 He is a safety engineer with the State Highway Department. He has been reading about a Highway Safety Improvement Program management system. According to the National Cooperative Highway Research Program report he is studying, this system will identify real problems instead of perceptions. The solutions will be cost-effective and in the event of legal action, the existence of the system will aid the defense. In the previous programs of this series, we followed Bill as he learned about the system's first four phases. They are Phase 1, identifying safety problems, Phase 2, identifying alternative improvements, Phase 3, evaluating alternatives and selecting improvements, and Phase 4, developing and implementing the Highway Safety Improvement Program. In this program, Bill will work his way through the last two phases, evaluating each safety improvement project and evaluating the overall safety program. The first step in evaluating a safety project is to obtain before and after accident data. Before means before any construction and after means only after the construction is complete. Bill will find out that in gathering information, he must consider the sphere of influence. Improvements like a curve warning sign or reduced speed sign can influence accident experience well beyond the physical location of the improvement. On the other hand, an improvement such as realignment of a curve or a new guardrail generally affects accident occurrence only within the physical limits of the improvement. Improvements at intersections are generally more difficult to evaluate. An intersection improvement may affect accident occurrence over one block away. However, as the distance from the intersection improvement increases, the possibility of other influences also increases. Also, Bill should exclude non-intersection related accidents altogether, like collisions with cars emerging from parking spaces or mid-block pedestrian accidents. There is still one other complicating factor for Bill to think about. The total number of accidents may not drop. It may even increase. But accident severity may drop enough to easily justify the improvement. For example, after adding a median barrier, more vehicles might collide with the barrier. However, fatal and injury accidents might drop because vehicles no longer cross over the median and have head-on collisions. In cases like this, safety benefits based on accident severity determine whether or not the barrier was cost-effective. So, before and after accident data must include the severity of accidents as well as total numbers of accidents. Bill will also have to get before and after traffic data for each improvement area. If it is not available, he can estimate it using these formulas. V sub B is the estimated traffic volume for the year prior to improvement in vehicles per day. V sub A is the estimated traffic volume for the year after improvement also in vehicles per day. V sub X is traffic volume for year X based on counts and adjusted to get average annual daily traffic in vehicles per day. R is the annual growth rate in percent for traffic on the road being studied. It may be positive, zero or negative. And N is the number of years from year X to the analysis year. The next step is to analyze the accident data and adjusted for overall traffic growth. Here is the formula that will work for both before and after data. Double A sub B is the adjusted annual number of accidents before improvement in number of accidents. A sub B is the annual number of accidents before the improvement in number of accidents. V sub B is the estimated traffic volume for the year prior to improvement in vehicles per day. And V sub A is the estimated traffic volume for the year after improvement in vehicles per day. Double A sub B is the number of accidents that would have been observed in the period after the improvement if the improvement had not been made. Now the before and after accident data can be evaluated strictly on the basis of the influence of the improvement. Now Bill can calculate the percent reduction in accidents due to the improvement. Pc is the percent change in number of accidents due to improvement. Double A sub B is the adjusted annual number of accidents before improvement in number of accidents. And A sub A is the annual number of accidents after improvement in number of accidents. When the percent change is negative, the total number of accidents following the safety improvement has dropped. A positive number means it increased. He can use the same formula for the adjusted annual number of fatal, injury, or property damage accidents before the safety improvement. Bill now knows how to figure the percent of change in number of accidents at a location. But how significant is the change? Is it only chance? He can find out with another simple equation. It is based on the theory of Poisson distribution. Pc sub C is the critical percent change in number of accidents. Double A sub B is the adjusted annual number of accidents before improvement in number of accidents. K is the standard normal variable on the x-axis of the standard normal distribution curve. Let's look at K more closely. K is related to the probability, P, that the critical percent change, Pc sub C, will not be exceeded due to chance. A 95% confidence level, or K equals 1.645, is recommended for most evaluations. For small numbers of accidents, say less than 20, a higher confidence level may be warranted. This job is easier with a standard chart. All Bill has to do is plop the point represented by the adjusted annual number of accidents before improvement, double A sub B, and the actual percent change in number of accidents, Pc. In this case, the point fell below the lower curve. So the change is significant at the specified confidence level, and the safety improvement has been a success. If Pc had fallen between the two curves, that would mean the change was probably due to chance and not to the safety improvement. If it had fallen above the upper curve, the situation was probably made worse by the safety improvement. The significance test is good