National Academies Press: OpenBook

Accident Modification Factors for Traffic Engineering and ITS Improvements (2008)

Chapter: Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data

« Previous: Chapter 2- Status of Existing AMFs and Identification of AMF Needs
Page 18
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 18
Page 19
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 19
Page 20
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 20
Page 21
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 21
Page 22
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 22
Page 23
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 23
Page 24
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 25
Page 26
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 26
Page 27
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 27
Page 28
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 28
Page 29
Suggested Citation:"Chapter 3 - Development of New AMFs through Analysis or Reanalysis of Crash Data." National Academies of Sciences, Engineering, and Medicine. 2008. Accident Modification Factors for Traffic Engineering and ITS Improvements. Washington, DC: The National Academies Press. doi: 10.17226/13899.
×
Page 29

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

18 This chapter provides a summary of each of the data analysis efforts to produce or enhance AMFs. The narrative includes a description of the treatment studied, the data used, the statisti- cal methodology, and the results of each evaluation. More detailed descriptions of these efforts are included in appendices to this report. Introduction Much of the Phase II effort involved new analysis or reanaly- sis of crash and other safety data for the treatments identified by the oversight panel and project team. Individual summaries are presented for the following four treatments: • Installation of a traffic signal at a rural intersection (new EB evaluation); • Conversion of undivided four-lane road to three lanes including a two-way left-turn lane—a “road diet” (reanaly- sis of data from two previous studies); • Increasing pavement friction on intersection approaches (reanalysis of previous study data); and • Increasing pavement friction on roadway segments (reanaly- sis of previous study data). The EB analyses of the following four urban signalized- intersection treatments will be described in one section, since the same reference group was used for all four: • Modification of left-turn signal phase (3 combinations), • Replacement of 8-in. signal head with 12-in. head, • Replacement of single red signal head with two signal heads, and • Replacement of nighttime flashing operation with regular signal phasing. Also, as noted in the Chapter 4 discussion, the analysis- driven expert panels recommended additional analysis for several AMFs. Many of the recommendations were based on a consensus that prior research was valid and applicable to the roadway class in question and thus that the AMFs from the research were appropriate. Other recommendations required additional analysis to be conducted by one or more of the NCHRP project teams. Two of those analysis efforts were undertaken in this project. The first analysis effort focused on travel speed. While travel speed is known to be a critical factor in both crash frequency and crash severity, limited effort has been focused on producing a relationship that would allow prediction of the crash-related effects of reducing average speeds by a given amount. If such a relationship could be developed, it could be used as an AMF for a wide spectrum of treatments with known effects on average speeds. For example, if the effects on travel speed of changing a speed limit or installing a neigh- borhood traffic calming device could be estimated, these effects could be converted into expected changes in crashes. Based on the panel recommendation, a reanalysis of data from a prior research study was conducted to confirm or modify the results and to determine if the findings, which were based on the many non-U.S. studies, were applicable to U.S. roadways. A summary of this analysis is provided in this chapter. The second analysis effort was conducted by the suburban/ urban panel and involved examining existing studies of the effects of median width on crashes. The panel was unable to come to a consensus on an AMF. At the panel’s recommen- dation, the project team conducted additional analysis of this issue using HSIS data from California. Note that because me- dian widths are not normally changed without changes in other critical roadway components (e.g., changes in number of lanes and/or shoulder width), a traditional EB before-after analysis was not possible. Instead, as described later in this chapter, cross-sectional regression analyses were conducted. Each of the summaries will provide information on the treatment, the methodology, the data used, and the results of C H A P T E R 3 Development of New AMFs through Analysis or Reanalysis of Crash Data

the analyses. Because all but two of the analyses conducted in this project utilized EB analyses, the following section pro- vides a brief description of that general methodology. A more comprehensive description is provided in Appendix B. Overview of the Empirical Bayes (EB) Methodology The general analysis methodology used in many of the following evaluations was the empirical Bayes (EB) before- after analysis as described by Hauer (7), which has become the standard of practice in recent years. This methodology does the following: • Properly accounts for regression to the mean, • Overcomes the difficulties of using crash rates in normal- izing for volume differences between the before-treatment and after-treatment periods, • Reduces the level of uncertainty in the estimates of safety effect, • Provides a foundation for developing guidelines for estimating the likely safety consequences of the contem- plated implementation of the evaluated treatment, and • Properly accounts for differences in crash experience and reporting practice in amalgamating data and results from diverse jurisdictions. To accomplish this, the EB analysis requires before- treatment and after-treatment crash and AADT data on the treatment sites and on a reference group of similar untreated sites. The similarity of untreated sites is determined on the basis of the geometrics of the sites (e.g., rural, four-leg, stop- controlled intersections and non-intersection locations on urban, undivided, four-lane, non-freeways), on similar AADT ranges, and on crash history in the before-treatment period. Using the reference group, a safety performance function (SPF) is developed—a regression equation that predicts an outcome variable (e.g., total crashes per year or injury right- angle crashes per year) based on either AADT only or on AADT plus other site descriptors (e.g., lane width or presence of a left-turn lane). (In the studies conducted for this project, generalized linear modeling was used to estimate model coef- ficients in the regression equations using the SAS software package and assuming a negative binomial error distribution, practices consistent with the state of research in developing these models.) A “time trend factor” based on the reference group data was also developed for each year in the before- treatment and after-treatment periods and is typically re- flected as an SPF multiplier. This multiplier is used to account for annual effects due to variations in weather, demography, crash reporting, and so forth, across the study period. The SPF outputs are combined with the observed crashes in the before- treatment period to produce an estimate of before-treatment crash frequency that is adjusted for regression to the mean. Using the SPF to control for AADT growth and the time trend factor, this adjusted before-treatment-period estimate is then projected to predict what would happen in each year of the after-treatment period if the treatment had not been imple- mented. The sum of these predictions (i.e., what would hap- pen without treatment in the entire after-treatment period) is compared to the observed crashes during that period (with the treatment implemented) to produce an estimate of the effect of the treatment. The procedure also calculates the standard deviation of this effect estimate, which makes it possible to determine if the measured effect is statistically significant or not at a specified level of significance. Installation of a Rural Traffic Signal Description of Treatment and Crash Types of Interest This analysis examined the safety impacts of converting rural intersections from stop-controlled operation to signal- controlled. The basic objective was to estimate the change in crashes. Target crash types considered included the following: • All crash types, • Right-angle (side-impact) crashes, • Left-turn-opposing (one-vehicle-oncoming) crashes, and • Rear-end crashes. The change in crash frequency was analyzed as well as the changes in overall economic costs, recognizing that different crash types and severity levels have different economic costs. Appendix B provides the details associated with this evaluation. Data Used Geometric, traffic-volume, and accident data for treatment and reference sites were acquired from HSIS for the states of California (1993–2002) and Minnesota (1991–2002) to facil- itate the analysis. In addition, the Iowa DOT provided a dataset of high-speed rural intersections in Iowa that were converted from stop-controlled to signal-controlled. The Iowa DOT also provided a reference group of similar sites. Because these data were limited, the time trend necessary to conduct an EB analysis could not be developed, but the results from a cursory analysis of the Iowa data were used to reaffirm the results from the analysis of California and Minnesota data. Methodology The general analysis methodology used was the EB before- after analysis, as previously described. The evaluation not 19

