« Disagreement Case Studies | Main | Marginally Revolved Biases »

February 26, 2007


"Why would car companies show ads to well-adjusted ad watchers?" You answered your own question with a much simpler explanation than this variance-increasing explanation already: "It can be reasonable to react positively to the fact that a car company choose to show you a car ad, but only if you react negatively when they choose not to show you an ad."
They show you the ad for the car because your reaction will be positive compared to not showing you the ad!

Foolish, yes I intended for that to be one of the possible explanations for ads. However, most people I describe it to do not find it very plausible.

Robin, how do you reconcile your view with the idea that an ad is a costly signal - If I'm lyin', I'm dyin'? That is, as info accumulates via reviews, prospective buyers who test drive it and then talk with their friends, etc., if the car is not better than unadvertised cars, the money for the ad is wasted.

Adrian, for an ad that is a costly signal, it is the existence of the ad that raises your value estimate, not its content.

But, I think I am rationally risk averse when making such a high-expense long-term purchase. So the price I am willing to pay tends to increase as I get more information. At relatively low search cost, I can choose a car that has a 95+% chance of ultimately being worth more than I have to pay for it. Buying a car blindly at my expected value of that vehicle runs a ~50% chance of disappointment, and my 95%-sure price is very low indeed.

Robin, fair point, up to a point. When I lived and worked in Japan I saw many ads on TV that had no product content; in these cases the signal did not extend beyond the ad's existence. However, when the ad makes assertions about the product's attributes, then it is warranting those attributes, and the positive signal effect with respect to those attributes should return, shouldn't it?

Adrian, it can make sense to tell people how to best use a product; this need not change their estimate of the value of the product they will achieve, but still provides useful info.

Robin, can you elaborate on why you'd expect -- or, more to the point, how an advertiser could design -- ads to actually increase the variance without increasing the mean? Increasing the variance without increasing the mean implies a symmetric effect across the population of consumers, but what kind of an advertisement does that represent? One that, for some inexplicable reason, carries both positive and negative information?

Also... Do you know of a model for information-acquisition on the assumption that new information has an expected position shift (and hence an expected value) of zero? If this were the case, it seems like it would be non-optimizing to ever invest in information (the best you could expect would be that it would reduce the variance in any given decision, but since one is making many decisions over a lifetime, the errors should cancel out anyway). But that just can't be right...

Paul, I said the point is very general, and follows from basic information theory. Variance can easily be good for decisions.

Why is it bias when a person reacts positively to the removal of uncertainty? In the stock market, uncertainty makes people value stocks less. Now, I'm not implying that the stock market doesn't have biases, but perhaps this reaction of the stock market isn't irrational. I am suggesting the removal of uncertainty regarding a car's attributes should lead on average to a more positive view without the presence of bias. Collecting information can have a non-negligible cost.

On the other hand, certain types of advertising should be viewed skeptically. When people are willing to give away a lot for the chance to sell something (such as timeshares), then the cost of that sale (accounting for failed sales) should be added to the estimated spread between "true market value" and price that is offered.

Jeff, there are different ways that an uncertainty might be resolved; the news could be good or bad. It would be a bias if they stock were to rise no matter how the uncertainty was resolved.

Robin, if the markets rise on average they may jsut be afraid of a black swan. Given this possibility, how can you see an average rise and conclude bias?

Also, am I correct to understand from your response to Paul that basic information theory does not allow for the information itself to be valued? Having information about a car from an ad reduces the information gathering costs associated with buying that car, and therefore should on average lead to an increase in its perceived value. Why is this wrong?

The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30