Community design variations in students' environmental walking supports

There has been a precipitous decline in the number of children who walk to school, an activity that can burn calories and provide healthy bouts of physical activity. This study explores community design among three neighboring communities in Salt Lake County (New Urban, Mixed, and Suburban), and the role that community design plays in the micro-level physical features that support walkability along estimated routes to school. Fifth grade students and their parents participated from two schools that shared a boundary but had different community design philosophies. Walkability was assessed block by block using trained raters and an environmental audit that measures micro environmental features, the Irvine-Minnesota Inventory (IMI). The IMI items were combined into six conceptually derived scales: accessibility, crime safety, density, diversity, pleasurability, and traffic safety. Measures included IMI scores for blocks, for routes to school, and for the traffic variability along a route to school. Estimated walking routes to school were based upon the shortest most direct route. Blocks for the estimated route to school were weighted and combined to characterize each student's walking route. Four different questions about walkability were answered. First, at the block level, New Urban blocks were more walkable than Suburban and Mixed blocks, but Mixed and Suburban blocks did not differ. Second, when parents and children perceived fewer path barriers and crime concerns for the walk to school then the six IMI walking route scales also indicated a more walkable route. Third, walking routes in the New Urban community were more walkable than routes in the Mixed and Suburban communities, and the Mixed community's walking routes were more walkable than routes in the Suburban community. Fourth, the Suburban community's walking routes had more traffic safety variability than either the New Urban or Mixed communities' routes. At both block and route levels, the New Urban community is more walkable than the Mixed and Suburban communities and is perceived that way by parents and children. This study provides one of the few community comparisons using micro-level environmental measures of walkability. The use of walking routes was a different technique that showed key features that related to walking to school and that related to community design Philosophy;. This study has implications for walkable community design at micro and macro levels.

COMMUNITY DESIGN VARIATIONS IN STUDENTS' ENVIRONMENTAL WALKING SUPPORTS by Jonathan Mark Gallimore A thesis submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Master of Science m Human Development and Social Policy Department of Family and Consumer Studies The University of Utah August 2009 V ARIA nONS In Copyright © Jonathan Mark Gallimore 2009 All Rights Reserved THE U N I V E R S I T Y OF UTAH G R A D U A T E SCHOOL SUPERVISORY COMMITTEE APPROVAL of a thesis submitted by Jonathan Mark Gallimore This thesis has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. Chair: Barbara Brown Co-chair: Carol Werner UNIVERSITY GRADUATE SCHOOL .~ G THE U N I V E R S I T Y OF UTAH G R A D U A T E SCHOOL FINAL APPROVAL To the Graduate Council of the University of Utah: I have read the thesis of Jonathan Mark Gallimore m [t s f m a i form and have found that (1) its format, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the supervisory committee and is ready for submission to The Graduate School. Date Approved for the Major Department Chair/Dean Approved for the Graduate Council Charles A. Wight Dean of The Graduate School Barbara Brown Chair: Supervisory Committee UNIVERSITY GRADUATE SCHOOL FIN AL READING APPROVAL in its final ChairlDean ABSTRACT There has been a precipitous decline in the number of children who walk to school, an activity that can burn calories and provide healthy bouts of physical activity. This study explores community design among three neighboring communities in Salt Lake County (New Urban, Mixed, and Suburban), and the role that community design plays in the micro-level physical features that support walkability along estimated routes to school. Fifth grade students and their parents participated from two schools that shared a boundary but had different community design philosophies. Walkability was assessed block by block using trained raters and an environmental audit that measures micro environmental features, the Irvine-Minnesota Inventory (IMI). The IMI items were combined into six conceptually derived scales: accessibility, crime safety, density, diversity, pleasurability, and traffic safety. Measures included IMI scores for blocks, for routes to school, and for the traffic variability along a route to school. Estimated walking routes to school were based upon the shortest most direct route. Blocks for the estimated route to school were weighted and combined to characterize each student's walking route. Four different questions about walkability were answered. First, at the block level, New Urban blocks were more walkable than Suburban and Mixed blocks, but Mixed and Suburban blocks did not differ. Second, when parents and children perceived bum to school. lMI). 1M! routes to school were based upon the shortest most direct route. Blocks for the estimated route to school were weighted and combined to characterize each student's walking route. fewer path barriers and crime concerns for the walk to school then the six IMI walking route scales also indicated a more walkable route. Third, walking routes in the New Urban community were more walkable than routes in the Mixed and Suburban communities, and the Mixed community's walking routes were more walkable than routes in the Suburban community. Fourth, the Suburban community's walking routes had more traffic safety variability than either the New Urban or Mixed communities' routes. At both block and route levels, the New Urban community is more walkable than the Mixed and Suburban communities and is perceived that way by parents and children. This study provides one of the few community comparisons using micro-level environmental measures of walkability. The use of walking routes was a different technique that showed key features that related to walking to school and that related to community design philosophy. This study has implications for walkable community design at micro and macro levels. v 1M1 communities, and the Mixed community's walking routes were more walkable than routes in the Suburban community. Fourth, the Suburban community's walking routes had more traffic safety variability than either the New Urban or Mixed communities' routes. environmental measures of walk ability. The use of walking routes was a different technique that showed key features that related to walking to school and that related to community design philosophy. This study has implications for walkable community design at micro and macro levels. v TABLE OF CONTENTS Page ABSTRACT iv LIST OF TABLES viii LIST OF FIGURES ix ACKNOWLEDGEMENTS x INTRODUCTION 1 Beneficial Walks to School Decline 1 Macro Environmental Walking Supports 3 Parent Perceptions of Walkability and Micro Environmental Walking Supports.. 4 Suburban and New Urban Macro and Micro Walking Supports 6 METHODS 12 Community Sites 12 Participant Inclusion and Segment Sampling 13 Micro Environmental Measures 14 Rater Training and Data Collection 15 IMI Scale Reliability 16 Creating IMI Block Scales, Route Scales, and Traffic Safety Difference Scores 18 Survey Measures 18 Planned Data Analyses and Tests of Assumptions 19 RESULTS 23 Community Differences in Walkability at the Block Level 23 Perceived Barriers and Walking Routes 27 Community Differences in Walkability at the Walking Route Level 32 Traffic Safety Variability 38 DISCUSSION 41 Page ............. ....... .................... .............. .................. ....... ............................. .. ........ ......................... .. .. ................ .... ............................................. .. .......... ........................................................... .......................................... ...... .. ................................ .............................................................. ............. ........................................ .......................................... ... ......... ... ...................................................... ................. 1 ...................................................... ......... Walk ability Supports .. Suburban and New Urban Macro and Micro Walking Supports ............................ 6 ... ........................................................................................................ .. .. ........ .. ........ ........... ........ .. ... .................... ... .......... ................................ ....................................................... ............ ..... .................... ....................................... ................................................................ ...... ............................................................................................. Traffic .............................. ............................ ............................. ..... ............................................................... ...................... ......... ...................... .......... .. .... .. ....... ..... ...... ...................... ...... ................... ........... .... ............................................. ... LeveL .. ............................ ...................................... .. .. ........ ............. LeveL ................. ................. ................ ..................................................... .... ... ..... .......... ..... ......... ..... .......... ..... .. ......... ............. ....... ..... ..... ........ ... .. ..... Page APPENDICES A. RULES FOR DEFINING STREET SEGMENTS 47 B. COMPARING THIS STUDY'S SCALES TO BOARNET ET AL. (2006) 53 C. DESCRIPTIVES FOR ALL VARIABLES 60 D. CORRELATIONS AMONG WALKING ROUTE SCALES, COMPONENTS, AND COVARIATES 61 E. UNIVARIATE ANALYSIS FOR SIX IMI WALKING ROUTE SCALES USING PLANNED CONTRASTS 62 REFERENCES 63 vii APPENDICES A. RULES FOR DEFINING STREET SEGMENTS ..................................................... 47 B. COMPARING THIS STUDY'S SCALES TO BOARNET ET AL. (2006) ........ ...... 53 ................................ .............................. .. ............ .. .. ..... ............... ...................................... .. .. .............. ...... UNIV ARIA TE ............... ................................. ...... ...... .... ........ .... ....... .. ..... ....................................................... ................................ .. .. .... ....... ..... .... .. VB LIST OF TABLES Table Page 1. Correlations among IMI block scales 24 2. IMI block scales by community: Univariate analysis of variance 3. Structure matrix and standardized discriminant function coefficients for block scales 26 4. Partial correlations among parent and child route walkability perceptions and IMI walking route scales 29 5. Correlations among estimated walking route IMI scales 33 6. Walking route scales - Rotated component matrix 33 7. Univariate planned contrast results for components 35 8. Route - Structure matrix and standardized discriminant function coefficients 37 9. Traffic safety variability for walking routes in three communities: Means for adjusted traffic safety difference scores, least traffic safe blocks, and most traffic safe blocks 40 10. Individual IMI items: Boarnet et al. (2006) and present study 55 11. Descriptives for variables used in all analyses 60 12. Correlations among walking route scales, components, and covariates 13. Planned contrast results for IMI walking route scales 62 ..... ..... .. ...... .. .... ....... .. .... ...... .. ..... .. ... .. ................ results ..... ..... ........ ..... .. ...................... .. ... .. ...... ... .. ...... ....... .. ... .... .... .. ... .... .. ...... .. .... ..... ... 24 scales.. ... . ... .. ..... .... ...... .. .. .. ...... ..... .. .. .... ..... ... ... ....... ...... ..... ........ perceptions and IMI walking route scales .. ...... ... .. .. ..... .. ... .... .. .... ... ..... ....... ................. 29 5. Correlations among estimated walking route IMI scales ........ .. .... .. .......... .. .... .. ...... ..... 33 6. Walking route scales - Rotated component matrix ....................... ....................... ........ 33 7. Univariate planned contrast results for components ..... ............ ........................ ..... ...... 35 8. Route - Structure matrix and standardized discriminant function coefficients .... .. .... .. ...... ......................... .. ...... ...... ..... .. ..... ..... .... .. ... .. .... ........ .. ....... ......... 37 ................................... .. ...... ..... .. .... ... .. ......... Boamet ........ .. ................. ..... .. 11 . .................... ...... .. .... .. ... ....... ... .... ....... covariates... .... .. ............................ .. ........ ............. .. ... .. .... .... ... .. .... .. ..... .. ... ...... ........... .. 61 ..... ... .... .. ... .... .. .... ....... .... .. .... .... LIST OF FIGURES Figure Page 1. Block level centroid plot of functions 28 2. Walking routes funneled through a highly dense arterial street 31 3. Walking route level centroid plot on both functions 39 ... ..... ..... ... .. .... ...... ...... .... .. ... ...... .... .. ......... .. .... .... street... .... .. .... ... .. ... ......... .... ............ ...... .. .......... ...... .. ........... .. ACKNOWLEDGEMENTS I would like to thank my committee members, Dr. Barbara Brown, Dr. Carol Werner, and Dr. Kevin Rathunde, for their continued support, insightful feedback, and patient guidance. I would like to thank Dr. Brown and Dr. Werner for the long meetings, help with statistics, and quick feedback on drafts. I could not have been successful without your help. You both demonstrated what mentors should be and what I can strive to become as a mentor. Thank you. INTRODUCTION Currently in the United States, low levels of physical activity and growing rates of obesity threaten the health of youth. Health professionals, researchers, and urban planners have recently considered how community design might support more physical activity and help prevent obesity (Centers for Disease Control and Prevention, 2000b; Frank, Engelke, & Schmid, 2003). According to the Centers for Disease Control and Prevention (CDC), more physical activity and less obesity would reduce the incidence of heart disease, type II diabetes, and some cancers (Centers for Disease Control and Prevention, 2000a; Dietz, 2004). One source of youth physical activity easily overlooked is walking to school. Walking to school, and other physical activity, could be discouraged or encouraged by a community's design. The present study examines how small-scale environmental features on walking routes to school differ for a community designed with a goal of being walkable and one that was designed to be less walkable. Beneficial Walks to School Decline Walking to school has declined dramatically all across the nation. Although 40.7% of children walked to school in 1969, only 12.9% of children walked to school in 2001 (McDonald, 2007). Across these same decades, overweight problems among children rose dramatically. In 2003 to 2004, 17.1% of 2- to 19-year-old children and adolescents were overweight (Ogden et al., 2006), a substantial rise from overweight rates of the 1970s. For example, the 6- to 11-year-olds overweight rates rose from 4% in & heart disease, type II diabetes, and some cancers (Centers for Disease Control and Prevention, 2000a; Dietz, 2004). One source of youth physical activity easily overlooked is walking to school. Walking to school, and other physical activity, could be discouraged or encouraged by a community's design. The present study examines how small-scale environmental features on walking routes to school differ for a community designed with a goal of being walkable and one that was designed to be less walkable. 