Transportation Research Record: Journal of the Transportation Research Board, No. 1922, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp. 52–61. 52 This paper uses optimization methodologies and project-level cost and effectiveness data to assess long-term safety needs for a network. The optimal solution specifies values of decision variables (locations, years, and safety improvement types) such that overall cost-effectiveness at the network level is maximized under budgetary constraints. The paper eval- uates the impact of alternative levels of safety funding on systemwide crash reduction and investigates the sensitivity of optimal funding levels to key safety management inputs. To demonstrate the methodology, data from Indiana’s state highway system are used. It is shown that increases in overall safety funding have an increasing effect on crash reduction, but such increasing benefit tapers off after a certain point. It was determined that over a 10-year period (2005 to 2015) the optimal annual average safety need for the network is approximately $450 per mile. Furthermore, it is shown that the overall network safety funding need is sensitive to the method for identifying hazardous locations and the criterion for eco- nomic evaluation. The results show that with currently available data it is possible for highway agencies to incorporate road safety proactively into their transportation planning processes in a comprehensive and system- wide context. Also, agencies can use the methodology to determine opti- mal safety funding levels on their networks for possible comparison with current levels. Highway asset management, of which safety management is an impor- tant component, advocates the combination of engineering principles with sound business practices and economic theory and involves the provision of tools that facilitate a more organized, logical, and inte- grated approach to decision making (1). The transportation sectors of state and local governments are therefore expected and encouraged to ensure appropriate use of public resources and operational account- ability (2). Consistent with such asset management trends is the esti- mation of funding needs for cost-effective long-term management of highway infrastructure. It is expected that in the near future highway agencies will seek safety need estimates often, not on the basis of past practice (historical spending levels) as done now, but on a more rational and accountable basis, such as one that involves maximizing some systemwide utility (such as average crashes saved per dollar). With knowledge of the optimal levels of funding for safety-related highway projects, highway agencies are afforded a rational means with which to incorporate road safety proactively into their short- and long-range transportation planning processes in a comprehensive and systemwide context. Highway safety enhancement efforts can be categorized as operator related, vehicle related, enforcement related, and environment related (including physical infrastructure). State and local highway agencies typically are responsible for addressing safety problems related to the physical road infrastructure, such as narrow lanes and shoulders. Most currently available safety needs analysis tools evaluate safety impli- cations of alternative geometric designs at the project level. To extend such analysis to the network level, the only recourse appears to be the use of current tools to repetitively carry out project-level needs assess- ment for each individual section on the network. This not only is laborious but also provides a solution that often is not optimal from a systemwide perspective. With current availability of more detailed inventory and crash data, cost data, and improved safety management methodologies (3–5), a potential exists to extend such research to address contemporary issues of network-level safety needs. Method- ologies useful for assessment of optimal network-level safety needs have been made available in past research (6–8) but have had little practical implementation, probably because of lack of certain analyt- ical tools and data. FHWA is developing a comprehensive highway safety improvement model and software tools, SafetyAnalyst, to be made available in 2006. The present paper investigates the optimal network safety needs and their sensitivity to key safety management inputs by using data from Indiana’s state highway network. ESTIMATION OF EXPECTED CRASH FREQUENCY A basic requirement of network-level safety management is to iden- tify sections on a road network that need some safety intervention now or will at some future time. This requires prior knowledge of estimated annual safety performance (crash frequency and severity) at various sections of the road network over an analysis period. Considerable research has been done on the prediction of expected safety perfor- mance of highway segments. For the present study, the crash predic- tion procedure is based on the empirical Bayesian (EB) method (4), which provides relatively unbiased estimates of the expected crash frequency. The EB method uses both a historical crash record and predicted crash frequency by using a multivariate crash prediction model. Crash records from 1997 to 2000 were obtained from a com- prehensive Indiana state highway safety database. To exclude outliers from the data set, only sections with lengths between 0.1 and 10 mi and lane widths of less than 15 ft were used. Also, sections within a distance of 200 ft from intersections were excluded. By using the Investigating the Sensitivity of Optimal Network Safety Needs to Key Safety Management Inputs Godfrey Lamptey, Samuel Labi, and Kumares C. Sinha G. Lamptey, Precision Engineering and Surveying, 2520 NW 97th Avenue, Suite 200, Miami, FL 33172. Current affiliation: Corzo Castella Carballo Thompson Salman, P. A. (C3TS), 901 Ponce de Leon Boulevard, Suite 900, Coral Gables, FL 33134. S. Labi and K. C. Sinha, School of Civil Engineering, Purdue University, 550 Sta- dium Mall Drive, West Lafayette, IN 47907.