Notes on the Interpretation of Traffic Accident Data and of Risk Homeostasis Theory: A Reply to L. Evans

A quick and summary comment is made on a paper by Evans, who set out to "definitively answer" the question "whether there is any validity" in risk homeostasis theory (RHT), and who arrives at the conclusion that the "theory should be rejected" because the data "provide the clearest evidence refuting the theory." It will be argued here that the seven cases Evans puts forward against RHT do not justify such a rather boldly-worded conclusion. On the contrary, the data presented by Evans can be seen to be either not inconsistent with RHT (Case 1), in error (Case 2), or fail to take account of relevant economic changes (Case 3) and of pertinent exposure variables (Case 5), incomplete or the product of guess-work (Case 4), irrelevant (Case 6), nonrepresentative (Case 7) and in all cases open to alternative interpretations that do not conflict with the theory under debate. Moreover, Evans' reasoning seems to betray not only careless reading of what RHT does in fact say, but also a fundamental misunderstanding of the notion of target risk in particular. Refuting a "refutation" does not, of course, imply that RHT is valid, or valid at least within as yet to be defined limits. Attempts to address the validity question should preferably take the form of well-controlled field experiments, instead of retrospective analyses of multi-interpretable archival data that can be debated ad infinitum.     

Risk Homeostasis Theory and Traffic Accident Data

The risk homeostasis theory posits, in essence, that a control mechanism analogous to the thermal homeostatic system in warm-blooded animals tends to keep risk per unit time constant, and, as a consequence, the number of traffic accidents per unit time of driving also tends to remain constant, essentially independent of changes in the traffic safety system. It is the purpose of the present research to examine the validity of this claim using a wide variety of traffic accident data. All the data examined are found to be incompatible with the risk homeostasis theory. The only specific field accident data offered in the literature to support the risk homeostasis theory are found to, in fact, refute the theory. The accident data provide evidence that a rich variety of user responses occur. While it is possible for users to collectively respond in such a way that safety benefits are completely cancelled, such a response is not particularly common; it is certainly not universally occurring, as suggested by the risk homeostasis theory. It is concluded that the risk homeostasis theory should be rejected because there is no convincing evidence supporting it and much evidence refuting it.     

The Case against Commercial Antivirus Software: Risk Homeostasis and Information Problems in Cybersecurity

New cybersecurity technologies, such as commercial antivirus software (AV), sometimes fail to deliver on their promised benefits. This article develops and tests a revised version of risk homeostasis theory, which suggests that new cybersecurity technologies can sometimes have ill effects on security outcomes in the short run and little-to-no effect over the long run. It tests the preliminary plausibility of four predictions from the revised risk homeostasis theory using new survey data from 1,072 respondents. The estimations suggest the plausible operation of a number of risk homeostasis dynamics: (1) commercial AV users are significantly more likely to self-report a cybersecurity event in the past year than nonusers, even after correcting for potential reverse causality and informational mechanisms; (2) nonusers become somewhat less likely to self-report a cybersecurity event as the perceived riskiness of various e-mail-based behaviors increases, while commercial AV users do not; (3) the negative short-run effect of commercial AV use on cybersecurity outcomes fade over time at a predicted rate of about 7.03 percentage points per year of use; and (4) after five years of use, commercial AV users are statistically indistinguishable from nonusers in terms of their probability of self-reporting a cybersecurity event as perceptions of risky e-mail-based behaviors increase.     

Deep Learning Models for Health and Safety Risk Prediction in Power Infrastructure Projects

Inappropriate management of health and safety (H&S) risk in power infrastructure projects can result in occupational accidents and equipment damage. Accidents at work have detrimental effects on workers, company, and the general public. Despite the availability of H&S incident data, utilizing them to mitigate accident occurrence effectively is challenging due to inherent limitations of existing data logging methods. In this study, we used a text-mining approach for retrieving meaningful terms from data and develop six deep learning (DL) models for H&S risks management in power infrastructure. The DL models include DNN_classify (risk or no risk), DNN_reg1 (loss time), DNN_reg2 (body injury), DNN_reg3 (plant and fleet), DNN_reg4 (equipment), and DNN_reg5 (environment). An H&S risk database obtained from a leading UK power infrastructure construction company was used in developing the models using the H2O framework of the R language. Performances of DL models were assessed and benchmarked with existing models using test data and appropriate performance metrics. The overall accuracy of the classification model was 0.93. The average R² value for the five regression models was 0.92, with mean absolute error between 0.91 and 0.94. The presented results, in addition to the developed user-interface module, will help practitioners obtain a better understanding of H&S challenges, minimize project costs (such as third-party insurance and equipment repairs), and offer effective strategies to mitigate H&S risk.     

