Causal Inference with Group-Based Trajectories and Propensity Score Matching: Is High School Dropout a Turning Point?

Thumbnail Image


umi-umd-3334.pdf (1.21 MB)
No. of downloads: 2260

Publication or External Link






Life course criminology focuses on trajectories of deviant or criminal behavior punctuated by turning point events that redirect trajectories onto a different path. There is no consensus in the field on how to measure turning points. In this study I ask: Is high school dropout a turning point in offending trajectories? I utilize two kinds of matching methods to answer this question: matching based on semi-parametric group-based trajectory models, and propensity score matching. These methods are ideally suited to measure turning points because they explicitly model counterfactual outcomes which can be used to estimate the effect of turning point events over time.

It has been suggested that dropout is the end result of a process of disengagement from school. In order to assess the effect of the event of dropout, it is necessary to separate dropout from the processes that lead to it. The extent to which this is accomplished by matching is assessed by comparing dropouts to matched non-dropouts on numerous background characteristics. As such, it is desirable to use a wide range of measures to compare the two groups.

I use the National Longitudinal Survey of Youth 1997 to address this question. Delinquency is measured in two ways: a six-item variety scale and a scale based on a graded-response model. Dropout is based on self-reports of educational attainment supplemented with official transcripts provided by high schools. Because of the breadth of topics covered in this survey, it is very well-suited to matching methods. The richness of these data allows comparisons on over 300 characteristics to assess whether the assumptions of matching methods are plausible.

I find that matching based on trajectory models is unable to achieve balance in pre-dropout characteristics between dropouts and non-dropouts. Propensity score matching successfully achieves balance, but dropout effects are indistinguishable from zero. I conclude that first-time dropout between the ages of 16 and 18 is not a turning point in offending trajectories. Implications for life course criminology and dropout research are discussed.