High probability of befuddlement

Statistics: The gullible, which includes most of us, should consider the source and method by which the numbers were obtained.

July 15, 2001|By Thomas Ginsberg

TWO YEARS AGO, a provocative study made headlines nationwide: Legalization of abortions in the 1970s had reduced crime rates in the 1990s.

Two months ago, a little-publicized study reached the opposite conclusion: Legalized abortion actually raised the adult murder rate in the 1990s.

Statistics, it seems, rarely have been more ubiquitous in public discourse, nor more questionable and confusing. Whether it's unborn criminals, census undercounts, or divorce rates, the American public and news media have been getting hard lessons lately in "numeracy" - literacy in numbers.

Last month, the Census Bureau started releasing figures showing a huge increase in the number of same-sex households from 1990 to 2000. Last week, it reversed course and told reporters not to compare the two sets of data because they had been "edited" differently.

In April, child-development researcher Jay Belsky ignited controversy by concluding that preschool children cared for by someone other than their mothers are more likely to become aggressive in kindergarten. He is the same researcher who had declared in 1986 that children in day care develop no differently than others.

Forty years ago, demographers projected that Philadelphia would grow by 1980 to between 2.5 million and 3 million residents. But at that moment it actually was shrinking, like most major cities, and hasn't stopped yet.

For officials, journalists and everybody else, the first lesson of numeracy is the same as for assessing anything: Consider the source and method by which numbers were obtained. Some statistics are in dispute, some wrong, some simply unknowable.

"The lesson for all of us is that talking only to people who have a vested interest in some numerical result or linkage can be beguiling," says John Allen Paulos, a Temple University mathematician and numeracy advocate. "Gullible journalism is often the result."

Most statistical myths are probably accidental, muddied through repetition or misunderstanding by uninformed people like a giant game of telephone.

Other "facts" may be repeated with a bias, validating the adage that "there are three kinds of lies: lies, damned lies, and statistics." (Note: Be wary of quotation sources, too. This one has been attributed to Benjamin Disraeli, Mark Twain and an unknown Victorian-era wit.)

Take the abortion-crime link. In 1999, law professors John J. Donohue and Steven D. Levitt said they had proved a statistical link between legalization of abortion and lower crime rates.

They said five states that had allowed abortion earlier than the rest of the nation in the 1970s also saw their crime rates decline earlier than the nation's. And states with high abortion rates also had experienced greater-than-average reductions in crime.

Their explanation: More abortions had led to fewer "unwanted" children, and "unwanted-ness" was a risk factor for becoming a criminal.

The study, sponsored by the private National Bureau of Economic Research, won a place in the headlines and stirred debate over abortion rights.

Now, it has been contradicted.

In May, law professor John R. Lott Jr. and economist John E. Whitley reported their findings that legalized abortion had helped loosen taboos against out-of-wedlock births, in turn leading to more single-parent families and more stressed families, and ultimately higher murder rates - by 0.5 percent to 7 percent.

"There are many factors that reduce murder rates, but the legalization of abortion is not one of them," said the report, published by Yale Law School.

Even if statistical findings are repudiated, they can still perpetuate statistical myths.

For example, the popular belief about marriage remains this: Half of them end in divorce.(Annoying buzzer sound.)

In fact, according to a new analysis of federal marriage data from 1995, the proportion of first marriages "disrupted" by divorce or separation was 20 percent after five years. It was 26 percent after seven years (so much for the "Seven Year Itch").

The disruption rate hit 43 percent after 15 years. But beyond that point the figures are unreliable, said Matthew D. Bramlett, author of the report for the U.S. Division of Vital Statistics.

"I cringe when I hear people say half of marriages end in divorce, because they don't know what they're talking about," Bramlett said.

So where did the half-of-all-marriages fable come from?

"It seems to come from a census report," said Christi Goodman, a marriage and family specialist at the National Conference of State Legislatures.

Goodman said that in some annual Census Bureau surveys, the number of divorces has been about half the size of the number of marriages for a particular year. Sociologists blame that shorthanded math for the misunderstanding.

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.