only included analysis of the effects of the treatment on crash frequencies for different accident types and severities indi- vidually; it also included analysis of the effects on the overall economic cost (or “crash harm”) of crashes before and after the treatment. Since different crash types are characterized by different severities (e.g., rear-end crashes are often less severe than angle crashes), the economic cost of a crash can be as- signed based on crash type and severity using type/severity economic costs from a recent FHWA study (72). Then, using an EB method that parallels the method described above for crash frequencies, the overall estimate of treatment effect when all crash types and severities are combined can be cal- culated. A much more detailed statistical description of the EB method for both crash frequency and economic costs is found in Appendix B. The analyses attempted to develop AMFs for two crash- severity levels (i.e., injury versus no-injury) within each of the four crash types (i.e., total, right-angle, left-turn-opposing, and rear-end) within three types of stop-controlled (before- treatment) intersections. This was not possible in both states. The intersection types and the treatment and reference group sample sizes are noted in Table 5. Iowa data included a total of 19 treatment sites and 59 reference sites (three- and four-leg combined). In addition to providing data on these stop-controlled sites, California and Minnesota provided data on other intersec- tions that were signalized throughout the entire before- treatment and after-treatment period (63 four-legged sites from California, 21 four-leg sites from Minnesota). These data were used in two ways. First, the signalized intersection datasets were used to verify that the treated intersections in the after-treatment period performed similarly to intersections that were already signalized during the after-treatment period. In general, the treated intersections did perform similarly, in- dicating that the treated sites were not an unusual group of intersections. This similar performance helps validate the cal- culated crash-related AMF. Second, the data on intersections that were signalized throughout the entire before-treatment and after-treatment period were used to develop a more de- tailed procedure for assessing whether a contemplated signal installation is warranted at a given location. In one sense, the initial EB analysis produces an AMF for each target crash type, but the AMF does not vary with AADT. The procedure provides a type of “AMF function,” allowing the engineer to input specific before- and after-conversion AADTs to estimate the expected effects. This procedure is documented in Appendix B. Results The initial EB evaluation of crash frequency by crash types indicated a significant effect of signal installation on angle, left- turn and rear-end crashes. Table 6 shows the individual and combined results of the California and Minnesota analyses— the AMFs and their standard deviation. As shown in Table 6, the Iowa results are similar to those from the other two states, providing some validation for those results. The best overall estimate of effect is shown in the bottom row where the California and Minnesota results are combined. There, for all intersection types combined, sig- nal installation is expected to reduce total crashes by 44 per- cent, right-angle crashes by 77 percent, left-turn crashes by 60 percent, and increase rear-end crashes by 58 percent. The analyses conducted indicated that while there were slight dif- ferences in effects for the different intersection types and crash severities, these differences were not statistically signif- icant. Thus, the overall AMFs shown above are appropriate for all rural site types. However, the significant increase in rear-end crashes raises questions concerning how much of the angle and left-turn savings are negated by this rear-end increase. An examination of the economic costs of the changes based on an aggregation 20 StateIntersection Type California Minnesota Three leg -- 2 treatment 522 reference Three leg, two lanes on major road 4 treatment 1,405 reference -- Four leg -- 15 treatment 736 reference Four leg, two lanes on major road 14 treatment 742 reference -- Four leg, four lanes on major road 10 treatment 183 reference -- -- = no sites. Table 5. Number of sites for treatment and reference groups. State Total Crashes Right-Angle (RA) Left-Turn (LT) Rear-End (RE) California 0.778 (0.061) 0.221 (0.036) 0.433 (0.065) 2.474 (0.373) Minnesota 0.488 (0.027) 0.228 (0.019) 0.374 (0.063) 1.300 (0.141) Iowa 0.950 (0.085) 0.265 (0.053) n/a 2.075 (0.323) California + Minnesota 0.559 (0.025) 0.227 (0.017) 0.401 (0.047) 1.579 (0.142) Table 6. Crash frequency AMFs (and standard deviations) by crash type for rural signalization.