17.1 % aI., ll-year-2 the 1971 to 1974 period (Ogden, Flegal, Carroll, & Johnson, 2002) to 18.8% in the 2003 to 2004 period (Ogden et al., 2006), which was the highest rate of increase in overweight among youth. In 2003 to 2004, the 6- to 11 -year-olds also had the highest incidences of overweight compared to other youth (aged 2- to 5- and 12- to 19-years-old). Accordingly, this research focuses on 5t h graders who are in an age group (10- to 11-year-olds) that is at high risk for being overweight. Daily habits of physical activity, such as walking to school, would contribute to a variety of benefits including obesity prevention, contributions toward overall daily physical activity recommendations, and societal benefits (Sothern, Loftin, Suskind, Udall, & Blecker, 1999). In a study of 14- to 16-year-old Filipino adolescents who walked to school, researchers calculated that walking to and from school for 60 minutes a day for 200 school days would require energy expenditures equal to the prevention of 2 to 3 pounds of weight gain per year compared with less active travel to school by car or bus (Tudor-Locke, Ainsworth, Adair, & Popkin, 2003). Although students in the United States might walk shorter distances to school than do Filipino students, any amount of moderate or brisk physical activity during the day can help. The additional daily minutes of moderate to vigorous physical activity (MVPA) due to walking to or from school vary from reports of an extra 3.5 MVPA minutes in Denmark (Cooper, Andersen, Wedderkopp, Page, & Froberg, 2005), to 8 to 14 minutes in England (Cooper, Page, Foster, & Qahwaji, 2003), to 2.2 to 4.7 minutes in a U.S., 6-state study (Saksvig et al., 2007), to 7 minutes in South Carolina (Sirard, Riner, Mclver, & Pate, 2005). Though modest, these minutes add to daily physical activity levels. Additionally, societal benefits from walking to school can include decreased automobile dependence and aI., 11-the incidences of Accordingly, this research focuses on 5th graders who are in an age group (10- to 11-year­olds) that is at high risk for being overweight. Sothem, school, researchers calculated that walking to and from school for 60 minutes a day for 200 school days would require energy expenditures equal to the prevention of (Tudor-Locke, Ainsworth, Adair, & States might walk shorter distances to school than do Filipino students, any amount of moderate or brisk physical activity during the day can help. The additional daily minutes of moderate to vigorous physical activity (MVPA) due to walking to or from school vary from reports of an extra 3.5 MVPA minutes in Denmark (Cooper, Andersen, Wedderkopp, Page, & Froberg, 2005), to 8 to 14 minutes in England (Cooper, Page, Foster, & Qahwaji, 2003), to 2.2 to 4.7 minutes in a U.S., 6-state study (Saksvig et aI., 2007), to 7 minutes in South Carolina (Sirard, Riner, McIver, & Pate, 2005). Though modest, these minutes add to daily physical activity levels. Additionally, societal benefits from walking to school can include decreased automobile dependence and 3 greenhouse gas emissions (EPA, 2003; Hubsmith, 2006), decreased time obligations for parents driving children to school (Ahlport, 2008), and increased opportunities for children to socialize with neighbors, friends, or siblings en route (Ahlport, 2008; Carver et al., 2005; Hubsmith, 2006). Given the many benefits of walking to school, it is important to examine community designs that deter or support walking to school. Macro Environmental Walking Supports Adults often walk more when they live in communities designed with macro level environmental features that represent the "3Ds" of walking supports: greater population Density, Diversity of land uses and destinations, and pedestrian-friendly Design (Cervero & Kockelman, 1997). Greater population density and diversity of land uses bring many individuals within close and convenient walking distances of attractive destinations. Research has shown that adults walk more when they live in areas with some combination of greater population densities, more nearby attractive destinations, or routes that are well-connected and accessible (Giles-Corti et al., 2005; Hoehner, Ramirez, Elliott, Handy, & Brownson, 2005; Leslie et al., 2005; Mota et al., 2007; Moudon et al., 2007; Saelens, Sallis, Black, & Chen, 2003). Studies have also shown that macro environmental supports, like the 3Ds, relate to children's walks to school, although fewer studies of children have been conducted. Some walking to school studies have used measures of density, land use diversity, and pedestrian-friendly street design that are typical of adult studies. These studies have found that more students walk in communities with greater population density (Braza, Shoemaker, & Seeley, 2004; Kerr et al., 2006; McDonald, 2008; Zhu & Lee, 2008), more land use diversity (Kerr et al., 2006; McMillan, 2007), and greater street intersection aI., ofland aI., aI., aI., aI., aI., aI., 4 density (Kerr et al., 2006; Schlossberg, Greene, Phillips, Johnson, & Parker, 2006). Instead of measuring density and diversity, other studies have used a simple measure of the road network distance between home and school to assess whether the community design provides convenient walking distances from home to school. When students had shorter distances from home to school or smaller school catchment areas, more children walked to school (Braza et al., 2004; Falb, Kanny, Powell, & Giarrusso, 2007; McDonald, 2007, 2008; McMillan, 2007; Nelson, Foley, O'Gorman, Moyna, & Woods, 2008; Schlossberg et al., 2006; Timperio et al., 2006; Ziviani, Scott, & Wadley, 2004). From this literature review, it can be seen that macro level environmental supports for walking, as assessed by measures similar to density, land use diversity, and pedestrian friendly design, have correlated with both adults' neighborhood walks and children's walks to school. Parent Perceptions of Walkability and Micro Environmental Walking Supports Macro environmental supports from density, land use diversity, and pedestrian-friendly design may not be sufficient to encourage children to walk to school; parent perceptions and micro level environmental supports have also been shown to be important. According to focus groups and survey data, parents have perceived many personal and environmental barriers to walking to school. Personal barriers have included conflicts between parents' work schedules and children's school schedules (Ahlport, 2008) and the convenience of dropping a child off at school (McDonald, 2007). Numerous studies have focused on parents' perceptions of environmental barriers to walking to school. For example, many parents have cited overly long distances between & & 2008; Schlossberg et al., 2006; Timperio et al., 2006; Ziviani, Scott, & Wadley, 2004). From this literature review, it can be seen that macro level environmental supports for walking, as assessed by measures similar to density, land use diversity, and pedestrian friendly design, have correlated with both adults' neighborhood walks and children's walks to school. Walk ability walking to school. For example, many parents have cited overly long distances between home and school as a barrier to walking to school (McDonald, 2007; Nelson et al., 2008; Timperio et al., 2006; Ziviani et al., 2004). However, previous research demonstrates that parents' perceptions of barriers to their children walking to school go beyond the measures of macro environmental walking supports of density, diversity, or interconnected street designs. Parent perceptions often focus on specific micro environmental features that pose barriers along the route to school. These perceptions focus on traffic dangers, walking path accessibility, and crime. These features of the route to school are typically not measured by the 3Ds of walkability. Parents' traffic safety concerns include perceptions of dangerous streets crossings (Hubsmith, 2006; Timperio, Crawford, Telford, & Salmon, 2004). Parents worry about their student's access to school because of missing or incomplete sidewalks (Ahlport, 2008). Additionally, parents perceive crime safety issues along their child's walking route to school (Ziviani et al., 2004), including the presence of hidden or unseen areas (Ahlport, 2008). Many studies show that children are less likely to walk to school when parents perceive these barriers (Ahlport, 2008; Frank et al., 2003; Hubsmith, 2006; Kerr et al., 2006; McMillan, 2007; Timperio et al., 2004; Ziviani et al., 2004). Parent perceptions suggest that children's walks to school are prevented by micro environmental conditions related to poor pedestrian access, crime concerns, and traffic dangers, although few studies have tested these relationships. An evaluation of California's Safe Routes to School (SR2S) program suggests that providing good micro environmental walking supports increases the number of children who walk to school. According to parents' retrospective reports, walking increased after sidewalks were installed, repaired, or completed and traffic lights replaced four-way stops (Boarnet, aI., aI., aI., 5 school. These perceptions focus on traffic dangers, walking path accessibility, and crime. These features of the route to school are typically not measured by the 3Ds of walkability. Parents' traffic safety concerns include perceptions of dangerous streets crossings (Hubsmith, 2006; Timperio, Crawford, Telford, & Salmon, 2004). Parents worry about their student's access to school because of missing or incomplete sidewalks (Ahlport,2008). Additionally, parents perceive crime safety issues along their child's walking route to school (Ziviani et aI., 2004), including the presence of hidden or unseen areas (Ahlport, 2008). Many studies show that children are less likely to walk to school when parents perceive these barriers (Ahlport, 2008; Frank et aI., 2003; Hubsmith, 2006; Kerr et aI., 2006; McMillan, 2007; Timperio et aI., 2004; Ziviani et aI., 2004). 6 Anderson, Day, McMillan, & Alfonzo, 2005a; Boarnet, Day, Anderson, McMillan, & Alfonzo, 2005b). However, this program also included educational outreach efforts and more enforcement of traffic safety by crossing guards and police officers (Hubsmith, 2006; McMillan, 2005), so it is impossible to attribute more walking to environmental changes alone. In correlational studies examining student's walking routes, the presence of a major arterial between home and school did not deter walking in a U.S. study (Schlossberg et al., 2006), although objective counts of higher traffic volume did relate to less walking to school in Australia (Timperio et al., 2006). Generally, these few studies have suggested that parent perceptions of micro level environmental barriers to walking to school are consistent with objective ratings of micro level environmental barriers. One study used both macro and micro measures to study ethnic disparities in walkability and safety around elementary schools. Zhu and Lee (2008) found that schools with many Hispanic students were more walkable because of macro environmental features of higher community population density and greater land use mix. But micro level audits of 200 meter street segments around the schools (using the Pedestrian Environment Data Scan (PEDS) audit tool) showed that these same schools had less perceived crime safety and more traffic safety problems. This study highlights the possibility that macro community design features can be inconsistent with micro level supports for walking. Suburban and New Urban Macro and Micro Walking Supports Community design can be seen as a tool that might promote or thwart children's walks to school. Many suburban communities, although often thought of as child-friendly places, in fact may limit children's independent mobility (Frank et al., 2003). In contrast, of a major arterial between home and school did not deter walking in a U.S. study (Schlossberg et aI., 2006), although objective counts of higher traffic volume did relate to less walking to school in Australia (Timperio et aI., 2006). Generally, these few studies have suggested that parent perceptions of micro level environmental barriers to walking to school are consistent with objective ratings of micro level environmental barriers. of200 child-aI., 7 designers of new urban communities are informed by a charter that explicitly values the walk to school. According to the charter, "Schools should be sized and located to enable children to walk or bicycle to them" (Congress for the New Urbanism, 2001). New urbanism is a rapidly growing alternative to standard suburban design, with 520 communities now listed as influenced substantially by new urbanism (Steuteville, 2008). A review of their specific macro and micro design features forms the basis of the hypothesized differences in children's walks to school in suburban and new urban communities. Typical suburban communities are often designed to provide macro environmental features of relatively low population densities, little land use diversity, and poorly connected street designs. The low density makes for long walking distances and the lack of diverse land uses means that many necessities are excluded from the residential-only character of the neighborhood, like close schools. Low levels of land use diversity may also create little visual variety in the built environment along walks through suburban communities making the walks less pleasurable. Additionally, the popularity of disconnected street forms in suburban design, such as cul-de-sacs and curvilinear streets, creates indirect routes that may be substantially longer than direct "crow-flying" distances (Southworth & Ben-Joseph, 1997). These design features were often adopted by designers to discourage automotive traffic through neighborhoods (Frank et al., 2003). Cul-de-sac street forms may provide some neighborhood benefits at the micro, or street, level such as perceived and actual safety from crime (Brown & Werner, 1985), low levels of traffic, and safe places for children to play on the street (Al- Kodmany, 1997). However, if these same micro level conditions characterize entire communities now listed as influenced substantially by new urbanism (Steuteville, 2008). A review of their specific macro and micro design features forms the basis of the hypothesized differences in children's walks to school in suburban and new urban communities. curvilinear streets, creates indirect routes that may be substantially longer than direct "crow-flying" distances (Southworth & Ben-Joseph, 1997). These design features were often adopted by designers to discourage automotive traffic through neighborhoods (Frank et aI., 2003). Cul-de-sac street forms may provide some neighborhood benefits at the micro, or street, level such as perceived and actual safety from crime (Brown & Werner, 1985), low levels of traffic, and safe places for children to play on the street (AI­Kodmany, 1997). However, if these same micro level conditions characterize entire 8 neighborhoods, in the form of multiple cul-de-sacs or disconnected streets, then all car traffic exiting from them gets tunneled onto collector streets and then high-traffic arterials. This funneled traffic may impose barriers to children walking to school. Thus, the suburban street forms that combine cul-de-sacs with low volumes of car traffic and arterials with high volumes of car traffic, may create an unacceptable level of variability in traffic safety for walking to or from school. In comparison to suburban communities, typical new urban communities are often designed to provide macro environmental features of relatively higher population densities, greater diversity of land uses, and well-connected street designs. Designing a community with higher density housing, such as townhomes and homes on small lots, may allow more families to live within walking distance of schools. The design philosophy behind new urbanism generally favors greater land use diversity of attractive designations and specifically favors walking distance to schools as part of that land use mix. Interconnected street forms are advocated by new urbanists as ways to enable fairly direct routes and short walking distances between destinations. Although gridded street forms may mean more residential street traffic than very low-traffic suburban cul-de-sacs, new urbanists argue that spreading out traffic more evenly across the grid allows for a manageable level of traffic that supports walkability for larger sections of the neighborhood and for more destinations outside the neighborhood (Calthorpe, 1993). New urban communities may mandate narrower roads, lower speed limits, and more traffic calming devices like bulb outs and traffic circles that allow pedestrians to share the roads safely with drivers (Calthorpe, 1993). fonn funneled high-fonns in traffic safety for walking to or from school. fonns new urbanists argue that spreading out traffic more evenly across the grid allows for a manageable level of traffic that supports walkability for larger sections of the neighborhood and for more destinations outside the neighborhood (Calthorpe, 1993). New urban communities may mandate narrower roads, lower speed limits, and more traffic calming devices like bulb outs and traffic circles that allow pedestrians to share the roads safely with drivers (Calthorpe, 1993). 