Author: Brian Everitt

Author: Brian Everitt

Risk Perception

Alistair Cooke in a Letter from America during the Gulf War of 1990 told the sad story of an American family of four who cancelled their planned holiday to Europe because of the fears of terrorist attacks on the country’s airlines. They decided to drive to San Francisco instead. At the last junction before their itinerary was to have lead beyond their small Midwest town, they collided with a large truck with fatal consequences for them all.

Life is a risky business, and deciding which risks are worth taking and which should be avoided has important implications both for an individual’s lifestyle, and the way society operates. The benefits gained from taking a risk need to be weighed against the possible disadvantages. Acceptable risk is proportional to the amount of benefits. For the individual, living life to the fullest means achieving a balance between reasonable and unreasonable risk, and this will often be a matter of judgement dependent largely on the individual’s personality. But in society as a whole, where the same balancing act is required, it has to be achieved through political action and legislation.

If risks could be assessed and compared in a calm and rational manner it would benefit both individuals and society. There is, however, considerable evidence that such assessment and comparison is not straightforward.

Defining and Quantifying Risk

Risk is defined in my dictionary as ‘the possibility of incurring misfortune or loss’. The word risk is derived from the Greek word ‘rhiza’, the hazards of sailing too near to the cliffs: contrary winds, turbulent downdraughts, and swirling tides. Risk is synonymous with hazard, and informal usages of the word may indicate the nature- there is a risk of brain damage to children having the MMR vaccination, or merely the existence, of the possible danger –I always avoid risks, etc, etc.

Quantifying risk and assessing risk involve the calculation and comparison of probabilities, although most expressions of risk are compound measures that describe both the probability of harm and its severity. Americans, for example, run a risk of about 1 in 4000 of dying in an automobile accident. The probability is 1 out of 4000 for injuries lethally severe. This probability is derived by simply noting that in the last few years there have been about 56,000 automobile deaths per year in a population of approximately 224 million. The figure of 1 in 4000 expresses the overall risk to society-the risk to any particular individual clearly depends on her exposure: how much she is on the road, where she drives and in what weather, whether she is psychologically accident prone, what mechanical condition the vehicle is in and so on. Because gauging risk is essentially probabilistic, it has to be remembered that a risk estimate can assess the overall chance that an untoward event will occur but is powerless to predict any specific event – just as knowing that the probability of tossing a head with a fair coin is one-half, but being unable to predict which particular tosses will result in heads and which in tails.

Risk Perception

Risk perception is one’s opinion of the likelihood of risk associated with performing a certain activity or choosing a certain lifestyle. Risk perception may be influenced by many factors some sociological, some anthropological and some psychological, but the result is that people vary considerably in which risks they consider acceptable and which they do not even when they may agree on the degree of risk involved. For example, to many people air travel represents the very model of a low-risk form of transport. For others, with no fear of travelling large distances by say car or train, the prospect of a flight, even with a well respected commercial airline, can be a nightmare, often requiring several trips to the airport bar before being able to board the aeroplane. But as the following figures demonstrate, flying is actually one of the safest forms of transport.

Perhaps one reason for some people’s excessive and clearly misguided fear of flying is the general view that being killed in a plane crash must be a particularly nightmarish way to die. Another possibility is that the flying phobic considers flying an alien environment, and this considerably distorts the perception of the risk involved. And a third reason is that flying accidents are more prominent in the

press than those involving automobiles, although the latter are far more common (see later). Perception of risk is very likely to be influenced by whether we feel in control of a perceived risk.

Research shows that people tend to overestimate the probability of unfamiliar, catastrophic and well-publicized events and to underestimate the probability of unspectacular or familiar events that claim one victim at a time. For example, in a study reported in [2], respondents were asked to rate a total of 90 hazards, each with respect to 18 qualitative characteristics: for example, whether the risk from such sources as mountain climbing or exposure to asbestos was voluntary or involuntary; personally controllable or not; and known to those exposed or not. A principal components analysis of the data identified two major components. The first component was labeled as “dread risk” and related judgments of scales such as uncontrollability, fear, involuntariness of exposure etc. Hazards, which rated high on this factor, included nuclear weapons, nerve gas and crime, hazards responsible for only very few deaths. Those that rated low on this factor included home appliances and bicycles. The second component, labeled ‘unknown risk’ related judgments of the observability of risks, whether the effects are delayed in time or not, the familiarity of the risk, and whether the risks are known to science or not. Hazards that rated high on this dimension included solar electric power, DNA research and satellites; those that rated low included motor vehicles, fire fighting and mountain climbing.

