The Problem with Opinion Polls

Back in December, when people actually had debates about whether or not Chrysler and GM would go bankrupt, one of the claims made by the anti-bankruptcy camp was that 80% of people would not buy a car from a bankrupt automaker. That number came from a CNW survey; here’s a summary from Motor Trend (hat tip Jane Hamsher):

A recent study from automotive market research firm CNW surveyed 6000 people intending to buy a new car within six months, and discovered that more than 80 percent of them would switch brands if the vehicle they wanted came from an automaker that went bankrupt. Breaking it down by company, Americans were more likely to abandon domestic automakers than foreign ones, with Chrysler faring the worst — a full 91 percent of buyers wouldn’t take home an Auburn Hills product if the company went bankrupt.

The theory was that bankruptcy would lead to an immediate collapse in sales which would lead to liquidation. (A later study, cited here, said that if the government were involved in the bankruptcy, the number of people who wouldn’t consider buying from GM would be 51%.)

This is what I said in December:

I strongly suspect that 80% is just a poorly worded and interpreted poll question. If you ask people in the abstract if they would buy cars from a bankrupt car company, of course they will say no. But in the real world, if the car they want is made by a bankrupt company, and they get a good deal, they will buy it. Just look at the November auto sales. GM was down 41%; Toyota, Honda, and Nissan were down 34%, 32%, and 42%, respectively. And everyone buying a car in November must have been aware that bankruptcy for GM was a serious possibility. (Besides, haven’t we been talking about a GM bankruptcy on and off for years?) Sure, bankruptcy will hurt sales a little. But 80% is just not credible.

Well, now we know. In May – during which Chrysler was in bankruptcy – Chrysler sales were down 47% from the year-ago period. Overall sales were down 34%, which means non-Chrysler sales were down around 33%. So as a crude estimate, if Chrysler were like the average automaker, for every 100 cars it sold last May, it would have sold 67 cars this May. Instead, it sold 53. That’s a 21% decrease – a lot less than the 91% predicted.

I’m not claiming I can predict auto sales better than other people – I know virtually nothing about auto sales. I’m just saying you shouldn’t rely on polls that ask people what they would do under some hypothetical scenario, since they don’t know what they would do.

• Become A Funded Futures Trader