9 New urban design features are intended to allow children to walk to school, but it is not clear whether walking will always occur as intended. Given that so many children do not walk to school any more, new urban designs might actually create dangerous traffic conditions near schools. This danger could occur at the new urban school sites because they may not have the dangerous streets that would merit school bus service. Without school bus options parents of children in new urban communities might drive their children to school. As McMillan (2005) argues, the number of private vehicles converging at the school for student drop off and pick up might create unsafe traffic congestion, thereby encouraging more parents to drive children to school, creating even worse traffic safety conditions. The new urban and suburban communities have reputations for crime safety, so there is no reason to hypothesize environmental differences in crime indicators. It is possible that new urban community designs, due to higher levels of neighborhood social contact (Brown & Cropper, 2001), might create perceptions of greater crime safety via greater local use of the area and "eyes on the street" (Jacobs, 1961). New urban communities might provide more sidewalks to enhance walkability than do suburban communities; in the present study, it is expected that the new urban community will have better sidewalk coverage than the suburban community. In sum, although new urban designs could be advocated to support walking to school if they create better walking conditions, research on micro environmental supports for walking are needed to substantiate these claims. This study contrasts two schools in a fairly safe middle class suburb in Salt Lake County. In the study area, some of the students who attend the school in the new urban community live right outside the new traffic conditions near schools. This danger could occur at the new urban school sites because they may not have the dangerous streets that would merit school bus service. Without school bus options parents of children in new urban communities might drive their children to school. As McMillan (2005) argues, the number of private vehicles converging at the school for student drop off and pick up might create unsafe traffic congestion, thereby encouraging more parents to drive children to school, creating even worse traffic safety conditions. ofthe better sidewalk coverage than the suburban community. 10 urban community boundaries, so that these students live on suburban blocks but their route to school includes both suburban and new urban blocks. This group will be referred to as the "mixed" community. The present study examines routes to school in communities that represent new urban and suburban design philosophies. For the new urban, mixed, and suburban communities the following four hypotheses are tested: 1. Block Hypothesis: New urban blocks will be more walkable than mixed or suburban blocks, with respect to micro level environmental walkability features. All six walkability features will be examined, but greatest differences are expected for traffic safety, accessibility, density, and pleasurability, and in the current community context, fewer differences are expected for diversity and crime safety. 2. Walking Routes Perceived Barriers Hypothesis: After controlling for sociodemographic variables, the micro level environmental walkability of the route to school will relate to fewer path barriers and crime concerns for both children and their parents. 3. Walking Routes Hypothesis: After controlling for sociodemographic variables, the new urban community's walking routes are predicted to be more walkable than the suburban community's walking routes, and the mixed community's walking routes will be somewhere in between the other two communities. All six walkability features will be examined and these features are expected to reveal differences among communities. 4. Traffic Safety Variability Hypothesis: After controlling for sociodemographic variables, new urban, mixed, and suburban communities will differ on traffic safety. walkability features will be examined and these features are expected to reveal differences among communities. traffic 11 safety variability. Children living in the new urban community are expected to have relatively consistent traffic safety features along their walking routes. Children living in suburban communities are expected to have both very safe and very unsafe traffic conditions along routes to school, creating greater variability in traffic safety across the route. METHODS Community Sites Two elementary schools were selected for this study that share a school boundary line and whose communities bordered one another. Most of the students at one school were eligible to ride school buses, due to greater distance to school or traffic safety concerns, and will be referred to as the less walkable school. Students who attended the less walkable school lived in a typical suburban community. The other school had no bus service and will be referred to as the more walkable school. The more walkable school was part of a new urbanist community development, with some students attending from the suburban community just outside the new urban community's border (the "mixed" community) and other students attending from the new urban community. The present study examines residential blocks and walking routes to school in these three communities: a new urban community, a mixed community, and a suburban community. Macro environmental features demonstrate that the selected communities are good representations of suburban and new urban forms. The new urban community has higher density housing forms, including townhomes and small lot single family detached housing with a median lot size of .12 acres (Napier, 2009). While the mixed community is composed of suburban blocks, it is just outside the new urban community with a median lot size of .25 acres (Napier, 2009). The suburban community is primarily communities: a new urban community, a mixed community, and a suburban community. 13 composed of single family detached homes with a median lot size of .35 acres (Napier, 2009). Other macro environmental features indicate two distinct community design philosophies. The new urban community has gridded street designs, small parks, protected open space, and land designated for a town commercial center. Please note that the new urban community was fairly new with the development opening to residents in June of 2004 and data collection beginning in May of 2007. Consequently, the business and commercial districts had not yet been built so that much of the planned land use diversity did not exist at the time of the study. However, others have argued that few "perfect" new urban communities exist yet, so research should simply evaluate the behavioral consequences of a wide variety of incomplete new urban communities (Brown & Cropper, 2001). The mixed community is composed of low traffic curvilinear streets and many cul-de-sacs with small scattered parks, pedestrian access points, and single family detached homes. The suburban community has fairly low traffic cul-de-sacs that tunneled cars into higher traffic arterials. It had no parks or protected open space and commercial facilities were sited along the arterials. Participant Inclusion and Segment Sampling Student's addresses were obtained from survey data reported by Napier (2009), and the location of each student's home was plotted onto maps. Then each school's location was marked on the maps and the child's estimated walking route to and from school was plotted based upon the shortest, most direct route. Addresses within 1.5 miles road network distance to school were selected for inclusion with the environmental audit, given that the school district considers this distance too far to be walkable and provides behavioral consequences of a wide variety of incomplete new urban communities (Brown & Cropper, 2001). The mixed community is composed of low traffic curvilinear streets and many cul-de-sacs with small scattered parks, pedestrian access points, and single family detached homes. The suburban community has fairly low traffic cul-de-sacs that funneled cars into higher traffic arterials. It had no parks or protected open space and commercial facilities were sited along the arterials. given that the school district considers this distance too far to be walkable and provides 14 bus service for students living this far from school (Jordan School District, 2009). One researcher then defined the street segments for data collection according to some basic rules to obtain internally consistent segments and deal with unique street forms (e.g., circular and curvilinear street forms). Appendix A contains a more detailed description of these rules with examples. Micro Environmental Measures The micro level environmental walkability of each community's design was measured with a field audit, the Irvine-Minnesota Inventory (IMI), to evaluate the environmental features related to students' estimated walking routes to school and to see if these features are related to the macro level designation of the student's community as new urban, mixed, or suburban. The IMI was developed and refined to have good inter-rater reliability by research teams in southern California and Minneapolis (Boarnet, Day, Alfonzo, Forsyth, & Oakes, 2006; Day, Boarnet, Alfonzo, & Forsyth, 2006). Other research has shown the IMI's ability to distinguish high-walkability routes from low-walkability routes in an urban setting (Brown, Werner, Amburgey, & Szalay, 2007). When the Irvine-Minnesota Inventory (IMI) was originally created the authors suggested that it assessed four conceptually derived scales: accessibility, pleasurability, perceived safety from crime, and perceived safety from traffic (Boarnet et al., 2006). The IMI authors suggested that future researchers might alter their four scales to fit the specific needs of their research project. Past research has conceptualized the macro level of walkability as involving density, land use diversity, and pedestrian friendly design (Cervero & Kockelman, 1997). Guided by this conceptualization, past research, with the IMI has found it useful to maintain the macro level concepts of density and diversity at lMI), new urban, mixed, or suburban. The IMI was developed and refined to have good inter­rater reliability by research teams in southern California and Minneapolis (Boarnet, Day, Alfonzo, Forsyth, & Oakes, 2006; Day, Boarnet, Alfonzo, & Forsyth, 2006). Other research has shown the IMI's ability to distinguish high-walkability routes from low­walkability routes in an urban setting (Brown, Werner, Amburgey, & Szalay, 2007). walk ability (Cervero & Kockelman, 1997). Guided by this conceptualization, past research, with the IMI has found it useful to maintain the macro level concepts of density and diversity at 15 the micro level by dividing the accessibility scale into three scales: density, diversity, and accessibility (Brown et al., 2007). The present study maintained that practice, splitting Boarnet et al.'s (2006) accessibility scale into three scales: diverse destinations (38 items), density of housing (6 items), and pedestrian accessibility (12 items). The other IMI scales used include pleasurability (54 items), perceived safety from traffic (52 items), and perceived safety from crime (14 items). Other minor changes were made to Boarnet et al.'s (2006) other three scales to assure that each item belonged to only one scale and to provide conceptual coherence. For a direct comparison of the present study's scales and the scales of Boarnet et al. (2006), please refer to Appendix B. Rater Training and Data Collection Raters were trained on the Irvine Minnesota Inventory (IMI) to gather environmental information for all of the blocks along possible walking routes to school. The IMI is composed of 160 unique items and asks raters about the presence of environmental features (yes/no) or their quantity (none, few, a lot). Raters were initially trained by viewing a slideshow presentation (Alfonzo, Day, & Boarnet, 2005) delivered by a group leader. After the initial training session, raters completed practice blocks using the IMI and a written list of definitions. An experienced group leader met with the raters and clarified questions about IMI items. Raters practiced auditing blocks independently and with the group leader until they achieved reliability according to the group leader. Data were collected during day light hours and in sunny conditions. Raters walked the street blocks filling out the IMI as they walked. Occasionally raters felt that street blocks were unsafe due to traffic, so they made their ratings by driving the block aI., Boamet al. 'Boamet et al. 's (2006) other three scales to assure that each item belonged to only one scale and to provide conceptual coherence. For a direct comparison of the present study's scales and the scales of Boamet et al. (2006), please refer to Appendix B. 1M I) Boamet, raters and clarified questions about IMI items. Raters practiced auditing blocks independently and with the group leader until they achieved reliability according to the 16 multiple times in a car and parking to fill out the IMI items. Driving a block was a method used during the refinement period of the IMI (Boarnet et al., 2006). Given limited resources, most data collection occurred from August to November of 2007 with some remaining blocks completed in October and November of 2008. IMI Scale Reliability Instructions for the Irvine Minnesota Inventory (IMI) recommended using percent agreement for interrater reliabilities in lieu of Kappa or Pearson's statistics because some IMI items occur infrequently. Percentage agreements were greater than .65 for all but eight IMI items that were dropped because of genuine rater disagreements on items that required judgments about the quantity of an environmental feature. Appendix B lists which variables were dropped due to disagreements. An example of a genuine rater disagreement involved assessing the amount of litter. One rater would mark no litter while another rater reported a little litter. The IMI scales used in this study are described below, with specific details included in Appendix B. Diverse destinations. This scale provides a comprehensive list of destinations in an area that together reflect land use diversity. Of 38 potential items, 21 were used in this study and 17 were never present so they were omitted. Examples of diversity items include religious buildings, green spaces, and schools. Density of housing. This scale assesses micro environmental measures of density by describing the type and size of buildings. Of six potential items, four were used in this study and two were never present so they were omitted. Examples of density items include single family detached homes, apartment buildings, and building height in storeys. 