TiVo's STOP||WATCH service is helping bridge the technology gap, and Comcast is working on a competing service.
By the way, the sample sizes Nielsen uses are not small. They're actually very large, with commensurately remarkably small margins of error. If you want to spend a semester in the applied mathematics class I taught 25 years ago, I could help you understand why that is the case.
The
legitimate concerns about the the sampling for ratings is
not with regard to the sample sizes. Rather, the legitimate concerns are that there is too much self-selection in the sampling. People are randomly selected to participate (that's good), and then they choose whether or not to do so (that's bad). The TiVo system has two sources of sampling bias: The fact that the population of TiVo subscribers, itself, represents a very substantial amount of self-selection (the decision to buy a TiVo), and then on top of that they allow subscribers to opt-out of the data collection, and that introduces another level of self-selection bias. The Comcast system would exhibit the same biases.
Remember, though: Ratings aren't really about how many people are watching programs -- there is no real value in that information -- but rather how many consumers are watching commercials and thereby making purchasing decisions based on that advertising. Therefore, the biases (mentioned above) that tend to make ratings less accurate in measuring "how many people are watching" actually tend to make ratings more accurate in measuring the "how many prospective customers are watching". So effectively, those biases tend to make the ratings less accurate in measuring the unimportant while more accurate in measuring the important.
2009 Ratings Discussion
Linear Mode
