Statistical Lies


Census Working Overtime The press loves to bury us in public opinion polls, but something about these pie charts tastes funny to me. By Daniel Price.


Statistics are easy to misuse. The only way to verify the use of statistics is to look at the entire study -- how data was collected (is it really random), how data was recorded and how it was analyzed. Many people do not understand statistical methods and misuse research, even if the research is valid.


"There's much undisputed evidence that married people are, on average, wealthier, healthier, and better educated than their single counterparts. Even the novice student of statistics will tell you that association does not mean causation. -- that being poor, unwell and uneducated is as likely to discourage marriage as the other way around. Nevertheless, the notion that tying the knot can alleviate poverty and bring about positive social change has become the central justification for the Bush Administration's push for low-income women to get and stay married... promot[ing] the institution in its narrowest sense...

"...the Fragile Families study offers evidence that income facilitates marriage; an increase of one dollar per hour in men's wages in the study increased the odds they'd marry by 5 percent. Men who earned $25,000 or more in the past year had more than double the rates of marriage of those who didn't...[hence, based on research, the best way to encourage marriage would be to raise the minimum wage]

"Nevertheless, the Administration is eagerly pushing ahead, willing to recruit any bit of research to its cause, no matter how flimsy or biased" (Sharon Lerner. "Marriage on the Mind." The Nation, July 5, 2004: 40-41).


"He concludes, rather dismissively, that "Mexican marry Mexicans." What Huntington does not report, although he has the data in his hands, is that the intermarriage rate for third-generation-plus Latinos in Los Angeles County was 57 percent! In other words, over half of the third-generation-plus in Los Angeles, the epicenter of his feared reconquista, is marrying outside the group. Huntington deliberately ignores contrary data" (David Montejano. "Who Is Samuel P. Huntington?" Texas Observer, 8/13/04: 12-13, 41).


"In his book Damned Lies and Statistics, sociologist Joel Best recalls the worst figure he ever came across on the printed page: that the number of American children gunned down has doubled each year since 1950.

"That may not seem so dubious at first glance, but Best knew better. He realized that if even just one child was gunned down in 1950 and you doubled that figure every year, the final death tally for 1995 would come in at exactly 34,359,738,638 American children. Of course thatís more kids than America, the world, and human history have ever produced, much less shot. Iím also pretty sure that if 34.3 trillion American children had been gunned down in 1995, we would have heard something about it from the news media, or at least from Rosie OíDonnell.

"Best got the statistic from a student's dissertation proposal. The student got it from a leading sociological journal. The journal got it from the Children's Defense Fund, who actually claimed that the number of American children gunned down each year has doubled since 1950. In other words, it only doubled once (unlike our population, which has more than doubled since 1950). Same words. Slightly different order" (http://www.abusedbythenews.com/columns/col004.htm, 8-24-04).


Polls

Polls have become a key instrument in the battle for public attention. Yet opinion polls are generally misleading. Even the best polls, those that use reliable methods, usually conceal vital truths from the public either by omitting, hiding, or oversimplifying important information (Phillips, Peter, and Project Censored. Censored 2004: The Top 25 Censored Stories. NY: Seven Stories Press, 2003: 211).

"It is commonly accepted in public opinion research that 70 percent or more of those contacted will refuse to participate in polls. With that single act, the refusers destroy the claim that a poll sampled people randomly because the results of any poll can honestly reflect the views of the general population only if the 70 percent who refuse to talk have near identical views with the 30 percent who agree to participate...

"Polls usually report out a statistical "margin of error" for their results.. But this margin of error statistic that makes polls look highly accurate is, in essence, a cover to hide the 70 percent who refused to participate.." (212).

"..self-awareness that her views were outside the "norm" [often] cause [people] to refuse to participate...

"Fewer African Americans and Latinos agree to be polled in most national samples.." (213).

Another key to polling fraud is that polls cost big money. Who pays? What influence do they have? "..there are strong financial incentives for polls to provide a simpler picture, one which validates the sponsors and the government" (Phillips, Peter, and Project Censored. Censored 2004: The Top 25 Censored Stories. NY: Seven Stories Press, 2003: 215).


Colby Glass, MLIS