1M1 1M1 Boamet aI., of2007 of2008. IM1) IM1 while another rater reported a little litter. The IMI scales used in this study are described below, with specific details included in Appendix 17 Accessibility. This scale assesses how easy it is to walk from one area to another along the block. Of 12 potential items, 9 were used in this study and 3 were never present so they were omitted. Examples of accessibility items include sidewalks, ditches (indicating a lack of access), and pedestrian access points in cul-de-sacs. Traffic safety. This scale gives a measure of traffic features that facilitate or discourage pedestrian travel. Of 52 potential items, 40 were used in this study and 12 were never present so they were omitted. Examples of traffic safety items include stop lights, crosswalks, and speed limit signs. Pleasurabilitv. This scale provides a measure of how pleasant a walking environment is for pedestrians. Of 54 potential items, 22 were used in this study, 25 were never present so they were omitted, 1 item was always present so it was omitted, and 6 items were dropped due to disagreements. Examples of pleasurability items include street trees, overhead electric wires, and buildings with garages. Crime safety. This scale assesses places and things that might make pedestrians more or less fearful while on a walk. Of 14 potential items, 8 were used in this study, 4 were never present so they were omitted, and 2 items were dropped due to rater disagreements. Examples of crime safety items include bars on windows, outdoor lighting, and abandoned buildings. After reliability checks, a single rater's data were used for analysis. Items were reverse coded as needed so that high values indicated more walkability. Items were then z-scored to allow each item to be measured on the same metric. oftraffic oftraffic Pleasurability. 18 Creating IMI Block Scales, Route Scales, and Traffic Safety Difference Scores In order to compute the six IMI scales for each block (for hypothesis 1), the individual IMI z scored items were averaged for each of the six micro environmental walkability composites: diversity, density, accessibility, traffic safety, pleasurability, and crime safety. A total of 224 block level segments were measured, including 40 for the new urban community, 75 for the mixed community, and 109 for the suburban community. To characterize each participant's estimated walking route (for hypotheses 2 and 3), the six IMI block scales were weighted to represent the contribution of each block to the total distance of all blocks for the participant's estimated walking route. For example, a 0.3 mile block would be weighted 0.25 to reflect its contribution to a 1.2 mile route to school (0.3 / 1.2 = 0.25). After weighting, the six IMI scales were summed across blocks following the shortest, most direct route to school describing each participant's estimated walking route. This process provided an overall score for each IMI scale, representing the average score for each participant's estimated walking route to school. The traffic safety variability for a participant's estimated walking route (for hypothesis 4) to school was calculated by subtracting their least walkable block (Min) from their most walkable block (Max). Larger numbers for the resulting Traffic Safety Difference score (Max-Min) indicate a wider range of traffic safety along the route. Survey Measures This study is part of a larger study that used surveys, and now environmental audits, to explore the relationship between the environment and walking to school. The ofthe of224 1.2 19 surveys described elsewhere (Napier, 2009) provided subscales that measured students' and parents' perceived path barriers and a question about crime concerns preventing a walk to school. The perceived path barriers subscale is composed of four questions describing the walk to school as too distant, too difficult, with unsafe crossings, and with traffic safety problems. The crime question asked participants about crime dangers making the walk to school unsafe. All questions used a 4-point scale where larger values indicated more perceived path barriers or crime concerns. The covariates were taken from the survey to address potential selection effects of residents into communities and other sociodemographic correlates of walking to school. They include: if the parent ideally wants their child to walk to school ("ideally walk;" 4-point scale; from "strongly against" to "strongly favor"), the number of rooms in the home (11-point scale; actual number), home owner (0 = no and 1 = yes), and parental education level (5-point scale; from "some high school" to "advanced college degree"; an analysis using a 2-point scale of some college or more reveals no differences in significance levels for subsequent statistical tests). Planned Data Analyses and Tests of Assumptions To test community differences on a block level (hypothesis 1), univariate planned contrasts weighted for unequal group size will be used for each IMI block scale. These weighted orthogonal contrasts will compare the new urban community to the other two communities (mixed and suburban), and also compare the mixed and suburban communities. Note that these weighted contrasts are used for all analyses that use planned contrasts (hypotheses 1 and 3). After planned contrasts, a multivariate analysis of variance (MANOVA) will compare communities on all IMI block scales with describing the walk to school as too distant, too difficult, with unsafe crossings, and with traffic safety problems. The crime question asked participants about crime dangers making the walk to school unsafe. All questions used a 4-point scale where larger values indicated more perceived path barriers or crime concerns. The covariates were taken from the survey to address potential selection effects of residents into communities and other sociodemographic correlates of walking to school. They include: if the parent ideally wants their child to walk to school ("ideally walk;" 4-point scale; from "strongly against" to "strongly favor"), the number of rooms in the home (II-point scale; actual number), home owner (0 no and I = yes), and parental education level (5-point scale; from "some high school" to "advanced college degree"; an analysis using a 2-point scale of some college or more reveals no differences in significance levels for subsequent statistical tests). I 20 discriminant analysis following up a significant multivariate effect. Partial correlations will assess the relationship between parent and child perceptions of walking routes and the IMI walking route scales, controlling for covariates (hypothesis 2). Community differences on the six IMI walking route scales will be evaluated by two sets of analyses (hypothesis 3). The first set of analyses (hypothesis 3) compares communities by combining the six IMI walking route scales into components. Univariate planned contrasts examined community differences on these component scores. The second set of analyses (hypothesis 3) will use a multivariate analysis of covariance (MANCOVA) to evaluate the communities on the six IMI walking route scales with discriminant analysis following up to a significant multivariate effect. Finally, an analysis of covariance (ANCOVA) will test for community differences in traffic safety variability along each walking route using the Traffic Safety Difference score (hypothesis 4) with Bonferroni pairwise comparisons following up a significant effect. Before data analysis, the covariates and variables were checked to ensure they met appropriate statistical assumptions for the planned statistical tests. The perceived path barriers and crime question (hypothesis 2) were normally distributed with no outliers for either parents or children. The covariates (hypotheses 2, 3, and 4) had some missing values and for each covariate missing values were replaced with that group's mean. After this mean replacement, two covariates were skewed, ideally would walk and number of rooms. LoglO transformations normalized these variables. The block level scales, used for hypothesis 1, were checked for normality but were skewed. Closer examination of histograms revealed outliers in all scales. Top and bottom coding were used to pull in outliers, with top and bottom code values equal to two IMl IMl analyses (hypothesis 3) will use a multivariate analysis of covariance (MANCOV A) to evaluate the communities on the six walking route scales with discriminant analysis following up to a significant multivariate effect. Finally, an analysis of covariance (ANCOV A) will test for community differences in traffic safety variability along each walking route using the Traffic Safety Difference score (hypothesis 4) with Bonferroni pairwise comparisons following up a significant effect. rooms. LogiO transformations normalized these variables. I, 21 standard deviations from the mean. No more than 5% of all scale values were changed as a result of top and bottom coding and the resulting variables were normal. Three of the six IMI block scales had significant results on Levene's test for equality of variance indicating different variances across communities. Type I sum of squares were used to adjust for unequal variance. For the block analysis, a Roy Bargmann step-down analysis was planned to assess the unique variance accounted for by each scale. However, the homogeneity of the regression coefficient assumption for step-downs did not hold across cells, indicated by a significant interaction between communities and dependent variables serving as covariates; heterogeneity of the regression coefficients across cells makes step-down analyses uninterpretable. Instead, discriminant analysis is the recommended follow up to a significant multivariate effect under these circumstances (Field, 2000; Tabachnick & Fidell, 2007). The walking route scales, used for hypotheses 2 and 3, were checked for normality but were skewed. Closer examination of histograms revealed outliers in all scales. This extreme variation in walking route scales could reflect consistent differences in the environmental features measured by each scale. Top and bottom coding were used to pull in outliers, with top and bottom code values equal to 2 standard deviations from the mean. No more than 6% of all scale scores were changed as a result of top and bottom coding and the resulting variables were normal. Covariates showed significant interactions with the communities. Analyses were conducted with and without the necessary adjustment for factor by covariate heterogeneity. Results were similar for both analyses, so the simpler analysis that assumes homogeneous covariate effects are presented (SPSS, 1988). indicating different variances across communities. Type I sum of squares were used to adjust for unequal variance. For the block analysis, a Roy Bargmann step-down analysis was planned to assess the unique variance accounted for by each scale. However, the homogeneity of the regression coefficient assumption for step-downs did not hold across cells, indicated by a significant interaction between communities and dependent variables serving as covariates; heterogeneity of the regression coefficients across cells makes step­down analyses uninterpretable. Instead, discriminant analysis is the recommended follow up to a significant multivariate effect under these circumstances (Field, 2000; Tabachnick & Fidell, 2007). differences the mean. No more than 6% of all scale scores were changed as a result of top and bottom coding and the resulting variables were normal. Covariates showed significant interactions with the communities. Analyses were conducted with and without the necessary adjustment for factor by covariate heterogeneity. Results were similar for both analyses, so the simpler analysis that assumes homogeneous covariate effects are presented (SPSS, 1988). Three of the six walking route scales had significant results on Levene's test for equality of variance indicating different variances across communities. Type I sum of squares were used to adjust for unequal variance. For the route analysis, the planned Roy Bargmann step-down analysis could not be conducted because heterogeneity of the regression coefficient was found across cells when using the dependent variables as covariates; instead, a discriminant analysis is planned to follow up a significant multivariate effect. The traffic safety difference variable was normally distributed, with no outliers, and covariates met the assumption of homogeneous regression coefficients between the covariates and communities across cells. For a significant effect, Bonferroni pairwise comparisons will summarize the pattern of differences between communities. For the block and walking route analyses (hypothesis 1 and 3), two analytic strategies are presented because violation of statistical assumptions made it impossible to conduct planned step-down analyses of unique effects of each dependent measure. First, univariate analyses of variance are conducted for each of the six IMI scales. To adjust for multiple comparisons, a Bonferroni adjusted significance level of p < 0.008 was used for each planned comparison. Second, multivariate analyses of variance (hypothesis 1) and covariance (hypothesis 3) with follow-up discriminant function analyses will examine how the six IMI scales cluster and how these clusters distinguish among the three communities. For the walking routes, substantial correlations among the six IMI walking route scales prompted a principal components analysis to detect underlying walkability components. Resulting components are tested with univariate tests with Bonferroni adjusted significant levels for multiple comparisons. 22 ofthe covariates; instead, a discriminant analysis is planned to follow up a significant multivariate effect. p three communities. For the walking routes, substantial correlations among the six IMI walking route scales prompted a principal components analysis to detect underlying walkability components. Resulting components are tested with univariate tests with Bonferroni adjusted significant levels for multiple comparisons. RESULTS Recall that two orthogonal planned contrasts that were weighted for unequal group sizes will test the hypotheses that communities differed in walkability, as reflected in their IMI block level and walking route scales. The first contrast compared the new urban community with the other two communities, and the second contrast compared the mixed and suburban communities. Community Differences in Walkability at the Block Level The correlation matrix in Table 1 shows bivariate correlations among the six IMI block scales. The small correlations among accessibility, density, and diversity support the decision to split the Day et al. (2006) access scale into these three separate scales. Additionally, the table shows that some of the IMI block scales are significantly correlated but values are not large enough to raise concerns about multicollinearity. For the univariate analyses of variance on individual block scales, the first planned contrast compares the new urban community to the other two communities (hypothesis 1). The new urban community is more walkable than the other two communities on four of the IMI block scales: density, diversity, pleasurability, and traffic safety (see Table 2). The second contrast indicates that when blocks are the unit of analysis, the mixed community was not different from the suburban community; this lack of difference is sensible, given that they were both composed of suburban blocks. The multivariate analysis of variance (MANOVA) on community type revealed 1M! of difference is sensible, given that they were both composed of suburban blocks. 