The Use of Propensity Scores as a Matching Strategy
Lindsay John, Ph.D.1,
Robin Wright, Ph.D.1*,
Eric K. Duku2,
and
J. Douglas Willms, Ph.D.3
1 McGill University
2 McMaster University
3 University of New Brunswick
* To whom correspondence should be addressed. E-mail: robin.wright{at}mcgill.ca.
 |
Abstract |
|---|
Objectives: This study reports on the concept and method of linear propensity scores used to obtain a comparison group from the National Longitudinal Survey of Children and Youth to assess the effects of a longitudinal, structured arts program for Canadian youth (aged 9 to 15 years) from low-income, multicultural communities. Method: This study compares 183 children in a community arts project to 183 children from a national longitudinal survey using propensity score matching. The variables included baseline scores of child-rated conduct problems, indirect aggression, emotional problems, self-esteem, and prosocial behavior and child gender, person most knowledgeable (PMK) education, PMK marital status, household income, and family functioning. Results: Mean score comparison showed that the groups were very similar on all covariates. Conclusions: Propensity score matching offers an alternative to true randomization that is cost-effective and convenient, particularly important for social work research in community-based organizations with a limited budget.
