37 / 414
臺大管理論叢
第
27
卷第
1
期
37
which 16 were discarded because of incomplete responses. Therefore, 406 questionnaires
were analyzed in this study. Non-response bias was not a factor because the t-tests of group
means revealed no differences between the nonrespondents and the respondents in the
sample (Armstrong and Overton, 1977).
Respondents varied in gender (male, 42.9%; female, 57.1%), age (less than or equal to
30 years of age, 22.4%; 31-40 years of age, 24.1%; 41-50 years of age, 27.6%; greater than
or equal to 51 years of age, 25.9%), marital status (single, 18.7%; married without children,
32.0%; married with children, 49.3%), education level (less than or equal to a high school
diploma, 7.4%; college degree, 24.4%; university degree, 48.7%; graduate school, 19.5%),
annual income (less than or equal to US$15,000, 11.6%; US$15,001-US$30,000, 18.7%;
US$30,001-US$60,000, 43.8%; greater than or equal to US $60,001, 25.9%), and
relationship length with their banks (less than or equal to 5 years, 22.7%; 6-10 years, 34.0%;
11-20 years, 29.8%; greater than or equal to 20 years, 13.5%).
3.2 Measure Development
All the measures used in this study were adapted from existing scales. All constructs
used a five-point Likert-type scale, with the descriptive equivalents ranging from Strongly
Disagree (1) to Strongly Agree (5). The three items used to measure asset specificity came
from Anderson and Weitz (1992) and Selnes and Sallis (2003). The measure of quality of
customer interaction included four items taken from Homburg and Stock (2004). Two items
to measure decision-making uncertainty were adopted from Gao et al. (2005). Confidence,
social and special treatment benefits were each measured with three items provided by
Gwinner et al. (1998). For the measurement of share of wallet, one item was adapted from
De Wulf, Odekerken-Schröder, and Iacobucci (2001), while the measure of co-production
included four items taken from Auh et al. (2007). A commonly used method for measuring
share of wallet is asking customers to report the percentage of the purchases of the focal
service provider that they normally make (De Wulf et al., 2001). Respondents indicated their
approximate share of wallet for the surveyed bank (measured on a continuous scale from 1%
to 100%). A descriptive analysis of the data revealed that 19.7 percent of customers held 1
percent to 25 percent of their investment products in the surveyed bank, 16.3 percent held 26
percent to 50 percent, 50.2 percent held 51 percent to 75 percent, and 13.8 percent held 76
percent to 100 percent. In particular, Wirtz et al. (2007) used a two-item self-reported scale
(Percentage Figure and Likert-type Scale) to determine share of wallet. The percentage
figure was converted to a seven-point Likert scale and then combined with the second item