A single close poll in GA-7 leaves our forecast unmoved

One new poll of Georgia's 7th District landed since our last run. A 20/20 Insight survey of 445 likely voters, fielded August 20-26 and reported by the Atlanta Journal-Constitution, put Republican incumbent Rich McCormick at 46 percent and Democrat Anthony Kozycki at 43 percent. On a sample that size, a three-point gap is small enough that the poll is best read as showing a close contest rather than a measured lead.

Our forecast did not move. It gives McCormick a 99 percent chance of winning and Kozycki a 1 in 100 chance, with a mean margin of R+21.0 and a 5th-to-95th percentile range running from R+36.7 to R+5.4. The run before this one, at 8:32 AM ET on September 8, had the same probabilities and the same mean margin, with a range of R+36.6 to R+5.4. The new number is the only survey we have in this district, and it is not enough on its own to pull the estimate toward it.

That gap between poll and model is worth stating plainly. The 20/20 Insight result sits well outside our 90 percent interval for the final margin. Our estimate leans heavily on district fundamentals and the national environment rather than on a single 445-person sample fielded two weeks before publication. If further polling in GA-7 comes in at similar margins, the model will move; one poll does not.

Nationally, our average of 24 generic-ballot polls over the past 45 days stands at D+6.2. Our model gives Democrats a 93 percent chance of reaching the 218 seats needed for House control, with a mean of 239.0 seats. One caveat on that figure is ours to own: our simulation draws correlated error at the national, census-division, state and district levels, but we have no data to add a term linking demographically similar districts across state lines, so the House control probability is marginally firmer than a fuller model would make it.