of rear-end, right-angle, and “other” crash costs for various severity levels, which is intended to cast light on this issue, is the subject of the economic analysis. Table 7 shows the results of that analysis. The results indicated that when crash types are combined using the costs of the different crash types, the rear-end increase does not negate the effect on angle and left-turn crashes. There are significant reductions in crash costs, with the top row in the table indicating that total costs are reduced by 73 percent (100[1-0.265]). The results also provide guid- ance on the intersection conditions under which signal instal- lation would be most beneficial. While there is little difference in the effects on three-leg sites versus four-leg sites and on sites with two lanes versus sites with four lanes on the major road, the results indicate that the benefits are greater on higher vol- ume intersections and are greater where the ratio of expected right-angle crashes to rear-end crashes is higher. Conversion of an Undivided Four-Lane Road to Three Lanes and a Two-Way Left-Turn Lane—a “Road Diet” Description of Treatment and Crash Types of Interest This analysis examined the safety impacts of converting four-lane roadways to three-lane roadways where the center lane is now a two-way left-turn lane. The site locations are in urbanized areas. The basic objective was to estimate the change in total crashes. A secondary objective was to use EB methodology to compare the results with results from a previous study that used full Bayes modeling and time series data for a group of treated and comparison sites matched on a one-to-one basis (44). Appendix C provides the details associated with this evaluation. Data Used Data were acquired from two sources. Geometric, traffic- volume, and accident data for 15 Iowa treatment sites used in a previous full Bayes study and data for an additional 296 ref- erence sites were provided by the Iowa DOT (44). The same types of data for 30 treatment sites and 51 reference sites in the cities of Bellevue and Seattle in Washington state and Mountain View, Oakland, Sacramento, San Francisco, San Leandro, and Sunnyvale in California were acquired from HSIS, whose staff had conducted the original evaluation of these installations (43). In the HSIS study, comparison sites were matched with treatment sites that were similar in terms of functional class, type of development, speed limit, inter- section spacing, and access control. Although the sites were located in different states, all sites were four-lane to three-lane conversions. Methodology The general analysis methodology used was the EB before- after analysis, as previously described. Separate AMF esti- mates were produced for the Iowa and the HSIS data, and the results were then aggregated to develop a combined AMF. While the Iowa data allowed the development of the annual factors to account for the time trend in the EB analysis, the HSIS data did not. The HSIS reference group was much smaller than the Iowa group (51 reference sites versus 296 sites), but that HSIS reference group had been chosen by local traffic engineers to be similar to the treated sites. Results The EB evaluation of total crash frequency indicated that the road-diet treatment had a significant effect in both datasets and when the results were combined. Table 8 shows the results from each of the two studies and the combined results—the AMFs and their standard error. As can be seen, the measured effects from the two databases differ markedly. The Iowa data indicate a 47-percent reduction in total crashes while the HSIS (California and Washington) data indicate a 19-percent decrease. The aggregated estimate is a 29-percent decrease in total crashes. The difference may be a function of traffic volumes and characteristics of the urban 21 Characteristic Cost AMF Total crashes 0.265 (0.001) California 0.315 (0.002) Minnesota 0.247 (0.001) Three leg 0.286 (0.004) Four leg 0.264 (0.001) Two lanes on major 0.265 (0.002) Four lanes on major 0.265 (0.001) AADT < 20,000 0.314 (0.003) AADT > 20,000 0.253 (0.001) Expected RA/Expected RE 4.5* 0.324 (0.002) Expected RA/Expected RE >4.5* 0.215 (0.001) * EB estimates for right-angle and rear-end when under stop-controlled conditions Table 7. Economic cost of AMFs (and standard deviations) for different before-treatment-period conditions, based on combined results from California and Minnesota. Dataset AMF Standard error Iowa 0.534 0.020 HSIS 0.811 0.025 All 0.707 0.016 Table 8. Crash frequency AMFs (and standard errors) for road diets.

environments where the road diets were implemented. The sites in Iowa ranged in AADT from 3,718 to 13,908 and were predominately on U.S. or state routes in small urban towns with an average population of 17,000. The sites in Washington and California ranged in AADT from 6,194 to 26,376 and were predominately on corridors in suburban environments that surrounded larger cities, with an average population of 269,000. In addition, in Iowa there appeared to be a calming effect as evidenced in a study (44) of one site that revealed a 4 to 5 mph reduction in 85th-percentile free flow speed and a 30-percent reduction in percentage of vehicles traveling more than 5 mph over the speed limit (i.e., vehicles traveling 35 mph or higher). The researchers’ speculation is that this calming effect would be less likely in the larger cities in the HSIS study, where the approaching speed limits (and traffic speeds) might have been lower to start with. The “new” Iowa results also seem to be incompatible with those in the earlier Iowa analysis of the same treatment site data (44). However, the 25-percent reduction reported in that study was based on average effects per mile derived by com- paring average crashes per mile after treatment with expected average crashes per mile without treatment. These results are not comparable to the “new” results since sites of different lengths were weighted equally. (The “new” results are overall effects that provide more weight to sites of longer length.) In addition, the “new” results use a much larger comparison group than the previous study, which used an equal number of treatment and comparison sites. Increasing Pavement Friction on Roadway Segments and at Intersection Approaches Description of Treatment and Crash Types of Interest This analysis examined the safety impacts of improving pavement skid resistance using data from the state of New York. The New York State DOT has implemented a skid acci- dent reduction program (SKARP), which identifies sections of pavement with a high proportion of wet-road accidents, per- forms friction tests on these locations, and treats those with both a high proportion of wet-road accidents and low friction numbers (below the Programmatic Design Target Friction Number, FN40R, of 32). The treatment generally involves a 1.5-in. resurfacing or a 0.5-in. microsurfacing using non- carbonate aggregates. This treatment is applied principally on the major road approaches at intersections, but is often extended some distance away from the intersection as well. The goal of this analysis was to develop separate AMFs for different crash types occurring at seven different intersection types (i.e., all intersections combined; three-leg signalized, stop-controlled and yield-controlled; and four-leg signalized, stop-controlled and yield-controlled) and on five types of roadway segments (i.e., all segments combined, rural two-lane segments, rural multilane segments; urban two-lane segments, and urban multilane segments). Appendix D provides the details associated with this evaluation. The target crash types of interest in the intersection analyses included the following: • Total, • Wet road, • Dry road, • Rear end, • Rear end wet, • Right angle, and • Right-angle wet-road. The target crash types considered for segments included: • Total, • Wet road, • Dry road, • Rear end, • Rear-end wet-road, • Rear-end dry-road, • Single vehicle, and • Single-vehicle wet-road. Data Used The data for this study were provided by the New York State DOT and included crash, geometric, and AADT data for treated and untreated intersections and segments during the period of 1994 to 2003. Data were included for 256 treated in- tersections and 3,993 untreated reference intersections, as well as for 36.3 miles (118 segments) of treated non-intersection locations and 1,242.4 miles (2,108 segments) of untreated reference locations. Methodology The general analysis methodology used was the EB before- after analysis, as previously described. SPFs and annual correc- tion factors were successfully developed for each of the site type/crash type combinations noted above. Results Intersection Treatments Estimates of the AMFs for the crash frequency analyses for intersection skid-reduction treatments are given in Table 9. Results that are statistically significant at the 95-percent level are in shown in boldface type. 22

