Arbitron's sampling methodology is FAR moreimportantthanMrorMrsDePetro's misdeeds
Sat Aug 23 13:45:05 EDT 2008
<<On Sat, 23 Aug 2008 12:56:25 -0400, "Dan Billings" <email@example.com> said:
> True. But I do not think the comparison to radio ratings is valid. It is
> one thing to ask people who they voted for minutes after they voted. It is
> another thing entirely to track radio listening through self-reporting.
Of course, the PPM is intended to combat this problem (the difference
between recall and actual listening), although it won't do anything
about the tiny sample sizes. (It looks even worse when you do
subgroup breakdown: what's the 95% confidence margin of sampling error
for men 18-34 who listen to AM drive in Portland?)
> In any case, my point was not on the polling methods but the flaw of the
> whole system being based on the honesty of radio and advertising people.
> There is too much money at stake for people not to cheat.
If the sample size were larger, the small proportion of
media-affiliated households in the overall population would prevent
the Mrs. DePetros from having any effect on the results. It would
probably require an order-of-magnitude larger sample. Arbitron's
success suggests that advertisers and sales departments are not
willing to pay for better samples.
More information about the Boston-Radio-Interest