by Dave Zornow
CSP Magazine, September 2001
Picture this: You just got a call from your plumber. You live in an old house, and you know your infrastructure has a few problems…but you've lived with them long enough and you know what to expect. Now, because of a change mandated by the city, you need to overhaul your heating and sewage systems and virtually everything else that travels through your pipes. Your plumber says not to worry…you'll probably never notice the change. But that comforting reassurance isn't leaving you feeling any more comfortable.
Network cable and TV is about to get a major plumbing overhaul. And cautious researchers want to be sure that Nielsen doesn't accidentally connect the waste outlet to the kitchen faucets.
Nielsen has proposed weighting the national people meter sample for the first time in US TV history. Weighting is necessary evil in most survey research. Although researchers use sampling to draw a small representative group of respondents from a larger population, there are bound to be disparities because it's hard to randomly draw a group that properly represents the diverse age/sex/race/geography/etc. of the population. Weighting adjusts for sampling shortcomings by creating weights for each respondent. The sum of each groups' weights, divided by the sum of all of the weights, should then match the group's proportion of the total population.
Up until now, Nielsen never believed that their national sample needed to be weighted. But all of this will change in September 2002.
Why does this matter to national cable ad buyers and sellers? If the same golf foursome played together for 35 years without a handicap, how would it affect scores and the competitive nature of this tightly knit group if we suddenly given handicaps? Now, lets put a little side money on the game, say several billion dollars in network TV advertising. Add the tough advertising environment of this year's upfront, and you'll see why researchers are scrutinizing Nielsen's methodology.
Is weighting a bad thing? Not according to the Media Ratings Council, the watchdog group that oversees research standards compliance for companies like Nielsen, Arbitron and MRI. They have been on record for several years recommending that Nielsen stop waiting and finally weight. Weighting promises to correct under-representation for hard to recruit segments (ethnic households, younger households, larger families) and adjust the weights for groups which are currently over-represented (for example, older households).
Nielsen's plan includes applying weights for at least eight different variables: These include county size and geography, the age of each head of household (less than 25 years old, 25-54 years old, Adults 55-64 and persons 65+), household size (whether a home has 1, 2 3-4 or 5+ individuals), presence of children, race and origin and if the telephone status (yes or no). They predict that the new methodology will only affect most ratings by about a tenth of a point. But that's small solace to a sales executive who is only selling a .3 rating.
What's Good, What's Unknown
Cable network researchers are praising Nielsen for their open approach to discussing the issue. Nicholas Schiavone, NBC's former chief research officer who now consults for the Fox News Channel, described a recent CAB/CONCAM meeting with Nielsen as "one of the most open, productive and methodologically disciplined dialogs" he's had with the ratings company. Nielsen has also hired a highly respected outside statistical consultant to comment on the process of implementation and suggest improvements to the proposed methodology.
The new process also promises to eliminate the "dual standards" of national cable penetration. Traditionally, the official Nielsen national cable penetration number was based on diaries, but cable networks always pointed to the higher people meter number for total cable homes. The new number will use data from Nielsen's national and local market meter households to set the standard.
But researchers remain concerned about the unseen devil in the details. Processes put in place in September 2002 may set standards for the next 35 years…if not longer. Is Nielsen's suggestion of deriving cable penetration data from a subset of the installed local household meter sample - with its 50% cooperation rate - a reliable source of cable penetration information? The weighting variables outlined above make sense today, but will they make sense twenty years from now? How will the composition of American TV viewers change over time, and will today's adjustments hold up over the long haul?
Cable networks will get a little closer to knowing how weighting will affect them this fall as Nielsen releases individual network comparisons showing the affects weighting would have had on February 2000 cable network viewing. A second release of comparative data is set for January 2002.
In your old reliable house with its old reliable plumbing, you could always count on the steam pipes to bang in the winter and the shower to get cold if the dishwasher is also running. It wasn't perfect, but you always knew what to expect. Network researchers feel the same way about their ratings: they know the strengths and weakness of Nielsen current methodology and what they should expect. What worries them are the potential new quirks that may come out a new system and how they will impact the consistency of the ratings currency Nielsen produces. ##
Dave Zornow is President/TNG Research, a media research consultancy and applications development company that works with media sellers and research providers.
Return to: Home • Published Articles