4.1. The Brokerage Patterns of Three Anchor Regions
Table 5 reports the GF frequencies of brokerage activities of three regions—Beijing, Shanghai, and Shenzhen—representing BGR, YRD, and PRD, respectively. Beijing recorded the largest number of brokerage activities, reaching 2369 in 2010 (
Table 5). As a national capital region, the presence of major public institutes and universities has highlighted Beijing as a brokerage region in the nation as well as BGR. The bifurcated roles are the liaison (43.5%) and consultant (43.4%), occupying almost 87% of the total brokerage activities. These two roles imply that Beijing mediates technologies outside of BGR, serving as a national-wide technology source. Then, the ratio is followed by a representative and gatekeeper. It is worth noting that both brokerage roles related to the neighbour regions in the megalopolis have lower significance. The comparison between the two roles is that Beijing tends to diffuse the accumulated technologies in the BGR toward the regions in the other super-regions. The ratio of a coordinator is substantially lower than the other roles. It seems to be clear that Beijing does not transmit the technologies among the neighbour regions in the BGR.
On the other hand, Shanghai, the anchor hub in YRD, has a more balanced brokerage pattern than Beijing has. The total frequency also marked the highest point (1718) in 2011, but sharply dropped in 2012. The highest role of Shanghai is a consultant (29.1%) followed by gatekeeper (21.9%), liaison (21.8%), and representative (20.8%). However, the dispersed distribution among these three roles imply that Shanghai serves as national and regional brokerage sources of technology. According to Huggins, Luo, and Thompson [
10], YRD has a longer tradition of openness than the other megalopoleis and established a strong industrial foundation for modern technology development. The hierarchical structure of knowledge transmission consisting of highly competitive city, such as Wuxi and Suzhou, is likely to affect the balance of the brokerage roles of Shanghai. The lowest ratio of a coordinator also corroborates that Shanghai is also not involved with the intra-megalopolis technology transmission.
Last, Shenzhen seems to have an intermediary pattern between Beijing and Shanghai. Shenzhen also has bifurcated roles—consultant and liaison—however, they are not as dominant as Beijing. The most frequent role is a consultant (35.5%), followed by a liaison (34.8%), a representative (14.6%), and a gatekeeper (13.7%). Consistent with the two other regions, the ratio of coordinator is the lowest. The rank order shows that Shenzhen also tends to play the role of a nation-wide broker, while the total frequency is lower than the other two regions. Even the highest frequency (662) in 2009 is less than half of the other two regions.
4.2. The Estimation of Network Dynamics
This section estimates the dynamics of the network and the contribution of the brokerage types by applying a ‘conditional method of moments’ for a longitudinal dataset [
59]. The convergence ratio value that compares the deviations between simulated values and observed values is 0.283, indicating that the goodness-of-fit of the model is appropriate for the model.
Table 6 reports the four models, starting from a structure effect-only (model 1) to the full model (model 4).
Let us turn to model (1), which includes only structural effect variables. The rate function explains the longitudinal progression of the technology transfer network. The rate refers to the expected frequencies with which regions have the opportunity to change a network tie from between
to +1. The rate (
) plunges from 27.11 to 0.007 (
); then it maintains at 0.007 at the end of the period (
). The network has a substantial change from 2009 to 2010, then a sharp decrease rate of the next period (2010 to 2011) indicates that fewer opportunities exist to change relationships in the last two periods than in the previous ones. After the relationship is established, then the nodes do not seem to easily change their relationships with their partners, which is interpreted as the path-dependency in the market-mediated technology. A similar pattern is consistently found through the other three models. The density of the network has a significant, but negative value (−1.435). On the other hand, the reciprocity has a positive coefficient (1.09), which is observed widely in the other empirical works [
59]. In the social network context, the higher density level is related to the higher opportunity cost in the establishment of a relation. If a node is positioned in the highly dense position, then the node is less likely to have an opportunity to change the previous relationship. Thus, given the high-density effect, the probability of changing its tie decreases and yields a negative sign of the variable. The positive and significant value of reciprocity also reflects that technology is transmitted with the partners that have already connected, denoting that those mutually-proven partners are likely to involve another transfer.
