5

DEVELOPING AND COMMUNICATING NEW MEASURES OF SAFETY

Consumers would benefit from predictive measures of the overall safety performance of new motor vehicles in comparing among purchase choices. Some consumers will also want an explanation of how such summary measures have been developed and more detailed disaggregated information and feature-by-feature comparisons. In this chapter, a strategy for the production and communication of such information is proposed and elaborated.

ATTRIBUTES OF GOOD SUMMARY MEASURES

Comprehensive measures of vehicle safety performance must meet several requirements. They should

  • Be related meaningfully to actual safety for the range of highway conditions in which the vehicle will be operated;

  • Provide a summary whose use or interpretation does not require extensive manipulation or combination with other information;

  • Be unambiguous and easy to understand and use;

  • Convey the degree of uncertainty associated with current knowledge and expert judgment;

  • Be transparent and flexible, allowing more sophisticated users to understand how summaries are produced and to apply different judgments to obtain their own summaries when that is desired; and

  • Allow the consumer to place the information in context.

The first of these attributes appears obvious. Producing more meaningful measures probably also presents the greatest long-term challenge.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information 5 DEVELOPING AND COMMUNICATING NEW MEASURES OF SAFETY Consumers would benefit from predictive measures of the overall safety performance of new motor vehicles in comparing among purchase choices. Some consumers will also want an explanation of how such summary measures have been developed and more detailed disaggregated information and feature-by-feature comparisons. In this chapter, a strategy for the production and communication of such information is proposed and elaborated. ATTRIBUTES OF GOOD SUMMARY MEASURES Comprehensive measures of vehicle safety performance must meet several requirements. They should Be related meaningfully to actual safety for the range of highway conditions in which the vehicle will be operated; Provide a summary whose use or interpretation does not require extensive manipulation or combination with other information; Be unambiguous and easy to understand and use; Convey the degree of uncertainty associated with current knowledge and expert judgment; Be transparent and flexible, allowing more sophisticated users to understand how summaries are produced and to apply different judgments to obtain their own summaries when that is desired; and Allow the consumer to place the information in context. The first of these attributes appears obvious. Producing more meaningful measures probably also presents the greatest long-term challenge.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information As indicated in Chapter 2, current knowledge can support the production of measures, particularly for crashworthiness, that have some correlation with actual safety performance. The most reliable predictions can probably be achieved through combining crash test results, statistical analysis of real-world crash outcomes, and expert engineering judgment. The resulting predictions will not be perfect, but just as with informed predictions about other uncertain systems, such as the weather, on the average they will be significantly better than guesses. Consumers are unlikely to have knowledge of the relative frequency with which safety features and characteristics affect crash likelihood or outcomes. Nor are they likely to adequately understand the variability among vehicles in such other key factors as weight and size. For these reasons, predictions of probable vehicle performance in specific crash modes must be placed in the broader context of the relative frequencies of crash modes and the variability in the relative performance of different vehicle types in each crash mode. Detailed comparisons of selected vehicle safety features and characteristics can provide consumers with only a portion of the information they need to make reasoned decisions about safety. Providing them with methods for weighting these features and characteristics by their relative importance would burden consumers with an extremely complex computational task. As discussed in Chapter 4, research suggests that consumers are unlikely to carry out this task and prefer that it be done for them. Hence, the second attribute: to be useful to most people, an ideal measure should provide a summary of overall vehicle safety. Whether a measure is “unambiguous and easy to understand,” the next attribute, is an empirical question. Risk communication research suggests that even experts may not be able to determine how best to explain and communicate a summary measure until they have tested several alternatives with typical consumers (NRC 1989; Roth et al. 1990; Morgan et al. 1992; Bostrom et al. 1994a) using read-aloud protocols (Ericsson and Simon 1993), focus groups (Merton 1956; Merton 1987; Morgan 1988; Stewart and Shamdasani 1990), and other appropriate methods. Conducting a series of such tests and iteratively refining the communications strategy is simple but essential. Without such tests one cannot be certain that a measure whose meaning appears obvious to experts carries the same meaning for laypeople. For the foreseeable future all estimates of the safety performance of vehicles will involve considerable uncertainty. As noted in Chapter 2,

