Multilevel Research

Many studies in psychology are concerned with phenomena and theories that are multilevel in nature. The central defining feature of multilevel research is that the data is hierarchically structured or nested, that is, the units of observation at one level of analysis are nested (meaning grouped) within units at a high level of analysis. In psychology, the most common examples of hierarchically structured data may be found in organizational and educational research. For instance, a study may consist of employees nested within work teams resulting in a two-level research with the individual employees at the lower level (Level 1) and teams at the higher level (Level 2). As another example, a study consisting of students nested within classes and classes in turn nested within schools would result in a three-level research with individual students at Level 1, classes at Level 2, and schools at Level 3. Many constructs and situations of interest to psychologists can be conceptualized and studied from a multilevel perspective.

From the perspective of multilevel research, the phenomenon under study is characterized as part of a dynamic system, with critical antecedents, processes, and outcomes conceptualized and measured at multiple levels of analysis (Chan, 1998b). For example, employee job satisfaction (Level 1) may be affected by variables at the same level, such as employee personality traits, or variables at the work team level (Level 2)—which is a higher level—such as team size, averaged tenure of the employees in the team, and the within-team variation in employee intelligence levels. Employee job satisfaction may also affect outcome variables at multiple levels, such as intent to quit (Level 1) and team performance (Level 2). More complex relationships, such as the interaction effect of a Level 1 variable and a Level 2 variable on a Level 1 (or Level 2) variable, may also exist. As explained later in this chapter, to the extent that multilevel effects exist, conceptualizations, measurement, analyses, and inferences that adopt a single-level approach and disregard the hierarchical nested structure of multilevel data are problematic.

Although the multilevel nature of much psychological research has been acknowledged for a long time (e.g., Fleishman, 1953), it is only in the last decade that more researchers have begun to systematically explicate the many complexities inherent in multilevel research and apply multilevel approaches and analyses to empirical studies in substantive areas, particularly those in educational and organizational psychology. Complexities in multilevel research involve problems in validation of multilevel constructs, decomposition of sources of variance in responses, and the associated issues of multilevel measurement and data analysis. One major reason for the recent increase in attention to multilevel research is that recent significant [Page 402]advances in data analytic techniques and development of statistical software have markedly increased our ability to model and empirically test complex multilevel relationships that were previously restricted to the conceptualization stage. Unfortunately, the majority of methodological work on multilevel research, particularly those addressing data analytic issues, are highly technical and not easily understood by beginning researchers or researchers who may not be methodological experts. Hence, the purpose of this chapter is to provide a nontechnical introduction to the major issues in multilevel research.

...