The results show statistically significant reductions at almost all types of intersections in total crashes; wet-road; rear-end; and rear-end, wet-road crashes. As expected, the largest effects were on total wet-road crashes (i.e., 40-percent to 78-percent reductions) and rear-end, wet-road crashes (i.e., 52-percent to 78-percent reductions). There was very lit- tle effect on wet-road, right-angle crashes. Overall, dry road crashes showed a statistically significant 14-percent increase. However, this did not negate the effects on wet-road crashes, as shown by the statistically significant 20-percent decrease in total crashes when all intersection and crash types were com- bined. To see if the principal benefits of improved skid resist- ance on wet-road crashes declined over time, the effect on wet-road accidents was analyzed by year after treatment. The analysis indicated no discernable decreasing trend over the 6 years of after-treatment-period data. Segment Treatments Estimates of the AMFs for the crash frequency analyses for segment-based skid-resistance treatments are given in Table 10. Results that are statistically significant at the 95-percent level are in bold. In general, the results show statistically significant reduc- tions in total crashes and in wet-road; rear-end; rear-end, wet- road; single-vehicle; and single-vehicle, wet-road crashes for most roadway categories. The only exception was for two-lane rural roads, where no significant decreases or increases in frequency were found. As expected, the largest statistically sig- nificant effects were on total wet-road crashes (i.e., 46-percent to 74-percent reductions), on wet-road, rear-end crashes (i.e., 36-percent to 66-percent reductions) and on wet-road, single-vehicle crashes (i.e., 38-percent to 71-percent reduc- tions). The only statistically significant increase found was for dry-road crashes on urban multilane roads (i.e., a 13-percent increase). However, that increase did not negate the overall treatment effect in that there was a 14-percent reduction in total crashes (i.e., dry plus wet) on these roads. A final analysis examined changes in the overall propor- tions of wet-road crashes before and after the treatments. It found a statistically significant reduction in the proportion of wet-road accidents at intersection locations (i.e., 40 percent 23 Intersection Type Total crashe s (s.e.) Wet-road (s.e.) Rear-end (s.e.) Dry (s.e.) Rear-end wet (s.e.) Right- angle (s.e.) Right- angle wet (s.e.) All 0.799 (0.028) 0.426 (0.030) 0.582 (0.034) 1.149 (0.051) 0.322 (0.041) 1.045 (0.078) 0.799 (0.123) Three-leg signalized 0.667 (0.050) 0.372 (0.053) 0.554 (0.065) 0.959 (0.093) 0.261 (0.066) 0.787 (0.125) 0.470 (0.161) Three-leg stop- controlled 0.819 (0.048) 0.355 (0.046) 0.586 (0.057) 1.302 (0.095) 0.335 (0.075) 0.828 (0.218) 0.828 (0.218) Three-leg yield- controlled 0.590 (0.114) 0.217 (0.103) 0.304 (0.086) 1.392 (0.321) 0.221 (0.161) n/a n/a Four-leg signalized 0.797 (0.052) 0.546 (0.070) 0.585 (0.068) 0.992 (0.081) 0.361 (0.084) 0.898 (0.117) 1.105 (0.294) Four-leg stop- controlled 1.271 (0.143) 0.597 (0.137) 0.943 (0.188) 1.754 (0.242) 0.482 (0.215) 1.687 (0.323) 0.829 (0.351) Four-leg yield- controlled 0.589 (0.216) 0.361 (0.371) 0.504 (0.248) 0.651 (0.273) n/a n/a n/a Table 9. Crash frequency AMFs (and standard error) by crash type for intersection skid-reduction treatments. Segment Type Total Crashes (s.e.) Wet-road (s.e.) Rear-end (s.e.) Dry (s.e.) Rear- end wet- road (s.e.) Rear- end dry- road (s.e.) Single- vehicle (s.e.) Single- vehicle wet- road (s.e.) All 0.764(0.023) 0.434 (0.024) 0.828 (0.043) 1.003 (0.043) 0.575 (0.055) 0.977 (0.068)) 0.698 (0.040) 0.399 (0.039) Rural 2 lanes 0.964 (0.073) 0.852 (0.126) 1.047 (0.149) 1.167 (0.114) 0.971 (0.256) 1.235 (0.219) 1.078 (0.141) 1.125 (0.287) Rural >2 lanes 0.684 (0.032) 0.346 (0.028) 0.776 (0.068) 0.875 (0.061) 0.474 (0.079) 0.838 (0.098) 0.588 (0.046) 0.292 (0.038) Urban 2 lanes 0.599 (0.082) 0.260 (0.066) 0.612 (0.142) 0.992 (0.195) 0.344 (0.145) 0.695 (0.216) 0.921 (0.232) 0.523 (0.247) Urban > 2 lanes 0.862 (0.038) 0.538 (0.045) 0.866 (0.059) 1.132 (0.065) 0.640 (0.084) 1.120 (0.099) 0.800 (0.083) 0.615 (0.115) Table 10. Crash frequency AMFs (and standard errors) by crash type for segment skid-reduction treatments.