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information the variability of crash test results based on a single test per vehicle is an important issue. In addition, the types of crash tests available do not adequately represent the wide range of real-world crashes, and crash dummies do not represent the range of human characteristics. Expert judgments involve additional uncertainty. An accurate impression of current knowledge about vehicle performance and safety cannot be conveyed to consumers unless information about the uncertainty associated with the current state of knowledge is given, the fourth attribute. Such information can be communicated in an easily understood way. The most promising approach is probably some form of graphical representation, such as shaded bars, to indicate a range of values (Ibrekk and Morgan 1987). Two illustrations of how these uncertainties might be portrayed are provided in the next-to-last section of this chapter. The design of any graphics to be used on an actual label should be a matter for empirical investigation. How uncertainty is represented deserves special attention in any such investigation, because the portrayal of uncertainty about a product attribute may reduce its importance in the eyes of the consumer or otherwise alter how it is used in decision making.1 Providing consumers with several separate measures of the safety performance of vehicles is of limited value unless guidance is given on how the measures can be combined into a summary assessment. Indeed, consumers are better off with summary measures that simplify comparisons of safety attributes among vehicles and assist decision making. At the same time, consumers should be able to examine the components of any summary measure and have access to a complete description of the weights and other assumptions underlying their combination into a single summary, the fifth attribute. Placing information in context, the final attribute, involves several factors. A good summary measure should allow comparisons among vehicles. Because a few attributes such as vehicle mass and size can have a profound influence on safety, consumers should be able to compare across vehicle classes as well as among vehicles in the same class. Moreover, they should be able to compare safety information with other important attributes such as price, performance, and reliability. Because motor vehicle crashes are not the only risks that people face, a vehicle safety information program should also help consumers make meaningful comparisons with other dangers to judge how much attention and how many resources they should allocate to dealing with motor vehicle crash risks.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information DESIGNING EFFECTIVE COMMUNICATIONS How consumers understand and interpret information depends on what they already know and believe. As explained in Chapter 4, very little research has been done to determine how people frame and think about issues of vehicle safety. Until such research is performed, it is premature to make precise recommendations about the contents of safety messages. The National Highway Traffic Safety Administration (NHTSA) should be able to conduct the necessary initial research fairly quickly (in less than 2 years) and at relatively low cost (for less than $300,000).2 An approach that has been used to develop other consumer-oriented risk communications holds promise. The premise is that people have knowledge and beliefs about a topic, which they use to filter and interpret new information (Morgan et al. 1992, 2050; Bostrom et al. 1994a, 789). Communicators need to understand these “mental models ” if they are to design messages that will not be dismissed or misinterpreted (Morgan et al. 1992, 2050; Jungerman et al. 1988). The first step is to construct a summary of decision-relevant expert knowledge. This expert knowledge is most usefully represented as a diagram that shows the causal influences among the primary factors affecting the decision at hand. Here the decision involves incorporating safety into a vehicle purchase decision. The next step is to conduct structured, open-ended interviews (Bostrom et al. 1992) of an appropriately selected sample of consumers to elicit their beliefs about the factors that affect the safety of automobiles. Audience segmentation should be taken into account in designing the sampling strategy because some groups, such as new drivers, may differ from others in conceptualizing automobile safety. The interview protocol should allow the expression of both accurate and inaccurate concepts but follow the overall structure of the expert diagram to permit easy comparison with expert knowledge. In-depth interviews are crucial in identifying key beliefs. However, they are also resource-intensive—both time-consuming and difficult to collect and analyze—making it unlikely that this kind of study will be large enough to reliably predict the relative frequency of different beliefs. So in the third stage of the research, the results from the interviews are used to design structured questionnaires, which can be administered to a larger sample of consumers to determine the prevalence of the beliefs encountered in the interviews.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information Alternative approaches to studying public understanding of risks differ from the mental models approach primarily in that they do not incorporate an explicit decision-making standard in the form of an expert model. For example, ethnographic interviews conducted by cognitive anthropologists on environmental risks have yielded results similar to those of the mental models approach (Kempton 1991; Kempton et al. 1995; Bostrom et al. 1994b). In another approach, investigators of consumer recall of product use information on household chemical products (Magat et al. 1988) used pretest interviews to develop a checklist of concept categories. Interviewers coded consumers ' responses directly into these categories during the interviews. This strategy is less resource intensive but appears less likely to identify systematically the full range of consumer beliefs about a complex risk process. Such studies of public understanding should not be confused with studies of attitudes, opinions, and self-judged levels of knowledge, which are common and do not provide information about the audience's substantive beliefs. Results from a full-fledged mental models study should be sufficient to support development of a draft vehicle safety message based on both the analysis of the information needed and the assessment of what that audience currently believes. However, experience suggests that appropriate expertise and formative research alone are insufficient to guarantee an effective communication. It is essential to iteratively test any communication empirically using multiple evaluation methods and revise the communication accordingly to ensure that it is effective (Morgan et al. 1992, 2054–2055; Bostrom et al. 1994a, 796). Different people will want different levels of detail. Some will only want simple summary information. Others will want additional information that explains generally how the summary was developed and provides a broader context for the information. A few will want much greater detail—a tutorial on current knowledge and an elaboration of the expert judgments on which the summary measures are based— so that well-educated consumers, or consumer interest and other groups, can tailor comparisons to their particular needs. These differences among consumers suggest the need for a hierarchically organized communication strategy. A three-level approach is recommended (Figure 5-1). The first level provides the simplest, most highly summarized information in the form of a vehicle safety label for all new passenger vehicles. More detailed information is provided at the next level down in the hierarchy in the form of a vehicle safety brochure. The brochure would explain in modest detail the fac-

