Social scientists have long been interested in the role that age, period, and cohort effects have in influencing longitudinal trends in a variety of research areas. However, because the three effects are linear derivatives of one another traditional statistical models cannot simultaneously isolate their unique effects due to their perfect confounding, an issue known as the identification problem. One recent solution to the identification problem is the estimation of age, period, and cohort effects through the use of cross-classified random effects modeling applied to repeated cross-sectional data. This approach takes advantage of the multilevel modeling framework and proposes that age can be treated as an individual-level (level-one) variable and period and cohort effects can be treated as categorical variables that define cluster membership at level-two, allowing all three effects to be simultaneously estimated. Using a Monte Carlo simulation study, this dissertation investigated two broad areas of methodological issues related to the performance of the model. First, four factors were manipulated that may influence the accuracy of estimates in the model: the number of survey years available for analysis; the cohort grouping employed in the model; the variability of the period effect; and the variability of the cohort effect. Second, four model fit indices were evaluated to determine their performance in detecting the cohort

grouping underlying the structure of the dataset used in the analysis. The results of the simulation study indicated that the cohort grouping used in the model heavily influenced the accuracy of many of the model parameters, while the number of survey years available for analysis most directly influenced the accuracy of the period-level predictor. To a lesser extent, variability in the period and cohort effects impacted the accuracy of a few of the model parameters, but tended to occur in scenarios where bias was exhibited due to the cohort grouping. Importantly, all four of the model fit indices performed well in detecting the cohort grouping underlying the dataset. Implications of these findings are discussed for applied researchers, funders and administrators of repeated cross-sectional surveys, and life course theorists. Areas for future methodological research are also provided.