ARTICLES

SOPHISTICATED POLLING TACTICS CAN GIVE CAMPAIGN A BOOST
Majority Rules! Magazine - March 1993

by Frank Noto

Women candidates have to work harder, longer, and smarter to get elected, and that includes the realm of polling.

One sophisticated analytical tool that a woman contender can use to her advantage is called causal modeling. This special tool can provide a candidate an advantage by enabling her to identify the compelling arguments that influence voters in the final months of a campaign.

Most candidates settle for what is known as cross-tabulation analysis-a system that identifies the popular arguments. But knowing which messages are compelling – that is, which messages have the power to change attitudes and influence voting behavior – can make the difference between winning and losing. That's a deeper level.

The popular arguments are easy to identify. All you have to do is look at the cross-tabulations to see which arguments voters agree with. Survey cross tabs report the issues that people say are important to them, as well as the arguments that they say they agree with. This system, common and very useful in the political analysis industry, will not tell you which arguments are the most compelling and persuasive.

Consider President Bush's effort to paint family values as a critical issue only to find voters-in the aftermath of the Republican National Convention-say that this is not as important as the economy and jobs and health care.

Few voters can accurately relate to a pollster the factors that influence their political attitudes and behavior. Fewer can provide a valid answer when asked what it would take to change their vote.

Hooking up with a polling firm which can give you this depth of analysis can add to your arsenal.

The system of causal modeling is akin to looking at the electorate with a different lens – an effort that many candidates and managers won't make. But women candidates are accustomed to working harder and smarter, and to making the extra efforts to gain a level playing field. This system fits right into that experience.

Casual modeling is based on what is known as multivariate regress techniques. This classroom term covers such technical fields as multiple regression and logit regression. But multiple and logit regression are single-step approaches to measuring the influence that issues have on voter intent. Causal modeling is a multi-stage technique that uses a variety of methods to measure influences on a voter.

In other words, if cross-tabs provide the essential arithmetic of survey research analysis, multivariate regression is the algebra approach. Causal modeling is trigonometry. Each approach is a higher level to the same game: math.

Cross-tabs might show that voters who favor a ballot measure are particularly likely to hold certain perceptions. Multivariate regression will then identify those issues that have the greatest direct impact on vote intent. Multiple regression sifts and winnows, separating the wheat from the chaff. The results identify the key perceptions that influence voters and rule out those that are coincidentally related to vote intent.

At the highest level, causal modeling goes one step beyond; this cutting edge analytic technique graphically depicts the influence process. Causal modeling will identify the key messages to be communicated and the supporting messages that are necessary to insure that the voter will act on those key messages. This could be viewed as the architecture involved in building a solid, winning campaign organization.

Casual modeling will give a candidate a deeper understanding of the opponent's vulnerabilities than can single stage analysis. While multiple or logit regression identifies the opponent's strongest and weakest arguments, causal modeling will reveal ways you can take advantage of those weaknesses best, and attack the strengths from a position of strength.

If the artistry involved in transforming the telephone calls made by your pollster into meaningful information sounds complicated, consider causal modeling as an even greater distraction.

Don't let it become that; go to an expert. Start by considering only those firms with a range of expertise in multivariate analysis as well as causal modeling. These pollsters will also know how to analyze cross-tabs. If you already have a pollster, have a serious discussion with them about what advantages you can attain by moving into a poll that shows causal modeling.

If you confront a close race, this additional boost of information can help you target your precious resources-financial and personnel - and perhaps eek out a victory.

If you opt for a causal analysis, realize that it is not just an analysis. It is something that should begin with the questionnaire. Make sure the firm has an experienced research analyst who has developed causal modeling polls. Remember, you can't turn a sumo wrestler into a ballet dancer. These fields are just that different.

In causal modeling, survey questions that use a numerical scale are generally more useful than the traditional "strongly agree" to "strongly disagree" range of answers. "Push" questions don't work at all; each argument should contain only one thought.

If properly applied, these simplified questions will enable a causal model to relate various ideas and messages directly to the "cause" and a voter is swayed. Causal modeling is the kind of technique that can only identify those arguments included in the questionnaire – not those omitted. Ask the questions you believe represent the range of possible arguments on an issue. Get to that point by working with the research analysts on background information, the opponent's literature, press clippings and everything else you can lay your hands on that focuses attention on meaningful issues in your race.

If this were early in a campaign, you may use a focus group to identify issues. Mid-stream, less than six weeks to Election Day, you are trying to refine your campaign – not launch it.

Once you start, you'll be hooked. Causal modeling can be used as part of the tracking studies to monitor the extent of attitudinal shifting as well as relating the information to the initial benchmark survey.

The cost of a survey with causal modeling is about $4,000 more than what some pollsters will charge for simpler forms of multivariate analysis. But, in a tight race, each dollar should be measured by the impact it will have on the bottom line: the number of voters who can be swayed. Once you believe that causal modeling can help you better target, and win the election, you will be a believer.


Frank Noto is the vice president of the San Francisco-based public affairs firm, GCA Strategies. His firm specializes in controversial land use projects across the nation. For more information, e-mail Noto, call him at 415-391-4100 or visit the GCA Strategies Web site at www.gcastrategies.com.