Comments on: Calibrate Your Ad Response http://www.overcomingbias.com/2007/02/calibrate_your_.html Overcoming Bias is economist Robin Hanson’s blog, on honesty, signaling, disagreement, forecasting, and the far future. Sat, 11 Feb 2012 23:23:58 +0000 hourly 1 http://wordpress.org/?v=3.1.3 By: Jeff Lonsdale http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421657 Jeff Lonsdale Tue, 06 Mar 2007 04:14:58 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421657 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? 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?

]]> By: Robin Hanson http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421656 Robin Hanson Thu, 01 Mar 2007 08:44:28 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421656 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. 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.

]]> By: Jeff Lonsdale http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421655 Jeff Lonsdale Thu, 01 Mar 2007 02:42:50 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421655 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. 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.

]]> By: Robin Hanson http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421654 Robin Hanson Tue, 27 Feb 2007 19:11:47 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421654 Paul, I said the point is very general, and follows from basic information theory. Variance can easily be good for decisions. Paul, I said the point is very general, and follows from basic information theory. Variance can easily be good for decisions.

]]> By: Paul Gowder http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421653 Paul Gowder Tue, 27 Feb 2007 14:05:02 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421653 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... 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…

]]> By: Robin Hanson http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421652 Robin Hanson Tue, 27 Feb 2007 02:42:50 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421652 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. 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.

]]> By: Adrian Tschoegl http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421651 Adrian Tschoegl Mon, 26 Feb 2007 20:50:14 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421651 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? 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?

]]> By: Mark Nau http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421650 Mark Nau Mon, 26 Feb 2007 16:19:11 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421650 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. 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.

]]> By: Robin Hanson http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421649 Robin Hanson Mon, 26 Feb 2007 15:10:09 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421649 Adrian, for an ad that is a costly signal, it is the existence of the ad that raises your value estimate, not its content. Adrian, for an ad that is a costly signal, it is the existence of the ad that raises your value estimate, not its content.

]]> By: Adrian Tschoegl http://www.overcomingbias.com/2007/02/calibrate_your_.html#comment-421648 Adrian Tschoegl Mon, 26 Feb 2007 14:41:34 +0000 http://prod.ob.trike.com.au/2007/02/calibrate-your-ad-response.html#comment-421648 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. 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.

]]>