24 Table 1 Access- Crime Pleasur- Traffic ibility Safety Density Diversity ability Safety Accessibility Crime Safety 0.283** Density 0.102 0.172** Diversity -0.138* -0.124 0.013 Pleasurability 0.151* 0.037 0.271** 0.402** Traffic Safety 0.091 0.001 -0.018 0.316** 0.282** *p < 0.05 (2-tailed). *V<0.01 (2-tailed). Note. The block scales were top and bottom coded to reduce skew. Table 2 IMI block scales by community: Univariate analysis of variance results Planned Comparison Mean Significance Level New Urban Versus Mixed Univariate New Urban Mixed Suburban F(2,221) Partial rf Mixed + Suburban Versus Suburban Accessibility 0.09 0.07 -0.03 3.53 0.03 0.214 0.048 Crime Safety 0.09 0.07 0.01 1.97 0.02 0.245 0.042 Density 0.28 -0.11 -0.03 17.79**' 0.14 0.000 0.051 Diversity 0.08 -0.06 -0.03 10.81*** 0.09 0.000 0.024 Pleasurability 0.14 -0.08 -0.03 15.62*** 0.12 0.000 0.094 Traffic Safety 0.07 -0.00 -0.05 5.15* 0.05 0.006 0.169 *p < 0.008. ** p< 0.001. ***p < 0.0002. Note. Bonferroni adjusted (a I number of tests). Type I sum of squares corrected for unequal variances. Correlations among IMI block scales IMl 0.283" 0.138' 0.151' 'p < 0.05 (2-tailed). "p < 0.01 (2-tailed). 0.172" 0.271" 0.402" 0.316" 0.282" o/variance , p < 0.008 . . , p < 0.001. "'p < 0.0002. F(2,221) = 17.79'" 10.81'" 15.62'" 5.15' 'l / 25 an overall significant multivariate effect using Pillai's criterion, Pillai's 0.31, F(12, 434) = 6.55, p < 0.001, partial rf = 0.15. The follow up discriminant analysis revealed two significant functions. The first function has an eigenvalue of 0.34 and accounts for 85.27% of the variance with a canonical R2 of 0.25, Wilks' Lamba = 0.71, F{\2, 432) = 6.81, < 0.000. The second function is also significant with an eigenvalue of 0.06 and accounts for 14.73% of the variance with a canonical R2 of 0.05, Wilks' Lamba = 0.95, F(5, 217) = 2.52, p < 0.03. The structure matrix in Table 3 reveals the nature of both functions. There are three block scales that load on the first function: density, pleasurability, and diversity. The first function appears to represent "Aesthetic Destinations". The second function is also characterized by three block scales: accessibility, traffic safety, and crime safety. The second function appears to represent "Safe Access." The standardized discriminant function coefficients in Table 3 show the unique contribution of each scale to that function. When the data are viewed this way, density, diversity, and pleasurability make the largest unique contributions to Aesthetic Destinations (function 1); and for Safe Access (function 2), all scales except density and diversity make large unique contributions. The standardized discriminant function coefficients contribute to understanding the functions because they show that on a block level pleasurability is also important for Safe Access (function 2) separating the mixed and suburban communities but the contribution is in the opposite direction of the other three block scales. This shows that when walking routes are accessible, traffic safe, and safe from crime that those walking routes might not contain pleasant environmental Pillai's = 434) 6.55,p < 0.001, partial r/ 0.15. The follow up discriminant analysis revealed two significant functions. The first function has an eigenvalue of 0.34 and accounts for 85.27% of the variance with a canonical R2 of 0.25, Wilks' Lamba 0.71, F(12, 432) 6.81, P < 0.000. The second function is also significant with an eigenvalue of 0.06 and accounts for 14.73% of the variance with a canonical R2 of 0.05, Wilks' Lamba = 0.95, F(5, 217) = 2.52, p < 0.03. The structure matrix in Table 3 reveals the nature of both functions. There are three block scales that load on the first function: density, pleasurability, and diversity. The first function appears to represent "Aesthetic Destinations". The second function is also characterized by three block scales: accessibility, traffic safety, and crime safety. The second function appears to represent "Safe Access." IMI features. Perhaps this finding is because safety and access features might not provide a Table 3 Structure matrix and standardized discriminant function coefficients for block scales Structure Matrix Standardized Coefficients Aesthetic Safe Aesthetic Safe Scales - Block Destinations Access Destinations Access Diversity 0.692 -0.043 -0.481 0.023 Pleasurability 0.646 -0.137 -0.335 0.374 Density 0.537 -0.115 -0.689 0.045 Accessibility 0.106 0.697 -0.051 -0.621 Traffic Safety 0.293 0.551 -0.129 -0.617 Crime Safety 0.108 0.489 -0.050 -0.351 Note. Variables are ordered by absolute size of correlation within function. function Coefficients IMI 26 pleasant view to pedestrian. The standardized discriminant coefficients are also used to compute discriminant function scores which generate the group's centroids for each function. Figure 1 shows the centroids for the IMI block scales in each community on both functions. Even though the centroid differences do not appear large, both functions are significant. The x-axis shows that Aesthetic destinations (function 1) separates the new urban community from the other two communities and is similar to the first planned contrast. The y-axis shows that Safe Access (function 2) separates the mixed and the suburban communities and is similar to the second contrast. Perceived Barriers and Walking Routes It was hypothesized that after controlling for covariates, parent and child perceptions of path barriers along the walk to school would relate to the six IMI walking route scales (hypothesis 2). For the perceived path barriers (walk too difficult, dangerous traffic, etc.), Table 4 portrays partial correlation coefficients between parents' and students' perceptions of walkability and the six IMI walking route scales controlling for the covariates. As expected, the general pattern is that when trained observers rate the blocks along the route as more walkable, parents and children also report more route walkability. The exception to this general pattern is the density walking route scale, which yields a pattern of more perceived path barriers for denser routes. This pattern was obtained for both parents and children, and may at first seem opposite from the typical pattern where high density is associated with more walkability (Braza et al., 2004; Kerr et al., 2006; McDonald, 2008; Zhu & Lee, 2008). However, it is important to consider how 27 the other two communities and is similar to the first planned contrast. The y-axis shows that Safe Access (function 2) separates the mixed and the suburban communities and is similar to the second contrast. walk ability walkability. aI., aI., & H r/i m u u 0 *s ,<L> CO -4 Aesthetic Destinations Figure 1 Block level centroidplot of functions 4- 2 - -2 - ·4- I -4 j ~', Mixed • • Suburban j (I • New Urban , Aestlletic centroid plot functions I 4 28 Table 4 Partial correlations among parent and child route walkability perceptions and IMI walking route scales IMI scales - Parent Perceptions Child Perceptions walking route Path Barriers Crime Path Barriers Crime Accessibility -0.476"' -0.299*** Crime Safety -0.383"* -0.139 -0.276** -0.327*** Density 0.462*** * * * Diversity -0.361*** -00..431010* ** Pleasurability -0.399** -0.428**' Traffic Safety -0.524*** -0.467*** V<0.05. /7<0.01. ***p< 0.001. Note. Higher numbers indicate more barriers, crime concern, or walkability. Covariates: Ideally walk, Rooms in home, Parental education level, and Home owner Partial correlations among parent and child route walkability perceptions and 1Ml walking route scales scales­walking Accessibility Density • p < 0.05 . •• p < 0.01. "'p < 0.001. 0.476'" 0.383·'- 0.462"­- 0.361-·- 0.399·· 0.524··' 0.299'" 0.276·· 0.327*·· 0.300··· 0.411·" 0.428'" 0.467*" education level, and Home owner 29 30 results for density at the block level are different from results for density at the route level. When single blocks are considered, density is highest on the new urban blocks (density mean: new urban = 0.28, mixed = -0.11, suburban = -0.04); however, when walking routes are considered, density is highest in the suburban community (density mean: new urban = 0.28, mixed = -0.11, suburban = -0.04) (Appendix C). To clarify how density is highest at the block level for the new urban community but highest at the route level for the suburban community, the survey, neighborhood maps, and walking routes were examined more closely. The surveys revealed that new urban community parents and children perceived few barriers to walking to school, with averaged values for all 4 path barriers equal to 1.31 for parents and 1.38 for children on a 4-point scale. Suburban parents and children perceived more barriers with averaged values of 2.84 for parents and 2.38 for children on a 4-point scale for both groups. Suburban neighborhood maps indicated that one arterial street with high density was on many walking routes (n = 31). Figure 2 shows this phenomenon whereby a dense arterial street has many walking routes "tunneled" onto it. Additionally, in the new urban community many routes (n = 20) passed blocks that featured green spaces or community gardens and no buildings, creating a lower overall density score for new urban walking routes. Mean scores for the IMI density walking route scale show that the suburban mean (0.45) is three times larger than the new urban (0.14), and over six times larger than the mixed (-0.07). This pattern appears to relate to differences in how pedestrian traffic is channeled in suburban and new urban communities, and will be discussed further in the discussion section. In surveys, parents and children also reported their perceptions of crime dangers. mean: new urban 0.28, mixed -0.11, suburban -0.04) (Appendix C). = funneled" = discussion section. Figure 2 Walking routes funneled through a highly dense arterial street 32 The children's perceptions of crime danger show significant partial correlations with their IMI crime safety walking route scales. The inverse relationship indicates that children's perceptions and IMI ratings are in accord about signs of crime safety. Somewhat surprisingly, parent perceptions of crime danger are not correlated with the IMI crime safety walking route scales. Further examination indicated that this lack of effect is related to overlap between the covariate "ideally would walk" and parents' perceptions of crime danger, simple r = 0.41, p < 0.001. As might be expected, when parents are concerned about crime on their child's route to school, they are less enthusiastic about their child walking to school. With "ideally would walk" removed, there is a significant inverse relationship between the IMI crime safety walking route scale and parent perceptions of crime danger, partial r = -0.235,/? < 0.01, showing that parents' perceptions and IMI ratings are in accord. Community Differences in Walkability at the Walking Route Level Children's estimated walking routes to school in the new urban community were predicted to be more walkable than those in the mixed and suburban communities, with the routes in the mixed community being more walkable than the suburban community (hypothesis 3). Recall that the routes in the mixed community include sections of a suburban and a new urban community. significantly correlated, with correlations up to r = 0.668. A principal components analysis with varimax rotation was employed to identify higher order components. The rotated component matrix displayed in Table 6 indicates that two components could characterize participants' estimated walking routes. The first component comprises = P danger,partial r 0.235, p Table 5 shows that almost all of the six IMI estimated walking route scales are = Table 5 Correlations among estimated walking route IMI scales IMI scales - Access­Crime Pleasur- Traffic Walking Route ibility Safety Density Diversity ability Safety Accessibility Crime Safety ** 0.512 Density - Diversity - Pleasurability -0.143 - Traffic Safety p 0.05 (2-tailed). **/?<Note. The walking route scales were top and bottom coded to reduce skew. Table 6 Walking route scales - Rotated component matrix IMI scales - Component Walking Route 1 2 Accessibility 0.192 0.803 Crime Safety 0.213 0.726 Density -0.107 -0.812 Diversity 0.811 0.321 Pleasurability 0.934 -0.013 Traffic Safety 0.831 0.386 Note. Varimax rotation with Kaiser Normalization was used. IMl scales 1M1 scales- Access- Crime Traffic ibilit~ Safet~ Densit~ Diversit~ abi l it~ Safet~ 0.512** -0.565** -0.409** 0.450** 0.489** -0.345** 0.317** 0.429** 0.612** 0.484** 0.439** -0.544** 0.668** 0.649** *< **p < 0.01 (2-tailed). matrix 1M1 33 34 diversity, pleasurability, and traffic safety and conveys that idea that these are pleasant, traffic safe walking routes with interesting destinations. The first component is called "Safe, Pleasant Destinations," which highlights key elements of new urbanism. The loadings on the second component suggest areas of low density, high accessibility, and high safety from crime. The second component is called "Low Density, Safe Access" and these walking route features are a mixture of new urban (crime safety and accessibility) and suburban elements (crime safety and low density). Both walking route components are normally distributed without outliers. However, the second component, Low Density, Safe Access, failed Levene's test for equality of variance, necessitating the use of Type I sums of squares. Two orthogonal contrasts compared the communities' estimated walking routes to school on each MANCOVA) Pillai's = 30.85, p < 0.000, rf= 0.55. Discriminant analysis was used to understand the linear combinations that MANCOVA created to separate the communities. A discriminant analysis on the six IMI walking route scales with covariates yields two significant functions. The first function has an eigenvalue of 2.53 and accounts for 80% of the variance with a canonical R2 of'0.72, Wilks' Lamba 0.17, F(12, 298) 34.83,/? < component using univariate tests with Bonferroni adjusted alpha levels. Results in Table 7 show that the new urban community is significantly more walkable than the other two communities on Safe, Pleasant Destinations (component 1) but not on Low Density, Safe Access (component 2). The mixed group was significantly more walkable than the suburban group on both walking route components. The multivariate analysis of covariance (MANCOV A) shows that the communities are different on the six IMI walking route scales, Pillai 's = 1.11, F(12, 300) = P r/ = of2.53 of = = 34.83,p Table 7 Univariate planned contrast results for components Planned Comparison Adjusted Mean significance levels New Urban Versus Mixed New Sub - Partial Mixed Versus Urban Mixed urban 154)= rf Suburban Suburban Safe, Pleasant 1.20 -0.04 -0.54 48.43 0.386 0.000 0.000 Destinations Low Density, 0.07 0.77 -0.89 76.19*** 0.497 0.45 Safe Access *p < 0.025. **p< 0.005. **V< Note. Bonferroni adjusted (a I number of tests). Type I sum of squares corrected for unequal variances. components Components " p "" p < """p < 0.0005. F(2, 154) = 48.43'" 76.19""" + r/ 0.000 / 35 36 0.000. The second function is also significant with an eigenvalue of 0.63 and accounts for 20% of the variance with a canonical R2 of 0.39, Wilks' Lamba = 0.61, F(5, 150) = 19.00,/? < 0.000. The structure matrix in Table 8 reveals the nature of both functions that separate the communities on the six walking route scales. Three walking route scales load on the first function: crime safety, accessibility, and density. The first function appears to represent "Low Density, Safe Access." The second function is characterized by two IMI walking route scales: pleasurability and diversity. The second function appears to represent "Aesthetic Destinations." These two functions are similar to what the principal component analysis revealed (see Table 6). The discriminant analysis and the principal component analysis results are slightly different because discriminant analysis creates linear combinations of the six IMI walking route scales to maximize group differences; whereas, principal components analysis seeks factors underlying the six IMI walking route scales, without regard to maximizing group differences. The standardized discriminant function coefficients in Table 8 show the unique contribution of each scale per function. For Low Density, Safe Access (function 1), all scales except traffic safety and diversity make large contributions. For Aesthetic Destinations (function 2), all scales except accessibility are big contributors. These findings add to knowledge from the structure matrix because traffic safety and diversity do not uniquely contribute much to Low Density, Safe Access but are big contributors in Aesthetic Destinations. function. =: =: 19.00,p IMI IMI function function differences. Recall that the standardized discriminant coefficients are also used to compute discriminant function scores which generate the group's centroids for each function. Table 8 Route - Structure matrix and standardized discriminant function coefficients Standardized Structure Matrix Coefficients Scales - Function Function Walking Route 1 2 1 2 Crime Safety -0.646 -0.139 Accessibility -0.088 Density 0.480 0.306 Diversity 0.651 Pleasurability -.0352 0.621 Traffic Safety -0.374 0.164 Note. Variables are ordered by absolute size of correlation within function. function IMI Scales­Walking -0.498 -0.131 -0.375 Coefficients -0.648 -0.348 -0.388 0.509 0.274 -0.068 0.818 -0.430 0.606 0.088 -0.322 function. 