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information FIGURE 5-1 Proposed hierarchically organized three-level communication strategy. In addition, the information will be available for summary publication by news magazines, consumer organizations, and similar groups. tors that contribute to crash avoidance and crashworthiness such as the important role of vehicle size and weight, discuss how these factors were used to produce the summary measures for the vehicle in question, and summarize the detailed information on which the summary measures are based. It might also briefly discuss alternative ways in which the summary information might be generated (e.g., for particu-

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information lar classes of drivers). The vehicle safety brochure would be available in the glove compartment of any new vehicle. As explained in the final section of the chapter, the information it contains would also be available to consumers in various other forms to allow access before they arrive in a showroom. At the most detailed level in the hierarchy, comparative information for all vehicle size and weight classes would be provided. A printed safety handbook would give a nontechnical but detailed explanation of the factors that contribute to vehicle safety, the algorithms used to produce the summary measures, and perhaps several other algorithms and associated results. It would also include a summary table, listing the safety ratings for all vehicles sold in the U.S. market. At a minimum, the table, if not the entire handbook, should be made available at the dealer showroom. Because much of the information will change on a regular basis, the handbook should also be available in computer hypertext form at a World Wide Web site that can be visited by consumers, automobile and insurance salespeople, reference librarians, and consumer groups. This version might even include software that would allow consumers to generate customized measures tailored to their own needs and driving habits. CONTENT OF COMMUNICATIONS The specific contents of these communications cannot be fully defined until the research on public understanding of vehicle safety that was outlined in the previous section has been completed, but the primary information to be communicated can be identified. Development of Summary Safety Ratings Given a particular driver, road, and traffic conditions, the overall safety of a vehicle depends on how well safety features help prevent a crash and how well the vehicle performs during a crash. Thus summary measures of both crash avoidance and crashworthiness are needed to provide consumers with information on the primary vehicle-related factors that affect safety. In producing a single measure for use by the general public, a reasonable starting point is to assume the average driving characteristics of the general population. In the safety brochure or handbook, it might be desirable to provide several additional summary measures for special populations (e.g., older drivers and teenage drivers).3

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information A single overall summary measure is a desirable long-term goal. However, the committee concludes that it is not currently feasible to produce a summary measure that combines both crash avoidance and crash performance information. The committee does believe that, if it is supplemented with the judgment of safety experts, current knowledge is sufficient to support the development of a meaningful single summary measure for vehicle crashworthiness and that such a measure could be made significantly more robust in the future if it is coupled with an appropriate program of continuing research (discussed in the next chapter). Whereas a similar summary measure might be produced for crash avoidance, the committee is not persuaded that the current knowledge is sufficient for experts to agree about how this should be done. The main problem is the relatively limited role that vehicle characteristics (versus driving behavior) currently play in predicting crash involvement. Because the effects of crash avoidance features are often small and observable in laboratory settings but not always in the field, it is difficult for the experts to reach agreement on their risk reduction potential.4 However, all else being equal, consumers will be better off with these safety features than without them. New vehicle technologies developed as part of the Intelligent Transportation Systems program, such as collision avoidance and night vision systems, may increase the importance of safety features in crash avoidance. For the near term, however, a checklist of key crash avoidance features rather than a summary measure must suffice. This strategy represents a compromise on several of the attributes outlined at the beginning of the chapter. The compromise is necessitated by incomplete current understanding. As knowledge is gained from refinements in testing methods and field crash data through the iterative process described in the next chapter, consideration should be given to adding a summary measure of crash avoidance and, ultimately, to combining the two measures. There is no unique way to construct a summary measure of the crashworthiness of a vehicle using the data now available or likely to become available in the foreseeable future. Nor can such a measure be constructed on strictly scientific grounds. The judgment of experts is needed to combine the various relevant data into a defensible estimate of vehicle crash performance. Expert judgment is commonly combined with formal analysis to support decision making under uncertainty. The methods and techniques that have been developed to support such analysis are referred

