After years of planning and some delay, Nielsen Media Research for the first time last week started weighting its national ratings sample to mirror more accurately the overall U.S. viewing universe.
Specifically, weight controls are now in place in the national ratings panel for roughly 50 demographic and geographic characteristics, including various education and income levels, race, age, and household and county size. Controls for cable and satellite penetration are also in place.
In overly simplistic terms, here's how weighting works. Say next Tuesday Frasier gets a 6 rating among women 18-34. But, for some reason, that demo was underrepresented by 2% in the national ratings sample compared with universe estimates of the actual viewing population. Nielsen's software would automatically add 2% to Frasier's rating in the demo.
In testing over the past year, says David Poltrack, executive vice president, research and planning, CBS, the weighting has had a slight negative effect on household ratings across the broadcasting and cable network universe. He puts the average decline at 1%-3%. But, he says, there has been virtually no impact on the demographic ratings, which are the basis for most TV ad sales.
For those sales that do involve a household rating, he said, everybody has had a year's worth of test data to adjust projections and guarantees, so the impact of the weighting on the sales process should be nil.
Although weighting isn't the preferred way to get a sample in line with the universe being measured, "it's now a necessary thing to do," says Poltrack.
A decade ago, such a move would have been seen as heretical in research circles and a grave distortion of the ratings service's randomly selected household sample (known collectively as the "Nielsen families"), which, in a perfect world, perfectly reflects the universe it is selected from.
But it's not a perfect world. And it has been made even less perfect over the years by the onslaught of pesky telemarketers, phone surveyors and others who have chipped away at the increasingly scarce downtime the average person has at home.
As a result, Nielsen has found, people are a lot crankier when they are called and less likely to agree to become a Nielsen family.
Thus, the research community has largely abandoned its opposition to weighting and, in fact, has been united in urging Nielsen to implement a weighting system for several years.
Researchers caution, though, and Nielsen agrees that weighting isn't a substitute for a good sample. It ought to be used largely to provide stability to the numbers when households fall out of the sample due to technical reasons such as equipment malfunctions or electrical storms.
Or even the recent blackout, although that's an extreme case, when 15% of the sample was knocked out and Nielsen effectively scrapped three days' worth of ratings.
"It just has to be watched and monitored very carefully," says Tim Brooks, executive vice president, research, Lifetime, "because, anytime you start adjusting numbers, there is the opportunity for mischief there."
Alan Wurtzel, president, research and media development, NBC, agrees. "The less you weight, the better," he says. "It's like seasoning in a good meal. A little goes a long way. If you start to overweight to correct for things you should have been correcting for in the field, that's when a problem develops."