AR-Sen: A Quick Review Of Poll Methodology

    24

    I was perusing some of the other blogs when I noticed this statement from Blake in regard to the DK/R2K poll:

    Lt. Gov. Bill Halter leads Sen. Blanche Lincoln 49% – 45% in a new poll conducted by Research 2000. The margin of error is +/- 4%, so it’s statistically a dead heat.

    That’s not exactly correct.

    A 4% MOE (assuming a 95% confidence interval) means that, if the poll were re-conducted, Halter would be somewhere between -4 and +12 against Lincoln (i.e. Halter would be somewhere between 45 and 53, while Lincoln would be between 41 and 49) 95% of the time.  Assuming an even distribution, the mid-point of Halter’s margin within the confidence interval is +4 (of course). This is important because that means the area to the left of the mid-point on your distribution contains +3, +2, and +1, all of which would translate to a Halter win.  Translation: If this poll were re-run multiple times (and excluding the 2.5% of the time that Halter scored below 45%) we would expect Halter to win more often than he loses.

    That’s not a statistical dead heat.

    The race is close, no doubt, and we cannot definitively say that voters prefer one candidate over the other, but we shouldn’t extrapolate from that lack of a concrete answer that Halter does not have an edge.  He most certainly does.