before treatment versus 16 percent after treatment) and seg- ment locations (i.e., 38 percent before treatment versus 16 percent after treatment). Signalized Intersection Treatments in Urban Areas Description of Treatment and Crash Types of Interest This analysis examined the safety impacts of four urban safety treatments implemented at signalized intersections in Winston-Salem, North Carolina. The treatments were the following: • Modification of left-turn signal phase (three combinations), • Conversion of nighttime flashing operation to steady operation, • Replacement of 8-in. signal heads with 12-in. heads, and • Replacement of single red signal head with dual red signal heads. The basic objective was to estimate the change in target crashes for each of the treatments. Target crashes, which dif- fered depending on the treatment, included left-turn crashes, nighttime angle crashes, and right-angle crashes. However, since the treatment might increase other types of crashes (e.g., the conversion back to regular nighttime phasing could increase rear-end crashes on the major road), additional crash types and total crashes were examined. The specific crash types for each treatment are presented below. Appendix E provides the details associated with this evaluation. Data Used Unlike many other jurisdictions, the City of Winston- Salem has documented the installation records for a large number of urban safety treatments implemented at intersec- tion and non-intersection locations and has systematically conducted simple before-after studies of those treatments. Documentation exists for over 70 individual treatments or combinations of treatments installed since the 1980s, along with target and total crash counts for before-treatment and after-treatment periods of 3 to 5 years. The City of Winston- Salem provided these files to the research team. The team then chose the four treatments for evaluation based on the following: • A statistical analysis of available crash sample size to ensure the possibility of statistically significant results; • The timeliness of the treatments to ensure a reasonable current driver and vehicle population (i.e., treatments in 1994 and later); • The quality of the existing AMF, based on the information described in Chapter 2 of this report; and • The availability of a group of similar but untreated inter- sections to be used as a reference group. The original Winston-Salem documentation did not in- clude information on before-treatment-period and after- treatment-period AADTs. Research team members traveled to Winston-Salem and extracted the AADTs for each treat- ment intersection approach for each before-treatment and after-treatment year from AADT books available at the City of Winston-Salem DOT. Winston-Salem does not have a computerized intersection inventory that is linkable to crash records. Thus, they could not provide the research team with data to be used in the development of a reference group of similar untreated sites. After consideration of other alternatives, a reference group was manually developed. The Winston-Salem Traffic Engineer and his staff identified 75 untreated signalized intersections that were similar to the treated sites in terms of traffic volume, num- ber of legs, number of approach lanes, and other characteristics during the study period (1990 to 2004). Crash data for all crashes in Winston-Salem for the full study period were ex- tracted from the North Carolina DOT crash files, which contain data on all crashes statewide, and these data were manually matched to the reference intersections based on street names found on the crash reports. AADT data for each year in the full period were then extracted for each reference intersection from the Winston-Salem AADT books noted above. After eliminat- ing intersections lacking in AADT or other data, 60 untreated intersections were available for developing the required SPFs. Methodology The general analysis methodology used was the EB before- after analysis previously described. SPFs and annual correction factors were successfully developed for total crashes. Efforts at es- timating SPFs for specific crash types were not successful. Hence, the proportion of crashes for each crash type and a recalibrated over-dispersion parameter were used in the EB analysis. Results Table 11 presents the results of the EB analyses for each of the four treatment types. For each treatment, results are presented for both the primary target crashes and for other important crash types. Statistically significant results are indicated by asterisks. Modification of Left-Turn Phase Three types of left-turn phasing treatments were identi- fied. In all cases, the target crashes for these treatments were 24

