Polls: More harm than good?

Most political polls this election season are untrustworthy. How can you take any of them seriously -- when we've seen such dramatic swings once the votes are counted? Take this week's primaries. According to CNN.com, here's where the race stands:

A "poll of polls" calculated by CNN indicates competitive races in both states. The Texas poll of polls of likely primary voters has Obama with a slight lead over Clinton -- 47 percent to her 45 percent, while 8 percent said they were undecided.

In Ohio, the poll of polls showed Clinton at 48 percent, Obama at 43 percent and 9 percent undecided.

"Poll of Polls ..." an interesting concept, until you read the fine print ...

The Texas poll of polls averages five surveys conducted February 26 - March 1: American Research Group, Reuters/C-SPAN/Houston Chronicle/Zogby, MSNBC/McClatchy/Ft. Worth Star-Telegram/Mason-Dixon, Fox News/Opinion Dynamics and Belo/Public Strategies.

The Ohio poll of polls averages four surveys, also conducted February 26 - March 1: American Research Group, Reuters/CSPAN/Houston Chronicle/Zogby, Cleveland Plain Dealer/Mason-Dixon and Fox News/Opinion Dynamics.

The problem being, some of the polls included are known to be complete and utter crap. American Research Group polls are notoriously misleading as are Zogby. You might as well invent your own numbers!

And then there is this analysis of past polls from primaries in comparison to the actual turn out from a blogger over at the dailykos.com:

State Clinton Clinton Obama Obama
(# polls) Avg Actual Avg Actual
SC(20) 27% 0 41% +14
AL (7) 45% -3 44% +12
CA (8) 46% +6 40% + 3
GA(10) 34% -3 50% +16
MO (6) 45% +3 44% + 5
NJ(12) 47% +7 40% + 4
NY (6) 53% +4 36% + 4
TN (6) 52% +2 33% + 8
MD (6) 34% +3 53% + 7
VA (7) 37% -2 55% + 9
WI (8) 43% -2 47% +11

Clinton performance vs average of polls= +1.36
Obama performance vs average of polls= +8.45

The real issue is not the fact that these polling companies are making millions based on surveys that have no real merit. The problem is the fact that people listen to them and begin to buy into whatever the poll "is telling them." Polls were meant to be tools for political campaigns to determine where they were in an election -- now they can be used as opinion leaders and end up giving voters the easy way out. Voting on the perception that your candidate is a "winner" rather than whether or not a candidate actually represents your ideological interests is not the best way to select a leaders of the free world. But most American voters love taking the easy way out.


Dr. Momentum said...

Prognostication is always a risky endeavor. Any method of doing so is going to have to rely on the past to decide what the future holds.

The polls were criticized yesterday on "Meet The Press" as being based on people who were likely voters last time around, for example. When a candidate is changing the playing field, polls based on last time's performances have less meaning.

As "the military is always fighting the previous war," so are the polls always polling the previous election.

Maggie said...

I'll bet polls have effects that aren't measurable. For example, nervous that Obama might not win one of those states, I might decide to make calls for his campaign. People might or might not choose to vote because of the poll (it's not an emergency so I won't go out of my way vs. crap! he's not going to win, I'd better get there vs. he's not going to win anyway, I won't bother). Apparently it doesn't all come out in the wash, though, since the numbers are so off from what they predict.