Animal shelters strive to permanently place dogs in homes with individuals and families. Yet, one long-term retrospective study estimates the percentage of dogs returned to shelters after adoption is approximately 9 percent. Returning shelter dogs impacts the resources available to care for additional animals. Identifying the root causes of returns to the shelter can inform programming targeted at reducing this phenomenon. 

Using a data set for a large shelter system in New York, comprised of 3,465 first-time shelter dogs tracked for six months following intake, we identified potential areas for interventions to reduce return rates of adopted dogs through multiple models. Logistic regression in combination with stepwise variable selection was the model reported in this research project to explore the relationship between the probability of a dog being returned to a large number of covariates. 

We found that a dog’s age, tendency for aggressive behavior, breed, length of stay, and shelter geography are related to the probability that a dog will be returned. Additionally, we found that transporting dogs between shelters was not related to return probability. JMP’s easy-to-use, interactive analysis and visualizations helped the client, who had little statistics training, understand and ultimately execute the analysis on her own. These findings will guide the allocation of resources to interventions, such as educational materials or training programs, that may help reduce overall return rates.