Panel Design

A panel design is used when researchers sample a group, or panel, of participants and then measure some variable or variables of interest at more than one point in time from this sample. Ordinarily, the same people who are measured at Time 1 are measured at Time 2, and so on. The successive measures are commonly referred to as waves. For example, a three-wave panel study would measure the same sample of participants on three separate occasions. The amount of time in between measurements is known as the interwave interval. The use of multiple measures on the same variable(s) over time allows for an assessment of longitudinal changes or stability on the variables of interest.

Perhaps the simplest and most common version of the panel design is the two-wave, two-variable panel, or cross-lagged panel design. In this design, two variables (X and Y) are measured at two discrete points in time from the same sample of participants. The procedure allows the researcher to test a series of different associations depicted in Figure 1.

The horizontal paths in the cross-lagged panel model represent stability coefficients. These values indicate how stable the X or Y variables are over time. If analyzed with regression, the larger these regression coefficients are, the greater the stability of the X or Y constructs over the course of the interwave interval. The diagonal paths are of particular interest as they are often compared to determine if X1 → Y2 is greater or less than Y1 → X2. This comparison often forms the basis for making causal inferences about whether X causes Y, Y causes X, or perhaps some combination of the two. Although cross-lagged panel designs are very common in the literature, there is a considerable controversy about their actual utility for testing causal inferences. This is because a valid causal analysis based on a cross-lagged panel design involves the absence of measurement error, correct specification of the causal lag between T1 and T2, and the assumption that there is no third variable, Z, that could exert a causal impact on X and Y. Rarely, if ever, are all of these assumptions met.

Despite its limitations, the cross-lagged panel study illustrates several key benefits of panel designs more generally. First and foremost, one of [Page 990]the fundamental prerequisites for establishing that X has a causal effect on Y is the demonstration that X occurs before Y. For example, if smoking is assumed to cause lung cancer, it is important to be able to demonstrate that people first smoke while otherwise free of cancer, and then lung cancer appears at a later point in time. Merely demonstrating that X precedes Y in a longitudinal panel study does not prove that X causes Y, but it satisfies this vital criterion, among several others, for establishing a causal relation. Another important quality of panel designs is that they do not confound interindividual growth with interindividual differences in intraindividual growth, the way that cross-sectional studies do. Because the same people are measured at multiple points in time, it is possible to statistically model changes within the same person over time, differences between people, and interindividual differences in the rate and direction of change over time.

Figure 1 Basic Structure of Cross-Lagged Panel Design

...