37 38 Figure 3 shows the centroids for each community on both functions. The x-axis shows that Low Density, Safe Access (function 1) is similar to the second planned contrast separating the mixed, and new urban, community from the suburban community. The y-axis shows that Aesthetic Destinations (function 2) is similar to the first planned contrast separating the new urban community from the other two communities. Traffic Safety Variability The suburban community's walking routes to school were expected to have the greatest traffic safety variability (i.e., have really safe and dangerous parts of the walk) and the new urban community the least variability (hypothesis 4). Planned comparisons were not used in this analysis rather pairwise comparisons were used to assess ANCOVA) difference, F(2, 154) = 10.12, p < 0.000, rf = 0.28. Bonferroni pairwise comparisons on adjusted means (Table 9) confirmed that walking routes in the suburban community have more traffic safety variability than the new urban community. Traffic safety variability along the route was also higher in the suburban community than the mixed community. For descriptive purposes, Table 9 shows the mean for the least and most traffic safe blocks along walking routes for each community. These scores indicate that the walking routes in the suburban community provided the least traffic safe blocks and the new urban community provided the most y­axis community differences. An analysis of covariance (ANCOV A) on the traffic safety difference score tested this hypothesis and shows a significant community difference, 154)= 1O.12,p<0.000, ,/=0.28. traffic safe blocks. C C5 C o 4> 5 H -4H -4 39 Mixed Low Density, Safe Access Walking route level centroid plot on both functions 4- 0 - -~ - 4 - I Suburban • I "'I I o • IVlixed New Urban • I "'I Figure 3 both/unctions I 4 Table 9 Traffic safety variability for walking routes in three communities: Means for adjusted traffic safety difference scores, least traffic safe blocks, and most traffic safe blocks. Community Traffic Safety Difference Mean Least Safe Mean Most Safe Mean New Urban 0.725a 0.03 0.72 Mixed 0.786a -0.17 0.61 Suburban 0.996b -0.43 0.59 Note. Covariates: Ideally walk, Rooms in home, Parental education level, and Owner. A larger group mean indicates more variability in traffic safety for the walk to school. Different subscripts in the column denote means differ atp 0.001 for Bonferroni pairwise comparisons. communities: traffic blocks. at p < 40 DISCUSSION communities based on three distinct analyses. Trained observer's ratings of blocks and walking routes showed that the new urban community had more micro level environmental features that support walking. These features include safety from crime, access to other areas, housing density, land use diversity, pleasant features, and traffic safety. When these features are present along the route to school, both parents and their children perceive the route to be more walkable. As expected in hypothesis 1, the new urban community was more walkable than both the mixed and suburban communities on four of the IMI block scales: density, diversity, pleasurability, and traffic safety. Additionally, there were no community differences on the IMI block scales between the mixed and suburban communities. The discriminant analysis results from this hypothesis also revealed that on a block level, the IMI scales clustered together into two functions: Aesthetic Destinations and Safe Access. Even on this basic, block level the new urban community was still different than the mixed and suburban communities. This study represents one of the times the IMI scales have been used to characterize communities. And this study also provides one of the few systematic explanations of the micro level environmental features of new urban and standard suburban communities. Overall, this study found substantial support for the walkability of new urban traffic 42 The second way that this study used the IMI to examine community design was by compiling estimated walking routes to school. Results showed that when trained raters found micro environmental features of a walking route supportive of walking, both parents and their children perceived the route to school to be more walkable, with fewer path barriers and lower levels of crime concerns (hypothesis 2). This shows that parents, children, and the IMI all had similar perceptions about the quality of environmental features along these walking routes to school. Analysis of the six IMI walking route scales (hypothesis 3) revealed that the new urban community had walking routes that were consistently more walkable than the mixed and suburban communities on all IMI walking route scales. In addition, the mixed community had more walkable routes than the suburban community on all IMI walking route scales (accessibility, crime safety, density, diversity, pleasurability, and traffic safety) because the mixed community's walking routes had both new urban and suburban blocks. The block level discriminant analysis indicated that the six IMI scales formed two functions that discriminated among communities: Aesthetic Destinations and Safe Access. Additionally, both the principal components analysis and discriminant analysis showed that the six IMI walking route scales cluster together to characterize two larger concepts related to walkability: Low Density, Safe Access and Pleasant Destinations. These findings suggest that the six concepts measured by the IMI scales are not stand alone concepts but rather interrelated concepts that cluster together to describe community differences. Finally, the third way that the IMI was used to examine community design is by testing the hypothesis that walking routes in the suburban community had more traffic ofa were consistently more walkable than the mixed and suburban communities on all IMI walking route scales. In addition, the mixed community had more walkable routes than the suburban community on all IMI walking route scales (accessibility, crime safety, density, diversity, pleasurability, and traffic safety) because the mixed community's walking routes had both new urban and suburban blocks. differences. traffic 43 safety variability than either the mixed community or the new urban community (hypothesis 4). These results suggest that suburban community designs can pose traffic barriers that prevent walking to school. Although suburban residents may live on cul-de-sacs with little traffic, walking to a daily destination like a school requires traversing more highly trafficked streets. This variability in traffic safety might necessitate bus service or parental chauffeuring for students to commute to school safely. Parents might also want to consider the traffic safety of routes to school when purchasing a house, instead of simply focusing on the conditions on the home block. When comparing the block and route analyses it is apparent that the IMI walking route scales had stronger effects than the IMI block scales. This walking route related amplification of environmental features helped stress and identify features that are important for walking to school that might not have been apparent with a simple block level analysis. These findings suggest that community differences are related to community designs and the underlying behavioral goals associated with those designs. The new urban community was designed with the goal of having all residences within a five minute walk of parks, community gardens, or other forms of open space. Although the nearby open space makes for convenient recreational access after work and school for residents, it also provides the benefit of creating pleasant walks to school. In suburban communities, residents often have to leave the low traffic neighborhood of cul-de-sacs venturing onto or across an arterial to get to school and other nonresidential destinations. In the new urban communities, the schools and other destinations may not require crossing arterials but may require walking within sight of pleasant neighborhood green space. Thus, new urban and suburban design philosophies may create very different traffic de­sacs different 44 conditions for pedestrians, tunneling the new urban pedestrians past community amenities, but banishing the suburban pedestrians to the arterials. Although this study does provide considerable evidence that new urban communities are more walkable than suburban communities, caution is needed when interpreting the results of this study. Recall that this study sampled from preexisting communities, so random assignment to community was impossible. As a partial counter to potential nonrandom selection of certain types of residents into communities, both sociodemographic controls and an attitudinal control (ideally I would like my child to walk to school) were used. If a parent who moved into the new urban community were motivated by the walkability of their child's route to school, this attitudinal variable may capture that effect and remove it from the effects of their perceptions of walkability. Another limitation is that the routes studied were the shortest routes to school and it was not possible to have children verify whether those routes were the ones taken. Furthermore, only three communities within one Salt Lake County suburb were examined and future research will be needed to establish the generalizability of these results. Additionally, the IMI items can offer imprecise measures of some important constructs. Although residential density varies along a continuum, density's scoring is the same for 5 large lot homes as for 5 small lot homes; blocks with 1 apartment complex are scored the same as blocks with 15 apartment complexes. When using environmental features to measure walking routes, researchers might encounter data analysis problems (e.g., redundancy, or violating assumptions) like this study did. As most researchers know, field data can be notoriously cumbersome because real life data does not always offer normal distributions or have the same controls as funneling walk ability. large lot homes as for 5 small lot homes; blocks with 1 apartment complex are scored the same as blocks with 15 apartment complexes. 45 laboratory data. This study altered the IMI data as little as possible to meet assumptions for data analyses. Environmental audits typically involve many items with dichotomous scoring and low base rates so that individual items may not be normally distributed and composites may require efforts to normalize. Few studies have examined micro level environmental features along walking routes with an audit instrument like the Irvine-Minnesota Inventory (IMI). By using this method, a wealth of information was added to understanding how community design is related to walking to school. This environmental information would not have been available with many macro level environmental walkability measures, such as population density, that are commonly assessed at large geographic levels of aggregation such as census tracts. This study's findings also have implications for future community designs or community revitalization efforts. Environmental features like safety from crime, traffic safety, aesthetic features, diverse land uses, density of housing, and access to other areas were important concepts that separated communities. This finding is consistent with previous research on high, medium, and low walkable urban routes that found that while walking participants made numerous references to traffic safety features, aesthetic features, and incivilities related to crime danger (Brown et al., 2007). Walking in a community, whether to school or for leisure, is an easy way to increase physical activity and most Americans need more physical activity to help circumvent one of the nation's largest health concerns, obesity. If residents' designs should be reexamined to support physical activity for all residents. The broader lMI). study'S traffic aI., Ifresidents' physical activity is hampered by community design, then community benefits of an active pedestrian community include less air pollution from reduced car travel, safer communities due to eyes on the street, and better social environments that give more opportunities to connect with neighbors. The IMI characterizes communities block by block and could be used to improve dangerous sections of walking routes much like safe routes to school has done (Boarnet et al., 2005a; Boarnet et al., 2005b; Hubsmith, 2006; McMillan, 2005). While improving the walkability of a route or a community, city planners might consider using an audit tool like the IMI because it provides a block by block analysis of micro environmental features to aid in revitalization. This could be particularly helpful because city improvements are typically done on a small scale. The current study found evidence that the new urban community was consistently more walkable than the other two communities. This study lends evidence to new urbanists' claims of designing more walkable communities. Walking in a community for leisure or to a destination like school is an overlooked source of physical activity. This research suggests that community design can influence micro level walkability features supportive of walking. Given the public concern over decreasing percentages of children who can walk to school, the use of new urban designs may allow communities to counteract this trend and re-instate walking to school as a normal and healthy activity. 