This study investigated the effects of exercise frequency, physical education class attendance, fruit consumption, and vegetable consumption on childhood obesity by grade level, controlling for gender. Analyses were conducted using data from the 2012 Youth Risk Behavior Surveillance System (YRBSS), a school-based survey in the United States. T-tests, factorial ANOVAs, and multilevel regression analyses were conducted to determine the relationships between body mass index (BMI) within a school (level 1) and the frequency of fruit consumption, vegetable consumption, exercise, and physical education class attendance, in addition to differences in those variables between schools (level 2). Exercise frequency and gender significantly predicted BMI within schools (level 1), while exercise frequency and the combination of all of the variables were the only significant predictors of BMI among all schools (level 2). Based on these results, we recommend that school-based childhood obesity intervention programs should include and promote a variety of physical activities.