We know the general public lies to us: during focus groups, in-store sampling, telephone polling, door-to-door canvassing, and in every other form of public opinion research. Marginal Revolution has been thumbing through (what we consider) an old university text, and unearthed an excerpt from Jean-Jacques Rousseau's public opinion theory that might shed some light. Here's his comment:
This is an interesting public opinion theory. It assumes that aggregate public opinion has two components - a selfish and biased component and a component that produces the "right" public policy. The difference between the optimal policy and the median voter's policy is due not to ignorance or systemic error, but to selfish desires that undermine the provision of public goods.
Ideally, public opinion research would prompt decisions that favour the "right" public policy. Unfortunately, the clients actually paying for the research frequently get in the way. As do genpop participants who lie about their motivations, goals, influencers, favoured goods and preferred policies.
"The correlation between stated intent and actual behavior is usually low and negative," writes Harvard Business School professor Gerald Zaltman in his influential book How Customers Think. After all, he notes, 80 percent of new products or services fail within six months when they've been vetted through focus groups.
That would explain how Coca-Cola thinks C2 will be a success.
Oh - and if you've drunk the POR Kool-Aid and are insufficiently cautious about extrapolating POR results, all those anomalies can be explained through sampling errors, measurement errors, margins of error, coverage error and non-response error. Here's a short primer.