization and 2.279 for innovation in marketing). The second most important relationship with knowledge components that foster the innovation (except innovation in the product) is the training, testing of products/processes, and technical assistance where the variable is significant for all the types of innovation except the innovation in product.
Adding the other important explanatory variables considered in the research and using
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
production remains important in the propensity to innovate however the network knowledge variable lose importance in the innovation in the product after the incorporation of the research and development activities. This last fact has to be interpreted with caution, and it is possible that the role of the knowledge network of innovation in the products of Peruvian companies is the linkage with the purpose of doing research and development activities, so the knowledge network and R&D activities are overlapping variables.
The last factor becomes highly important for the four types of innovation. Thus, for instance, the odds of innovation in products for firms that carry out activities in research and development is 10.94 times higher than the odds of innovation in product for firms that do not carry out those activities. In the same way, the odds of innovation in the process, organization, and commercialization for the companies that carry out research and development activities are more than three times the odds of companies that do not carry out such activities. Also, the measure of absorptive capacity shows that a basic infrastructure that is destined to manage and create knowledge increases the probability to innovate in all aspects.
Table N° 18
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Analysing the second model with the improved measure of the network, R&D activities and absorptive capacity remain important in the propensity to innovate. It is also possible to see that the most obvious change is the decreasing in significance and in the value of the odds ratio of the degree of the knowledge network.
Table N° 19
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
However, it is possible to observe that knowledge network is still important in innovation performance, especially the relationships related with joint studies, research and development and engineering. This fact shows the limitation of the first measure of network, and it give us a better perspective about the role of networks in the decision to innovate within the firm.
Advancing in this analysis, the research adds the control variables that were discussed and defined in the previous chapters: year, technology intensity and size of the firm.
The model reveals that the variable year is significance on the innovation in the organization and it also has an odd ratio minor to 1. Then it is necessary to use the
Table N° 20
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
inverse of the coefficient to be able to give a correct interpretation. Thus, if the age is decreased by one unit and all the values of the other variables of the model remain constant, the odds of the firms to innovate increase by 1.01 times more than the odds if they did not decrease that year. That means that there is a negative relationship between year and innovation in the organization. This is aligned with other empirical papers.
The explanation of this fact is that young companies tend to innovate more due to the more flexible rules and costs they have to face. In term of size, it is interesting that the odds of innovation for mediums firms are 2.56 times larger than the odds of innovation for microenterprise. Another interesting finding is that the odds of innovation for mediums firms is1.94 times larger than the odds of innovation for microenterprise. This result is aligned with a wide range of empirical evidence. The explanation is that medium and big companies are more willing to reorganize their areas to find a more optimal labor division.
Also, the model reveals that technology intensity (measured by industry) is not a significant variable in our model of innovation. Although the empirical evidence in other countries indicates that companies from highly technology-intensive sectors innovate more than companies from sectors of low technological intensity, Peruvian companies innovate equally regardless of sector.
Adjusting the previous model with the improved measure of the network, the research establishes a more complete version of the model proposed to study the impact of networks along the different dimensions of innovation. The model includes the other important explanatory variables and the more acceptable control variables in the empirical evidence. In general, the model shows that networks play an important role in the propensity to innovate. The production network seems to be the more important network for innovation because after examining the significance of the different degree
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
of relations for this variable in the model, it is possible to see that the interaction between a firm and another firm is significant from the most basic interaction such as asking information until the most complex interaction such as making joint studies or making together research and development activities (except the impact of training in the innovation in process).
Table N° 21
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In the case of knowledge, the network is interesting that relationships with a high degree of interaction and related with knowledge creation are positive and retain high significance with the different types of innovations (production, process, organizational and commercial). It is also interesting that educational relationships (such as training and technical assistance) with knowledge organization have a positive and significant effect on innovation in process and organization.
And, although the network with customers seems to be a weaker variable, it can be important for the innovation in commercialization where the Research and Development collaboration has a high significance with this innovation. It might be understandable due to the fact that this collaboration would tend to focus on the customization of the appearance more than the customization in the function of the product itself. Confirming the theory, the decision to perform a Research and Development activity and the presence of a good absorptive capacity has a high significance for the four dimensions of innovation. Also, there is a negative relationship between the innovation in organization and the years of operation, so younger firms have a greater possibility to change their form of operation. And medium and big firms tend to innovate in organization more than microenterprises.
Finally, in order to answer the second research question, the study makes a model that reformulates the variable networks in the variable origin. As it was mentioned in the previous chapter, the variable origin has five categories: non-relationship (0), only relationship with National Organization (1), only relationship with Asia Pacific Organization (2), only relationship with Non-Asia Pacific Organization (3), and relationship with more than one source.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In the model, the relationship and importance of the rest of the independent variable and control variable are still the same. The importance of each origin is contrasted with the lack of relationship and it is possible to see that the relationship with Asia Pacific organizations is highly significant and positive. However, this is also true to the relation
Table N° 22
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
with national organizations and Non-Asia Pacific organizations. To compare if there is a significant difference between Asia Pacific Organization and the other two sources, the study will test the coefficients of each origin between them.
Although the coefficients of Asia Pacific Relation has a greater impact than the coefficients of Non-Asia Pacific Regions in the innovation in process and innovation in the organization, the test shows that the coefficient is not statistically significantly different. The same is possible to observe if we compare the relationship with Asia Pacific Organizations and national organizations. In this sense, all the origins of relations have importance in innovation but none of them is dominant over the others.
This fact is already significant because the positive impact of Asia Pacific linkages
This fact is already significant because the positive impact of Asia Pacific linkages