INFORMATIVE OPINION

Selection Bias in Computer Bulletin Boards
Posted by Martin on Tuesday, 16 December 1997, at 6:59 p.m.

To read this Board one would think that Boxsters are loaded with mechanical problems, Porsche dealers and salespersons are scoundrels, and the ordering and delivery process is fraught with hazards and mistakes. The actual experiences of people are almost certainly much more positive than this, though. The likely truth is that most Boxster owners have few mechanical difficulties, most customers have a pleasant and nonproblematic relations with their dealers, and for most people the ordering and delivery process goes smoothly.

The reason for the difference is what research methodologists call "selection bias". It is well known in survey research that when people self select into an opinion poll--that is, *they* decide to participate instead of the interviewer calling them--the number of problems reported is far greater than what is really the case. This makes sense if you think about it. For instance someone is far more likely to post that "My car is delayed, what's going on?" than they would be to post a message saying "I just wanted to let you know that everything about my delivery was on time."

Most studies on this selection bias show the effect to be very strong. In other words self selected samples are not just a little more negative, instead they are typically *way* more negative. This point should be kept in mind in reading this Board. It's an obvious point but one that is easy to forget when you hear so much negative stuff. If we heard from the folks who *aren't* posting the picture would be a lot brighter than the near constant stream of problems we see reported on this Board.

[ Home ] [ User INFO ] [ Boxster Board ] [ Archives ] [ Sponsors ] [ Museum ]
[ Events ] [ Polls ] [ Links ] [ Classifieds ] [ Boutique ] [ Pete's Page ]