Of all the various ways that people complain about national media suckitude, there is no type of political media coverage that sucks more than election coverage.
I bet this could be proven as an objective fact. There is an entire cottage industry of useful, accurate election coverage that has sprouted online over the last few years entirely because national political media coverage of election analysis is so downright sucky. Large websites such as Real Clear Politics, Pollster, 538, Electoral-Vote, Daily Kos, MyDD and even Open Left would be a shadow of their current size if not for the abject incompetence in election horserace coverage provided by larger news outlets. And this isn't even to mention the "professional" electoral forecasts provided by CQ, Cook, Rothenberg and others.
What the larger news media primarily lacks compared to the online electoral analysis cottage is context and detail. That is, the broad range of polls and other electoral factors and rarely discussed by media outlets that pay for their own polls, and the drill-down detail on individual states and congressional campaigns.
A perfect case in point came yesterday, when the Gallup poll showed Republicans ahead by 4% in the national House ballot for 2010. This individual poll generated hundreds of news stories across the national media, almost all of them claiming that Republicans were now ahead in the race for the House in 2010. However, what this crappy media coverage failed to tell readers is that the Gallup poll was one of only 18 national house ballot polls produced over the last month, 13 of which still showed Democrats ahead.
(More in the extended entry)
|
One poll, taken on its own, has much more inaccuracy than polling averages. A study conducted by the National Council on Public Polls after the 2008 election indicated that, on average, the 507 polls taken during the last 20 days of the election missed the final margin of the campaign they surveyed by 4.0% (see the candidate margin, and multiply it by two). However, if someone--say, an enterprising online, amateur analyst--had just averaged the polls in each campaign over the last 20 days of the election, they could have produced a mean error of only 2.6%.
Empirical evidence demonstrates that poll averaging reduces error, no matter what deductive arguments against poll averaging are given from time to time. As such, providing the broad context on all polling would mean more accurate election horserace coverage. However, given that many news outlets are paying for their own polls, and that more dramatic headlines are viewed as a means to increase audience share, many new outlets simply will not provide this broader context. Had they done so for the National House Ballot, they would have been forced to tell readers that Republicans have gained ground, but that Democrats still hold a substantial, roughly 4.2% advantage. Like so:
Nov 12: Democrats +4.17
The complete methodology behind this table can be found here. Also, I update the numbers every time a new national House ballot appears. You can see the topline result in the button in the middle column of Open Left.
Republicans have gained, but they are still behind. Further, worries about Republicans outperforming their poll numbers in the national House ballot are unfounded. Across the 150 campaigns for which I have looked at polling averages since 2004, Democrats, not Republicans, actually gain a couple tenths of percent from the final poll to the final result on average.
Since polling methodologies for the National House Ballot are the same as they are for other campaigns, there is no reason why Republicans would perform better from the final poll to the final result in that specific campaign, but not in other campaigns. It a conclusion is akin to flipping a coin seven times, having it come up heads six times, and then concluding that heads is the result on 80% of coin flips. To date, the better performance of Republicans from the final poll to the final result in the House ballot is simply a statistical fluke from a too small data set. When only a dozen or so polling averages are examined, such a fluke can easily occur. However, over time, the results will even out.
I am somewhat torn about pointing this all out. On the one hand, I would actually like it if national election horserace coverage provided more context and detail in an attempt to improve accuracy. On the other hand, as long as they keep sucking, the more likely it is that I am able to make a living.
Speaking of which, our fundraiser is still ongoing. Tim Tagaris and Robert Greenwald have each pledged to give $1 to Open Left for every new follower they have on Twitter, up to $250 each. So, not only will it help Open Left, but you will have two new must reads on your Twitter feed. Please, start following Tim and Robert and Twitter now! |