46 aI., aI., APPENDIX A RULES FOR DEFINING STREET SEGMENTS The Irvine Minnesota Inventory is designed to assess the walkability of each street segment, which includes both sides of a street between intersections. many cases, there is a clear definition of the beginning and ending points of a street segment. However, this Appendix addresses those instances where rules that are more specific are needed to provide consistent definitions of a street segment. 1. A segment was stopped at any intersection or 90 degree bend in the road if no participant lived beyond this intersection or bend, or if this area was not part of a direct route to school. For an example of this rule, see the picture below. According to this rule we would not rate student J's entire street, from street termination D to intersection A, because half of the street after intersection B is not part of the student's route to school. Furthermore, if there were no other students to the south of student J, then when measuring arterial C, we would only measure from intersection A northward, that is, the student's street to the school's street. School Intersection A Q Student J Arterial C Street Termination D Intersection B In if Artelial / Sire et T ermmation D • Inters e ction N 48 If a street ended at another street, or intersection, then that is considered the end of the segment. In example 1, segment 731 stops at segment 735 because the street ends at segment 735. Additionally, if a segment intersected with, or terminated at, a major arterial road, then that is considered the end of the segment. In example 2, segment 728 ends because it crosses arterial A. Example I Example 2 731 7 • Aite rial 72< / 3. A street form was not split into multiple segments if one of the two resulting segments were exceedingly small (less than 300 feet), or if the resulting segments were very similar in their geographic/IMI features. The decision about segments being too small or the similarity of features was made by the group leader through site visits. See example 1 for an illustration of the resulting segment being too small. Segment 574 comes to an intersection at segment 572 but segment 574 does not stop at segment 572 because the new segment would be only three houses long. 2. Segment 31 I Segment 735 Segment 731 ends / Exampl(' ~ Segment 728 Arterial A ! Segment 728 ends / I ~""''''''' ........... ...•.. ....•.. 49 resulting segments would be similar in geographic/IMI features. Segment 795 could geographic/IMI features and were left as one segment. Segment 574 Ends Too small Segment 572 Segment 574 ~<* Segment 796 i >'~ Segment 795 Segment 795 ends 7^ 4. A segment did not stop because of an intersection with a cul-de-sac. For an illustration of this rule, see example 1 below. Segment A is defined as the distance from one four-way intersection to another four-way intersection. Instead of stopping and starting at every cul-de-sac to form four different but very small segments. Curvilinear streets were the exception to this rule and in some communities, there were multiple streets that made up one curvilinear street form. So the rule used on a curvilinear street was if a cul-de-sac lined up with a street name change, then the cul-de- sac was used as the starting and stopping point between segments. For an illustration of this rule, see example 2 below. The curvilinear street form D is really three different streets, but where do they start and stop? This rule says that if cul-de-sac A lines up with a street name change on curvilinear street D, then cul-de-sac A is In example 2, there is a good example of not breaking up a segment because the 1M I end at segment 796 but the two segments this would create do not differ in their Example 1 I , ____ l --:---... / ----_', ,i Example 2 I . >- I \ )< 50 Example 1 Cul-de-sacs / i. Segment A Example 2 Cul-de-sacs A - J B / C \Sfreet form D \ 5 . Curvilinear streets were difficult to cut up into segments because, at times, the starting and stopping points of segments were not very clear and did not start or stop at intersections or cul-de-sacs. The IMI authors suggested developing and applying a consistent rule when the starting and stopping points of segments were not clear (Appendix A in Boarnet et al., 2006). For this study, whenever the starting and stopping points of a segment were not clear then the geographic layout of the neighborhood was used to help determine the starting and stopping points of the segments. The rule in defining where segments started and stopped was that the new segment started and the old segment stopped at the first building after the open space. Some neighborhoods have paths, small parks, and other open space features that make segments distinctive in character. When one road has houses on both sides, the starting and stopping point for segments AC and AE. Notice that the segment would not stop for cul-de-sac B since the end of segment AC is at the collector, street C, that leads to an arterial road. Exmnpl(' 1 I " 5. aI., 51 a segment. On-site inspection by the group leader is needed to make this judgment call, based on how much of the street walkability would be changed by the open space. that make up this circular street form. Notice that on the left hand side of the picture there is written "Where the open space ends." Segment 781 does not become segment 782 at the segment 820, but rather "Where the open space ends." Segment 781 ends after the open space and segment 782 starts at the first building after the open space. Segment 782 Building Segment 820 Space Ends Connecting Street / / Segment 781 •^Segment 780 The previous picture from rule 5 (above) illustrates this point very well. Notice the division between segments 781 and 780. These two segments started and stopped at then a stretch of open community space, the open community space marks the end of An example of this rule is illustrated in the picture below. There are three streets 6. Curvilinear streets were split up according to intersections with connecting streets. 52 this intersection because this is where a connecting street leads into this curvilinear street form. If a street form made a 90 degree bend in the road and retained the same street name after the bend, then the 90 degree bend in included as part of the segment. Example 1 shows segment 716 and it that has two 90 degree bends in it. The street has the same name before and after the 90 degree bends, and is considered one segment. But if a street form made a 90 degree bend and the street name changed as a result of the bend, then the 90 degree bend was considered the end of the segment. In example 2, segment 770 and 771 show a street form that has a 90 degree bend in it and is really two different streets. After the 90 degree bend the street form has a new street name. Because the street form was given a new name, this rule states that a new segment begins the bend. Example 1 Example 2 Segment 771 End Segment" Segment 770 Ends (the bend) 7 " Segment 716 Begin Segment J 7. E d S t .~ n egmen l I ,I I I I ---. \ ...•. ---. ........................ . " \1 i · •··•·· .... 1 Segme~t , " , , . /';' ,/ . , . APPENDIX B COMPARING THIS STUDY'S SCALES TO BOARNET ET AL. (2006) When the Boarnet et al. (2006) created their original four scales some items were double coded indicating that they could belong to one of two scales. Two double coded items were moved from accessibility into traffic safety, bike lanes and bike markings. Two items were moved, pedestrianized streets and street direction (i.e., one or two way street), from the Boarnet et al. (2006) accessibility scale into our traffic safety scale because we felt that these items were tapping into the concept of traffic safety more than pedestrian accessibility. The pleasurability scale (54 items) used in this research is similar to the Boarnet et al. (2006) pleasurability scale with the exception of 2 items. First, predominant building height was moved from the pleasurability scale into the density scale because, conceptually, building height was more strongly related to density than pleasurability. Second, the IMI authors double coded the presence of a sidewalk buffer meaning it could belong to pleasurability or traffic safety and was assigned to traffic safety. Also note that this study converted the three different types of slopes from three single dichotomous variables into a one-item, categorical variable for slope. The crime scale used in this research (14 items) is the same as the Boarnet et al. (2006) safety from crime scale. The traffic safety scale (52 items) is the same as the Boarnet et al. (2006) perceived safety from traffic scale (please note that 16 of these items are counted APPENDIXB Boamet Boamet Boamet traffic safety. Also note that this study converted the three different types of slopes from three single dichotomous variables into a one-item, categorical variable for slope. The crime scale used in this research (14 items) is the same as the Boamet et al. (2006) safety from crime scale. The traffic safety scale (52 items) is the same as the Boamet et al. (2006) perceived safety from traffic scale (please note that 16 of these items are counted 54 twice, once for each end of the segment). Note that an asterisk (*) by the variable name indicates it was omitted due to lack of variance meaning the items were either always or never present. Variables with a plus sign (+) indicate items that were deleted due to raters' disagreements that could not be resolved. And a minus sign (-) indicates that the variable was dropped because the version of the IMI used did not ask about this item. 1M! 55 Individual IMI items: Boarnet et al. (2006) and present study Boarnet Present Study Description S=Crime Safety TS= Traffic Safety AC=Accessibility PL=Pleasurability CR=Crime Safety DN=Density DV=Diversity Item Scale Item Scale 1 Monument T 1 Monument TS Monuments/markers 2a Crosswalk T 2a Crosswalk TS Crosswalk T 2b Whitline TS White line 2b Colrline T 2b Colrline* TS Colored line 2b Zebrastp T 2b Zebrastp TS Zebra striping Diffrdsf T 2b Diffrdsf15 TS Different road surface T 2b Cmother TS Other type of traffic calming Curbcuts T 3 Curbcuts TS Curb cut Traffsig T 4 Traffsig TS Traffic signal T 4 Stopsign TS Stop sign 4 Yieldsgn T 4 Yieldsgn TS Yield sign T 4 Pedactsg TS Pedestrian activated signal T 4 Pedcrssg TS Pedestrian crossing sign T 4 pdunovps* TS Pedestrian overpass/underpass/bridge Safecros T 5 Safecros TS How safe is it to cross T 6 Convcros TS How convenient it is to cross segment 7 Banners P 7 Banners PL Banners 8a Pedstree A 8a Pedstree* TS Pedestrianized street 8b Streetdir A 8b Streetdir TS Street direction P 9 Alley* PL Presence of alley T 10 Vehlanes TS Number of vehicle lanes llaOpenview P 1 laOpenview+ PL Open view 1 lbViewattr P 1 lbViewattr+ PL Attractiveness of the view 12aSfhatach A 12aSfhatach DN Single family home detached 12aSfhdtach A 12aSfhdtach DN Single family home attached A 12aTownhome DN Town home A 12aMoblhome* DN Mobile home 12aResother A 12aResother* DN Other type of residential use 12aSchool A 12aSchool DV School 12aHighschl A 12aHighschl* DV High school 12aCollege A 12aCollege* DV College 12aSchother A 12aSchother DV Other type of school Table 10 IndividualIMl study Boamet et al. (2006) T= Traffic Safety A=Accessibility P=Pleasurability Monument Monument 2b Whitline 2b Diffrdsf Diffrdsf* 2b Cmother 3 Curb cuts 4 Traffsig Traffsig 4 Stopsign 4 Pedactsg 4 Pedcrssg 4 pdunovps underpasslbridge 5 Safecros Safecros 6 Convcros Convcros 9 Alley 10 Vehlanes 11 aOpenview 11 aOpenview+ IlbViewattr 11 b Viewattr+ 12aSthatach 12aSthatach 12aSthdtach 12aSthdtach 12aTownhome 12aMoblhome 12aHighschl * 56 Table 10 continued Item Scale Item Scale Description 12aPubspace A 12aPubspace DV Public space 12aPbspothr A 12aPbspothr DV Other type of public space 12aGymfitns A 12aGymfitns DV Gym/fitness center 12aMovieth A 12aMovieth DV Movie theater 12aRecoher A 12aRecoher DV Other type of recreational 12aComctlib A 12aComctlib DV use Community center/library 12aMustheat A 12aMustheat* DV Museum/theater 12aPubcivic A 12aPubcivic DV Post office, police station, courthouse, DMV 12aCivother A 12aCivother DV Other type of public/civic space 12aReligion A 12aReligion DV Religious institution 12aMedicine A 12aMedicine* DV Medical facility 12aInsother A 12aInsother* DV Other type of institutional 12aRetrest A 12aRetrest DV Retail stores/restaurants 12aFinancl A 12aFinancl DV Financial institution 12aHotelhos A 12aHotelhos* DV Hotel/hospitality use 12aCardeal A 12aCardeal* DV Car dealership 12aGasserv A 12aGasserv DV Gasoline/service use 12aCommothr A 12aCommothr DV Other type of commercial 12aOffices A 12aOffices* DV use Offices 12aService A 12aService* DV Service 12aOfseroth A 12aOfseroth* DV Other type of office/service 12aLghtind A 12aLghtind* DV Light industrial use 12aMdhvind A 12aMdhvind* DV Medium/heavy industrial 12aIndsothr A 12aIndsothr* DV uses Other type of industrial 12aHarmain A 12aHarmain* DV Harbor/marina 12aUndevlnd A 12aUndevlnd DV Undeveloped land 12aAgriclnd A 12aAgriclnd DV Agricultural land 12aNature A 12aNature DV Nature feature 12aOther A 12aOther* DV Other land use 12bVermixus A 12bVermixus* DV Vertical mixed use 12c Bigbox A 12c Bigbox DV Big Box store 12c Shopmall A 12c Shopmall* DV Shopping mall 12c Strpmall A 12c Strpmall DV Strip mall/strip store 12c Drivthru A 12c Drivthru DV Drive thru 13aParkplay P 13aParkplay PL Park/playground 13aPlaysprt P 13aPlaysprt PL Playing or sports field 13aPlazasq P 13aPlazasq PL Plaza/square 13aPubgardn P 13aPubgardn Public garden DescriEtion use Community center/library 12aHotelhos * Hotellhospitality l2aCardeal use 12aOffices * Offices office/Mediumlheavy uses Other type of industrial 12b V ermixus p layground PL 57 Item Scale Description 13aBeach* PL Beach 13aPubother PL Other type of public space 13bAccesspb PL Public space accessibility 14 Barsclub* CR Bars/clubs s CR Adult use s 14 Chckcash* CR Check cashing store s CR Liquor store Restarnt p 15 Restarnt PL Restaurant p PL Coffee shop p 15 Libbkstr* PL Library/bookstore Cornerst p Cornerst* PL Corner store p 15 Artgllry* PL Art gallery Farnermk p 15 Farnermk* PL Farmers' market p PL Lake/pond p PL Open field/golf course p 16 Fountain* PL Fountain p 16 strmrivr PL Stream/river p 16 Ocean* PL Ocean p PL Forest p PL Mountain p PL Desert A AC Highway A Railroad* AC Railroad 17 Implndus A 17 Implndus* AC Impassable land use A 17 river AC River drnditch A 17 drnditch AC Drainage ditch A 17 Sixlnrd AC Six lane road 17 Barrothr AC Other type of barrier A & 18aSidewalk AC Sidewalk 18bSdwkcomp AC Completeness of sidewalk network P 18c Sdwkcndt PL Condition of sidewalk 18dSdwkpave* PL Decorative/unique sidewalk paving PL Arcades PL Awnings PL Other type of sidewalk protection 18fSdwkbuff P & Sdwkbuff TS Sidewalk buffer 19 Othrsdwk 19 Othrsdwk AC Path other than sidewalk A & TS Bike lane A & 20bBklntype TS Type of bike lane 21aMidblock 21aMidblock TS Midblock crossing Table 10 continued Item Scale DescriEtion 13aBeach P 13aPubother P 13bAccesspb P 14 Barsclub S 14 Adultuse S 14 Adultuse* 14 Chckcash S 14 Liquorst S 14 Liquorst* 15 Restamt P Restamt 15 Coffshop P 15 Coffshop 15 Libbkstr P Library !15 Comerst P 15 Comerst* Comer 15 Artgllry P 15 Famermk P Famermk* 16 Lakepond P 16 Lakepond* 160penFGC P 160penFGC* 16 Fountain P 16 strmrivr P 16 Ocean P 160cean* 16 Forest P 16 Forest* 16 Mountain P 16 Mountain* 16 desert P 16 desert* 17 highway 17 highway* 17 Railroad 17 Railroad * 17Implndus 17 river 17 dmditch dmditch 17 Sixlnrd 17 Barrothr A 18aSidewalk A&T 18bSdwkcomp A 18c Sdwkcndt 18dSdwkpave P 18eArcades P 18eArcades* 18eAwnings P 18eAwnings* 18eSwptcoth P 18eSwptcoth sidewalk 18f Sdwkbuff P&T 18f Sdwkbuff buffer 190thrsdwk A 190thrsdwk sidewalk 20aBikelane A&T 20aBikelane 20bBklntype A&T 21 aMidblock T 21 aMidblock crossing 58 Item Scale Description T 21bMbwtln TS Midblock crossing - white line T 21bMbclln* TS Midblock crossing - colored line 21bMbzebra T 21bMbzebra TS Midblock