Nov. 10, 2002, Rocky Mountain News
by David Kopel
Among the biggest losers on election nights were political pollsters and the media who love them.
The Rocky Mountain News (teaming up with News4) and The Denver Post (teaming up with 9News and 850 KOA radio), produced daily tracking polls in the weeks leading up to the election. While these polls were pretty accurate for some races - such as the Gov. Bill Owens blowout and the U.S. House District 7 nail-biter - they were quite wrong on some other contests, including Sen. Wayne Allard's comfortable re-election margin.
The polls also showed Amendments 28 (mail ballots) winning (plus-12 in the final Post poll, plus-18 in the News), but it lost by 16 percent. Overall, the results illustrate why readers should not put too much faith in polls.
The News reported on more contests, and its last poll showed close races for secretary of state (plus-1 Republican) and state treasurer (plus-6 Republican), but those races finished with 19- to 20-point Republican landslides. Each paper used an excellent firm with extensive experience in Colorado politics: Talmey Drake for the News, and Ciruli Associates for the Post, so the problem can't be ascribed to poor quality work.
Yet the fact is that - despite improvements in computer modeling - even the best pollsters are having trouble accurately gauging the electorate.
One of the core problems is the "nonresponse rate" - the number of people surveyed who refuse to talk to the pollster. Nonresponse rates have risen to very high levels, and with Caller ID and Voicemail, it's easier than ever for households to reject a pollster's attempt at contact. (If a household won't pick up the phone, the refusal doesn't technically count in a nonresponse rate, although the effect is the same.)
Some pollsters try to make estimates to address the nonresponse rate, but this is difficult. In the Denver papers, the pollsters never report the nonresponse rate. This skews the statistical accuracy of reported results. For example, a newspaper article on a poll might say that "the poll surveyed 400 Colorado residents, with a margin of error of 4 percent." This means that there is 90 percent probability that the results from a random sample of 400 Coloradans is within 4 percent of what a survey of every Coloradan (or every eligible/likely Colorado voter) would find. But the problem is that the sample isn't really random; it's a sample of that subset of Coloradans who are willing to talk to a pollster.
The News and the Post would do a service to their readers by insisting that every poll they print include the nonresponse rate, and by always reminding readers how much things can change between polling time and election time.
U.S. Olympic Committee head Lloyd Ward is a member of the men-only Augusta National Golf Club, a private club in Georgia. Ward is being pressured to pressure Augusta to change its membership rules. News coverage (Oct. 21) of the USOC meeting in Colorado Springs reported on the committee's battle over "a firmer ethical stance."
This phrasing assumes that the USOC, in forcing Ward to exert pressure on Augusta to change its rules, is the "ethical" thing to do. Whether men-only clubs like Augusta, or women-only clubs (such as the ladies investment club which the News touted on Oct. 5) are "unethical" is a subject of legitimate debate, but it's not appropriate for a news article to presume to resolve that debate.
Speaking of ethics, my nomination for most annoying columnist in a Denver daily is Randy Cohen, "The Ethicist," whose New York Times column runs in the Sunday paper. Consider last Sunday's column, in which a disillusioned campaign worker reports that her former boss is "rude and condescending." She wonders whether to vote for his opponent, although she disagrees with the opponent on the issues.
"The Ethicist" responds that the writer has "a civic obligation" to vote for the nasty guy who shares her views on the issues. He points to President Lyndon Baines Johnson, who was personally racist, yet whose public policies promoted racial justice.
Cohen's reasoning is plausible, but it's preposterous to call it an ethical imperative. Last Tuesday, Arkansas voters dumped Sen. Tim Hutchinson (who was quite in tune with the state's voters on the issues) because they thought he was personally a bad man, since he dumped his wife of 29 years and married a young former staffer.
Similarly, in New Jersey, polls showed that a large majority of voters were ready to vote against Sen. Robert Torricelli (whose voting record very carefully followed New Jersey opinion) once the voters discovered his habit of shaking down people for lavish personal gifts, and then lying about it.
There's a reasonable debate about whether voters should be guided by ideology or by character. For Cohen to suggest that such voters - including the majority of voters in Arkansas and New Jersey - are unethical people who violate their "civic duty" is absurd. If you want ethics guidance, stick to the common-sense advice columns written by Jeanne Philips in the News and "Kathy & Marcy" and Harriet Cole in the Post.