A plot of the data in the space of these two components demonstrated that perceptions of risk were clearly related to the resulting position of an activity, particular in respect of the ‘dread’ factor; the higher a hazard’s score on this factor, the higher its perceived risk, the more people wanted to see its current risks reduced, and the more they wanted to see strict regulation employed to achieve the desired reduction in risk. It seems that the risks that kill people and the risks that scare people are different.

The findings in [2] are well illustrated by the results of polls of college students and members of the League of Women Voters in Oregon. Both groups considered nuclear power, their number-one ‘present risk of death’, far ahead of motor vehicle accidents, which kill 50,000 Americans each year; cigarette smoking which kills 150,000 and handguns, which kill 17,000. “Experts” in the same poll considered motor vehicle accidents their number one risk, with nuclear power below the risk of swimming, railroads and commercial aviation. Here the experts seem to reached the most defendable conclusions. Average annual fatalities expected from nuclear power, according to most scientific estimates, are fewer than 10. Nuclear power does not appear to merit its number-one risk rating. It appears that the two well-educated and influential segments of the American public polled in Oregon seem to have been misinformed. The culprits are not difficult to identify. Journalists report almost every incident involving radiation. A truck containing radioactive material is involved in an accident, a radioactive source is temporarily lost, a container leaks radioactively – all receive nationwide coverage, whereas the three hundred Americans killed each day in other types of accidents are largely ignored. Reports in the media concentrate on issues and situations that frighten – and therefore interest-readers and viewers. It fills its coverage with opinions (usually from interested parties) rather than facts or a logical perspective. In terms of nuclear power, for example, phrases such as deadly radiation and lethal radioactivity are common, but the corresponding deadly cars and lethal water would not sell enough newspapers, although thousands of people are killed each year in automobile accidents and by drowning. The problem is highlighted by a two-year study of how frequently different modes of death become front-page stories in the New York Times. It was found that the range was from 0.02 story per 1000 annual deaths from cancer to 138 stories per 1000 annual deaths from airplane crashes.

Misperception of risk fuelled by the media can lead to unreasonable public concern about a hazard, which can cause governments to spend a good deal more to reduce risk in some areas – and a good deal less in others. Governments may, for example, spend huge amounts of money protecting the public from, say, nuclear radiation but are unlikely to be so generous in trying to prevent motor vehicle accidents. They react to loudly voiced public concern in the first case and to the lack of it in the second. But if vast sums of money are spent on inconsequential hazards, little will be available to address those that are really significant.

Examples of where such disparities make little sense are not hard to find. In the late 1970s, for example, the United States introduced new standards on emissions from coke ovens in steel mills. The new rules limited emissions to a level of no more than 0.15 milligram per cubic meter of air. To comply with this regulation, the steel industry has to spend $240 million a year. Supporters of the change estimated that it would prevent about 100 deaths from cancer and respiratory disease each year, making the average annual cost per life saved $2.4 million. It is difficult to claim that this is money well spent when a large scale program to detect lung cancer in its earliest stages, for example, might be expected to extend the lives of 7000 cancer victims an average of one year for £10,000 each and when the installation of special cardiac-care units in ambulances could prevent 24,000 premature deaths a year at an average cost of a little over $200 for each year of life.

Politicians will often sanction huge expenditure to save say an identifiable group of trapped miners, but not to improve mine safety or to reduce deaths among the scores of anonymous miners who die from preventable work related causes each year. Politically, at least, Jo Stalin might have got it just about right when he mused that a single death is a tragedy, but a million deaths is a statistic.

Risk Presentation

The ability to make a rationale assessment of risk is important both for the individual, but particularly for a society which hopes to be governed by sensible and justifiable policies and legislation. It is unfortunate, therefore that, in general, people have both a limited ability to interpret probabilistic information and a mistrust of experts (sadly, statisticians in particular). But rather than dismissing public understanding of technical issues as being insufficient for ‘rationale’ decision making, experts (including statisticians) need to make as much effort as possible to increase the public’s appreciation of how to evaluate and compare risks. Risks might be presented in different ways to make them more transparent. For example, risks presented as say annual fatality rates per 100,000 persons; fail to reflect the fact that some hazards (pregnancy, motor cycle accidents etc) cause death at a much earlier age than others (lung cancer due to smoking for instance). One way to overcome this problem is to calculate the average loss of life expectancy due to exposure to the hazard. Table 1 gives some examples of risks presented in this way.

