"Stephanie Weil" wrote in message
oups.com...
On Mar 23, 8:50 pm, "David Eduardo" wrote:

Yep. They could probably do it with much less and still get MRC
certification, but because the People Meter is new, they are using a very
large sample.

David, but for a huge city like New York, where we have eight million
plus people....don't you think a better sampling of the public's
listening habits can be made by increasing the amount of people
surveyed? I'd say a minimum of a million?

The first issue is cost. To do the current sample costs each station
somewhere around a half-million a year, give or take. To do one million
people meters would cost several thousand dollars a year per participant,
or, in essence, more money than the total ad revenue for the whole market.

The reason we use polling techniques where we take a sample of the
population is sometimes compared to a blood test. It only takes a few drops
to determine the characteristics of all the blood in the body. If you
oversample, there is no increase in accuracy in both the blood test and in
radio ratings.

A sample works if you get a few people of every kind. That means in every
age cell, each sex in every age cell, each ethnicity in every age and sex
cell, etc. If your sample is proportional to the number of people in each
group, all you have to do is project the results into the total population.
That means if 2% of NYC Metro is Hispanic females between 25 and 34, then 2%
of your sample should be in that group.

One of the tests of any statistical sample is whether if done again (called
"replication") with the same technique on the same dates with the same
sample design but different people, will the results be the same. The answer
is that the Arbitron sample replicates easily within the acceptable margin
of error. Remember, this is not a horse race. Being "first" in radio does
not mean the other stations lose... there are multiple winners, and a few
percent margin of error is acceptable to advertisers and stations.

The New York radio market is 15.2 million persons 12+, but when you sample
such a universe, you look for proportionality by geography (counties and
boroughs), age, sex, ethnicity and, among Hispanics, language dominance. If
each subset is proportionally polled, then the result will reasonably
represent everyone to the extent required by the end user of audience
research, the advertiser.

Radio and TV ratings are audited annually by the Media Research Council, a
group of statisticians and academics who principally report to the ad
industry. In addition, any change in methodology requires MRC review to
maintain accreditation.