How Safe is Safe?

Guest Post by Bill Pomfret (first posted on CERM ® RISK INSIGHTS – reposted here with permission)

Just how safe is “safe”? Should working in a chemical plant have the same level of risk as skydiving (which kills about 40 people per year in the U.S.A Should working in a plant be as safe as driving your car? Or should it be as safe as flying in a plane, which is safer than driving a car by two orders of magnitude?

While the term FAR may be simple to understand and may represent a useful yardstick, many companies, especially in the U.S., are unwilling to put such targets in writing. Imagine walking into company XYZ’s plush world headquarters office and on the wall in the reception area is a sign that reads, “We at XYZ consider it tolerable to kill 4 people per 100-million-man hours.” The lawyers would have a field day! However, as we shall see, some organizations have established such quantified risk targets.

People’s perception of risk varies depending on their understanding or familiarity with the risk. For example, most people are familiar with driving. The perceived level of risk is relatively low to most people, even though approximately 45,000 people die every year in the U.S. alone due to traffic accidents. If a new chemical facility is being proposed nearby a residential area, the level of understanding of the residents regarding the chemical process will probably be low, and their discomfort, or perceived level of risk will no doubt be high, even if the process may have a very good historical safety record.

Perception of risk will also vary in proportion to the number of possible deaths associated with a particular event. For example, of the 45,000 traffic fatalities every year, the deaths usually occur one, or a few, at a time. Even with this surprisingly high number of deaths, there’s little (if any) public outcry that something be done to lower this figure. Yet when there’s an accident with a school bus involving injuries to many children there typically is an outcry. The same could be said about relatively high-risk sports such as skydiving, hang gliding, and ultralight aircraft. Although these sports involve relatively high risk, it’s rare that one hears of multiple fatalities. The people involved also made their own conscious choice to partake in the activity and outsiders are generally not exposed to the risks. Accident’s involving the chemical industry, however, frequently do involve multiple fatalities. Bhopal, India was the worst to date with over 3,000 deaths and 200,000 injuries. The overall risk associated with working in a chemicals plant can be shown to be less than the risk of driving (at least in the U.S.), yet the public’s perception of the risks of the two activities is typically reversed.

Voluntary vs. Involuntary Risk

There’s a difference between voluntary and involuntary risks. Examples of voluntary risks would be driving a car, smoking cigarettes, and so on. Examples of involuntary risks would be having a chemical plant built near your home after you’ve lived there several years or secondary smoke from other peoples’ cigarettes.

People can perceive similar risks differently. For example, the Jones own a house in the country. One day, company XYZ builds a toxic chemical plant nearby. After the plant is built, the Smiths buy a house next to the Jones. Both households face the same risk, but they’ll probably each have a different perception of it. To the first couple (the Jones) who lived there before the plant was built, the risk is involuntary (although they obviously could move). To the second couple (the Smiths), who bought their home after the plant was built, it’s voluntary.

People are usually willing to face higher voluntary risks than involuntary ones. For example, when one of the authors was younger, he was willing to accept both the risks of riding a motorcycle in a large city and of skydiving. The risks were voluntary, and he considered himself to be the only one at risk at the time. (One could argue the finer points of that.) As a married father, he no longer wishes to accept those risks (never mind the fact that he can no longer afford them).

Another factor involved in the perception of risk is control. For example, the wife of the author does not like flying. (In fact, flying is the number two fear among Americans. Public speaking is number one!) Her stated reason for discomfort is that she doesn’t feel “in control.” When you’re sitting behind the wheel in your car at a stop sign and a drunk driver plows into your car, you weren’t in control then either. After all, no one goes out planning to have an accident.

Tolerable Levels of Risk

The concept of acceptable or tolerable levels of risk is not solely a technical issue; it involves philosophical, moral, and legal matters. Deciding how safe is safe enough can’t be answered by algebraic equations or probabilistic evaluations. Alvin Weinberg has termed these “trans scientific questions,” for they transcend science.

One issue that presents difficulties is trying to statistically estimate extremely unlikely events. Estimating very rare events, such as a severe chemical accident, cannot have the same validity as estimates for which abundant statistics are available. Because the desired probabilities can be so small (e.g., 10-6 per plant per year), there is no practical means of determining the rate directly. One can’t build ten thousand plants and operate them for one hundred years to tabulate their operating histories. Putting it a simpler way, measuring something that doesn’t happen very often is difficult.