Model 2 is dedicated to explaining the longitudinal change of a network through the individual attribute of each region. The interpretation of the coefficient is as non-standardized coefficients, similar to that of the logistics regression model. Each coefficient is basically a log odds-ratio, and it indicates how the log-odds of relation formation changes with a unit change in the corresponding independent variable. The exogenous foreign technology has a positive and significant, but weak value (0.0011). The patent number of a region, however, has a negative effect (−0.083). Then, regions have the tendency to transmit technology to other regions when they have already accumulated technologies from overseas countries. At the same time, even if the region has a higher level of the domestic patent, they do not seem to be involved in exchanging technology with others. It also corroborates the idea that nodes exposed to direct relationships with foreign nodes, through formal technology agreements or informal know-how contact, are expected to gain preferential access to knowledge [
66].
The next two variables examine whether a region’s previous licensing activity matters in the evolution of the network. While the observed license-in value has not gained statistical significance, license-out has a positive effect, with a significance level of 0.01 (0.0016). The result confirms that the regions prefer to transmit technology to other regions that already have previous experiences of a technology licensing-out contract, which is consistent with Belso-Martínez et al. [
67]’s work in that the previous knowledge mediating experience facilitates the creation of partnerships, thus fostering brokerage. The influence of accumulated licensing-out experiences also seems to be determined by the strategic risk-aversion decisions of licensor regions. The interpretations of the contradictory significance level of two licensing variables would be that the potential licensing-in regions do not consider how many technologies are imported to the partner regions, rather they seem to consider that the licensing-out records matter more.
Model 3 captures the contribution of the GF brokerages of three regions except for a coordinator, which has a substantially low frequency. Considering the differences between the four coefficients, it confirms the idea that all the brokers have an identical effect on the network growth process. In this case, the brokerage roles played by the anchor regions influenced the network evolution in either a positive or negative way. The interesting result is the different signs of the coefficients between a gatekeeper and a representative (−0.012 and 0.025, respectively). The consultant has a small, but positive effect (0.008). Nevertheless, liaison was estimated to have a negative effect (−0.012). Reflecting the types of the brokerage on the basis of the social network theory, the brokerage role as a liaison appears to have the opportunity to benefit from intermediating between heterogeneous groups [
53].
The common feature of the two positive estimation values (representative and consultant) is that an anchor region has more influence on the regions outside of its own super-region. The representative, which acquires technology knowledge from other co-located regions, acts as the technology source for other regions. As a consultant role, an anchor region connects only with regions located outside of the super-region, thus connecting other regions. Contrarily, the gatekeeper, acquiring from the regions outside of the super-region, acts as a technology source for the regions inside of its own super-regions. The liaison connects different regions in different super-regions. To sum up, the main contribution of anchor regions, in the perspective of the national innovation system, is directed to the external regions outside rather than inside of the super-regions [
65].
Model 4 maintains a similar result of the rate function as the previous models. The first difference in the objective function is the negative sign of the number of patent purchases (License_In), which was already anticipated in the previous model. It is clear that, given the other variables, a region’s experiences in purchasing technology have little, but a negative effect on the evolution. The second point is that the two variables (gatekeeper and representative) that are related to the anchor regions’ megalopolis are estimated with no significance power. Nonetheless, the effect of the other two variables (consultant and liaison) turned to reverse effects. It is plausible that the two major roles have a significant effect on the evolution of a whole network, while the effects are in opposite directions. Following the idea of [
53], who hypothesized the positive relation between the liaison role and innovative network, the positive coefficient of the liaison corroborates that it gains benefit from brokering between different super-regions, taking advantage of diverse technology sources. The role of a consultant, however, is estimated to be a negative effect on the evolution process. Despite its highest frequency, given the control of the other variables’ effect, the mediation of technology between the regions is in another megalopolis.