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information to as decision analysis (Raiffa 1968; Keeney 1982; Von Winterfeldt and Edwards 1986; Watson and Buede 1987). They have been applied in a variety of fields. In introducing the idea of using judgment to automobile safety experts and providing them with a framework for exercising this judgment, the committee found it useful to personalize the problem in the following form: Suppose that your daughter has taken up residence in a land that is similar in its technologies, behaviors, and regulations to our own. However, this land has a completely different set of automobile manufacturers and models. Your daughter asks you to help her pick a vehicle that will offer her and her family good crash safety performance. How do you proceed? Vehicle weight and size are the first things experts typically want to know about. Although they may argue that various other data, such as highway crash fatality statistics and crash test performance information, have less predictive power, they typically want to include such data in the information they will consider in advising their daughters. The problem reduces to getting safety experts, statisticians, and decision analysts to work together to refine a consensus about the algorithm they would use to advise their daughters. Whereas the committee has concluded that such a subjective combination is possible, choosing the actual algorithm is a matter for extensive research—an effort beyond the limited time and resources of this committee. Four sets of data should be considered for inclusion in the construction of a summary measure of vehicle crashworthiness. First is the statistical relationship between the size and weight of a vehicle and its crash performance. Second, crash test results for a specific vehicle are available from the automobile manufacturers, NHTSA, and the insurance industry. With regression analysis, statistical relationships between crash test performance and highway crash performance can be established. The correlation coefficients that can be achieved with current test designs are low, which means that the amount of variance that can be explained on the basis of the test data is small, but it is not zero. Faced with making a choice, consumers on the average will be better off by using the crash results in their decision than they will be by ignoring them (and guessing). Combining the size and weight data with available crash test data will require judgment about both the functional

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information form of the algorithm to be used and the values of various weighting coefficients on the basis of factors such as statistical data on the frequency of various crash types. Third, a number of engineered features (e.g., structure of the occupant compartment, presence of special energy-absorbing components) influence the overall crashworthiness of a vehicle. Sufficient statistical data may not be available to allow a formal incorporation of these factors into the measure, but they can probably be incorporated through the use of expert judgment. Once a new car model enters the market, the final type of data—actual highway crash performance—begins to accumulate. Methods are available (DeGroot 1970) that will allow the new data to be combined with prior performance estimates in a process known as Bayesian updating. The result is that over time the predictive power of the crash performance estimate can be improved as more real-world performance data are incorporated.5 The summary measure that results from this process, even though it will be updated annually, will involve significant uncertainty. Honesty and fairness to both manufacturers and consumers require that this uncertainty be communicated. Otherwise, a very crude estimate with limited predictive power may be misinterpreted by users as a precise indicator of safety. This means that results must always be reported as ranges, preferably in a graphical display. Once the creation of a summary estimate of vehicle crashworthiness has begun, improvements should become apparent. A number of crash tests, adding to or replacing those now being conducted, are likely to be found that yield greater predictive power. Through the use of advanced “black box” instrumentation added to the computers that are now on board virtually all new vehicles, field crash data might be refined to yield much greater predictive power than is now possible. An institutional strategy and the activities and resources required to support such a process of continuous refinement and improvement are outlined in the next chapter. Reporting Safety Ratings How best to report a summary measure of expected vehicle safety performance is a question that cannot be answered without conducting experimental studies with groups of laypeople. In general, however, some