Improving Health Risk Assessment as a Basis for Public Health Decisions in the 21st Century

One-fifth of the way through the 21st century, a commonality of factors with those of the last 50 years may offer the opportunity to address unfinished business and current challenges. The recommendations include: (1) Resisting the tendency to oversimplify scientific assessments by reliance on single disciplines in lieu of clear weight-of-evidence expressions, and on single quantitative point estimates of health protective values for policy decisions; (2) Improving the separation of science and judgment in risk assessment through the use of clear expressions of the range of judgments that bracket protective quantitative levels for public health protection; (3) Use of comparative risk to achieve the greatest gains in health and the environment; and (4) Where applicable, reversal of the risk assessment and risk management steps to facilitate timely and substantive improvements in public health and the environment. Lessons learned and improvements in the risk assessment process are applied to the unprecedented challenges of the 21st century such as, pandemics and climate change. The beneficial application of the risk assessment and risk management paradigm to ensure timely research with consistency and transparency of assessments is presented. Institutions with mandated stability and leadership roles at the national and international levels are essential to ensure timely interdisciplinary scientific assessment at the interface with public policy as a basis for organized policy decisions, to meet time sensitive goals, and to inform the public.     

The Concept of Antifragility and its Implications for the Practice of Risk Analysis

Nassim Taleb's antifragile concept has been shown considerable interest in the media and on the Internet recently. For Taleb, the antifragile concept is a blueprint for living in a black swan world (where surprising extreme events may occur), the key being to love variation and uncertainty to some degree, and thus also errors. The antonym of “fragile” is not robustness or resilience, but “please mishandle” or “please handle carelessly,” using an example from Taleb when referring to sending a package full of glasses by post. In this article, we perform a detailed analysis of this concept, having a special focus on how the antifragile concept relates to common ideas and principles of risk management. The article argues that Taleb's antifragile concept adds an important contribution to the current practice of risk analysis by its focus on the dynamic aspects of risk and performance, and the necessity of some variation, uncertainties, and risk to achieve improvements and high performance at later stages.     

Dimensions of Risk Perception for Financial and Health Risks

This study of 29 MBA students compares two models of risk perception for both financial and health risk stimuli. The first, inspired by Luce and Weber's Conjoint Expected Risk (CER) model, uses five dimensions: probability of gain, loss and status quo, and expected benefit and harm. The second, inspired by the Sovic et al. psychometric model, employs seven dimensions: voluntariness, dread, control, knowledge, catastrophic potential, novelty, and equity. The CER-type model provided a better fit for most subjects and stimuli. Adding the psychological risk dimensions from the Slovic et al. model explained only modestly more variance. Relationships between the dimensions of the two models are described and the construction of a hybrid model explored.     

The Management of Risk: Application to the Welding Industry

The management of health risk in the welding industry is considered based on a discussion of the major sources of harm to welders arising from their employment (e.g., accidents and inhalation of fumes and gases). It is shown that present methods neither enable the assessment of the societal and human costs involved, nor permit the specific association of delayed health effects to occupational fume exposures. Reported accidents usually occur early in the working experience and contribute to a large number of working days lost, while fume exposures may contribute to a reduction in life quality which is poorly defined. It is concluded that risk management can only be attempted after much more information is made available concerning the origin, nature, and duration of health effects, especially as related to individual welding technologies and applications.     

Major Hazard Information Policy in the European Community: Implications for Risk Analysis

The main impetus to the development of information about major industrial hazards in the European Community comes from the so-called Seveso Directive, which defines an information network and requires the generation and transmission of information as the basis for accident prevention and risk management. This important policy development, which calls for the formal identification and analysis of major hazards and the communication of risk information to members of the public, presents new opportunities and challenges to risk analysis and research in Europe. This paper briefly reviews the accidents that gave rise to the Directive and shaped its content, and then summarizes its requirements. The status of its implementation in the EC Member States is discussed, with special emphasis given to the comparison of safety analysis practices, the Major Accident Reporting System (MARS), and risk communication. Some new research directions stimulated by the Directive are identified.     

Policy Goals for Health and Safety

Health management and safety regulation are separate disciplines but share the aim to extend expectancy of life in good health. The need to improve cost-effectiveness calls for their co-ordinated management according to a unified rationale. Three guiding principles of accountability, demonstrable net benefit and a uniform measure of performance, have been laid out in Canada by the Joint Committee on Health and Safety. They call for open accounting in terms of (health-related quality-adjusted) life expectancy. The principles are utilitarian in format but, it is argued, inequity is naturally diminished in the process of optimizing cost-effectiveness through maximum marginal returns. Comments are made on practical implementation. The need for public consent in practice calls for two additional principles reflecting fair procedure and sovereignty of the citizens. It is concluded that public health and safety measures should be surveyed, documented for cost-effectiveness and prioritized for improvement.     

