Behavioral targeting: A new polling tool?

In any democracy, opinion polls serve the purpose of informing the public on how people, or certain groups of people, think about the key issues of the day, and about the public figures who make the decisions that affect their lives. In any free society, the information that polls generate provides the basis for making good judgments and reasoned individual choices.

That being said, polls—and the organizations that conduct them—are not without their detractors. For example, it is often claimed that pre-election polls tend to influence people’s decisions on whom to vote for, or indeed, on whether or not to vote at all. It is also argued that polls tend to be self-fulfilling (due to what is known as the “bandwagon effect”). These criticisms cast pollsters as powerful influencers of public opinion and disruptors of the political dynamics by which free individual choices translate into public policy.

To be fair, there seems to be meager empirical evidence to support these contentions, and legitimate pollsters and census takers continue to enjoy their deserved credibility.

I do have one serious misgiving about polls, however.

Polls attempt to gauge public opinion on the raging issues confronting society by surveying random samples of individuals, mainly through personal interviews. Surveys are conducted over specific spans of time, and the survey results reflect conditions that obtained during these polling periods.

In today’s complex, dynamic and highly unpredictable world, these so-called “parametric conditions” tend to shift by the hour. There is therefore a distinct possibility that survey results will turn out to be “wrong,” not because of faulty analysis or flawed field procedures, but because they have been overtaken by events.  This is a major reason why many newly developed products turn out to be duds, or why lackluster political candidates emerge as surprise big winners.

Even the most sophisticated polling procedures cannot predict what public opinion will be at any time in the future. They can only establish with reasonable certainty, or “confidence levels,” what it was over some discrete period of time in the immediate past. True enough, past poll results may also indicate discernible trends or patterns over time. However, projecting these observed trends into the future is an extremely risky business for the reason that complex social systems have a tendency to veer—violently at times—in some unexpected directions. As Steve Jobs famously said, you can connect the dots looking backwards, but not looking forward. It is for this reason that making plans on the basis of past observations may lead to unexpected consequences.

The Promise of Behavioral Targeting

Behavioral Targeting is a technique used by big merchandisers and advertising firms to determine the effectiveness of their marketing campaigns. The enormous amounts of data needed for this purpose are developed by mining digital information on the search behavior of millions of unwitting individuals and by surreptitiously latching onto their social-media postings, email messages and Internet downloads and uploads. It is also used extensively by police and national security organizations in tracking down the behavior of unsuspecting individuals in order to predict potential criminal or terroristic acts.

Used in conjunction predictive analytics and highly sophisticated computer simulation models, researchers can evaluate the potential effects of seemingly innocuous behavior of individual actors on the larger system of which they are a part.  For example, a class of computer simulation models called agent-based models (ABMs) is used extensively by academic researchers in studying the potential impact of individual actions on the behavior of markets, political movements, supply chains, and other forms of social networks. This software can be used to test how actions and choices at the individual level transform into emergent patterns of behavior of entire social, political and economic systems.

It is our belief that polling organizations can vastly enhance the accuracy and usefulness of their craft by combining their time-tested methodologies and procedures with these newly emerging data-gathering and analytical techniques. By so doing, not only will they be able to give accurate representations of how people perceived their worlds at certain point in time in the past, but effectively give a running picture of how their opinions are currently emerging.

In a word: gauge public opinion in real time.

(The article reflects the personal opinion of the author and does not reflect the official stand of the Management Association of the Philippines. The author is a former professor of Management in UP Mindanao. Feedback at [email protected] and [email protected]. For previous articles, please visit map.org.ph)