Article in Group & Organization Management · January 014 citations reads 13,031 authors: Some of the authors of this publication are also working on these related projects

Job autonomy. Job autonomy was assessed using the nine-item job autonomy 
subscale from Morgeson and Humphrey’s (2006) work design questionnaire. 
A sample item is “the job allows me to make my own decisions about how to 
 by guest on July 31, 2014
gom.sagepub.com
Downloaded from 

---

Ho and Nesbit 
399
schedule my work.” Responses were given on a 6-point response scale where 
1 = strongly disagree and 6 = strongly agree. The alpha coefficient was .94.
Supervisors’ performance rating. Five items developed by Goodale and Burke 
(1975) were used to measure five performance dimensions related to organiz-
ing and planning, reliability, adaptability, productivity, and quality of work. A 
sample item is, “This subordinate produces a quantity of work that meets the 
established standards.” Furthermore, one self-developed item was used to 
measure the dimension of initiative, “This subordinate is willing to accept 
extra assignments and originates action without constant supervision.” The 
supervisor was asked to rate his or her subordinates on each of these six per-
formance dimensions on a scale from 1 (poor performance) to 4 (average 
performance) to 7 (excellent performance). The alpha coefficient was .92.
Objective work performance. Objective work performance was measured in 
terms of the annual premium for insurance policies sold from March to May 
2010. The annual premium is the amount of money a customer pays for insur-
ance coverage of 1 year. Every insurance organization in Hong Kong collects 
data on this measure, which are used as the basis for calculating sales reve-
nue. Hence, annual premium is a standard objective performance measure 
used by each of the five insurance companies involved in this study. All of 
these insurance companies are similar in size and also have similar insurance 
products, product variety, organization structures, and policies as well as 
have the same sales territory.
Job satisfaction. Four items developed by Manz (1981) were used to measure 
job satisfaction. A sample item is, “My work gives me a sense of satisfac-
tion.” Participants responded using a response scale ranging from 1 (strongly 
disagree) to 6 (strongly agree). The reliability was .90.
Control Variables
Organizational tenure (measured in years), gender, educational level, and age 
were included as control variables in our analyses. Previous research has 
shown these variables to be related to the outcome variables of work perfor-
mance and job satisfaction (e.g., Hom & Griffeth, 1995; Spector, 1997; 
Sturman, 2003). As our respondents worked in four different industries in 
Hong Kong and mainland China, industry categories and location were also 
considered as control variables. Manufacturing, Insurance, and Public 
Transportation were dummy variables, with Engineering as the comparison 
category. As all respondents from the manufacturing industry only worked in 
 by guest on July 31, 2014
gom.sagepub.com
Downloaded from 

---

400 
Group & Organization Management 39(4)
mainland China, a dummy variable created for location (0 = Hong Kong; 1 = 
China) is exactly the same as that of manufacturing industry (0 = other indus-
tries in Hong Kong; 1 = manufacturing industry in China). To avoid such 
redundancy, location was omitted as control variables in our study.
Conscientiousness. Conscientiousness was assessed using the subscale within 
the Big Five Inventory (BFI; John, Donahue, & Kentle, 1991). This subscale 
consists of nine items. A sample item is, “I make plans and follow through with 
them.” Responses to each item were anchored by a 6-point scale ranging from 
1 = strongly disagree to 6 = strongly agree. The alpha coefficient was .81.
Results
Prior to testing the hypotheses, we conducted a series of CFAs to examine 
the discriminant validity of the measures. To examine the distinctiveness of 
the study variables, we conducted a CFA to examine the five key variables 
in our model, namely self-leadership, job autonomy, job satisfaction, con-
scientiousness, and performance ratings. To reduce the number of parame-
ters in the structural equation modeling, the item parceling method 
recommended by Bogozzi and Edwards (1998) was used on the variable of 
job autonomy as it consisted of more than seven items. Both job autonomy 
and conscientiousness were modeled using three parcels, with items 
assigned to each parcel randomly. Ten parcels were also created for self-
leadership by summing and averaging the items within each subscale to 
create 10 composite indicators. Next, the five-factor model with the items 
or parcels assigned to the five corresponding variables was used as the 
baseline model. Five alternative models were examined against the baseline 
five-factor model. As shown in Table 1, this baseline five-factor model fit 
the data reasonably well, χ
2
(289) = 840, CFI = .92; GFI = .88; RMSEA = 
.07; SRMR = .07, whereas the other alternative models all exhibited signifi-
cantly worse fit than the baseline model. These results justified the exami-
nation of self-leadership, conscientiousness, job autonomy, job satisfaction, 
and job performance as distinct constructs.
Common Method Variance
Because we collected the data for the measures of self-leadership and job 
satisfaction from the same participants at the same time, we used the 
marker variable approach (Lindell & Whitney, 2001) to determine the 
impact of common method variance (CMV). Information reported by our 
respondents on a theoretically unrelated construct (i.e., external locus of 
 by guest on July 31, 2014
gom.sagepub.com
Downloaded from 

---

Ho and Nesbit 
401
control; Spector, 1988) was used as a marker variable. One can conclude 
that CMV is not an issue in the study if the significant zero-order correla-
tions for the variables remain statistically significant even when the marker 
variable is partialled out. Based on our analyses, there were no statistically 
significant changes in the zero-order correlations of self-leadership with 
job satisfaction even after partialing out the marker variable from the cor-
relation matrix. As such, we concluded that CMV was unlikely to have 
been a problem in our study.
Descriptive Statistics
Table 2 presents means, standard deviations, and zero-order Pearson correla-
tions among all variables in this study. As expected, self-leadership was posi-
tively and significantly related to the outcome variables of supervisor 
performance rating (r = .21, p < .01), objective work measure (r = .20, p < 
.05), and job satisfaction (r = .29, p < .01). Conscientiousness was also posi-
tively and significantly associated with self-leadership (r = .29, p < .01), as 
well as supervisor performance rating (r = .21, p < .01) and job satisfaction 
(r = .23, p < .01).