The experimentally-controlled evaluation of BikeTexas Safe Routes to School, the Coalition’s school-based education and encouragement program, is two-tiered. The first goal was to determine the immediate effects of the intervention on psychosocial variables such as motivation and self-efficacy. The next phase is to determine whether or not mediational change translates into behavior change as defined by an increase in walking and bicycling to school.

A post-test directional ANOVA on knowledge at time 2, controlling for knowledge at time 1, indicates a significant difference in knowledge between students in the experimental condition and students in the waitlist condition, F(1) = 16.50, p < .001. Post hoc t-test analyses indicate that there is a significant increase in knowledge in students in the experimental condition [t(65) = -5.57, p < .001]but not for students in the waitlist condition.

Similarly, a post-test directional ANOVA on motivation at time 2, controlling for motivation at time 1 reveals a significant difference in motivation, F(1) = 3.13, p = .04. Post hoc t-test analyses indicate that there is no significant change in motivation within condition. Still, although not significant students in the experimental condition showed a slight increase in motivation [t(60) = 1.40, p = .08] while students in the waitlist condition showed minimal change (p = .75).

There is no significant interaction for change in bicycle self-efficacy, F(1) = .61, p = .22. Both the experimental students [t(58) = 1.98, p = .03] and the waitlist students [t(41) = 3.23, p = .001] displayed a significant increase in bicycle self-efficacy. There is no interaction for physical activity self-efficacy or for total self-efficacy. The fact that both conditions exhibit change in efficacy could indicate that the significant results are spurious or change as a factor of age.

To examine the second phase, directional post-test ANOVAs were conducted on outcome variables. First, mode of arrival was calculated based on mode of transportation to school over a ten-day period. There was no significant difference in walking to school at time 2 based on condition, after controlling for walking at time 1, F(1) = .09, p = .38. There was no significant difference in bicycling to school at time 2 based on condition, after controlling for bicycling at time 1, F(1) = .18, p = .34. When bicycling and walking were collapsed together, there was still no difference in walking and bicycling to school at time 2 after controlling for bicycling and walking at time 1, (p > .05).

The outcome variable was also assessed as a continuous variable on the student survey when students respond individually to mode of arrival over the past five days (0=never, 1=1 to 2 days, 2=3 to 4 days, 4=every day). There is a trend towards a difference in walking and bicycling to school based on condition after controlling for walking and bicycling at time 1, F(1) = 1.84, p = .09. Disappointingly, post hoc t tests indicate that, while there is no significant change in walking and bicycling to school for students in the experimental condition (ps > .05), students in the waitlist condition exhibit an increase in bicycling and walking to school, t(27) = 2.12, p = .04.

When overall weekly bicycling is examined, there is a significant difference in weekly bicycling as exercise at time 2 based on condition after controlling for bicycling at time 1, F(1)=5.84, p = .02. Post hoc independent sample t-test analyses indicate that students in the experimental condition and students in the waitlist condition displayed no difference in bicycling at time 1 [t(114) = .10, p = .92] but students in the experimental condition were bicycling significantly more often at time 2, t(105) = 2.61, p = .01. These results may indicate that the intervention increased interest in bicycling but may not have the breadth to increase ability to bicycle or walk to school. Variables outside of the intervention’s control such as parent’s perception of safety and traffic conditions may moderate the impact. Strengthening this possibility is the fact that there is also a significant difference at time 2 in asking parents to bicycle or walk to school after controlling for asking at time 1, F(1) = 3.26, p = .04. Post hoc paired sample t tests reveal that students in the experimental condition significantly increased asking their parents if they could walk or bicycle to school [t(62) = 2.95, p = .004] while students in the waitlist condition did not [t(41) = -.42, p > .05].

In examining secondary gains, there is a trend for students in the experimental condition to display less time spent in sedentary behavior at time 2 compared to students in the waitlist condition, after controlling for sedentary behavior at time 1, F(1) = 1.32, p = .13. Similarly, there is a small trend for more healthy eating at time 2 in the experimental condition than in the waitlist condition after controlling for eating at time 1, F(1) = 1.40, p = .12.

It remains important to examine the role of socioeconomic status. Independent t-tests were conducted to determine if socioeconomic status, as measured on a school level by the proxy variable of percentage subsidized school lunches, impacted key variables. A school was determined to be low SES if > 50 percent of school lunches were subsidized while a school was determined to be high SES if < 50 percent of school lunches were subsidized. Across condition, there was a significant difference in knowledge at time 2 with students in low socioeconomic status schools (M = .62, SD = .15) displaying more errors in bicycle knowledge than students in medium and high socioeconomic status schools (M = .71, SD = .15; t(107) = 3.05, p = .003. Perhaps explaining why physical activity self-efficacy was not influenced by the intervention, there was a difference in physical activity self-efficacy based on SES. Students in the low SES schools displayed greater physical activity self-efficacy (M = 11.75, SD = 2.91) than students in the high SES schools (M = 10.66, SD = 2.46; t(106) = 2.07, p = .94. There was a trend for students in the low SES schools to display greater total self-efficacy compared to students in the high SES schools t(103) = 1.74, p = .09. There was also a trend for students in the low SES schools to exhibit more walking and bicycling to school (M = 3.64, SD = 1.92) compared to students in the high SES schools (M = 3.08, SD = 1.34; t(92) = 1.62, p = .11).

In examining the influence of weight, students were classified as normal weight if their BMI was below the 85th percentile in terms of age and gender. Students were classified as at risk if their BMI was between the 85th and 95th percentiles and overweight if their BMI was above the 95th percentile. Two categories of weight were used for analyses: normal and overweight (at risk for overweight and overweight collapsed together). Parsing out weight like this creating nearly equivalent sample sizes. There was a trend for people who were overweight to bicycle to school more over a ten-day period at time 2 (M = .5, SD = 1.97) compared to people who were normal weight (M = .06, SD = .31, t(98) = 1.6, p = .11). When the two modes of transportation, bicycling and walking, were collapsed together over the ten-day period at time 2, overweight students were more likely to walk and bicycle to school (M = 2.21, SD = 3.92) than normal weight students (M = 1.17, SD = 2.94; t(98) = 1.5, p = .14). People who were normal weight had parents who modeled physical activity more than did parents of those overweight, t(106) = 2.04, p = .04. At time 1, there was a trend for people who were normal weight to display less sedentary behavior than people were who were overweight, [t(111) = 1.85, p = .067] but at time 2, this difference became significant, t(106) = 2.83, p = .01.