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information form of graphical display should be used that allows consumers to compare the crashworthiness (and in the future, the crash avoidance potential) of a specific vehicle with all new vehicles on the market as well as with other vehicles in the same class. A comparison with the median of all vehicles on the market is probably best. The mean can be substantially influenced by a few outliers. The vehicle safety label should provide simple summary graphics. Because some people find graphics confusing and prefer text, the label should also provide a verbal summary of the result and an explanation of the two or three most influential factors. In the interest of simplicity, it probably should not give detailed data on the value of the component variables. Such information would be provided in the safety brochure that accompanies the vehicle. The algorithm used to produce the summary measure should reflect the behavioral characteristics of the average population of drivers. If the expert panel that produces the measures believes that significantly different algorithms should be used to reflect different types of drivers (e.g., teenage drivers and older drivers), the results should be briefly summarized in the vehicle safety brochure. A complete nontechnical explanation of the algorithms and how they work should be provided in the safety handbook. As noted earlier, it will not be possible to specify the design or content of the vehicle safety label, the vehicle safety brochure, or the safety handbook until research on consumer knowledge and beliefs has been completed and draft communications have been field tested and iteratively refined. To illustrate how the middle stages of the safety label production process might proceed, two examples of possible vehicle safety labels were developed and received limited testing (Figure 5-2). Resources did not allow for appropriate evaluation and redesign of these labels. However, the preliminary evaluations illustrate how the process of refinement would work. A structured interview instrument was used to collect read-aloud protocols from a sample of 10 adults in Pittsburgh (see Appendix D).6 Interviewees were given first one label, then the other, with the order counterbalanced. For each label, the interviewee was asked to provide first impressions of the label, to read through that label aloud saying anything that came to mind, to comment on what he or she liked or disliked about the label, and finally to explain the graphics in the lower half of the label and provide suggestions for improving the label. Last, in-

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information FIGURE 5-2 Examples of possible vehicle safety labels for a hypothetical compact car called the 1997 XYZ300. As discussed in the text, labels must be designed on the basis of experimental studies involving the public. There is no other way to reliably determine what does and does not work.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information terviewees were given both labels to compare, then were asked to complete a short written questionnaire (see Appendix D). Each interview lasted 20 to 25 min. The results of the interviews illustrate the kinds of design guidance such studies can provide and are largely consistent with the focus group study results described in Chapter 4. Interviewees provided suggestions for the arrangement of the graphical display, the composition of the label, and wording or language choices. The comments interviewees made during the read-aloud protocols provide useful insights about several aspects of the label, especially the source of the information on the label, that is, who produced the label; the content of the top statement on safe driving behavior; specific crash avoidance features or the scales used for the features; the graphical crashworthiness displays; the text provided on crashworthiness; and preferences for one label over the other overall, or for some section of one label over that section in the other. Each of these types of comments is discussed in more detail in the following paragraphs. As in the focus groups, interviewees wanted to know who was responsible for the label and expressed concern that it might have been produced by a manufacturer. Again, this suggests the appropriateness of a clearly indicated government source for vehicle safety information. Several interviewees made negative comments about the opening statement on driver behavior, referring to it as a “public service commercial ” and suggesting that it be dropped or moved. Because driver behavior is a dominant factor, the committee believes that such a statement provides essential information. However, the extensive comments from interviewees suggest that the location and content of any such statement need further study. Several interviewees did not understand what some of the specific crash avoidance features listed were, such as daytime running lights. Several complained about the use of differing scales. The scale for reporting stopping distance used in the left-hand label was preferred by several respondents, because it made it clear that there was a standard for stopping distance. A few stated that the explanatory text at the top of this section on the left-hand label was helpful. Almost all the interviewees succeeded in characterizing the graphical displays of crash performance appropriately, if not always completely. Some of the interviewees were able to describe correctly what the graphical display of crash performance was saying after just a brief

