And it's mostly bungled:
... Note especially the final line in the excerpt below. It’s a lovely, concise summary of the difference between declared and revealed preferences — and why, if you’re at all interested in describing how the world actually works, you should put all your energy into chasing the latter and ignoring the former....
Sales of Green Works have fallen to about $60 million a year, and those of other similar products from major brands like Arm & Hammer, Windex, Palmolive, Hefty and Scrubbing Bubbles are sputtering. “Every consumer says, ‘I want to help the environment, I’m looking for eco-friendly products,’ ” said David Donnan, a partner in the consumer products practice at the consulting firm A. T. Kearney. “But if it’s one or two pennies higher in price, they’re not going to buy it. There is a discrepancy between what people say and what they do.”
Here is what I wrote when I read that line:
Differences between what people say and what they do is called hypothetical bias. The cool thing is there are ways to mitigate and even eliminate the bias so that data from what people say they will do can be used to make more accurate forecasts.
Yet, the offending line is partly correct. If all you care about is understanding how people behave under usual conditions then you want revealed preference data. But if you are interested in how people might behave under unusual conditions (e.g., when policy might change environmental quality) and there is no variation in usual conditions then you need stated preference data. And, I know Freakonomics is pushing field experiments as the only way to understand behavior under unusual conditions and that is where the "all your energy" stuff comes from but that is either narrow minded or plain ol' rent seeking. Open minded social scientists understand that a picture becomes more clear as the types of data and theories applied to a problem increases. Field experiment, stated preference, attitudinal, etc, data are complements, not substitutes. If stated preference data is cheaper than conducting field experiments then the rational researcher should choose stated preference data.
Here is another statement of that in my SuperFreakonomics review, motivated by the guys dumping on lab experiments.