I was watching my favorite Bubblegum Journalism channel – CNN – yesterday, when I saw a report indicating that Barack Obama’s strong lead has been diminishing while he has focused the past week on foreign affairs. The key data point put forth as evidence was Obama tanking in the polls in Minnesota. According to the story, his 17 point lead from June had shrunk to only 2 points.
I found this a curious development, even for a state that elected Jesse Ventura to its highest post, so I thought I’d look at the polling data. A good source for polling data is RealClearPolitics.com (RCP), a site that does a nice job of providing daily updates of all kinds of polls. They track national polling and also state-by-state polling of the presidential race. They show individual poll results, and then also provide the “RCP Average” of a selection of recent polls, to give what hopefully is an accurate ongoing snapshot. Turns out, CNN has been conveniently cherry-picking poll results.
[UPDATES! See the bottom of this post for late-breaking updates.]
Looking at Minnesota, the current RCP Average is Obama at +5.3 (meaning Obama leads by 5.3 percent). Examining the polling breakdown reveals the Quinnipiac poll dated 7/22 giving Obama a 2 point lead. This contrasts with the Quinnipiac poll dated 6/24 where Obama had a 17 point lead. Apparently, CNN had it right! But wait: RCP also lists a Rasmussen poll dated the same day (7/22) as the Quinnipiac poll, with Obama at +13. The last 5 Rasmussen polls, going back to April, have put Obama at +14, +15, +13, +17, and +13, a pretty consistent set of results given the margin of error. It seems the bizzare Quinnipiac fluctuation might say something more about Quinnipiac’s methods between its two polls than anything else.
But the story here really isn’t either Quinnipiac’s polling methods or the fortunes of the Obama campaign. The story is CNN’s apparent habit of selecting polling results in order to create ongoing drama in a critical presidential election. Surely the professional journalists at CNN had access to the same Rasmussen results mentioned above, but made a deliberate decision to ignore them, presumably because it would take some of the “pop” from their story if they gave the full picture.
The Minnesota result is only the beginning of what appears to be a concerted effort by CNN to “keep things interesting”. Looking at the CNN national polling map, we see Virginia identified as red, meaning McCain is winning, and the poll connected to that designation (VCU) shows McCain up by 8 points. Yet the RCP Average for Virginia reveals Obama at +1. The VCU survey cited by CNN was done on May 12-18. RCP lists no less than six polls that are more recent. Five of them give Obama the lead, while one gives McCain the lead by a single point. Clearly this is a very close race, but why did CNN feel it necessary to quote a poll that is over two months out of date and well outside the statistical norm of the other polling? Are they too lazy to have looked at other polls, or are they intentionally selecting a certain poll merely for effect?
What about Colorado? The CNN map paints it red, and shows McCain up by 2 points on the strength of a Quinnipiac poll dated 7/22. To CNN’s credit in this case, the Quinnipiac poll is at least the most recent one identified by RCP. However, the RCP Average puts Obama at +1.7. RCP identifies nine other polls, eight of which show Obama with the lead except a Rasmussen poll in March where Obama and McCain were tied. The most recent Rasmussen poll, dated 7/22, had Obama at +3.
What about Nevada? CNN paints it red based on a Mason-Dixon poll dated June 11. They ignored the most recent poll identified by RCP (again, Rasmussen) with Obama at +2. This is a 5 point improvement from the previous Rasmussen poll from June.
Are we seeing a pattern here?
What about the critical state of Florida? CNN paints it red on the basis of an ARG poll dated 7/21 giving McCain a 2 point edge. The RCP Average for Florida reveals a tie. The most recent Rasmussen poll (7/22) shows Obama at +2. The most recent Quinnipiac (6/16) shows Obama at +4. To be fair, the RCP site is not yet showing the most recent ARG poll cited by CNN, and only shows ARG’s poll from June which had Obama at +5. The question this begs is: why does CNN have the most up to date polling only when it favors McCain? Why does CNN ignore more recent polling only when that polling favors Obama?
What about Indiana? This is another close race that needs an honest analysis. CNN paints Indiana red on the basis of a poll done in APRIL by the South Bend Tribune (a journalistic juggernaut?) giving McCain an 8 point lead. Yet the RCP Average for Indiana is Obama +0.5. The surveys used in that average include the poll cited by CNN, but also three other polls, all of which give Obama an edge, by +8, +1, and +1.
What about New Mexico? Here, CNN simply punts, stating “No Polling Information Available”. RCP gives eleven different polls conducted by two pollsters (SurveyUSA and Rasmussen). The most recent 5 polls, dating back to May, give Obama +6, +3, +8, +0 (tie), and +9, respectively. Why could The Most Trusted Name in News not find these polls as easily as I could?