identified as those involving at least one left-turning vehicle on the treated roadway. The first treatment involved replac- ing a permissive left-turn phase with a permissive/protected phase at three sites. There was little change in either the target or total crashes, but the small sample size suggests cau- tion in concluding “no effect.” The second treatment involved replacing a permissive left- turn phase with a fully protected phase at eight sites. Here, the target left-turn crashes were reduced by approximately 98 percent, a statistically significant reduction, and total crashes changed very little. The third treatment type involved replacing a permis- sive/protected phase with a fully protected phase at four sites. Here, the left turn crashes were eliminated while the total crashes changed little. Since the results of the second and third treatments refer to conversions to a fully protected phase and since the results were similar, they were combined into one group to increase the sample of sites to 12. As shown, the combined AMF was 0.014 for the left-turn crashes (a statistically significant result), and the total crashes again were virtually unchanged. Since the left-turn crashes decreased substantially and total crashes did not, it is evident that there must have been an increase in non-left-turn crashes of the same order as the decrease in left-turn crashes. Unfortunately, these data did not allow the research team to examine changes in other spe- cific crash types. Further research is necessary to determine the specific reasons for the effect on non-left-turn crashes. However, it seems reasonable to speculate that introducing a protected left-turn phase tended to increase rear-end crashes more than others because of the increased number of phases (and therefore dilemma zone opportunities) and the increase in queues that can result from the reduced green time avail- able for all traffic not protected by the introduced phase. If this is the case, the implication is that this is still a very safety-effective measure from a total-harm perspective since left-turn crashes tend to be of the side-impact variety and therefore are more severe than rear-end crashes. The results also imply that the treatment would be most effective overall where there is a relatively high frequency of left-turn crashes. Conversion of Nighttime Flashing Operation to Steady Operation There were 12 intersections where nighttime (9 p.m. to 6 a.m.) flashing operation was replaced with regular phasing. The EB analysis indicated that nighttime angle crashes (the ones most likely to be positively affected) were reduced by approximately 34 percent, a statistically significant reduction at the 0.10 level of significance. Total nighttime crashes also saw a significant reduction of approximately 35 percent Replacement of 8-in. Signal Heads with 12-in. Heads There were 26 intersections where 8-in. signal heads were changed to 12-in. heads. The EB analysis indicates that right- angle crashes experienced a statistically significant decrease of approximately 42 percent. Total crashes experienced virtually no change. This implies an increase in non-angle crashes of approximately the same size as the decrease in angle crashes. While it is not possible to determine the specific crash types for the non-angle crashes, one could hypothesize that they are predominately rear-end crashes, which might be increased by more drivers stopping rather than proceeding through the signal—the same effect seen for red-light cameras at intersec- tions. Since a “tradeoff” occurred and since the severities of angle and non-angle crashes can differ, an economic analysis 25 Treatment Type No. of Treatment Sites Crash Type AMF(standard error) Replace permissive left-turn phasing with permissive/protected (1) 3 Left-Turn All 0.978 (0.277) 1.045 (0.135) Replace permissive left-turn phasing with protected (2) 8 Left-Turn All 0.021 (0.021)** 0.975 (0.085) Replace permissive/protected left-turn phasing with protected (3) 4 Left-Turn All 0.000 (0.006)** 1.020 (0.123) Replace permissive or permissive/protected with protected (combination of 2 and 3) 12 Left-Turn All 0.014 (0.014)** 0.992 (0.070) Convert nighttime flash to normal phasing (4) 12 Nighttime Angle All Nighttime 0.659 (0.180)* 0.651 (0.145)** Replace 8-in. signal heads with 12-in. heads (5) 26 Right-AngleAll 0.580 (0.070)** 0.970 (0.060) Add second signal head (6) 8 Right-AngleAll 1.050 (0.130) 1.180 (0.110) ** Statistically significant at the 0.05 significance level * Statistically significant at the 0.10 significance level Table 11. Crash frequency AMFs for urban signalized intersection treatments by treatment type.

was conducted in order to accurately estimate the effect of this treatment on overall “crash harm.” Based on a recent report from FHWA (72), the comprehensive cost per crash is $47,333 for angle crashes and $26,735 for rear-end crashes. Assuming that the overall severity of non-angle crashes is quite similar to that of rear-end crashes, the economic analysis revealed a reduction of about $11,800 per intersection-year in the over- all crash harm due to this treatment. Replace Single Red Signal Head with Dual Red Signal Heads A second red signal head was added to the existing head at eight intersections in Winston-Salem. The EB analysis indicates a slight, but not statistically significant, increase in both right- angle and total crashes after installation. Given the limited sam- ple of sites that were evaluated, these results suggest that the installation of double red signal heads does not appear to be an effective strategy for reducing total or angle crashes. Speed Change and Crashes Description of Treatment and Crash Types of Interest The objective of this analysis was to develop an AMF that would relate speed change caused by a treatment to a change in crash frequency for a certain level of crash severity (i.e., fatal, injury, no-injury). Thus, if one could estimate the effect of a treatment on mean speed, the AMF would then translate this change in mean speed to an estimated change in crash fre- quency. With this speed change AMF, if the change in mean speed can be anticipated so can the potential safety effect. If the change in mean speed can be measured, the future safety effect can be estimated without waiting for crashes to materialize. The treatments could theoretically include “passive” treatments such as changes in speed limits or increased speed enforcement (i.e., “passive” in the sense that the driver can choose to react or not react to the treatment) and “active” treatments that could include changes to the roadway such as residential traffic- calming speed tables (i.e., “active” in the sense that the driver is “forced” to reduce speed by the treatment). A study by Nilsson (73) hypothesized a “power model” relating the ratio of before-treatment and after-treatment crash frequency to the ratio of before-treatment and after-treatment mean speed raised to some power, with the power changing for different crash severities. Elvik et al. (74) further developed this power model using a large set of data extracted from pub- lished research reports. The objective of the reanalysis of the Elvik et al. data done in this research was to determine if such a relationship exists, and if so, whether the power model or some alternative model form best describes the relationship be- tween speed changes and crash frequency. Thus, this AMF (or series of AMFs) is related to any treat- ment that is associated with a changed mean speed. Appendix F provides the details associated with this evaluation. Data Used The data used in this reanalysis were supplied by Elvik from his earlier study (74). His research team had extracted data on mean speed change and the related crash-frequency change from 97 published international studies containing 460 results. Each result contained information on mean speed and crash frequency before treatment and mean speed and crash frequency after treatment. Some studies included information on mean speed and crash frequency for a set of comparison locations where no treatment was installed. Each result also included study details such as the type of study (e.g., simple before/after or time series), the country where the study was conducted, the type of location (urban/rural) and the road system. The country data were important since a secondary objective of this reanalysis was to determine if the results from non-U.S. studies could be applied to U.S. roads. The research team then reexamined each of the study results and clarified possible errors through a series of con- versations with Elvik. The research team also calculated revised estimates of standard errors of the crash and speed changes based on the study types, using procedures that will be used in the upcoming Highway Safety Manual. Methodology The final dataset was then used to examine the power model results and to develop alternative model forms. One alternative model form was based on the “maneuver time needed to avoid a crash,” which is a function of initial speed and the distance to an obstacle or vehicle. The other alterna- tive model form was based on the underlying physics of speed versus crash frequency and crash injury. The two alternative model forms were then compared and used to develop a final set of proposed AMFs. Results Like the results of the earlier study (74), both alternative models indicated that the data supported the existence of a rel- atively strong relationship between speed change and change in fatal and non-fatal injury crash frequency. The relationship with property-damage-only (PDO) crashes was not distinct. Both of the alternative model forms indicated that the speed- versus-crash relationship in the foreign studies was similar to that in the U.S. studies. Quality of fit statistics indicated that both model forms were slightly more accurate than the power model. As expected, the two alternative models produced 26

