Earlier this month, the Democratic Leadership Council released a new poll purporting to show that Democratic voters support the DLC’s “centrist” agenda. This is the latest salvo in a pissing match that’s been going on for months between the DLC and its “progressive” rival, the Campaign for America’s Future. The two camps are battling over the soul of the Democratic Party: Should it become more conservative, or stand resolutely on the left?
Not content to argue on merit, each side has hired a pollster to prove its position is not only right, but also popular. The CAF has hired Stan Greenberg, one of the “liberal” consultants who helped elect Clinton in 1992 and then got the boot after the Democrats lost big in 1994. The DLC has hired Mark Penn, one of the “moderate” consultants who replaced the “liberals” in 1995 and helped re-elect Clinton. Pollsters have a whole bag of tricks with which to conduct contests like this one. Here’s how Penn and Greenberg play some of them.
1 Rig the questions. Overt bias, such as asking people whether they oppose “corporate welfare” or “special rights for homosexuals,” is kid stuff. The true art is to rig the question so deeply that no one can see it. There are several ways.
Stack the deck. If you’re afraid that people won’t deem the benefit of your proposal to be worth the cost, don’t mention the cost. Example: Greenberg asks people whether they’d like to “create” bonds to improve infrastructure. Better yet, while ignoring the cost of your proposal, assert the cost of the alternative. Example: Penn asks, “Do you think it would be better to stay within the existing [Social Security] system–for example, raise payroll taxes and cut benefits–or would it be better to move toward more structural changes to Social Security like letting people control portions of their own retirement savings?” While neglecting to mention the risks involved in privatization, Penn defines the alternative as a tax hike and (not or) a benefit cut.
D ictate the dilemma. If your opponent says his policy will achieve a popular goal, force people to choose between the policy and the goal. Example: Penn asks, “Do you think the role of government is to redistribute existing wealth or to foster conditions that enable everyone to have a chance to make a higher income?” If you favor opportunity for all, you have to oppose redistribution. Penn also asks, “Do you think that a better trade policy would limit our trade with other countries and be protectionist, or would a better trade policy be more aggressive at opening up markets for our goods and increasing trade among nations?” If you favor opening foreign markets, the question forces you to oppose trade restrictions–even though the best argument for trade restrictions is that they’re necessary to pry open foreign markets.
Skew the spectrum. To prove that Americans want a government of the center, Penn repeatedly asks people to choose between a left-wing idea, a right-wing idea, and an idea in between. Guess which idea wins? This is like being asked how you want your steak cooked: Most people take the middle option. Much of the dispute between the DLC and the CAF comes down to whether there are two political choices (left, right) or three (left, right, center). By offering three alternatives, the DLC poll begs that question. It certainly doesn’t address the reality of the current two-party system.
Define the alternative. Greenberg asks people why they voted for Clinton instead of the Republicans. This causes them to focus on the Republican themes Clinton rejected (cuts in Medicare, education, and the environment) and overlook the ones he co-opted (crime, family values, budget-balancing). Penn asks people about Clinton’s differences with old-style Democrats. This causes them to overlook the traditional Democratic positions they count on Clinton to defend.
2 Massage the data. Once the survey is completed, the pollster can organize the results in a way that favors his client. Since the public favors “moderation” but doesn’t agree on what it entails, the DLC uses the word without defining it: “Do you generally consider yourself liberal, moderate, or conservative?” In answer, 38 percent of Democrats say they’re moderate, 16 percent conservative, and 36 percent liberal. From this, Penn concludes, “Most Democrats are moderates and conservatives.” It could equally be concluded that three-quarters of Democrats are moderates or liberals. But since the DLC has already succeeded in defining “moderate” as the opposite of “liberal,” its spin prevails.
Rather than let voters decide whether they’re “moderate,” Greenberg offers them various policies and sorts their answers into categories, defining the “moderate” group as narrowly as possible. People who say they chose Clinton on the basis of his support for welfare reform, fighting crime, or a balanced budget–31 percent of the sample–are classified as voting for “moderation.” Those who chose Clinton because he supports Medicare, education, or the environment–59 percent–are classified as voting for “domestic programs.” Greenberg spins this as a 2-to-1 mandate for liberalism. What about the 16 percent who voted for Clinton because of “his support for family values and parents,” or the 8 percent who liked “his support for tax cuts for college”? Weren’t those voters endorsing “moderation”?
Penn plays the same trick in reverse, by inventing “Suburban Values” Democrats, who care primarily about “better, safer schools, safer streets, and the need for strengthening traditional family values.” Penn distinguishes these “values issues”–including family leave–from the “economic issues” that supposedly preoccupy liberals. These “Suburban Values” voters double the DLC’s constituency from 28 percent to 56 percent of the party. But aren’t education, family leave, and the breakdown of community “economic” issues? And are voters who worry about such issues really rejecting government and liberalism? Penn’s taxonomy cheats the “liberal” vote just as Greenberg’s taxonomy cheats the “moderate” vote.
3 Grease the presentation. Any point you failed to win by rigging the questions and categories can be cleaned up in the “executive summary” (the pollster’s spin) and the press release and news conference (the client’s spin on the pollster’s spin). Greenberg’s survey, for example, asks people who voted for Clinton to pick from a list of possible reasons why they did so. Thirty percent cite, among other reasons, “his support for education.” In the CAF press release, this magically becomes a mandate for “increased spending on education.”
When the data contradict his theory, a pollster knows what he must do: Dump the data. Penn and the DLC have been peddling the theory that Clinton did better than congressional Democrats in 1996 because he was less liberal. So in his survey, Penn asks people whether they regard Clinton and the congressional Democrats as liberal, moderate, or conservative. Surprise! More people apply the term “liberal” to Clinton than to the Democrats, and more apply the term “conservative” to the Democrats than to Clinton. This unfortunate result goes unmentioned in Penn’s summary, and the DLC’s press release begins as though it never happened: “Democratic rank and file voters are following President Clinton into the vital center of the national debate, according to a new survey.” The next day’s Los Angeles Times goes with the DLC spin: “Clinton’s Centrist Path: Big Draw Among Party Faithful.”
The point isn’t that polls are dishonest. The point is that they disguise argument as science. Which has a curious–and perhaps salutary–effect. Politicians skew their politics to suit the polls. Meanwhile, their pollsters skew the polls to suit their clients’ politics. One profession’s sophistry foils the other’s cowardice. Just don’t tell them.