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information glance.7 The modal responses on the written questionnaire showed that interviewees could use the displays to compare the typical values for that specific vehicle with those for its class and for all vehicles and could use the range information for comparative purposes. However, in the interviews the graphical displays were the source of the largest number of comments of all kinds, indicating that interviewees spent most effort on this section of the labels. Some interviewees called the displays “confusing” and “hard to understand,” and some stated that they did not know what the values or scales in the graphics represented. Both positive and negative comments were made about the numbers in the right-hand display. Some interviewees said that they are “nice” and make the display easier to use; others said that they did not know what they mean. Whereas a few interviewees preferred the right-hand display because it was less cluttered, many interviewees stated that they preferred the left-hand label specifically because it provided “the picture explained in text,” which they liked. But one interviewee disliked this text as well as the graphical display and spent little time attempting to understand either. He stated emphatically that he did not want to “go to math class to buy this car” and that the comparison is to a “typical car” and not to “anything real.” This interviewee did not like the use of qualifying words such as “probably,” remarked “no probabilities please,” and stated that a description of crash features, such as the “engine will drop down to the ground and not end up in your lap in a crash,” would be more useful and had influenced his most recent car purchase.8 These interview results alone cannot be generalized, but they illustrate that a graphical display can convey the necessary information to consumers who are concerned about safety. They illustrate the need for additional studies of prior beliefs about vehicle safety and alternative label designs. Whatever the final design, the content of the safety label should probably not be incorporated into the current vehicle window sticker. A separate label, where all the information relevant to vehicle safety can be concentrated, is desirable. A separate label has a better chance of attracting consumer interest and attention, given the amount of information provided on the existing window label. The safety label should be prominently displayed on the vehicle, but the window may not be the best location because of limitations on window space for some vehicles and concerns about visibility in driving test vehicles.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information The design and content of the vehicle safety brochure and the safety handbook should be built on studies of consumer understanding and perception of the determinants of vehicle safety. Once an initial draft has been prepared, it should be subjected to read-aloud protocol analysis (Ericsson and Simon 1993) and other appropriate evaluation (Schriver 1989). As with the safety label, an iterative approach will be essential to produce a communication that is clear and serves consumers' needs. GETTING THE MESSAGE OUT There is no one best way to get vehicle safety information to consumers. The safety label ensures that consumers will have relevant information at the point of sale. The safety brochure and instructions for obtaining it and other information through the NHTSA hotline and electronically via the World Wide Web should be available at the dealer showroom. Dealers could give the brochure and summary information from the handbook to prospective buyers at the same time they provide the NHTSA-required booklet on collision losses. At some stage the latter information might be incorporated into the safety brochure. The marketing literature clearly indicates that the information should be available to consumers well before they get to the dealer showroom to be most timely and useful. A multichannel approach is recommended. Summary information could be included in selected mailings by insurance companies and organizations such as the American Automobile Association and the American Association of Retired Persons. It could also be advertised, reprinted, or summarized in trade and consumer journals. Portions of the material could be accessed by fax or mail by calling the NHTSA hotline. Finally, the full spectrum of materials should be made available electronically on a hierarchically organized World Wide Web page. To reach younger audiences, a curriculum segment could be developed for use in driver's education courses that explains how to use the safety information in choosing a safe car. Of course, a public service advertising campaign, at least initially, would help increase consumer awareness of the safety label and the accompanying brochures. Once awareness becomes widespread, many safety-conscious consumers can be expected to seek the information at an early stage when they begin to consider purchasing a new car. These strategies for development and communication of improved vehicle safety information should be viewed as part of a continuing process to yield both improved information and safer cars.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information Organizational arrangements for ensuring such a permanent process are considered in the next chapter. NOTES 1. Hsee (1995) has found that portrayal of uncertainty for a critical option or attribute (e.g., vehicle safety) can result in a decision maker placing more weight on a less important but more tempting attribute (e.g., vehicle styling). 2. This cost estimate reflects the judgment of several study committee members who have conducted similar research projects. 3. If conditional summary measures of both crash avoidance and crashworthiness could be produced, combining them into a single overall measure would require additional behavioral knowledge as well as an understanding of how those behaviors combine to affect the overall risk of fatality and injury in motor vehicle crashes. 4. Of course, what is known about the crash avoidance potential of vehicle safety features can be made available to consumers in the more detailed safety handbook. 5. Of course, it will not be possible to add this additional information on real-world crash experience when new vehicle models are introduced. 6. The interviews were conducted by Shane Frederick, a Ph.D. student in the Department of Social and Decision Sciences at Carnegie-Mellon University. Interviewees were paid a nominal amount for their participation. All were high school graduates, three had completed college, and two had completed an advanced degree. None were physical scientists or engineers. In comparison, roughly 90 percent of the U.S. population have graduated from high school and almost 25 percent have graduated from college (Rothberg 1995). 7. These interviewees were among the best educated in the sample. 8. Those who preferred the right-hand label were among the more highly educated in the sample. Of those with a high school education, all but this interviewee (who liked neither) preferred the left-hand label with its verbal explanation of the graphics. REFERENCES Abbreviation NRC National Research Council Bostrom, A., B. Fischhoff, and M.G. Morgan. 1992. Characterizing Mental Models of Hazardous Processes: A Methodology and an Application to Radon. Journal of Social Issues, Vol. 48, No. 4, pp. 85–100. Bostrom, A., C.J. Atman, B. Fischhoff, and M.G. Morgan. 1994a. Evaluating Risk Communications: Completing and Correcting Mental Models of Hazardous Processes, Part II. Risk Analysis, Vol. 14, No. 5, pp. 789–798. Bostrom, A., M.G. Morgan, B. Fischhoff, and D. Read. 1994b. What Do People Know About Global Climate Change? 1. Mental Models . Risk Analysis, Vol. 14, No. 6, pp. 959–970.