crossing - zebra striping 21bMbdfrdsf T 21bMbdfrdsP TS Midblock crossing - different road surface T 21bMbother TS Midblock crossing - other P 22 Flatslpe PL Flat/gentle slope P 22 Modrslpe Moderate slope P 22 Stpslpe Steep slope P 23 Dining PL Outdoor dining area 24 Benches PL Benches 24 Busstops* PL Bus stops 24 Heatlamp* PL Heat lamps P 24 Bikerack PL Bike racks P 25 Pubrestr* PL Public restroom P 26aStrttree PL Street trees P 26bSdwkshde+ PL Sidewalk shade P 27 Bldgstry DN Building height S 28 Abndbldg Abandoned buildings GO 29 Prctbldg Percent of segment with buildings S 30 Windbars Bars on windows Frntprch P 31 Frntprch+ PL Front porch P 32 Blnkwall PL Percentage of segment with blank walls 33aGarages PL Number of buildings with garages 33bPrmntgar PL Prominence of garages 34aParkstrc* PL Parking structure P 34bPrdusprk* PL Predominant use of first floor of parking structure P 35 Parklot- Parking lot Driveway P 36 Driveway PL Prominence of driveways Mntbuild s 37 Mntbuild+ Maintenance of buildings Mntlndsc 00 38 Mntlndsc- Landscape maintenance Grafitti S 39 Grafitti Graffiti s 40 Litter+ Litter Dumpster s 41 Dumpster Visible dumpster Elecwire p 42 Elecwire PL Overhead electrical wiring Lighting s 43 Lighting Outdoor lighting Table 10 continued Item Scale Descri,etion 21bMbwtln 21bMbclln 21 bMbzebra 21 bMbdfrdsf 21 bMbdfrdsfl' 21bMbother 22 Flatslpe 22 Modrslpe 22 Stpslpe 23 Dining 24 Benches P 24 Busstops P 24 Heatlamp P 24 Bikerack 25 Pubrestr 26aStrttree 26bSdwkshde 27 Bldgstry 28 Abndbldg CR 29 Prctbldg S CR 30 Windbars CR 31 Fmtprch Fmtprch+ 32 Blnkwall 33aGarages P 33bPrmntgar P 34aParkstrc P 34bPrdusprk structure 35 Parklot 36 Driveway driveways 37 Mntbuild S Mntbuild+ CR 38 Mntlndsc S maintenance 39 Grafitti Grafitti CR 40 Litter S Litter+ CR 41 Dumpster S Dumpster CR dumpster 42 Elecwire P Elecwire wiring 43 Lighting S Lighting CR lighting 59 Table 10 continued Item Scale Description A Presence of freeway T 45 Speedlim Speed limit 46 Spdbump TS Speed bump 46 Rumblesp* TS Rumble strip 46 Curbbulb TS Curb bulb out T TS Traffic circle 46 Median Median 46 Trcmpark Traffic calming - parking 47aCuldesac T 47aCuldesac Cul de sac 47bPedacspt AC Pedestrian access point 48 Attrtve+ PL Attractiveness of segment 49 Historic* PL Historic buildings 50 Intrstng 50 Intrstng+ PL How interesting the segment 51 Stvendor* PL is Street vendors 52 Pubart* PL Public art 53 Billbrd PL Billboard S 54 Safepple How safe you feel walking s 55 Dogs CR Dogs p 56 Dmntsmll* PL Dominant smell Monumnt2 1 Monumnt2 TS Monuments/markers 2 2 Crosing2 Presence of crosswalk 2b Whtline2 White line 2 T 2b Colline2* TS Colored line 2 T 2b Zebra2 TS Zebra striping 2 Dfrdsf2 T 2b Dfrdsf2* Different road surface 2 T Crsothr2* Other type of crosswalk marking Curbcut2 T 3 Curbcut2 TS Curb cut 2 Trafsgn2 4 Trafsgn2 TS Traffic signal 2 4 Stopsgn2 TS Stop sign 2 Yield2 4 Yield2 TS Yield sign 2 T 4 Pedact2 TS Pedestrian activated signal 2 Pedestrian crossing sign 2 T 4 Pedbrgd2* Presence of pedestrian bridge Trfsafe2 T 5 Trfsafe2 TS How safe is it to cross 2 Trfconv2 6 Trfconv2 How convenient it is to cross segment 2 Note. + indicates item was deleted due to irresolvable raters' disagreements. research Item Scale DescriEtion 44 Freeway 44 Freeway AC 45 Speedlim TS 46 Spdbump T 46 Rumblesp T 46 Curbbulb T 46 Trafcrcl 46 Trafcrcl 46 Median T TS 46 Trcmpark T TS 47 aCuldesac 47aCuidesac TS CuI 47bPedacspt A 48 Attrtve P 49 Historic P 50Intrstng P is 51 Stvendor P Street vendors 52 Pubart P 53 Billbrd P 54 Safepple CR 55 Dogs S 56 Dmntsmll P 1 Monumnt2 T Monumnt2 2 Crosing2 T TS 2b Whtline2 T TS 2b Colline2 2b Zebra2 2b Dfrdsf2 Dfrdsf2* TS 2b Crsothr2 2b Crsothr2 * TS 3 Curbcut2 4 Trafsgn2 T Trafsgn2 4 Stopsgn2 T 4 Yie1d2 T 4 Pedact2 4 Pedcrs2 T 4 Pedcrs2 TS 4 Pedbrgd2 TS 5 Trfsafe2 Trfsafe2 6 Trfconv2 T Trfconv2 TS * indicates item was omitted because it was either always or never present. - indicates item was dropped because the IMI version downloaded for this research did not ask about this item APPENDIX C DESCRIPTIVES FOR ALL VARIABLES Table Descriptives for variables used in all analyses New Urban Community Mixed Community Suburban Mean (Std Dev) Mean (Std Dev) Mean (Std Dev) Covariates (n= 161) Ideally Walk Number of Rooms Owner or Renter Education Level Survey Items (n=129) Parent Path Barrier Parent Crime Question Student Path Barrier Student Crime Question Scales - Block (n=224) Accessibility Crime Safety Density Diversity Pleasurability Traffic Safety Scales - Walking Route (n= Accessibility Crime Safety Density Diversity Pleasurability Traffic Safety Traffic Safety Difference Score Component Score 1 Component Score 2 3.862 (0.571) 7.652 (2.486) 0.8 (0.332) 3.8 (0.695) 1.31 (0.382) 1.310(0.604) 1.229 (0.294) 1.375 (0.647) 0.09 (0.242) 0.094 (0.125) 0.28 (0.416) 0.08 (0.22) 0.137 (0.190) 0.072 (0.213) 161) -0.012(0.141) 0.09 (0.071) 0.142 (0.179) 0.473 (0.135) 0.18(0.054) 0.291 (0.146) 0.69 (0.245) 1.417(0.708) -0.054 (0.534) 3.333 (0.95) 8.149 (2.01) 0.962 (0.168) 3.39 (0.875) 1.71 (0.704) 1.55 (0.769) 1.967 (0.743) 1.813(0.941) 0.074 (0.297) 0.069 (0.239) -0.109 (0.322) -0.063 (0.108) -0.08 (0.194) -0.004 (0.211) 0.021 (0.152) 0.122 (0.196) -0.067 (0.226) 0.239 (0.113) 0.062 (0.067) 0.194 (0.103) 0.773 (0.289) -0.019 (0.665) 0.76 (0.526) 2.618(1.094) 9.461 (1.824) 0.98 (0.127) 3.411 (0.954) 2.841 (1.026) 1.912(0.851) 2.511 (0.923) 2.375 (1.071) -0.034 (0.359) 0.006 (0.335) -0.035 (0.325) -0.03 (0.162) -0.029 (0.21) -0.045 (0.186) -0.314(0.195) -0.265 (0.148) 0.45 (0.409) 0.091 (0.132) -0.01 (0.075) 0.001 (0.182) 1.027 (0.248) -0.664 (0.675) -0.82 (0.917) Note. The group mean was substituted for missing demographics. The values are NOT adjusted for covariates when applicable but have been top and bottom coded. APPENDIXC 11 analyses Communit~ Community {Dev2 {Dev2 {Dev2 2.618 (1.094) 1.310 (1.912 (1.8l3 (0.2l3) 0.012 (0.141) 0.314 (0.195) 0.l35) 0.1l3) 0.l32) 0.18 (0.054) 0.0 I 1.417 (Com,eonent {0.5342 {0.5262 {0.9172 APPENDIX D CORRELATIONS AMONG WALKING ROUTE SCALES, COMPONENTS, AND COVARIATES Table 12 Correlations among walking route scales, components, and covariates Number Owner Access Crime Pleasur Traffic Compo Compon Ideally of or ibility Safety Density Diversity ability Safety nent 1 ent 2 Walk Rooms Renter Accessiblity - Crime Safety 0.512** - Density 0.565** 0.409** - Diversity 0.450** 0.489** 0.345** - Pleasure -0.143 0.612** - Traffic Safety ** 0.484 0.439** 0.544** ** 0.668 ** 0.649 - Component 1 0.271** 0.472** -0.082 ** 0.802 0.923 0.733** - Component 2 0.799** 0.558** 0.894** 0.311** 0.030 0.461** 0.000 - Ideally Walk 0.320 -0.314** ** -0.450 0.436** ** -0.400 ** -0.471 0.175** - Number of Rooms 0.274** 0.295** -0.042 0.294** 0.180* 0.153** 0.210** - Owner or Renter -0.035 -0.153 0.018 ** -0.355 ** -0.239 ** -0.248 ** -0.338 0.016** 0.082 0.261** - Parent Education 0.038 -0.117 0.054 0.065 0.133 -0.005 0.094 -0.083 -0.061 0.025 p < 0.05 level (2-tailed). *V<0.01 level (2-tailed). Note. Two covariates, Ideally Walk and Number of Rooms, were log 10 transformed and the IMI Scales were top and bottom coded. APPENDIXD CORRELA TrONS COY ARIATES co variates ibilit~ Safet~ Densit~ Diversit~ abilit~ 0.512" -0.565" -0.409" 0.450'* 0.489" -0.345" 0.317** 0.429** 0.612'* 0.484'* 0.439*' -0.544" 0.668" 0.649*' 0.271" 0.472" 0.802" 0.923" 0.733" 0.799" 0.558" -0.894" 0.311" 0.461" -0.320" 0.314" 0.158* -0.450" -0.436" -0.400" -0.471*' -0.175" 0.274" 0.295" 0.294" 0.180' 0.179* 0.245* 0.153" -0.210*' -0.355** -0.239*' -0.248*' -0.338*' 0.016*' -0.261*' -0.176* '.. . p < 0.01 taIled). log10 APPENDIX E UNIVARIATE ANALYSIS FOR SIX IMI WALKING ROUTE SCALES USING PLANNED CONTRASTS Although the component analysis is the most conservative test, it does not allow comparison of the three communities on all six IMI walking route scales. Univariate tests with the same planned contrasts were used for this analysis. The first contrast in Table 13, new urban versus the mixed and suburban communities, shows a difference on all walking route scales except density. The second contrast, mixed versus suburban, revealed that these communities differed on all six walking route scales. Table 13 Planned contrast results for IMI walking route scales Planned Comparisons Adjusted Mean Significance Levels New Urban Mixed IMI walking New Sub­Partial Versus Mixed Versus route scales Urban Mixed urban F(2, 154) = rf Suburban Suburban Accessibility -0.01 0.01 -0.31 49.26*** 0.390 0.0013 0.0000 Crime Safety 0.10 0.12 -0.27 82.26*** 0.517 0.0000 0.0000 Density 0.09 -0.08 0.49 49.55*** 0.392 0.1145 0.0000 Diversity 0.44 0.24 0.11 48.17*** 0.385 0.0000 0.0000 Pleasurability 0.17 0.06 0.00 43.03*** 0.358 0.0000 0.0000 Traffic Safety 0.27 0.20 0.01 28.55*** 0.270 0.0000 0.0000 *p < 0.008. *>< 0.001. p < 0.0002. - Bonferroni Adjusted (a / number of tests) Note. Corrections for unequal variance were made to the sum of squares. APPENDIXE I MI scales Si~.nificance Sub- Partial Fe2, 154t, = rI + 49.26* • 82.26'" 49.55'" 48.17*" 43.03'" Safet~ 28.55'" 'p < 0.008. "p < '''REFERENCES Ahlport, K. (2008). Barriers to and facilitators of walking and bicycling to school: Formative results from the non-motorized travel study. Health education & behavior, 55(2), 221-244. Al-Kodmany, K. (1997). Comparison between Cul-de-sac and Through Streets. Paper presented at the 28th Annual Environmental Design Research Association (EDRA). Alfonzo, M., Day, K., & Boarnet, M. (2005). Irvine-Minnesota Inventory for observation of physical environment features linked to physical activity. Training protocol. Retrieved June 4, 2007, from https://webfiles.uci.edu/kday/public/index.html Boarnet, M. G., Anderson, C. L., Day, K., McMillan, T., & Alfonzo, M. (2005a). Evaluation of the California safe routes to school legislation - Urban form changes and children's active transportation to school. American Journal Of Preventive Medicine, 28(2), 134-140. Boarnet, M. G., Day, K., Alfonzo, M., Forsyth, A., & Oakes, M. (2006). The Irvine- Minnesota Inventory to measure built environments: Reliability tests. American Journal of Preventive Medicine, 30(2), 153-159. Boarnet, M. G., Day, K., Anderson, C , McMillan, T., & Alfonzo, M. (2005b). California's safe routes to school program - Impacts on walking, bicycling, and pedestrian safety. Journal of the American Planning Association, 71(3), 301-317. Braza, M., Shoemaker, W., & Seeley, A. (2004). Neighborhood design and rates of walking and biking to elementary school in 34 California communities. American Journal of Health Promotion, 19(2), 128-136. subdivisions - Evaluating psychological and social goals. Journal of the American Planning Association, 67(4), 402-Brown, B. B., & Werner, C. M. (1985). Social cohesiveness, territoriality, and holiday decorations: The influence of cul-de-sacs. Environment and Behavior, 77(5), 539- 565. 35(221-AI-de-sac K, https://webfiles.uci.edu/kday/public/2), K, Irvine­Minnesota American 2), K, c., American Brown, B. B., & Cropper, V. L. (2001). New urban and standard suburban subdivisions­Evaluating Planning 4),402-419. 17(64 Brown, B. B., Werner, C. M., Amburgey, J. W., & Szalay, C. (2007). Walkable route perceptions and physical features - Converging evidence for en route walking experiences. Environment and Behavior, 39(1), 34-61. Calthorpe, P. (1993). Next American metropolis: Ecology, community, and the American dream. Princeton, NJ: Princeton University Press. Carver, A., Salmon, J., Campbell, K., Baur, L., Garnett, S., & Crawford, D. (2005). How do perceptions of local neighborhood relate to adolescents' walking and cycling? American Journal of Health Promotion, 20(2), 139-147. Centers for Disease Control and Prevention. (2000a). Healthy People 2010: Objectives for Improving Health (PartB: Focus Areas 15-28). Atlanta, GA: Centers for Disease Control and Prevention. Centers for Disease Control and Prevention. (2000b). Promoting better health for young people through physical activity and sports. Silver Spring, MD. Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: Density, diversity, and design. Transportation Research Part D-Transport and Environment, 2(3), 199-219. Congress for the New Urbanism. (2001). Charter of the new urbanism. Chicago, IL: Congress for the New Urbanism. Cooper, R., Andersen, L. B., Wedderkopp, N., Page, A. S., & Froberg, K. (2005). Physical activity levels of children who walk, cycle, or are driven to school. American Journal of Preventive Medicine, 29(3), 179-184. Cooper, A. R., Page, S., Foster, L. J., & Qahwaji, D. (2003). Commuting to school - Are children who walk more physically active? American Journal of Preventive Medicine, 25(A), 273-276. Day, K., Boarnet, M., Alfonzo, M., & Forsyth, A. (2006). The Irvine Minnesota Inventory to measure built environments: Development. American Journal of Preventive Medicine, 30(2), 144-152. Dietz, W. H. (2004). Overweight in childhood and adolescence. New England Journal of Medicine, 350(9), 855-857. EPA. (2003). Travel and Environmental Implications of School Siting. Retrieved from http ://www.epa. gov/smartgrowth/schooltravel .htm. Falb, M. D., Kanny, D., Powell, K. E., & Giarrusso, J. (2007). Estimating the proportion of children who can walk to school. American Journal of Preventive Medicine, 33(4), 269-275. 1. 1), American 2),139-147. Objectives Part B: young Transport 2(3), A. 3), A. 1., Preventive 4),273-9), http://school_travel.A. Preventive 4),269-65 Field, A. P. (2000). Discovering statistics using SPSS for Windows: Advanced techniques for the beginner. London: Sage Publications. Frank, L. D., Engelke, P. O., & Schmid, T. L. (2003). Health and community design: The impact of the built environment on physical activity. Washington, DC: Island Press. Giles-Corti, B., Broomhall, M. H., Knuiman, M., Collins, C , Douglas, K., Ng, K., et al. (2005). Increasing walking - How important is distance to, attractiveness, and size of public open space? American Journal of Preventive Medicine, 28(2), 169-176. Hoehner, C. M., Ramirez, L. K. B., Elliott, M. B., Handy, S. L., & Brownson, R. C. (2005). Perceived and objective environmental measures and physical activity among urban adults. American Journal of Preventive Medicine, 28(2), 105-116. Hubsmith, D. A. (2006). Safe routes to school in the United States. Children, Youth and Environments, 16(1), 168-190. Jacobs, J. (1961). The death and life of great American cities. New York: Vintage. Jordan School District. (2009). Jordan School District Policy Manual. Retrieved, from http://www.jordandistrict.org/policymanual/p. php?id=197. Kerr, J., Rosenberg, D., Sallis, F., Saelens, B. E., Frank, L. D., & Conway, T. L. (2006). Active commuting to school: Associations with environment and parental concerns. Medicine and Science in Sports and Exercise, 38(4), 787-794. Leslie, E., Saelens, B., Frank, L., Owen, N., Bauman, A., Coffee, N., et al. (2005). Residents' perceptions of walkability attributes in objectively different neighbourhoods: A pilot study. Health Place, 11(3), 227-236. McDonald, N. C. (2007). Active transportation to school - Trends among US school children, 1969-2001. American Journal of Preventive Medicine, 32(6), 509-516. McDonald, N. C. (2008). Children's mode choice for the school trip: The role of distance and school location in walking to school. Transportation, 35, 23-35. McMillan, T. E. (2005). Urban form and a child's trip to school: The current literature and a framework for future research. Journal of Planning Literature, 19(4), 440- 456. McMillan, T. E. (2007). The relative influence of urban form on a child's travel mode to school. Transportation Research Part a-Policy and Practice, 41(\), 69-79. SPSSfor techniques O., c., 2), and Retrieved. id= 197. J. 4),787-walk ability different 3),227-35,23-41(1),69-66 Mota, J., Gomes, H., Almeida, M., Ribeiro, J. C , Carvalho, J., & Santos, M. P. (2007). Active versus passive transportation to school - differences in screen time, so