Background: A lack of physical activity can lead to long-term health problems for youths (aged 5–18) worldwide. Built environment characteristics are increasingly being recognized as important factors affecting transport walking, a reliable source of overall physical activity for youths. However, the relationship between built environment characteristics, especially residential density, and youths’ walking for transport purposes (transport walking) remain largely inconclusive, due to limited variation in built environment variables and an assumption of linear association. Methods: In this study, we explore the dose-response relationship between built environment characteristics and transport walking for youths in Hong Kong, a city with large variations in residential density. Detailed transport walking behaviors, such as the number of trips and walking duration, were extracted from the 2011 Hong Kong Travel Characteristics Survey (N = 13,287; aged 5–18). Neighborhood socioeconomic status, age, gender, household income, and household vehicle ownership were controlled in the generalized additive mixed models (GAMMs) of the built environment-transport walking associations. Results: We found inverted U-shaped associations between population density and both the odds and total minutes of transport walking. Population density within 30,000–60,000 persons/km2 is optimal to promote transport walking for youths. In addition, the number of recreational facilities and retail shops were positively associated with likelihood of engaging in transport walking, and number of bus stops was negatively associated with transport walking. Conclusion: We add new empirical evidence on the significant and non-linear relationship between urban density and transport walking. Although increasing urban density in already densely developed cities may not be an effective intervention strategy to increase transport walking in youths, such strategy may still be effective in other less dense areas.