Brian Kernighan on Millions, Billions, Zillions

KernighanNumbers are often intimidating, confusing, and even deliberately deceptive—especially when they are really big. The media loves to report on millions, billions, and trillions, but frequently makes basic mistakes or presents such numbers in misleading ways. And misunderstanding numbers can have serious consequences, since they can deceive us in many of our most important decisions, including how to vote, what to buy, and whether to make a financial investment. In this short, accessible, enlightening, and entertaining book, leading computer scientist Brian Kernighan teaches anyone—even diehard math-phobes—how to demystify the numbers that assault us every day. Giving you the simple tools you need to avoid being fooled by dubious numbers, Millions, Billions, Zillions is an essential survival guide for a world drowning in big—and often bad—data.

Why is it so important to be able to spot “bad statistics?”

We use statistical estimates all the time to decide where to invest, or what to buy, or what politicians to believe. Does a college education pay off financially? Is marijuana safer than alcohol? What brands of cars are most reliable? Do guns make society more dangerous? We make major personal and societal decisions about such topics, based on numbers that might be wrong or biased or cherry-picked. The better the statistics, the more accurately we can make good decisions based on them.

Can you give a recent example of numbers being presented in the media in a misleading way?

“No safe level of alcohol, new study concludes.” There were quite a few variants of this headline in late August. There’s no doubt whatsoever that heavy drinking is bad for you, but this study was actually a meta-analysis that combined the results of nearly 700 studies covering millions of people.  By combining results, it concluded that there was a tiny increase in risk in going from zero drinks a day to one drink, and more risk for higher numbers. But the result is based on correlation, not necessarily causation, and ignores potentially related factors like smoking, occupational hazards, and who knows what else. Fortunately, quite a few news stories pointed out flaws in the study’s conclusion.  To quote from an excellent review at the New York Times, “[The study] found that, over all, harms increased with each additional drink per day, and that the overall harms were lowest at zero. That’s how you get the headlines.”

What is an example of how a person could spot potential errors in big numbers?

One of the most effective techniques for dealing with big numbers is to ask, “How would that affect me personally?” For example, a few months ago a news story said that a proposed bill in California would offer free medical care for every resident, at a cost of $330 million per year. The population of California is nearly 40 million, so each person’s share of the cost would be less than $10. Sounds like a real bargain, doesn’t it? Given what we know about the endlessly rising costs of health care, it can’t possibly be right. In fact, the story was subsequently corrected; the cost of the bill would be $330 *billion* dollars, so each person’s share would be more like $10,000. Asking “What’s my share?” is a good way to assess big numbers.

In your book you talk about Little’s Law. Can you please describe it and explain why it’s useful?

Little’s Law is a kind of conservation law that can help you assess the accuracy of statements like “every week, 10,000 Americans turn 65.” Little’s Law describes the relationship between the time period (every week), the number of things involved (10,000 Americans), and the event (turning 65). Suppose there are 320 million Americans, each of whom is born, lives to age 80, then dies. Then 4 million people are born each year, 4 million die, and in fact there are 4 million at any particular age. Now divide by 365 days in a year, to see that about 11,000 people turn 65 on any particular day. So the original statement can’t be right—it should have said “per day,” not “per week.” Of course this ignores birth rate, life expectancy, and immigration, but Little’s Law is plenty good enough for spotting significant errors, like using weeks instead of days.

Is presenting numbers in ways designed to mislead more prevalent in the era of “alternative facts” than in the past?

I don’t know whether deceptive presentations are more prevalent today than they might have been, say, 20 years ago, but it’s not hard to find presentations that could mislead someone who isn’t paying attention. The technology for producing deceptive graphs and charts is better than it used to be, and social media makes it all too easy to spread them rapidly and widely.

Brian W. Kernighan is professor of computer science at Princeton University. His many books include Understanding the Digital World: What You Need to Know about Computers, the Internet, Privacy, and Security. He lives in Princeton, New Jersey.