slightly different values for the injury-crash AMF and the fatal-crash AMF. The decision was then made to combine the AMF values from the two models to produce the recom- mended AMFs. Table 12 provides these combined AMFs. To illustrate the use of these AMFs, consider a road on which the mean speed is 60.0 mph. If some measure is ex- pected to increase the mean speed by 2.0 mph, injury acci- dents are expected to increase by a factor of 1.10 and fatal accidents by a factor of 1.18. Thus, what may appear to be a small change in mean speed has a large impact on accidents. It is expected that these AMFs would be usable for treat- ments associated with change in mean speed on freeways and rural highways. Their usefulness for urban-street treatments is less certain. There were some indications in the data that speed change related to passive treatments on urban streets had less effect on crash frequency than is shown in Table 12 (i.e., an AMF closer to 1.0) and that the effect of active speed control on these streets (i.e., road humps, traffic circles, chi- canes, and so forth) may not be fully captured here. However, neither of these indications was found to be fully confirmed in the statistical analysis. In the absence of other knowledge, it is concluded that the tabulated AMFs can be applied to both active and passive treatments on urban streets. How- ever, the user should understand that there is less certainty about the AMFs when they are used for urban streets than when they are used for freeways and rural highways. Effect of Median Width Description of Treatment and Crash Types of Interest Studies on the effect of median width have shown that in- creasing width reduces cross-median crashes, but the amount of reduction varies across studies. The effect of median width on median-related or all crashes is even less clear. The basic objective of the research was to develop AMFs for median width for different types of roads. Methodology The preferred method for developing an AMF is to conduct a before-after study in which the treatment installation/ removal/change date is known, and thus the safety before and after this date can be tracked. The current state-of-the-art methodology for conducting such studies makes use of an EB approach, which helps to account for issues such as regression to the mean, changes in traffic volumes, and changes in crashes over time that are due to other factors (e.g., weather). However, there are a number of treatments in the roadway environment that are not “installed” or changed in a manner that allows for a before-after study. Median width is one such treatment. It is very unlikely that the median width on a highway will ever be changed without making other significant changes to the geo- metric cross-section. For example, the most common change in median width would occur when additional travel lanes are being added to the left-hand side of a roadway, thus narrowing the median. In this case, the fact that there is a significant change other than the change in median width makes it more difficult to isolate the effects of the change in width in an EB before-after evaluation. In this case, a cross-section model that predicts safety on the basis of varying median widths, traffic volumes, and other factors is still the most feasible option for determining the expected safety benefits as median width changes. In this evaluation, negative binomial (NB) regression models were developed with crash frequency as the dependent variable and site characteristics such as traffic volume, shoul- der width, and median width as independent variables. The pa- rameter estimates from the NB models were used to develop AMFs. The analysis focused on total crashes and cross-median 27 Non-fatal Injury Crashes Fatal Injury Crashes 0v (mph) 0v (mph) v (mph) 30 40 50 60 70 80 v (mph) 30 40 50 60 70 80 5 0.57 0.66 0.71 0.75 0.78 0.81 5 0.22 0.36 0.48 0.58 0.67 0.75 4 0.64 0.72 0.77 0.80 0.83 0.85 4 0.36 0.48 0.58 0.66 0.73 0.80 3 0.73 0.79 0.83 0.85 0.87 0.88 3 0.51 0.61 0.68 0.74 0.80 0.85 2 0.81 0.86 0.88 0.90 0.91 0.92 2 0.66 0.73 0.79 0.83 0.86 0.90 1 0.90 0.93 0.94 0.95 0.96 0.96 1 0.83 0.86 0.89 0.91 0.93 0.95 0 1.00 1.00 1.00 1.00 1.00 1.00 0 1.00 1.00 1.00 1.00 1.00 1.00 1 1.10 1.07 1.06 1.05 1.04 1.04 1 1.18 1.14 1.11 1.09 1.07 1.05 2 1.20 1.15 1.12 1.10 1.09 1.08 2 1.38 1.28 1.22 1.18 1.14 1.10 3 1.31 1.22 1.18 1.15 1.13 1.12 3 1.59 1.43 1.34 1.27 1.21 1.16 4 1.43 1.30 1.24 1.20 1.18 1.16 4 1.81 1.59 1.46 1.36 1.28 1.21 5 1.54 1.38 1.30 1.26 1.22 1.20 5 2.04 1.75 1.58 1.46 1.36 1.27 0v = initial mean travel speed v = change in mean travel speed 0v = initial mean travel speed v = change in mean travel speed Table 12. Crash-frequency AMFs for injury and fatal crashes based on initial speed and speed change.

