In order to avoid the multicollinearity issue, the research adopted the latent factors derived from previously PCA as new variables and ran cluster analysis independently for both periods. The research deployed two methods and Euclidean distance for the cluster analysis. Firstly the research ran several cluster analysis to compare the results of the clustering. According to rules of thumb to decide the better number of clusters, the larger CCC and Pseudo F statistic the better, the optimal cluster number for both two periods was inferred to be three in this research based on table 6.
Next, the wards method was applied to derive the centroid of each cluster. Wards method maximized the variance among groups but minimized the variance within groups.
Therefore its centroids were most appropriate to be the seeds for K-means analysis. After assigning the seeds, the research followed up applying K-means method to calculate the Euclidean distance between each firm and seeds and to dispatch sample firms into different groups based on the distance.
Table 6 Criteria of optimal number of clusters
Below table presented the average scores of each factor of each cluster in two periods. The nomination for the clusters was determined on the features of each clusters. In order to picture out the underlying characteristics of the groups, the research also compared categorical information for each cluster and aimed to give an appropriate name for each cluster. The list of the subgroups was shown in appendix 1.
Table 7 Averaged factor scores in each cluster before crisis
Cluster Pseudo F A Rsquare CCC Cluster YEAR A Rsquare CCC
2 27.87 19.56 -6.29 2 43.20 25.75 -5.32
3 35.32 34.57 -6.26 3 47.76 45.47 -7.98
4 31.39 46.75 -10.62 4 51.03 61.41 -12.03
5 38.64 56.93 -10.60 5 45.27 67.09 -13.09
Before Financial Crisis After Financial Crisis Criteria of selecting optimal numbers of clusters
Process
Efficiency Assets
Utilization Technology Exploration
103 75 26
A1 A2 A3
Factor1 Administration Process
Efficiency 0.558223 -0.297808 -1.352358 Factor2Assets Utilization
Arrangement -0.258562 0.408263 -0.153381 Factor3 External
Relationship 0.361814 -0.336624 -0.462312 Factor4 Technology
Exploration 0.287631 -0.834550 1.267892 Cluster
Before Financial Crisis
15
Figure 3 Comparison of factor scores mean among clusters before crisis
Before financial crisis:
(1) Cluster A1: Process Efficiency Group
More than half of the firms were classified into this group. The capability of managing Administration Process Efficiency and the capability of External relationship illustrated strong influence on the cluster. While further evaluating the variables mean of each cluster, it was found that this group had the highest C_S ratio but still maintained certain level of margins. The contrast indicated that there had some forces other than products helping the firms of the group to create margins. The higher Administration Process Efficiency factor scores projected that the group enjoyed the competitive advantage in managing administration process efficiency and prevented any unnecessary of resource wastes. However, in the economic downturn, the efficient advantage cannot guarantee positive ROIC. Even though the group preserved certain level of operation margin, its low margin was offset by the high portion of DA expenses and TAX costs.
(2) Cluster A2: Assets Utilization Group
The assets utilization group here was composed by smaller size of companies. The cluster showed strong capability of utilizing assets. The group was featured by the highest FATO rate, and lowest DA expense. Moreover, the group was the only cluster who
maintained the highest operation margins and enjoyed positive ROIC even in the economic downturn. The group had higher C_S but maintained most profitable margins. In sum, the group was currently harvesting its previous investment and hard efforts from the point of its lower DA_S ratio and R&D cost.
(3) Cluster A3: Technology Exploration Group
This group was classified according to its eminent capability of technology exploration. In figure 3, the group (A3) presented the highest SGA_S and RD_S among three groups, and the two high values indicated that the group was trying to balance resource in exploration and exploitation activities. This group was composed by the larger firms
16
among all 3 groups because its employee size was the largest. Its INT_S was remarkable high but TAX_S and the C_S was the lowest. The substantial investment in the RD and SGA seemed significantly offset the group’s profits because it had the lowest ROIC and EBIT_S.
Table 8 Averaged factor scores in each cluster after crisis
Figure 4 Comparison of factor scores mean among clusters after crisis
After financial crisis:
(1) Cluster B1: Process Efficiency Group
Half of the firms was classified into this group which demonstrated competitive advantage in managing Process Efficiency. More than 80% of the members were notably overlapped with those was in cluster A1. Despite the group had the highest C_S, it had the lowest SGA cost and highest INTO rate. The group enjoyed a moderate ROIC rate. These data implied that the group managed lower product margin but they succeeded in exerting control over process efficiency to squeeze profits.
(2) Cluster B2: Ambidexterity Prone Group
This group was composed by large size firms and presented potential to generate the highest ROIC and profit margins, even though its administrative efficiency and asset
utilization were not salient among groups. The notable RD_S and SGA_S depicted the features of ambidextrous organizations. The group illustrated lowest C_S ratio and highest EBIT_S.
Process
Efficiency Ambidexterity
Prone Assets
Utilization
104 53 47
B1 B2 B3
Factor1 Ambidextrous
Value Creation -0.377905 1.079662 -0.381277 Factor2 Assets Utilization
Arrangement -0.326341 -0.286999 1.045754 Factor3 Relationship
Leverage -0.509585 0.264011 0.829879 Cluster
17
(3) Cluster B3: Assets Utilization Group
The assets utilization group here was composed by smaller size companies. The capability of Assets Management was standing out to feature the group. The group presented the highest FATO, ARTO and INT_S but the lowest RD_S, DA_S and TAX_S.
Obviously, the group presented the poorer performance due to its lowest EBIT_S and ROIC.
The group was inferred to enjoy the benefits of its previous investment due to its lowest DA_S ratio and high FATO ratio.