• 검색 결과가 없습니다.

V. Data analyses and Results

1. Measurement Model Validation

In order to validate measurement model, this study generated tables of measurement model with item reliability, convergent validity and discriminant validity by using PLS. Convergent validity consists of composite reliability values for each constructs, average variance extracted values for each construct and factor loading value for each measurement instrument.

Table 26 shows the result of reliability and convergent validity of this sample set. All factor loading values in Table 26 are above 0.6, and several measurement items are dropped while validating factor loading value.

There were five measurement items in the exploitation. E1 and E5 are dropped. Due to low factor loading for every possible item combination, E1 is dropped. E1 item ask organization’s degree of process innovation. Even though there were footnotes describing the definition of process innovation, many respondents were asked meaning of process innovation. this factor could lower the factor loading value. E5 is dropped because factor loading value of E5 shows low factor loading value for every possible item combination. E5 item question asked the extent of the entry barrier of organization’s industry.

Many respondents did not understand the meaning of entry barrier. This could also affect to the factor loading value.

Table 26. Reliability and convergent validity

Construct Measure Loading C.R AVE

Exploitation E2 0.7670

0.893 0.735

There are five measurement items in business governance decentralization. BIZ1, BIZG3 and BIZG4 are dropped during the factor loading validation process. Due to low factor loading

values for every possible item combination, BIZG1 is dropped.

BIZG1 item asks about business unit’s decision freedom. Original questionnaire is written in English. During the translation, some context lost occurred which could lower the factor loading value.

BIZG3 and BIZG4 item asks about budget rights freedom and evaluation freedom. BIZG3 and BIZG4 showed the low factor score for every possible item combination, but reason why the factor loading value is low is uncertain.

There were eight measurement items in Environment instability [Competition]. COMP1, COMP3, COMP4, COMP8 are dropped during the factor loading validation. COMP1 item asks about the competitor count level. However, many respondents did not know how many competitors are supposed to be considered high.

This could lower the factor loading value. COMP3 item asks about the highness of fixed cost and inventory cost. However, many respondents were confused the definition of fixed and inventory cost. This could lower factor loading value. COMP8 ask “Exit barrier” of their industry. However, many respondents confused definition of “exit barrier”. This could lower the factor value.

There were 3 measurement items in information intensity. Only IA1 is dropped while factor loading validation. IA1 ask what types of the task the organization usually process (uniqueness or repeatability). Many respondents were confused about the concept of unique task and repeated task. This could lower the factor loading value

There are six measurement items in business competency, and only BIZCOM4 is dropped during factor loading validation.

BIZCOM4 asks about the organization’s ethic issue, and this issue was later considered to be sensitive issue for the respondents. This survey was done under face-to-face condition.

The respondents could have felt the uneasiness when answering this question item. This could also lower the factor loading value.

There were five measurement items in Participative Decision-Making Scale, and only PART1 is dropped during the factor loading validation. PART1 asks about the decision making delegation level. In terms of context level, the context of delegation is relatively far from participative decision making scale than other variables. This could lower the factor loading value.

There are seven measurement items in absorptive capacity, and only ABS2, ABS3 are dropped during the factor loading validation. ABS2 and ABS3 asked the degree of cultural misunderstanding level and its frequency. This was also later considered to be a sensitive matter to answer, this survey was conducted under face-to-face condition as well, likely the respondents could have felt uneasiness when answering this item.

This could lower factor loading value.

Recommended values for construct reliability is greater than 0.7.

The construct reliability values are all above 0.7 levels.

Recommended values for average variance expected is above 0.5. All AVE variable is greater than 0.5.

Loading and Cross loading table is generated to validate measurement model. Appendix 1 shows loading and cross loading result. All cross loading values are in suitable range except POWR1 item.

There were three measurement items in power no-sharing. Only POWR1 is dropped during the loading and cross loading validation. The cross loading result shows that this item is more related with participative decision making scale, but the cause of this problem is uncertain.

Measurement items of IT governance decision areas consist of single item. Therefore, IT principle, IT architecture, IT infrastructure, Business application need and IT investment are not subject to test factor loading and cross loading.

Each IT decision area is coded 1 to 5 based on Davenport and Prusak (1997) (See figure 6). Greater IT governance arrangement values represent higher decentralization. Since business monarchy and IT monarchy are have same centrality, those archetypes are coded as 1. Duopoly archetype describes the form of decision making that made by IT personnel and joining executives or business unit manager. Therefore, Duopoly archetype can be categorized as more centralized form rather than federal archetype. Duopoly archetype is coded as 2. Federal archetype is coded as 3, Feudal is coded as 4 and Anarchy coded as 5. However, in this study sample, there was no case for anarchy archetype.

After the construct validation process, there was improvement on measurement instrument consistency and validity. All screened measurement items are accepted as suitable constructs.