Risk Analysis for Environmental Health Triage

The Homeland Security Act mandates the development of a national, risk-based system to support planning for, response to, and recovery from emergency situations involving large-scale toxic exposures. To prepare for and manage consequences effectively, planners and responders need not only to identify zones of potentially elevated individual risk but also to predict expected casualties. Emergency response support systems now define "consequences" by mapping areas in which toxic chemical concentrations do or may exceed Acute Exposure Guideline Levels (AEGLs) or similar guidelines. However, because AEGLs do not estimate expected risks, current unqualified claims that such maps support consequence management are misleading. Intentionally protective, AEGLs incorporate various safety/uncertainty factors depending on the scope and quality of chemical-specific toxicity data. Some of these factors are irrelevant, and others need to be modified, whenever resource constraints or exposure-scenario complexities require responders to make critical trade-off (triage) decisions in order to minimize expected casualties. AEGL-exceedance zones cannot consistently be aggregated, compared, or used to calculate expected casualties and so may seriously misguide emergency response triage decisions. Methods and tools well established and readily available to support environmental health protection are not yet developed for chemically-related environmental health triage. Effective triage decisions involving chemical risks require a new assessment approach that focuses on best estimates of likely casualties, rather than on upper plausible bounds of individual risk. If risk-based consequence management is to become a reality, federal agencies tasked with supporting emergency response must actively coordinate to foster new methods that can support effective environmental health triage.     

The Effect of Product Safety Regulation on Safety Precautions

This paper explores the means by which lighter safety regulations alter the precautionary behavior of 200 subjects in a field test of cigarette lighters with a child-resistant feature. The new lighter design leads respondents to believe the lighters are safer, but there is no clear-cut evidence that the effect on perceived risk levels is excessive. Using the estimated relationship between cigarette lighter risk perceptions and a variety of measures of precautions, this paper provides explicit estimates of the effect of regulations on precautionary behavior and on lighter safety. On balance, the child-resistant feature will reduce fire-related injuries by much more than any diminished precaution taking.     

Integrated Uncertainty Analysis for Ambient Pollutant Health Risk Assessment: A Case Study of Ozone Mortality Risk

The U.S. Environmental Protection Agency (EPA) uses health risk assessment to help inform its decisions in setting national ambient air quality standards (NAAQS). EPA's standard approach is to make epidemiologically‐based risk estimates based on a single statistical model selected from the scientific literature, called the “core” model. The uncertainty presented for “core” risk estimates reflects only the statistical uncertainty associated with that one model's concentration‐response function parameter estimate(s). However, epidemiologically‐based risk estimates are also subject to “model uncertainty,” which is a lack of knowledge about which of many plausible model specifications and data sets best reflects the true relationship between health and ambient pollutant concentrations. In 2002, a National Academies of Sciences (NAS) committee recommended that model uncertainty be integrated into EPA's standard risk analysis approach. This article discusses how model uncertainty can be taken into account with an integrated uncertainty analysis (IUA) of health risk estimates. It provides an illustrative numerical example based on risk of premature death from respiratory mortality due to long‐term exposures to ambient ozone, which is a health risk considered in the 2015 ozone NAAQS decision. This example demonstrates that use of IUA to quantitatively incorporate key model uncertainties into risk estimates produces a substantially altered understanding of the potential public health gain of a NAAQS policy decision, and that IUA can also produce more helpful insights to guide that decision, such as evidence of decreasing incremental health gains from progressive tightening of a NAAQS.     

Relative Risk Perception for Terrorism: Implications for Preparedness and Risk Communication

Terrorism presents a significant risk that is often approached at public policy, infrastructure, or emergency management level. Public perceptions of the likelihood of terrorist events, and how this may relate to individual preparedness, are not always extensively examined. The tendency to think that negative events are less likely to happen to oneself than to the average person is known as optimism bias. Optimism bias is relevant to perceptions of terrorism, because it is thought to be related to a reduction in precaution use. Using an online survey of 164 participants, this study aimed to determine whether Sydney residents thought they had a lower likelihood of experiencing terrorist events than other Australians. Significant optimism bias was observed for witnessing terrorist events, but not for personally experiencing terrorist events. In addition, Sydney residents tended to think that terrorist attacks were more likely to occur in Sydney than another major Australian city in the next five years. At the same time, household and workplace preparedness for terrorism was quite low, as was awareness of emergency strategies in the central business district. Perceptions of high likelihood of terrorism happening in one's own city, yet low preparedness present a challenge for risk communication and emergency management strategies. The diversity of possible terrorist targets, and the simple plans that can moderate the effects of a disaster may need to be emphasized in future anti‐terrorism initiatives.     

