The application of HLM to the analysis of the dynamic interaction of environment, person and behavior - hierarchical linear modeling - A Special Issue: Focus on Hierarchical Linear Modeling

Journal of Management, Nov-Dec, 1997 by Jeffrey B. Vancouver

Over time, a person's behavior in an environment changes the environment and the person. These changes affect subsequent behaviors. Static conceptual and empirical models of environments, persons and behavior cannot capture the processes involved. Dynamic models of the complex interaction between environments, people and behaviors, and analytic techniques for testing these models are required. Conceptually, models have emerged that describe humans as self-regulating living systems, which both affect and are affected by their environments (Bandura, 1991; Carver & Scheier, 1981; Ford, 1987; Karoly, 1993; Latham & Locke, 1991; Powers, 1973; Vancouver, in press). While these models provide a way to think about the phenomenon, little empirical data has been generated that directly test many of the propositions in the models because of the complexity of the data analysis requirements. A key source of complexity is found in the influence between multiple levels of analysis (Nesselroade & Ford, 1987). Fortunately, hierarchical linear modeling (HLM) is an analytic technique specifically developed to analyze multi-level models (Bryk & Raudenbush, 1992). In this study, I apply HLM to analyze models derived from a self-regulating framework to examine some of the interactions between the environment, person and behavior. The results of the application demonstrate the importance of both the self-regulating models and the use of appropriate analytic techniques for assessing such models.

The self-regulating framework's underlying model is the negative feedback loop, in which a task is monitored by comparing a goal for that task with perceptions of task progress. If a discrepancy exists between the perception and the goal, an individual will attempt to reduce the discrepancy by either altering the goal or applying resources to task-related behavior (Austin & Vancouver, in press; Klein, 1989; Lord & Hanges, 1987). Each discrepancy-reducing loop is one among many that defines a complex hierarchical structure (Carver & Scheier, 1981; Miller, Galanter & Pribram, 1960; Powers, 1973). The simple loop has been used as an explanatory mechanism for single, goal-striving processes (Campion & Lord, 1982; Naylor & Ilgen, 1984). Consideration of multiple loops has been advocated as an explanatory mechanism for more complex goal-striving processes (Keman & Lord, 1989; Klein, 1989; Lord & Hanges, 1987), but is rarely examined due to analysis difficulties (Nesselroade & Ford, 1987). Specifically, Nesselroade and Ford note that:

In living systems, where it is the organization of the variables in patterns of mutual influence that sustains life and development, the power and generality of a multivariate approach is not only attractive; it is essential. Moreover, because living systems are open systems, it is impossible to understand their functioning separate from their contexts. This requires a multivariate approach involving combinations of person and environment variables (p. 59).

As the complexity of interacting negative feedback loops increases, the consideration of loop processes and the dynamic interaction between the environment, person and behavior becomes critical (Bandura, 1986; Kernan & Lord, 1989). Individuals 1) act on and monitor many goals, 2) react to the observable consequences of those actions and the opportunities and constraints provided by the environment, and 3) assess strategies and resources needed to achieve the goals (e.g., Naylor, Pritchard & Ilgen, 1980). Together, this interaction produces a dynamic but potentially tractable, interplay of factors (Austin & Vancouver, 1996). In this study, I examined that dynamic interaction by manipulating or measuring aspects of the environment, person and behavior over time as they relate to goal striving.

Past Research

Past research on goal striving has focused on performance in single goal situations (see Locke & Latham, 1990, for a review). This research has clearly shown the merit of defining a specific and challenging goal to focus attention and other resources toward performance on that goal. Yet, the role of feedback (Erez, 1977), goal acceptance and commitment (Hollenbeck & Klein, 1987), strategies (Wood & Locke, 1990), and conflict among goals (Locke, Smith, Erez, Chah & Schaffer, 1994) have arisen as important variables when attempting to understand motivated behavior in complex settings. In fact, most work situations require individuals to balance many goals and tasks (Kernan & Lord, 1989; Tsui & Ashford, 1994). Less is understood about the trade-offs individuals make in terms of resources and the interaction of the environment, person and behavior in these situations (Kernan & Lord, 1989, 1990; Locke & Latham, 1990; Wood, Bandura & Bailey, 1990).

Studies using multiple goals are rare (Locke & Latham, 1990), and single goal studies often belie the processes involved. For example, Locke (1982) reported an asymptotic relationship between goal level and performance. Specifically, he found that as the goal level reached impossible levels, the positive goal level-to-performance relationship tended to flatten at an ability ceiling. This indicated that subjects continued to strive for the goal. However, in a multiple goal situation, the impossible goal might be rejected in favor of an attainable goal that needs the resources used by the original goal. For instance, a common distinction is made between quantity and quality (e.g., Bavelas & Lee, 1978; Woodworth, 1899) aspects of a task. A common paradigm is to set a goal related to the number of satisfactory products, thus combining the quantity and quality aspects (Locke & Latham, 1990). Yet, when they are separated, many studies have found that a specific, difficult quantity goal will often cause quality to suffer (Locke & Latham, 1990), or that a specific, difficult quality goal will cause quantity to suffer (Terborg & Miller, 1978).