Tolerable Risk in the Process Industries

It’s common to view personal risk in a subjective, intuitive manner. Many people won’t consider driving a motorcycle, no matter how wonderful their biker friends say it may be. The author’s wife who doesn’t like flying believes it’s all right for her family to drive to the airport in the same car but not to fly in the same airplane (even though she understands that flying is two orders of magnitude safer). Logic doesn’t always apply when evaluating relative risk.

We should not, however, have the same subjective attitude about risks in the process industry. Usually, the people making the risk decisions (e.g., engineers) are not the ones who will be facing the eventual risk (e.g., workers or nearby residents). Although none of the more famous accidents in the process industry would ever be considered “acceptable,” the companies involved did not go out of business the next day. Therefore, the losses must have been considered “tolerable.” How many accidents might the industry be willing to consider tolerable? How many accidents must take place before there’s public and political outcry?

There are approximately 2,300 petrochemical plants in the United States alone. If an average of one were to have a catastrophic event involving over 20 fatalities every year (which represents an individual plant risk of 1/2,300 per year), how long would it be before there was a public outcry and the government stepped in? What if such an accident only happened once every ten years (1/23,000 per year)? There is no such thing as zero risk, but it’s very difficult to decide what level should be considered “tolerable.”

The relevant statistics can be rather confusing. An individual risk of 1/ 2,300 per year means that out of 2,300 plants, on average, one might go “boom” every year. It’s important to realize, however, that you can’t predict which plant and you can’t predict when one will go “boom.” But since people don’t build 2,300 plants all at once, or live next to 2,300 plants, they want to know the risk of the one plant they’re associated with. The risk for an individual plant remains the same, 1/2,300 per year. However, some are just not comfortable with such a number. Some twist things around a bit and say the risk of an accident is “once every 2,300 years.” This causes even more confusion. Some then assume it will be 2,300 years before there’s an accident and, therefore, they have nothing to worry about. Nothing could be further from the truth. For example, approximately 1 in 4,000 people in the U.S. die in a car crash every year. If you go to a sports event with 4,000 people present, one can assume that someone will die in an automobile accident within the next 365 days. However, you can’t predict which person and you can’t predict which day. The error associated with inverting the number and stating that you’ll live 4,000 years before you die in a car crash should now be obvious.

Deciding what level of risk is tolerable could be compared to choosing your weapon in Russian roulette—how many barrels do you want in your gun? Would you choose an automatic pistol that always had a round in the chamber? (I hope not, although there have been such Darwin award winners!) Or would you choose a revolver with six chambers? What if you could choose a gun with fifty barrels or one with five thousand barrels? Obviously, the more barrels there are, the lower the risk of hitting the bullet when you pull the trigger. Either way, you must play the game. You do, however, at least get to choose your weapon. There is no such thing as zero risk.

One in 2,300 per year, or the incorrect reciprocal of 2,300 years between accidents, may initially sound so remote as to be of no concern to most people. A more intuitive answer might initially be fifty years. (The reasoning being that someone’s working life is fifty years, and they don’t want anything to happen during their life.) But how long will the plant be around? Let’s say 25 years. So, take a gun with fifty barrels and pull the trigger once a year. What’s the likelihood of hitting the bullet? 50%. (Although based on another set of assumptions and simplifications, the answer is 40%. Even the statistics involved are not inherently obvious!)

Would you want to have your name on the drawings of a plant if you knew there was a 50% chance of a catastrophic accident happening during the life of the plant? Probably not.in the range of 1/1.000,000

What if instead of 50 years, you choose 500 years? Then the risk would now be 25/500, or a 5% chance. Should that be tolerable? What about 5,000 years? Now it becomes 0.5%. There is no such thing as zero risk, but just how low must one go for the risk to be considered tolerable? That’s the proverbial $64,000 question to which there is no scientific answer. U.S. industry cannot afford to blow up an entire plant, involving dozens of fatalities, once every year (1/2,300). The negative press and public outcry would be ruinous. After all, OSHA stepped in and produced 29 CFR 1910.119 after several major Gulf Coast accidents happened over approximately a ten-year period. A serious accident once every ten years might be viewed as tolerable (1/23,000). Risk targets in the range of 1/10,000 have been documented. In fact, in The Netherlands, the government even publishes what it considers to be tolerable fatality rates. Switzerland and Singapore have done the same.

BIO:

Dr Bill Pomfret; MSc; FIOSH; RSP. FRSH;
Founder & President.
Safety Projects International Inc, &
Dr. Bill Pomfret & Associates.
26 Drysdale Street, Kanata, Ontario.K2K 3L3.
www.spi5star.com      pomfretb@spi5star.com
Tel 613-2549233