All of these examples show CNN identifying states as red where the preponderance of polling suggests they should be blue (or neutral in the case of Florida). Are there counter-examples where CNN reported a state as blue when RCP data suggest it should really be red? No, there is not a single such example. Assuming CNN is not politically biased, this evidence suggests they are trying to help the underdog in order to keep the story as “hot” as possible. This is news creation, not news reporting. It is sensationalism, not journalism.
On a national level, the CNN “poll of polls” gives Obama a +3 edge. By contrast, the RCP national polling average gives Obama a +5 edge, identifying the following polls dated between July 13 and July 26 (all in favor of Obama): ABC/WaPo +3, CBS/NYT +6, NBC/WSJ +6, Fox News +1, Rasmussen +5, Gallup +9.
RCP also has a nice electoral map which allows the reader to play “what if” games. Based on the polling data, they identify each state as solid blue, leaning blue, sold red, leaning red, or tossup. On this basis, ignoring tossups, Obama leads in electoral count by 238 to 137. The magic number is 270, leaving Obama 32 votes short of victory. Of the 12 identified tossup states, Obama would reach the magical 270 by taking as few as two of the 12, e.g., Florida plus New Mexico, or Ohio plus Michigan, or Ohio plus Virginia, or Florida plus Colorado. If all voting went according to current polling in these tossup states, Obama would win 322 to 216 (that’s even giving Florida to McCain). For McCain to win, he will need a clean sweep of all of the major tossup states – Florida, Ohio, Michigan, Indiana, Virginia, North Carolina, and Missouri.
Perhaps not surprisingly, based on what is revealed above about CNN’s practices, their own electoral map works hard to try to keep things looking more competitive. They give 221 votes to Obama and 189 to McCain (38% higher for McCain than RCP’s data would seem to justify). Their basis for this allocation is self-servingly vague: “CNN’s Electoral College estimate is based on several factors, including polling, voting trends, and ad spending.” A more honest answer, perhaps, is that CNN bases its reporting less on objective facts and more on whatever it deems is most useful in keeping the public tuned into their infotainment channel long enough for them to see their advertisers’ wares. Perhaps they should change their slogan to: “The most sold out name in news.”
UPDATE #1: As of this morning, CNN has updated its “poll of polls” to show Obama’s lead increasing to 6 points. Meanwhile USA Today, my other favorite Bubblegum Journalism source, is giving McCain a 4 point lead. It seems that both claims can be true at the same time thanks to the wonders of sponsored polling.
USA Today points to a new Gallup/USA Today poll of 791 “likely voters”. This poll was conducted from 7/25 to 7/27. A Rasmussen tracking poll conducted over the same period, and also of “likely voters”, has Obama at +3. Rasmussen polled 3000 likely voters, a sample size almost four times larger than the Gallup poll, but of course USA Today won’t report on that competing poll because they were co-sponsor of the Gallup poll.
Interestingly, Gallup also conducted another poll over the exact same period, of 2674 registered voters, and that poll gives Obama an 8 point lead! How to explain an amazing 12 point swing between two Gallup polls conducted at the same time? Obviously it must hinge heavily on the definition of “likely voters” versus “registered voters”.
Who is a likely voter? In this poll … that was determined by how much thought people have given to the election, how often they say they vote and whether they plan to vote in the election in November.
In other words, a likely voter is whatever the polling body decides is a likely voter. This kind of fast and loose approach to polling makes it very easy to tweak the results to show whatever your sponsor (USA Today, in this case) wants to see in its headline.
Journalism takes another kick in the groin, this time thanks to USA Today instead of CNN.
Update #2: It gets funnier. USA Today asked Gallup editor Frank Newport about the polling differences. Newport said “registered voters are much more important at the moment, because Election Day is still 100 days away”. So, naturally, USA Today decided to lead with the less important likely voter figure rather than the” much more important” registered voter figure. Amazing.
Newport went on to explain the polling difference this way:
As for the difference between the [Gallup] tracking and USA TODAY/Gallup polls, [Newport] says not to read too much into it. “Statistical noise” may be largely to blame.
Statistical noise? Statistical noise??? Gallup reports the error margin for the likely voter poll at +/- 4 points, and the registered voter poll at +/- 2 points. That means their professional statisticians believe the largest possible statistical differential between the two polls is 6 points. Yet, Gallup’s own editor shrugs off a 12 point differential as merely statistical noise? Doesn’t that suggest that the stated error margins are pure nonsense?
Update #3: In order to clarify the point of this post, the CNN Electoral Map image has been replaced by the RCP Electoral Map image, and a link has been provided to the CNN Polling Map. See the attached commentary stream for further explanation.
Update #4: Another good site offering broader polling data may be found here:
Note that this site supports the points made in this post.
(Recommendation by gummitch)