Senate sits at 52-48 in our model as Texas moves toward Talarico

Sixty-four days out, our generic ballot average is D+6.7, drawn from 22 polls fielded within the last 45 days. That is a shade wider than the D+6.6 we carried yesterday.

Our Senate numbers give Democrats a 52 percent chance of winning a majority and Republicans 48 percent, with a mean of 50.3 Democratic seats. The asymmetry in the thresholds matters: Democrats need 51 seats for a majority, Republicans 50, because the vice president is a Republican and a 50-50 chamber goes to the Republicans. In the House, our model gives Democrats a 94 in 100 chance of winning a majority, against 6 percent for Republicans, on a mean of 241.1 seats where 218 are needed.

One note on how that House figure is built. Our simulation draws correlated error at four levels — national, census division, state, and each district on its own. It does not include a term linking demographically similar districts in different places, because we have no cluster data for one. That missing layer means our House control probability is marginally firmer than a fuller model would produce. We are not putting a number on the gap, because this run does not measure it.

Six races moved on the board. Our AK-1 estimate went from a 75 percent Democratic win probability to zero, a drop of 74.8 points and the largest single change in the set. Texas Senate rose 11.5 points, from 55 to 66 percent, and TX-15 crossed the midline, from 35 to 51 percent. Moving the other way, our Ohio Senate estimate fell 10.3 points to 42 percent, and OH-7 slipped 8.6 points to 49 percent. Iowa's Senate race, which we wrote about yesterday, now sits at 66 percent after a 42.1-point move.

Recent state polling is thickest in the Senate. Texas Public Opinion Research, surveying 1,000 likely voters August 21-24, had James Talarico at 48 percent and Ken Paxton at 42. EPIC-MRA released two results from the same August 22-28 field period among 600 Michigan likely voters, with Abdul El-Sayed at 47 to Mike Rogers's 43 in one and 50 to 45 in the other. In South Carolina, Impact Research found the Republican and Democrat Annie Andrews level at 41 percent apiece in samples of 700 and 800 likely voters fielded August 18-24, with 5 percent for others in the smaller version. The University of New Hampshire, polling August 20-24, had Jack Reed at 51 to Raymond McKay's 31 among 750 Rhode Island likely voters and Ed Markey at 49 to John Deaton's 30 among 881 in Massachusetts. In OH-1, Quantus Insights polled 509 likely voters August 21-24 and reported Greg Landsman at 48 to Eric Conroy's 45 in one version and 47 to 43 in the other; both are close.