Research Methods, Longitudinal Studies

Longitudinal studies examine change over time. Studies that measure the same people, organizations, media, or other entity repeatedly over time are longitudinal, such as a study of the same children who are sixth graders this year and were fifth graders last year. Studies that measure new individuals or entities at different time points are cross-sectional, for example, a study of this year's sixth graders compared to last year's sixth graders.

How change occurs over time is critical to many theories and questions regarding children, adolescents, and the media. Taking a developmental perspective involves noting how the children change over time in their relationship with the media. For example, as children age, how do their reactions to violent movies change?

Theories that aim to explain processes or understand causality also benefit from attempting to model how the process changes over time. Take, for example, the question of whether exposure to risky sex in media portrayals affects the likelihood of teens engaging in risky sex. If a researcher collects data at one time point and finds a correlation between media exposure and risky sex behavior, the process is unclear. Did exposure to risky sex portrayals lead adolescents to model that behavior? Or, did a propensity to engage in risky sex lead teens to seek out media likely to contain sexual content? With longitudinal data, it may be possible to test both hypotheses to see which explanation best fits the data.

To conduct a longitudinal analysis, a researcher needs at least two waves or rounds of data; four or more are better for understanding the process of [Page 725]change. The data could be generated from, for example, repeated observations, surveys, physiological measures, or analyses of media content over time.

Longitudinal studies also require a measure of time that makes sense in the context of the study, such as seconds, weeks, years, sessions, age, or grade. It is important is to measure around the time when the researcher expects change to be detectable. Some analysis methods require the same time interval between measurements (e.g., every 3 months). Other methods allow unique intervals per person, such as the time elapsed since each child's natural exposure to a media message. Some methods also allow explanatory variables to vary over time.

There have been advances recently in the methods available to analyze longitudinal data and the statistical packages that include those techniques. Multilevel modeling can examine simultaneously the predictors for change over time within individuals, differences between individuals, and differences in groupings of individuals. For example, a national study examined the effects of alcohol advertising exposure over time, individual characteristics, and levels of alcohol advertising present in each market to understand which factors contributed to drinking over time. Another analysis method is event history analysis (also known as survival analysis and failure rate analysis), which examines when specific events occur to individuals and what predicts them. For example, when do youth begin and subsequently cease complying with a campaign message?