Essay: THE SCIENCE & SNARES OF STATISTICS

AMERICANS believe in numbers. As a democracy, the U.S. chooses its leaders statistically, so to speak, by the simple process of counting votes. Numbers measure the economy, record social progress, identify people on credit card rolls and bank accounts. “In a numerically conscious society,” says Rand Corp. Researcher Amrom H. Katz, “progress is measured by numbers, not by quality.”

In fact, the American appetite for statistics seems insatiable, and the statisticians obligingly crank out an unending supply, ranging from the annual per-capita consumption of paper (540 Ibs.) to the number of dishes (nine) that the typical family breaks in the course of a year. Sports fans are longtime lovers of the well-tempered statistic. To know that Roger Maris replaced Babe Ruth as the home-run king through a fluke in total games played, is to be an aficionado instead of an amateur. For the average American, to be told that a lofting astronaut has threaded a celestial needle of time and place and reached orbit is to be faced with the incomprehensible. But to know that he is traveling at 17,500 m.p.h. is a measure that means something to an earthling who must watch the “60 m.p.h.” speed-limit signs.

But this dedication to numbers has created its own pitfalls for the innocent—and opportunities for the purveyors.

There is an air of certainty about the decimal point or the fractionalized percentage—even in areas where the measurement is statistically absurd or the data basically unknowable. A classic example is a survey made some years ago, which solemnly reported that 331% of all the coeds at Johns Hopkins University had married faculty members.

True enough. Johns Hopkins had only three women students at the time, and one of them married a faculty member. The American Medical Association announces not that very few people dream in color, but that “only 5% of Americans” dream in color. New York City has 8,000,000 rats. How does anybody know? Statisticians have a phrase for this, borrowed from the computer industry on which they now rely. The phrase is “garbage in, garbage out”—meaning that the result that comes out is only as good as the material that is fed in.

Sales & Strikes

For the sake of drama or publicity, numbers are slapped on nearly everything—and the bigger the number the better. During July’s two-day rail strike, the Chicago Association of Commerce and Industry issued an instant statistic that the city was losing $40 million to $60 million a day, into which total were cranked lost railroad fares and freight revenues, reduced restaurant and hotel receipts, smaller store sales, and presumably the money that visiting butter-and-egg conventioneers or traveling salesmen might spend on tours and girls. Overlooked was the probability that most of the businessmen made their visit anyway the minute the strike had ended. “What can you say about a strike,” says DeVer Sholes, the association’s director of research and statistics, “except that they’re striking? But the news media are anxious to build up the story, so you have to fluff it up some way.”

Fluffing for or by news media is the root cause for many an abused statistic. Newsmen during the Detroit race riots pressed a harried fire chief for damage estimates. His guess:

$500 million. So far, in the cooling aftermath of riot, insurance companies are processing only $84 million worth of damage claims, and the overall loss is now put at $144 million. For newsmen, the National Safety Council issues forecasts of expected highway deaths over holiday weekends, usually with a prediction tacked on of “record fatalities.” What the forecast never says is that the record is due to population increases and wider use of automobiles, and that the fatality rate is usually just as high proportionately on other weekends—holiday weekends are just a bit longer.

“The knowledge industry,” as education is now called, is touted as a $200 billion industry—one of the nation’s biggest. Presumably every expenditure, down to janitors’ salaries and the cost of the new gymnasium (also computed as part of the construction industry), is figured in. But then there is the $126.7 billion of Government spending, the $189 billion service industry, the $21 billion annual economic loss through crime, and the $25 billion that Vance Packard says is spent on disposable packages each year. The grand total has soon soared past the gross national product.

Hungry to Bed

In an irreverent study of the numbers game called How to Lie with Statistics, Author Darrell Huff coined the word “statisticulation”—the art of lying with statistics while seeming objective. One trade that statisticulates regularly is advertising. A nostrum for piles or pyorrhea is endorsed by eight cut of ten physicians because the ad agency has tirelessly spent time and money assembling panel after panel until it finds one that shows an eight-out-of-ten result.

Politicians are equally susceptible to the selected statistic.

John F. Kennedy’s campaign claim that “17 million Americans go to bed hungry every night” was based on nothing more than a 1955 Department of Agriculture study on eating habits, which reported that, along with predictable diet deficiencies in lower-income groups, 13% to 17% of U.S.

households with incomes of $10,000 or more also suffered nutritional shortages. In fact, it can easily be proved that most people in a given group make less than the average income. Sample: In a group of ten, nine make $10,000 a year and one has an income of $1,000,000. Their average yearly income is $109,000. This reminds many statisticians, who have some humor about their shortcomings, of the fellow with his head in a refrigerator and his feet in an oven, who declares, “On the average, I feel fine.”

Since it is obviously impractical to poll the nation on anything less important than the selection of a President, one cherished statistical tool is the sample. Not even statisticians can agree on how big or good a sample can be relied upon as representing the whole. Dr. Alfred C. Kinsey’s celebrated reports were criticized by statisticians not so much for their moral implications but because they made sweeping presumptions on the basis of too small a sample (in the male study, only 5,300 men provided data). The Nielsen ratings, by which television programs live or die, have been justly attacked because Nielsen recorders are necessarily hooked to the sets of those viewers willing to have a recorder—a special class by definition, whose tastes may or may not correspond with those of the unpolled millions of the total TV audience.

Still the state of the art of statistics has come a long way since 1661, when its founding father, London Haberdasher John Graunt, began a careful count and found that more boy babies died in infancy than girls, and concluded that therefore there must be more women than men in Britain. Today’s scientists who no longer believe that anything is absolutely certain, also believe that many things are predictably probable. And it is the computer, fed with vast amounts of past data, that can project or at least outline the alternatives of several possible futures. “The computer has enshrined statistics,” says M.I.T.’s Professor Harold A.

Freeman. “Without it, statistics would still be a grubby business.” Where once all they had to do was count, and perhaps draw graphs, statisticians are now “programmers,” with a mystique all their own. Unquestionably, for the moment, numbers are king. But perhaps the time has come for society to be less numerically conscious and therefore less willing to be ruled by statistics.