Dealing with the White Death: Avalanche Risk Management for Traffic Routes

This article discusses mitigation strategies to protect traffic routes from snow avalanches. Up to now, mitigation of snow avalanches on many roads and railways in the Alps has relied on avalanche sheds, which require large initial investments resulting in high opportunity costs. Therefore, avalanche risk managers have increasingly adopted organizational mitigation measures such as warning systems and closure policies instead. The effectiveness of these measures is, however, greatly dependent on human decisions. In this article, we present a method for optimizing avalanche mitigation for traffic routes in terms of both their risk reduction impact and their net benefit to society. First, we introduce a generic framework for assessing avalanche risk and for quantifying the impact of mitigation. This allows for sound cost-benefit comparisons between alternative mitigation strategies. Second, we illustrate the framework with a case study from Switzerland. Our findings suggest that site-specific characteristics of avalanche paths, as well as the economic importance of a traffic route, are decisive for the choice of optimal mitigation strategies. On routes endangered by few avalanche paths with frequent avalanche occurrences, structural measures are most efficient, whereas reliance on organizational mitigation is often the most appropriate strategy on routes endangered by many paths with infrequent or fuzzy avalanche risk. Finally, keeping a traffic route open may be very important for tourism or the transport industry. Hence, local economic value may promote the use of a hybrid strategy that combines organizational and structural measures to optimize the resource allocation of avalanche risk mitigation.     

Perception of Risk and the Attribution of Responsibility for Accidents

Accidents, one often hears, “happen”; we accept, and even expect, that they will be part of daily life. But in situations in which injury or death result, judgments of responsibility become critical. How might our perceptions of risk influence the ways in which we allocate responsibility for an accident? Drawing from attribution and risk perception theory, this study investigates how perceived controllability and desirability of risk, in addition to perceived danger and recreational risk‐taking, relate to attributions of responsibility for the cause of unintentional injury in a unique setting: U.S. national parks. Three parks, Mount Rainier, Olympic, and Delaware Water Gap, provide the setting for this survey‐based study, which considers how park visitors (N = 447) attribute responsibility for the cause of a hypothetical visitor accident. Results suggest that respondents tended to make more internal (i.e., related to characteristics of the victim), rather than external (i.e., related to characteristics of the park, or park management) attributions. As respondents viewed park‐related risk as controllable, they were more likely to attribute the cause of the accident to the victim. Moreover, among other significant variables, having experienced a similar accident predicted lower internal causal attribution. Opportunities for future research linking risk perception and attribution variables, as well as practical implications for the management of public outdoor settings, are presented.     

Risk Perception and the Value of Safety

This paper examines the relationship between perceived risk and willingness-to-pay (WTP) for increased safety from technological hazards in both conceptual and empirical terms. A conceptual model is developed in which a given household's WTP for risk reductions is a function of traditional socioeconomic variables (i.e., income and base level of risk) and perceived characteristics of the hazards (i.e., dread, knowledge, and exposure). Data to estimate the model are obtained through a combined contingent valuation and risk perception survey that considers 10 technological hazards, five of which are well-defined (e.g., death rates are known and the risks are relatively common) and five are less well-defined. Econometric results, using TOBIT estimation procedures, support the importance of both types of variables in explaining WTP across all 10 hazards. When the risks are split into two groups, the results show that WTP for well-defined hazards is most influenced by perceived personal exposure, while WTP for less well-defined risks is most influenced by levels of dread and severity.     

Threshold Evaluation of Emergency Risk Communication for Health Risks Related to Hazardous Ambient Temperature

Emergency risk communication (ERC) programs that activate when the ambient temperature is expected to cross certain extreme thresholds are widely used to manage relevant public health risks. In practice, however, the effectiveness of these thresholds has rarely been examined. The goal of this study is to test if the activation criteria based on extreme temperature thresholds, both cold and heat, capture elevated health risks for all‐cause and cause‐specific mortality and morbidity in the Minneapolis‐St. Paul Metropolitan Area. A distributed lag nonlinear model (DLNM) combined with a quasi‐Poisson generalized linear model is used to derive the exposure–response functions between daily maximum heat index and mortality (1998–2014) and morbidity (emergency department visits; 2007–2014). Specific causes considered include cardiovascular, respiratory, renal diseases, and diabetes. Six extreme temperature thresholds, corresponding to 1st–3rd and 97th–99th percentiles of local exposure history, are examined. All six extreme temperature thresholds capture significantly increased relative risks for all‐cause mortality and morbidity. However, the cause‐specific analyses reveal heterogeneity. Extreme cold thresholds capture increased mortality and morbidity risks for cardiovascular and respiratory diseases and extreme heat thresholds for renal disease. Percentile‐based extreme temperature thresholds are appropriate for initiating ERC targeting the general population. Tailoring ERC by specific causes may protect some but not all individuals with health conditions exacerbated by hazardous ambient temperature exposure.