Science junk hits the Washington fan

-- Back in December 2000, President Clinton and Vice President Gore were busy fellows- what with dishes to pack, furniture to ship and an election to contest. So busy were they that they neglected to read some of the fine print in a cascade of administration-ending paperwork. One of these was an obscure item called the "Federal Data Quality Act" (FDQA), which was dutifully signed by the president.

Put simply, the FDQA prohibits the use of junky science in the promulgation of federal regulations and laws. And, now that the new hats are in town, it shouldn't be much of a surprise that the FDQA is being turned against the "science" of the Clinton-Gore team, particularly concerning the global environment.

Specifically, it has been turned against the "U.S. National Assessment of the Potential Consequences of Climate Variability and Change" (USNA), a document that breaks the cardinal rule of science: If a hypothesis doesn't work, throw it out. The Assessment can't pass the simplest of scientific tests.

The Assessment began with a 1997 letter from Gore to all the federal agencies, and was published 10 days before the 2000 election. If the Office of Management and Budget chooses to apply FDQA, the Assessment will be redacted down the Memory Hole soon.

And none too soon. The power of the USNA's bad science can be seen in recent drafts of Sen. Tom Daschle's (D-S. Dak.) energy bill, where the USNA provides the findings necessary to induce new fuel economy measures and prohibit drilling for domestic oil--all in the name of global warming and its pernicious effects on America.

In fact, that it serves as the basis for legislation is the reason that the USNA has run afoul of the law. The FDQA requires scientific objectivity and normal reproducibility of positive results in any simulation or scientific experiment that underpins prospective regulations. The Assessment has neither.

The Assessment purports to project the consequences of United States warming, produced by two computer models. One is from Canada and the other from the United Kingdom. Both models are extreme outliers. Unlike the consensus of the dozens of available models, the Canadian model produces an exponentially increasing heating. The result is a ridiculous rise of 8.1ºF in projected U.S. temperatures this century. The UK model predicts greater precipitation changes than any other model the USNA looked at.

A horde of peer reviewers--some from federal laboratories that have a track record of global warming doomsaying--told the USNA that the use of these two models was wrong. Even the greens at the United Nations agree that these models can't be used to make local and regional climate projections with any reliability.

How does even the rankest climate amateur know the Canadian model is a joke when applied to the United States? Because it "predicts" that U.S. temperatures should have changed 300 percent more than they did in the last 100 years. In fact, neither the Canadian model nor the British can beat a table or random numbers when it comes to predicting U.S. temperature for the last century.

A climate model is nothing but a statement of scientific hypothesis: What we "think" should happen based upon currently fashionable theory. When a hypothesis doesn't work (i.e., performs worse than a bunch of darts thrown at the Dow Jones), the ethic of science requires that it be thrown out. In this case, it means that the USNA should have used better models, or, absent a defensible model, it should have used none. If a computer simulation of climate can't beat a table of random numbers over the United States, it borders on scientific malpractice to continue to apply it.

It wasn't that the politically chosen leaders of the USNA didn't know there was a problem. In fact, the USNA's politically handpicked steering committee was so disturbed about the finding of the peer-reviewers that it commissioned its own study. Guess what? The USNA's own scientists verified that the temperature models didn't work over the United States. And yet the report went forward, now serving as the basis for the most sweeping energy legislation ever introduced in this nation's history.

Well, anyway, all of these shenanigans are precisely what the Federal Data Quality Act was designed to prevent. The irony is that the obscure piece of legislation that slipped through when Clinton and Gore weren't minding the store is about to throw the USNA and its global warming hysteria into its well-deserved dustbin.

This looks good; perhaps it could be applied to this:

Reading Between the Numbers

Tuesday, February 19, 2002
By Wendy McElroy

A Jan. 23 congressional study claimed that salaries for women managers in seven out of 10 industries examined had declined from 1995 to 2000. Uncritical newspapers rushed to announce that the wage gap between men and women had widened.

But National Review columnist Betsy Hart took the time to examine the study commissioned by Reps. Carolyn Maloney, D-N.Y., and John Dingell, D-Mich. She found it to be a "biased and highly-emotionalized reinterpretation" that the "creative" staff of Maloney and Dingell had imposed upon otherwise straightforward data. The reinterpretation allowed Maloney to label the study explosively as "a wake-up call" for America and to hint at the need for more federal regulation in the workplace.

