Determining the appropriate unit of analysis is an important issue in organizational research. This often requires considering multiple, nested levels. Ignoring the nested structure of workforces violates statistical assumptions and ignores factors which may influence outcomes. It also leads to group level constructs such as leadership, climate and culture potentially being considered only at the individual level which may not be theoretically appropriate. However, the complexity of the modern workplace often makes identifying meaningful groups and subgroups difficult.
This presentation describes the methods used by the Australian Public Service Commission as it attempted to identify organizational subgroups suitable for multi-level analysis in a large data set: The 2015 Australian Public Service (APS) Employee Census. While this dataset contains a large number of individuals (n=91,937), it contains only limited information about their position in their organizations structure and hierarchy. It also covers employees from 100 independent organizations, each of which has shaped its internal structure around its own needs and context. The anonymous nature of the employee census also makes it impossible to accurately identify relationships between employees, such as identifying supervisors and their subordinates.
To identify workgroups suitable for further analysis, researchers needed to identify what level of the organization represented a meaningful unit of analysis. This was shaped by theoretical, practical and business concerns. Identifying such units in the data required a number of considered assumptions to be made about individual respondents. Decisions needed to be made which would influence the robustness of the analyses conducted using the data, such as the minimum size of the work units to be used in future analyses. Finally, attempts needed to be made to verify the validity of these units.
The end result of this process is a number of discrete organizational units being identified which allows the APS to look at work-place factors in a way not previously possible.