173 Transportation Research Record: Journal of the Transportation Research Board, No. 2280, Transportation Research Board of the National Academies, Washington, D.C., 2012, pp. 173–182. DOI: 10.3141/2280-19 Department of Urban and Regional Planning, University of Hawaii at Manoa, Saunders 107, 2424 Maile Way, Honolulu, HI 96822. Corresponding author: K. Kim, karlk@hawaii.edu. shown to change physical activity patterns (9); and other correlates between land use, urban form, and transport decisions exist (10). As the mix of different land uses increases, travel behavior should change. With increased entropy (land use mix), greater complexity, more competing agents, and increased activities would be expected, and these developments would in turn be expected to increase inter- actions and promote an increased risk of accidents, collisions, injuries, fatalities, and costs. Increased entropy brings a greater assortment of driver, vehicle, roadway, and environmental factors that come into play, exacerbating roadway hazards. The entropy effects would also be incurred at the boundaries and edges of districts where movements and transitions from one set of conditions to another would increase accident risk. Although entropy may be difficult to model, a reasonable hypothesis is that entropy exacerbates accidents. On the basis of earlier research in Hawaii on land use (11, 12) and accessibility and accidents (13) through the use of a variety of tools, including geographic information systems and geospatial–temporal models (14, 15) and structural equation modeling (16), it is evident that human factors and roadway factors such as volume and road- way configuration play stronger roles than land use or the mix of activity generators in explaining accident risk. Yet, Hawaii provides a unique and compelling environment with which to explore the subtle and challenging connections between land use and safety. Although Hawaii conjures up images of swaying palms and sandy beaches, it is an ideal setting in which to explore the relationships between land use and accidents. First, because of its remote loca- tion and island setting, it provides an isolated laboratory in which to focus on the interactions between driver, vehicle, roadway, and environmental factors. Second, with only four counties and a high degree of centralization for reporting of accidents to police, accident data are generally of higher quality and consistency than those from other more distributed systems. Third, the research presented in this paper builds on more than two decades of experience working with accident, land use, demographic, and transportation interactions. OVERVIEW OF RESEARCH The paper is structured in the following manner. After a description of the data sources and methods, the accident types and independent variables, including land use, traffic volume, roadway characteris- tics, traffic volume, transit use, and roadway patterns, as well as land values and entropy measures, are presented as standard descriptive statistics (means, variance, skewness, kurtosis, etc.). Then, the acci- dent data and entropy measures are mapped and described. Efforts to build, test, and validate the negative binomial regression models are recounted. The best-fitting models are presented along with a discussion comparing the various models with each other and sum- marizing some of the similarities and differences. In a concluding section, the implications and contributions of the work for traffic safety programs and future research endeavors are described. Entropy and Accidents Karl Kim, Pradip Pant, Eric Yamashita, and I Made Brunner This study explored the relationships between land use entropy (the extent to which land uses are mixed, heterogeneous, and nonuniform) and motor vehicle accidents. Two aspects of entropy were considered: (a) the mix of jobs and housing and (b) the diversity of jobs in addition to the mix of jobs and housing. These measures were developed and tested with census data and geographic information system technologies combined with comprehensive police accident reports from the city and county of Honolulu, Hawaii. A grid-based approach was adopted with accident counts and negative binomial regression. Various types of acci- dent counts were considered, including total, daytime, and nighttime accidents, as well as accidents involving tourists, nonuse of seat belts, and driving under the influence of alcohol. Grid-based characteristics were also considered, such as distance from the urban center, traffic volume, roadway length, transit use, land values, and roadway configu- ration (intersections versus dead ends). Although entropy plays a sta- tistically significant role, especially for total accident counts, daytime accidents, and accidents involving tourists, the relationships involving effects such as volume, roadway length, distance to the central busi- ness district, and transit use are generally more readily detected than entropy effects. Although the research shed additional light on the com- plex and subtle relationships between land use and accidents, implica- tions for both traffic safety and modeling of spatial phenomena were also apparent. Rather than examine accidents without consideration of driver characteristics and vehicle and roadway factors, this study esti- mated interactions between human and vehicle factors while also taking into account differences in environmental conditions and land uses that affect crashes at different spatial resolutions. This paper explores the relationships between land use entropy and motor vehicle accidents. The research builds on efforts to explain the relationships between urban form and travel behavior advanced by Handy (1) and extended by Cervero and Kockelman (2) and others who have tackled the questions of how density, diversity, and design influence transportation. Although current research focuses on the linkages between urban form and walking, biking, and public transport use (3–5) and effects on obesity and public health (6, 7), this paper investigates the connections between entropy and acci- dents. Similar to the definition provided by Cervero and Kockelman (2), land use entropy is defined as the extent to which land uses are heterogeneous, mixed, and nonuniform. Sufficient evidence exists to suspect that urban form and residen- tial choices affect travel behavior (8). The presence of parks has been