Hart phoned Maloney's office, identified herself, and spoke directly to the congresswoman who mistakenly assumed the journalist was also a liberal feminist. Maloney explained that the existence of wage discrimination was considered a fact and the study had been a search for the supporting data. Then, in Hart's words, "Maloney ... shared with me her intention to keep the Right from finding out what she and Dingell are up to." After all, she didn't "want to scare the right wing so that they stop collecting data" on women and the workplace.

Anyone familiar with what passes for statistics within feminism will not be surprised by the willful corruption of data. An underlying assumption of data-manipulators is that people are too stupid to notice the sleight-of-hand. The media exacerbates the problem by not asking the most basic questions about statistics, even ones with surprising conclusions. For example, journalists rarely ask, "What is the margin of error?" or "Does the average reflect a mean or a median?" Public schools contribute their share by failing to teach the fundamentals of statistical analysis.

It is important to guard against those who twist data and, then, wield the results as political weapons. Every statistic should be required to answer several questions before you accept it:

1. Who says so? This inquires into the possible bias of the researchers. For example, Maloney's staffers might well be biased toward processing data in a manner that supports legislation the congresswoman favors. The source of their income doesn't invalidate what they say, but it does call for taking a closer look at their data.

2. How do they know? Unbiased researchers may employ a sloppy methodology that comes from laziness or error. For example, a much-cited study entitled Prostitution, Violence and Post-Traumatic Stress Disorder, collected data from streetwalkers in four "strolls" that were notorious for drug use and violence. Yet the conclusions of the study comment on all prostitutes, including high-paid call girls. In short, it uses an unrepresentative sample to draw broad conclusions about a general population.

3. What's missing? Always place the data within a proper context. The GAO data upon which the congressional "study" is based openly states its limitations: It does not control for highly significant wage factors such as "years of continuous presence in the workforce." As Hart comments, "studies which do control for all relevant factors continually show that the wage gap between men and women virtually or totally disappears."

4. Does the conclusion make sense? Do not let statistics displace your common sense. Consider a "fact" popularized several years ago by feminist Naomi Wolf: 150,000 American women die each year of anorexia. According to the Centers for Disease Control and Prevention, this would make anorexia the fourth-leading cause of death in both males and females. Yet the CDC missed that data. The grossly inflated number had been taken from a newsletter of the American Anorexia and Bulimia Association, which claimed 150,000 to 200,000 women "suffered" from anorexia nervosa. The actual death rate is closer to 100.

5. Did someone change the subject? Researchers often redefine terms in such a manner as to produce desired results. For example, by the word "rape" most people mean forced intercourse. But feminist studies frequently include all sexual assault under that label. In turn, sexual assault is sometimes expanded to include harassment. Popular statistics — e.g., "one in four female college students will be victimized by rape or attempted rape" — must include the definition of "rape" being used in order to be meaningful.

Finally, and most importantly, remember that a correlation does not indicate cause and effect. A correlation is a mutual relationship between A and B — for example, if one goes up, then the other goes down. A cause-and-effect relationship means that A causes B. Consider the claim that women make 75 percent as much as men for doing the same job. The statement draws a correlation between being a woman and earning power but it says nothing about cause and effect. The 75 percent (if true) may be caused by other factors not weighed by the study. For example, women often leave the workplace to have children. This factor alone may cause much of the wage gap.

In his definitive yet delightfully simple book How to Lie With Statistics, Darrell Huff observes, "The secret language of statistics, so appealing in a fact-minded culture, is employed to sensationalize, inflate, confuse, and oversimplify."

Yet statistics are too useful to dismiss. Instead, the secrecy should be removed.

With Huff tucked under your arm, unafraid "right-wingers" should approach Maloney's statistics, ask how they are funded, whether the conclusions are overbroad, what is their context, how does she define all relevant terms, and is it a correlation rather than a cause-and-effect? Develop this level of skepticism toward data, and four out of three times you won't go wrong.

Wendy McElroy is the editor of ifeminists.com. She is the author and editor of many books and articles, including the forthcoming anthology Liberty for Women: Freedom and Feminism in the 21st Century (Ivan R. Dee/Independent Institute, 2002). She lives with her husband in Canada.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.