Multi-Quest International for John Fleming (10/6-8, likely voters):
Paul Carmouche (D): 43
John Fleming (R): 48
(MoE: ±8.3%)
Check out that margin of error — it’s a monstrous 8.3 points. Now, the pollster claims that the MoE is lower (6%, according to Roll Call), but the poll’s sample size is a minuscule 140, making a 6% MoE mathematically impossible. (Update: In the comments, MichiganLiberal notes that it would be possible to have an MoE that low for this poll, but doing so would require using a significantly lower confidence interval than what is normally used by reputable pollsters. Either way, it’s a junky poll.)
Republicans are hyping this poll in response to a Kitchens Group survey for Carmouche that showed Dems in the lead by 10 points. If this is their best evidence of Fleming’s strength, they have a lot to feel embarrassed about.
Meanwhile, Roll Call gives us a taste of the air war in the 4th:
While the Democratic Congressional Campaign Committee has released an ad attacking Fleming’s Social Security plan and his support for a national sales tax, the National Republican Congressional Committee is going after Carmouche’s record as a district attorney.
The NRCC’s attacks are focused on the case of John Pilinski Jr., who Republicans say was allowed to go free after numerous arrests and convictions because Carmouche failed to prosecute him under Louisiana’s habitual offender law.
“Paul Carmouche, soft on crime, wrong for us,” the NRCC ad states.
The Pilinski stuff reeks of typical of out of context Republican BS, but it remains to be seen whether or not it’ll stick. The Carmouche campaign is pretty livid over this one.
and the sample size is 140? That means they found only 67 people voting for Fleming and 60 voting for Carmouche?
This poll really tells us only one thing: the GOP is scared they will lose one more seat.
That if a campaign released an ad or statement that is a complete lie about their opponents, the said opponents have the legal right to go over to the offices of campaign 1 and punch the guy in the mouth, period. Trading words of outrage back and forth really won’t stop anyone from continuing the lies. I’d go 24 on their ass.
That said, again my opinion of internal polling has been made before. But if there is evidence that it’s mathematically impossible for the Fleming camp to get such a number, I would tend to believe that.
So my statistical explanation:
Margin of error is usually computed on a 95% confidence level – the distribution will tend to follow a bell curve, but 95% of the time, its within a certain interval. After some work that I won’t get into here, it turns out that MOE = 0.98/sqrt(n) where n is the sample size. So mathematically, its 8.3% (or more precisely 8.28%) for a sample size of 140.
So in this case, the pollster probably decided to inflate the value of their poll by using a much smaller confidence level. For example, if we use a 90% confidence interval instead, there margin of error is 6.9% – but the chance that its outside that MOE is double that of a 95% interval.
To get a margin of error of 6.0% with a sample size of 140, you have to use a confidence interval of ~85%. So there’s a full 15% chance that the actual numbers are outside that 6% margin of error.
Of course, this assumes the pollster actually did the above instead of just making up numbers. Or being terrible at basic math.
Compared to what Paul Broun of Georgia just said about Obama. Man, Georgian’s sure can pick ’em.
Fleming supports a national sales tax and privatizing social security. Carmouche and the DCCC should hit him over this. A national sales tax would kill him in the business community. Remember he is like the third string candidate. He beat Jeff Thompson, the man who was endorsed by the Chamber of Commerce and outgoing incumbent Jim McCrery, so this leaves Carmouche a big opening.
The sample size was 400 not 140 like Roll Call reported. Look, the fact is Fleming can and probably will beat Carmouche. Fleming just keeps on winning so keep up the attacks on the blogs like everyone else has since day one.