for fatal, injury, and property damage accidents as well. The next question Bill has to ask is, was the improvement worth doing from an economic viewpoint, or should the agency have spent the money on some other safety improvement? He can answer that question by calculating the average annual economic loss at the location before and after the safety improvement. There is a formula for each. EL sub B is the annual economic loss before the improvement in monetary units. EL sub A is the annual economic loss after the improvement in monetary units. Double A sub B is the adjusted annual number of accidents before. A sub A is the annual number of accidents after. C is the societal cost of each accident. F is the subscript indicating fatal accidents. I is the subscript indicating injury accidents. And P is the subscript indicating property damage accidents. He can also calculate the equivalent uniform annual benefit, EUAB. He can also calculate the net annual benefit, NAB, and the benefit cost ratio, B over C. The greater the net annual benefit, the more desirable the improvement. A positive NAB means society will experience a monetary net benefit from the improvement. If the NAB is zero or negative, then the money will be better spent on a different type of improvement or location. The benefit cost ratio is an indicator of the cost effectiveness of the project. The benefit is related to the reduction in accidents. A ratio greater than 1.0 indicates the project was cost effective. On projects involving small numbers of accidents, the ratio may be very large, but the net annual benefit value may be relatively small. The benefit cost ratio is a better indicator of the effectiveness of various types of improvements, while the net annual benefit value is more useful for evaluating the overall benefit of the safety program. Bill should record the NAB and B over C values for each safety project, even if they indicate the project was not successful. Safety analysts will use these records to learn how different types of improvements perform. Now Bill is ready for the last phase, phase six, evaluating the safety program. The first step is to evaluate categories of improvements. Classification codes can make this job easier. For example, channelization of turning lanes is code 10, and site distance improvement is 13. When the improvements are all in categories, Bill must aggregate the annual accident data, benefits, and costs. For example, he should treat all pavement marking projects as one category. However, if a project involves a combination of pavement marking and signing, he should not combine it with the pavement marking only projects. He should create a separate category indicating the specific combination of improvements. Also, he should not combine projects at highly dissimilar locations. In other words, signing and marking projects at railroad crossings should be separated from signing projects at other locations. Now for each project in each category, Bill must calculate the overall percent change in accidents, total benefits, total costs, total net benefits, and overall benefit cost ratio. Here are the formulas. PC sub O is the overall average percent change in number of accidents for all projects in the improvement category. Sum AA sub B is the summation of the adjusted annual number of accidents before improvements for all projects in the improvement category. Sum A sub A is the summation of the annual number of accidents after improvement for all projects in the improvement category. EUAB Sub-O is the overall Equivalent Uniform Annual Benefit for the improvement category. EUAC Sub-O is the overall Equivalent Uniform Annual Cost for the improvement category. Some EUAB is the summation of Equivalent Uniform Annual Benefits for all projects in the improvement category. And some EUAC is the summation of Equivalent Uniform Annual Costs for all projects in the improvement category. Now, Bill can calculate the overall performance of a specific category of improvements, the overall net annual benefit, NAB Sub-O, and the overall benefit cost ratio, B over C Sub-O, for all projects. The greater the net annual benefit, the more desirable the improvement category. A positive NAB Sub-O means society will generally experience a monetary net benefit from this type of improvement. If the NAB Sub-O is zero or negative, then money would be better spent on a different type of improvement. The overall B over C Sub-O ratio is an indicator of the cost effectiveness of the improvement category because the benefit is related to the reduction in accidents. A B over C Sub-O ratio greater than 1.0 indicates this category of improvement is cost effective. Once he has summaries for each category, Bill can aggregate them to obtain the combined overall result of all projects in the safety program each year. There are forms for this also. In future years, he can combine annual worksheets. Over a period of years, the agency will develop valuable data on most safety improvement categories. The agency must be sure to update costs and benefits from previous years with appropriate inflation factors. This entire process is complete when Bill writes an annual evaluation report on the Traffic Safety Program. The report should describe program activities during the year and summarize the results of program evaluations. It should include worksheets with categories for the year and with cumulative totals for the last five years. The report should highlight the overall accident reduction and the net annual benefits of the program or suggest actions to improve performance in the future. And perhaps most important, this report should circulate to safety program managers and top management. It will help to focus attention on traffic safety program management in the future. In this program and the previous two of the series, traffic engineer Bill Granke has worked his way through the six phases of the Highway Safety Improvement Program process. For more information on this or other IRF videotapes, write to the International Road Federation or call the numbers on your screen.