Work groups and employee attitudes: issues in the analysis of aggregated data

The present study investigated units of analysis issues in the measurement of work group climate. The responses of over 2,500 employees in 241 work groups to both a set of 20 descriptive work group climate items and 13 affective job satisfaction items were decomposed into between and within group components. The impact of the nesting in the data upon item dimensionality was examined by performing iterated principal factor analyses on each set of items at each level of analysis (total sample, between groups, and within groups). Results demonstrated that all three factor patterns were similar for the climate items, while, for the satisfaction items, the between and within group patterns were quite different.

A (persons:groups) x items generalizability analysis conducted on each set of items demonstrated larger variance due to groups for the climate than for the satisfaction items. Within-group internal consistency estimates were slightly lower than the total sample estimates for the climate items; they were essentially identical for the satisfaction items. For between group analyses, the decision to treat persons as a fixed versus random facet was found to have a much larger impact upon estimated generalizability than the corresponding decision for the item facet. Even with persons random, the climate items demonstrated acceptable generalizability, while the satisfaction items displayed adequate generalizability only with fixed persons.

Work group membership was demonstrated to be significantly related to both employee attitudes and self ratings of performance, replicating previous research. This research was extended by demonstrating that the group effects were found even after removing the effects due to a number of employee individual differences.

The pattern of results obtained from correlating the climate and satisfaction scales with other variables at the three levels of analysis demonstrated the discriminant validity of the climate perceptions. Work group climate, as a situational variable, explained a sizable proportion of the relationship between group membership and (self-reported) performance.

It is concluded that investigation of components of variance and covariance within and between groups has relevance to both practical and conceptual issues in I/O psychology, and further use is encouraged.