Tuesday, November 03, 2009

Arithmetically Challenged at the New York Times

If Fox Is Partisan, It Is Not Alone

reads the headline on a New York Times article by John Harwood.

The analysis, such as it is, is here

Four years later, amid the Iraq war and President George W. Bush’s re-election campaign, the audience data had shifted. Fox News viewers had become 51 percent Republican and just 30.8 percent Democratic ...

By 2008-9, the network audiences tilted decisively, like Fox’s. CNN viewers were more Democratic by 50.4 percent to 28.7 percent; MSNBC viewers were 53.6 percent to 27.3 percent Democratic; Headline News’ 47.3 percent to 31.4 percent Democratic; CNBC’s 46.9 percent to 32.5 percent Democratic; and Comedy Central’s 47.1 to 28.8 percent Democratic.

Whiskey Fire notes that the fact that audiences are partisan doesn't show that the networks are partisan. It is possible that Republicans just don't like to watch non partisan news.

Harwood didn't write the title but he did present raw numbers without comparing them to proportions in the population. Note that the fraction of self identified Republicans in all the audiences is greater than the fraction in the general population (22%).

The numbers must be scaled to be meaningful. The correct calculation is the probability that a given Reopublican watches Fox divided by the probability that a given Democrat watches Fox. At the moment there are more than 1.5 Democrats for every Republican (says pollster.com).

The article doesn't include current numbers for Fox. The most recent are from 2004 (their audience was trending Republican) I don't know why more recent numbers weren't presented (not really) Assuming the current numbers are equal to the old numbers (Fox viewers 51 R to 30.8 D) and the current population I get a Republican is 2.68 times as likely to watch Fox as is a Democrat.

Using the latest numbers a given Democrat is 9% more likely to watch CNN than is a given Republican and that a given Democrat is 21% more likely to watch MSNBC than is a given Republican. 1.09 and 1.21 are not similar to 2.68. Clearly Harwood doesn't understand arithmetic.

The much more amazing thing is that he claims that CNBC's audience has a Democratic partisan tilt. It is 46.9 D to 32.5 R. So, given the headline, evidently CNBC is partisan (sure) and Democratic (huh ?!?). Arithmetically challenged or not Harwood has to understand that there is something wrong with his analysis.

Adjusting for the different numbers of self identified Democrats and Republicans in the population, I find a given Democrat is 89% as likely to watch CNBC as a given Republican. Harwood got the sign of the tilt wrong.

Look this is really obvious. In the USA more people self identify as Democrats than as Republicans. Therefore if one's audience is half Democratic and half Republican one appeals more to Republicans than Democrats. Raw numbers about audiences do not show "partisan fragmentation" they show that more people in the USA are Democrats and so, unsurprisingly, more people watching cable news are Democrats.

The worthless analysis also describes increased "partisan fragmentation." due to shifts in audiences which are partly the result of the declining fraction of self identified Republicans in the population.

1 comment:

Anonymous said...

Can you pl. have a link for archives in your blog. (I feel one of your Sep. blogs is very profound and I would like to have the opportunity to refer to it later)