OCR for page 110
Shopping for Safety: Providing Consumer Automotive Safety Information DeGroot, M. 1970. Optimal Statistical Decision. McGraw-Hill, New York. Ericsson, K.A., and H. Simon. 1993. Protocol Analysis: Verbal Reports as Data (revised edition). MIT Press, Cambridge, Mass. Hsee, C.K. 1995. Elastic Justification: How Tempting but Task-Irrelevant Factors Influence Decisions. Organization Behavior and Human Decision Processes, Vol. 62, No. 3, pp. 330–337. Ibrekk, H., and M.G. Morgan. 1987. Graphical Communication of Uncertain Quantities to Nontechnical People . Risk Analysis, Vol. 7, No. 4, pp. 519–529. Jungerman, H., H. Schutz, and M. Thuring. 1988. Mental Models in Risk Assessment: Informing People About Drugs. Risk Analysis, Vol. 8, No. 1, pp. 147–155. Keeney, R.L. 1982. Decision Analysis: An Overview. Operations Research, Vol. 30, No. 5, Sept.–Oct., pp. 803–838. Kempton, W. 1991. Lay Perspectives on Global Climate Change. Change, June, pp. 183–208. Kempton, W., J.S. Boster, and J.A. Hartley. 1995. Environmental Values in American Culture. MIT Press, Cambridge, Mass., 320 pp. Magat, W.A., W.K. Viscusi, and J. Huber. 1988. Consumer Processing of Hazard Warning Information. Journal of Risk and Uncertainty, Vol. 1, pp. 201–232. Merton, R.K. 1956. The Focused Interview: A Manual of Problems and Procedures. Free Press, Glencoe, Ill. Merton, R.K. 1987. The Focussed Interview and Focus Groups. Public Opinion Quarterly, Vol. 51, pp. 550–566. Morgan, D.L. 1988. Focus Groups as Qualitative Research. Sage Publications, Newbury Park, Calif. Morgan, M.G., B. Fischhoff, A. Bostrom, L. Lave, and C.J. Atman. 1992. Communicating to the Public. Environment, Science, Technology, Vol. 26, No. 11, pp. 2,049–2,056. NRC. 1989. Improving Risk Communication. National Academy Press, Washington, D.C., 332 pp. Raiffa, H. 1968. Decision Analysis: Introductory Lectures on Choice Under Uncertainty . Addison Wesley, Reading, Mass. Roth, E., M.G. Morgan, B. Fischhoff, L. Lave, and A. Bostrom. 1990. What Do We Know About Making Risk Comparisons? Risk Analysis, Vol. 10, No. 3, pp. 375–387. Rothberg, I.C. 1995. Myths About Test Score Comparisons. Science, Vol. 270, No. 5241, Dec., pp. 1,446–1,447. Schriver, K. 1989. Evaluating Text Quality: The Continuum from Text-Focused to Reader-Focused Methods. IEEE Transactions on Professional Communication, Vol. 32, No. 4. Stewart, D.W., and P.N. Shamdasani. 1990. Focus Groups: Theory and Practice. Sage Publications, Newbury Park, Calif. Von Winterfeldt, D., and W. Edwards. 1986. Decision Analysis and Behavioral Research. Cambridge University Press, New York. Watson, S.R., and D.M. Buede. 1987. Decision Synthesis: The Principles and Practice of Decision Analysis . Cambridge University Press, New York.