So, for example, the average age at death for unmarried males is 3500 days younger than the corresponding average for men who are married. This does not, of course, imply a cause (marrying) and effect (living 10 years longer) applicable to every individual, although it does, in general terms at least, imply that the institution of marriage is ‘good’ for men. And the ordering in Table 1 should largely reflect society’s and government’s ranking of priorities for increasing the general welfare of its citizens. Thus, rather than introducing legislation about the nuclear power industry or diet drinks, a rational government should be setting up computer dating services, stressing the advantages of marriage (particularly for men), and encouraging people to control their eating habits. It is hard to justify spending any money or effort on reducing radiation hazards or dietary hazards such as saccharin.

Perhaps the whole problem of the public’s appreciation of risk evaluation and risk perception would diminish if someone could devise a simple scale of risk akin to the Beaufort scale for wind speed or the Richter scale for earthquakes. Such a ‘riskometer’ might prove very useful in providing a single number that would allow meaningful comparisons among the risks from all types of hazards, whether they be risks due to purely voluntary activities (smoking and hang gliding), risks incurred in pursuing voluntary but virtually necessary activities (travel by car or plane, eating meat), risks imposed by society (nuclear waste, overhead power lines, legal possession of guns), or risks due to acts of God (floods, hurricanes, lightening strikes).

A possibility for such a scale is suggested in [2], based on the numbers of people who die each year pursuing various activities. If one person in say, N people dies, then the associated risk index is set at log10 N; ‘high’ values indicate hazards that are not of great danger, whereas ‘low’ values suggest activities to be avoided if possible. (A logarithmic scale is needed because the risks of different events and activities may differ by several orders of magnitudes.) If everybody who took part in a particular pursuit or was subjected to a particular exposure died, then Paulos’s suggested risk index would assume the value zero, corresponding to certain death (life is an example of such a deadly pusuit). In the U.K., 1 in every 8000 deaths each year results from motor vehicle accidents; consequently the risk index for motoring is 3:90. Here are some examples of values for other events:

Paulos’s suggestion would clearly need refinement to make it widely acceptable and of practical use. Death, for example, might not be the only concern; injury and illness are also important consequences of exposure to hazards and would need to be quantified in any worthwhile ‘index’. But if such an index could be devised, it might help prevent the current, often-sensational approach, to hazards and risks adopted by most journalists. A suitable riskometer rating might help improve both media performance and the general public’s perception of risks.

Summary

Given a chance, people would rather not have to confront the gambles inherent in life’s risky activities. Psychological research shows that whenever possible,

people attempt to reduce the anxiety generated in the face of uncertainty by denying that uncertainty, thus making the risk seem either so small that it can safely be ignored or so large that it clearly should be avoided. They are uncomfortable when given statements of probability rather than fact; they want to know exactly what will happen. Tell people that their probability of developing cancer from a seventy year exposure to a carcinogen at ambient levels ranges between 0.0001 and 0.000001 and their response is likely to be ‘yes, but will I get cancer if I drink the water?’ They fail to appreciate just how low are chances of 1 in 10,000 and 1 in 1,000,000. Such values give rise to what Douglas Hofstadter has called ‘number numbness’ – the inability to fathom, compare, or appreciate very big or very small numbers.

The general public’s perceptions of risk are often highly inaccurate, but by underestimating common risks while exaggerating exotic ones, we may end up protecting ourselves against the unlikely perils while failing to take precautions against those that are far more dangerous. For example, people may be persuaded by sensational news stories that chemicals and pesticides considerably increase the risk of certain types of cancer. Perhaps they do, but the three main causes of cancer remain smoking, dietary imbalance and chronic infections. Statisticians, psychologists and others need to put more effort into finding ways of presenting risks that enable them to be more rationally appraised and compared. But this is unlikely to be easy since people tend to form opinions rather quickly, usually in the absence of strong supporting evidence. Strong beliefs about risk, once formed change very slowly and are extraordinarily persistent in the face of contrary evidence. Risk communication research must take up this challenge if the public are ever going to be persuaded to be rational about risk.

References

[1] Paulos, J. (1994). Innumeracy, Penguin, London.

Further reading

Walsh, J. (1996). True Odds, Merritt Publishing, Santa Monica, California.

The BMA Guide to Living with Risk. (1990). Penguin, London.