10.07.2015 Views

Organizational Behaviour Comportement Organisationnel

Organizational Behaviour Comportement Organisationnel

Organizational Behaviour Comportement Organisationnel

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

track assistant, associate, and full professors employed in 12 departments in each of six prominentCanadian universities. To ensure that the chosen departments were typical in terms of paradigmdevelopment, universities with very active research profiles, determined by grant and contractawards, were chosen. In total, approximately 2201 questionnaires were distributed and 649 werereturned for a response rate of 29%. By university, response rates ranged from 25% to 38%. Bydiscipline, response rates ranged from 22% (math) to 40% (geology and anthropology). As therewere no significant demographic differences across universities, and across the pattern ofintercorrelations among variables within each university, the data were combined and examinedas a single sample. The average age of this sample was 48.28 years, with 92% male respondents.Ninety percent of the sample indicated that they were tenured, with 60% at the rank of fullprofessor, 30% at the rank of associate professor, 9% at the rank of assistant professor, while theremaining two percent responded as belonging to the “other” category. As there were norespondents from three departments, 69 departments were available for group-level analyses. Thenumber of respondents from each department ranged between 2 and 21, with an average of 9.07.MeasuresDepartment conflict. Perceived conflict within a respondent’s academic departmentwas measured with Rahim’s (1983) eight-item intragroup organizational conflict scale. For eachitem, the word group was replaced by the word department. Although Rahim’s scale assumesconflict to be unidimensional, later literature (Jehn, 1995; 1997) suggests the presence of threedistinct types of conflict: task, relationship, and process. As a result, an exploratory factoranalysis was conducted to discern the dimensionality of our measure of conflict. Correspondingclosely to Jehn’s descriptions of the nature of conflict, four items (e.g., “there is friendlinessamong members of my department,” reverse coded) were found to load on a factor which welabeled relationship conflict, three items (e.g., “there is difference of opinion among members ofdepartment”) loaded on a second factor labeled task conflict, and one item (“we have lots ofbickering over who should do what job”) loaded on the third factor, labeled process conflict.Support for these exploratory findings was determined by comparing three models usinga confirmatory factor analysis (CFA): single factor, correlated two-factor, and a correlated threefactormodel of conflict. An examination of the overall fit indices indicated that the correlatedtwo-factor model (χ² = 71.81, df = 13, CFI = 0.98, NFI = 0.97, NNFI = 0.96, RMSEA = 0.09) fitthe data better than the single factor model (χ² = 207.42, df = 20, CFI = 0.93, NFI = 0.92, NNFI =0.90, RMSEA = 0.12) and the correlated three-factor model (χ² = 107.22, df = 17, CFI = 0.97,NFI = 0.97, NNFI = 0.94, RMSEA = 0.09). All items in the two-factor model had significantfactor loadings. The internal consistency reliabilities of the relationship and task conflict subscaleswere .88 and .77 respectively, with a zero-order correlation of .70 at the individual level.Based on these findings, two dimensions of conflict, task and relationship, were examined in allsubsequent analyses.To justify aggregation at the department level, we examined within-group agreement ondepartmental task and relationship conflict for each department in our sample using James,Demaree, and Wolf’s (1984) estimation approach (r wg ). For departmental relationship conflict,the median r wg was .70, while that for departmental task conflict was .60, with 50% of thedepartments having values of r wg greater than .60 on both measures. While these estimates maybe less than optimal, they must be examined in light of the reduced number of items in eachmeasure (James et al., 1984). Nevertheless, an ANOVA test (cf. George & Bettenhausen, 1990)suggests that there are discernable between-department differences in relationship conflict (F =2.93, p

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!