Risky Business: When Numbers Mislead

Nicholas Freiling

Think about your neighbors, your classmates, your coworkers, and your extended family. If this brings more than 112 people to mind, odds are at least one of them will be killed in a motor vehicle incident.

That’s according to the National Safety Council, which calculates the average American’s odds of dying from various causes.

But those odds seem high. Many of us have hundreds of acquaintances without knowing any traffic fatalities personally. Can it be true that every American with more than 112 friends is likely to see at least one them killed this way?

Not exactly. That’s assuming your neighbors, classmates, coworkers and extended family represent a fair sample of the American population — that your “average friend” is an “average American.” For any single person, this isn’t likely to be true. Most people’s friends will have some common characteristics — similar religious beliefs, similar lifestyles, similar levels of wealth. These characteristics, among many others, determine actual, personal exposure to risk.

If, for example, your friends are mostly women, you’re less likely to see one of them die in a car accident. Same goes if they’re mostly college educated, over age 50, or possess one of a host of other characteristics that correlate with reduced risk of dying in a car accident.

This caveat applies to all of the National Safety Council’s statistics — and to most other statistics on risk.

Statisticians and risk analysts can’t possibly account for all factors in the likelihood of some event occurring. Regarding motor vehicle incidents, for example, your personal risk would be far below the National Safety Council’s average odds if you rarely drive, or far higher if you never wear a seat belt. So instead of spending years thinking of every possible factor in the likelihood of dying on the road, they take the total number of incident-related deaths, divide by the total number of people, et voila!

But what’s the big deal? Everyone knows statistics like this are just averages. The National Safety Council itself makes that very clear. That’s why we have P-values, margins of error, disclaimers, and so on.

The problem arises when government uses such statistics to create rules and regulations that interfere with people as they go about their private lives.

Mitigating the risk of harm is one of the biggest justifications government regulators use for their interference. Agencies like the FDA, EPA, and OSHA are supposed to examine Americans’ exposure to such risk from sources that fall under their respective auspices. When some level of risk gets too high, the agencies are supposed to propose regulations that mitigate that risk and justify these solutions in terms of costs and benefits to the public.

One example is a proposed FDA regulation that requires food product manufacturers to take measures that limit the likelihood of food contamination. The FDA justifies this rule by citing the fact that more than one million cases of foodborne illness occur in the United States every year. Illness like this costs Americans both time and money, and sometimes even their lives.

But as with any risk, those already taking measures to limit their exposure — in this case, to riskier foods — suffer less often than those who don’t. They pull the average risk down. Americans who don’t eat meat, for example, are far less likely to experience poisoning from foodborne E. coli. Such behavior entails some cost, however. It may require more expensive meals when dining out. It might mean going hungry at a friend’s barbeque and having to buy dinner elsewhere later on. Or it simply might be hard for someone who particularly loves meat.

So when an administration approves this particular regulation, which will raise the cost of producing certain food products, those who already take steps to protect themselves will pay the most. They already pay by forgoing riskier foods. Now they must pay extra when they do occasionally indulge. Even if this regulation cuts the incidence of foodborne illness in half, a disproportionately large percentage of those “saved” will be those who did not take precautions themselves before the rule was implemented — people who did not already pay a premium for food safety. Meanwhile, those who already paid a premium will pay twice.

This same story applies to other regulations. It explains why responsible 20-year-olds can’t enjoy a beer, or why mature 17-year-olds can’t vote. It explains laws outlawing certain types of firearms for even trained, experienced users. It’s the reason (or at least it's the stated reason) behind prohibition of marijuana and other drugs for even those who might benefit from their medicinal uses.

Of course, some will say that inequities like these are the prices we pay to live in a civilized, prosperous society. If we want social stability and sustainable growth, so the story goes, we must improve the average person’s outcomes and lower the average person’s risks — not leave the reckless masses to their own poor decision making. This involves transferring some consequences of risky behaviors from the careless to the careful.

But coerced transfers aren’t the only, or the best, way to reduce average exposure to risk. Market forces are often far better at yielding more safety than regulation. Insurance providers, for example, discourage risky behavior by offering discounts to policyholders who avoid excessive risk. Demand for side-impact and rollover airbags grows every year without further government mandates. And information-sharing, especially in the digital age, almost immediately ostracizes any firm whose deficient products or services, such as contaminated food, put its customers at risk.

These forms of risk management are different because they are adopted voluntarily by individuals who alone have knowledge of what risks they face, and who mitigate risk only to the extent that it improves their unique circumstances. They aren’t based on statistical averages. They impose no cost on people who don’t want them, and allow for innovations that override prior methods of mitigating risk.

That’s something no one-size-fits-all regulator can ever claim.

Related Articles


{{relArticle.author}} - {{relArticle.pub_date | date : 'MMMM dd, yyyy'}} {{relArticle.author}} - {{relArticle.pub_date | date : 'MMMM dd, yyyy'}}