crashes (definitive and probable). Whether a crash was cross- median was deduced based on the location of the crash and the movement preceding the crash. Data Used Ten years of data (1993 to 2002) on divided roadway sections in California were obtained from HSIS. HSIS has a crash file providing detailed information about individual crashes, a roadway file that has data on traffic volume and other site char- acteristics, and an intersection/ramp file that shows the location of intersections and ramps. Data for about 27,131 mile-years of divided roadway sections without median barriers were ex- tracted from HSIS. Sites where the two sides of the roadway were on separate grades were eliminated. To the extent possi- ble, only “traversable” median locations were included in the dataset. A preliminary analysis of the dataset revealed that me- dian widths of 100 ft or larger were coded as 99 ft in the dataset. Hence, all sections with median width coded as 99 ft or larger were removed. Sections with “variable median width” were also removed. In addition, whenever the type of access control changed for a particular year, data were eliminated for that sec- tion for that year. Eliminating these sections resulted in 19,933 mile-years. Table 13 shows the number of mile-years by access control, number of lanes, and type of area (i.e., rural or urban). For roads with partial or no access control and more than four lanes, the number of mile-years was minimal, and hence, this group was not considered for the analysis. Table 14 shows the total number of crashes and cross-median crashes for the different roadway types. Cross-median crashes represent be- tween 3 percent and 6 percent of total crashes on roads with full access control and about 12 percent of total crashes on roads with partial or no access control. Roads with full access control experience relatively fewer cross-median crashes probably because these roads generally have larger median widths. In the sample for this research, the average median width for roads with full access control ranged from 55 to 60 ft, whereas the average median width for roads with partial or no access control ranged from 29 to 40 ft. Full access control roads in rural areas with more than four lanes had relatively few cross-median crashes (i.e., 548), and thus it was not possible to develop satisfactory models for this group. Hence, AMFs were not developed for this group. Results Tables 15 and 16 show the AMFs for median width derived from the NB models for all crashes and cross-median crashes. The AMFs were calculated by using a 10-ft median width as the base case. It is clear that increasing median width is asso- ciated with a reduction in total crashes as well as with cross- median crashes. Here are the findings regarding the AMFs: • As expected, median width has a larger effect on cross- median crashes than on total crashes; • The AMFs for cross-median crashes are very similar for the two urban roadway types with full access control (i.e., with four lanes and five or more lanes); • The AMFs for cross-median crashes are very similar for the two rural roadway types; and • The AMFs for total crashes are very similar for the two four-lane urban roadway types (with full access control and partial or no access control). Overall, the AMFs are quite similar to those obtained from previous studies that were also based on cross-sectional models (48, 49, 75, 76, 77). However, this study used a much larger sample of mile-years and crashes in arriving at the AMFs. Separate AMFs were also developed on the basis of area type (rural or urban), level of access control (full or partial/none), and type of collision (total or cross-median). Hence, the AMFs produced in this effort are those recom- mended in Chapter 5. 28 Area Type Access Control Number of Lanes Rural Urban 4 3,258 1,549Partial or No Access Control 5+ 70 107 4 8,331 3,037Full Access Control 5+ 1,604 1,970 Area Type Rural Urban Level of Access Control No. of lanes Total Cross-Median % Cross- Median Total Cross-Median % Cross- Median Partial or No Access Control 4 13,255 1,593 12.0 28,185 3,438 12.2 4 33,009 1,961 5.9 35,690 1,554 4.4 Full Access Control 5+ 12,624 548 4.3 43,385 1,507 3.5 Table 13. Mile-years by roadway type. Table 14. Number of crashes (total and cross-median) by roadway type.

29 Rural, 4 Lanes, Full Access Control Urban, 4 Lanes, Full Access Control Urban, 5+ Lanes, Full Access Control Median Width (ft) Total Crashes Cross- median Crashes Total Crashes Cross- median Crashes Total Crashes Cross- median Crashes 10 1.00 1.00 1.00 1.00 1.00 1.00 20 0.96 0.86 0.95 0.89 0.93 0.89 30 0.93 0.74 0.90 0.80 0.86 0.79 40 0.90 0.63 0.85 0.71 0.80 0.71 50 0.87 0.54 0.80 0.64 0.74 0.63 60 0.84 0.46 0.76 0.57 0.69 0.56 70 0.81 0.40 0.72 0.51 0.64 0.50 80 0.78 0.34 0.68 0.46 0.59 0.45 90 0.75 0.29 0.65 0.41 0.55 0.40 100 0.73 0.25 0.61 0.36 0.51 0.35 Rural, 4 Lanes, Partial or No Access Control Urban, 4 Lanes, Partial or No Access ControlMedian Width (ft) Total Crashes Cross- median Crashes Total Crashes Cross- median Crashes 10 1.00 1.00 1.00 1.00 20 0.95 0.84 0.95 0.87 30 0.91 0.71 0.90 0.76 40 0.87 0.60 0.85 0.67 50 0.83 0.51 0.81 0.59 60 0.79 0.43 0.77 0.51 70 0.76 0.36 0.73 0.45 80 0.72 0.31 0.69 0.39 90 0.69 0.26 0.65 0.34 100 0.66 0.22 0.62 0.30 Table 15. AMFs for median width for roads with full access control. Table 16. AMFs for median width for roads with partial or no access control.

Next: Chapter 4 - Development of New AMFs through Expert Panels »
Accident Modification Factors for Traffic Engineering and ITS Improvements Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB's National Cooperative Highway Research Program (NCHRP) Report 617: Accident Modification Factors for Traffic Engineering and ITS Improvements explores the development of accident modification factors (AMFs) for traffic engineering and intelligent transportation system improvements. AMFs, also known as crash reduction factors, are designed to provide a simple and quick way of estimating the safety impacts of various types of engineering improvements, encompassing the areas of signing, alignment, channelization, and other traffic engineering solutions.

The following appendices to NCHRP Report 617 are available online:

* Appendix A: Methodology for Determining Crash-Harm Rating for Treatments

* Appendix B: Effects of Converting Rural Intersections from Stop to Signal Control

* Appendix C: Safety Effects of Four-Lane to Three-Lane Conversions

* Appendix D: Safety Effects of Improving Pavement Skid Resistance

* Appendix E: Evaluation of the Safety Effectiveness of Urban Signal Treatments

* Appendix F: An Empirical Examination of the Relationship Between Speed and Road Accidents

* Appendix G: Accident Modification Factors for Median Width

Two AMF treatment summaries that appear in the printed version of NCHRP Report 617 contain incorrect information. These treatments are "Add Intersection Lighting" and "Add Roadway Segment Lighting." The information in these two AMF treatment summaries has been corrected in the online version of the report.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!