Sound-Bite Statistics: 'Yes, but' Is the First Victim

President Clinton brags about a 9 percent decrease in violent crime. Robert Dole blames him for doubled drug use among teenagers. Statistics and politics make a heady mixture. It has been said that figures don't lie, but liars figure.

But this is more a case of figures that are right up to a point and political slogans that oversimplify complicated facts. The 9 percent decrease in violent crime comes from a Census Bureau study called the National Crime Victimization Survey. That differs from FBI figures, based on police reports, that show a 4 percent decrease in crime. As the name suggests, the "victimization survey" surveys not criminals but victims; thus, for example, it does not report on murder for the obvious reason that the victims cannot speak.

But both surveys register an improvement. The question is whether the president deserves the credit or, as Mr. Dole asserts, Republican governors do.

Or do other factors come into play?

For one, police have developed more-effective preventive strategies. For another, more criminals are in prison - three times as many as 15 years ago. And, perhaps most relevant is the aging of the population, with baby boomers past the prime years for committing crimes. But that means that, with the expected 20 percent increase in teenagers, we may see more crime in coming years.

The increase in drug use, reported by the Department of Health and Human Services, led Dole to accuse the president of "a naked failure of leadership." Mr. Clinton, in turn, charged him with short-changing the administration's antidrug program as Senate majority leader.

Here again other factors should be considered. For one, survey questions were drastically rewritten in 1994, and the department is not sure it successfully adjusted for that. Some numbers are not statistically significant. For example, heroin use by teenagers superficially doubled - from 0.3 percent in 1994 to 0.7 per cent in 1995. But the actual number of heroin users in the sample of 4,600 went from 14 to 32.

In several other cases the sample is too small and the margin of error too great to trust the results. Then there is the fact that many of those questioned give confusing answers. For example, some who say they did not use marijuana in the past year then say they did use it in the past month.

In the heat of a campaign, candidates are interested in sound bites, not statistical uncertainties. "Yes, but" is the first victim.

*Daniel Schorr is senior news analyst for National Public Radio.