Disregarding third-party candidates, Professors DellaVigna and Kaplan found that towns that offered Fox by 2000 increased their vote share for the Republican presidential candidate by 6 percentage points (to 54 percent, from 48 percent) from 1996 to 2000, while those that did not offer Fox increased theirs by an even larger 7 percentage points (to 54 percent, from 47 percent).
You would think that with the title of "Fair? Balanced? A Study Finds It Does Not Matter" the New York Times article would go on to say that it does not matter even if a news source is biased people will still get information from the news cast. Well, if you thought that then you were expecting way too much from the New York Times.
This article does a craptacular job of analyzing flimsy data to draw a conclusion that is not even implied by the data. What the data boils down to is that in cities where Fox News is aired on cable it turns out that one percent more of the population voted for Kerry than it did for Gore.
So, at best, it shows that Fox News is in fact not biased (although not really), But really the data just means that more people voted for Kerry than Gore in these cities, and that is it. There was no replication in the study, and there is NO connection to say whether a biased news channel makes a difference in the Presidential voting anyways. There are so many flaws that even a person that failed my stat class could point them out.
In short this article would, at best, get a D in my class, and that is mainly do to effort. Obviously the author has no education in statistics, or if he does then he must have bought (?) his way through the class.
The kicker? The author is a Professor at Princeton. I thought they employed smart people there...
p.s. Statistics is like power, in the wrong hands it is a very bad thing.
3 comments:
This "study" (har har) is just full of holes.
My prof has a great quote on his computer... "Data will tell you anything if you torture it enough". I thought that was a fabulous quote...
That is true, but there is right and wrong things to do in statistics.
Deceptive or ignorant people can do anything they want with data, but (for the most part) every set of data has only one conclusion, and truthful statisticians should always agree on that conclusion. So you can't just draw any conclusions with your data, you can be wrong.
Post a Comment