One of the ways people can analyze national polls is whether the state polls that are being released are in the same universe as the national polls – whether a national environment of a given size is being reflected in the state polling. Take the New York Times/Siena polling released last month, and you’ll see the perfect example of state polls that make sense with their national result – a D+14 National saw double digit leads for Biden in Michigan, Wisconsin, and Pennsylvania, meaning those states remain a few points right of the nation, as they were in 2016. That makes sense, and those state polls were about what you’d expect for that kind of overwhelming national lead. You can quibble on the margins, but all told, that set of state polling was in the same universe as the national polling. But what does the state polling since say, and is there any evidence of a weakening Biden position in there?
According to an analysis of every* state poll released with July field dates, there is an unweighted 9.1% swing to Joe Biden as compared to results in the state in 2016, which would be in line with a national lead of just over 11%. The asterisk is this excludes Gravis polling conducted for One America News Network and Democratic internal polls conducted for campaigns, because those polls fail to meet standards around functionally and believability. If you expand the exclusion list to include anything from either Public Policy Polling (D), Civiqs (D), or Trafalgar Group (R), which are in the grey zone of partisan public pollsters, you get a 9.3% swing. If you take the best and worst poll from the more restricted average, as they do with Olympic figure skating or diving, you get a 10% swing. All of this points to a double digit lead for the former Vice President nationally, right in line with the national polling we are currently seeing in national polls.
The state polls aren’t really getting any better for the President, either – if you just take the polls with at least one polling date since July 10th, you can squint and get the swing down to 8.4% (or 7.5% with the Olympic averaging method), but that’s partially a function of the CBS/YouGov Texas poll released on July 12th being eligible but the University of Texas at Tyler poll released that same day not being eligible. Small variations will come from how you deal with cutoffs, and the single worst state poll for Trump – Monmouth’s Pennsylvania release – is the most recent state poll listed on Fivethirtyeight by field dates, so even if there was some amount of momentum, it is arguably gone.
The data doesn’t get a lot better for the President if you limit the swing to states decided by 10% or less in 2016, but it does tighten a bit, as double digit swings against the President in Missouri, Alabama, and Alaska go away, but even then you’re still looking at an environment where the incumbent President is losing nationally by 8% – and all of this cherry picked progress for the President is wiped out when you go back to the last two sets of live caller state polls, Fox News’ set released on June 25th and the NYT/Siena polls of the same day, which showed a combined average swing of 9.65% from 2016. (Fox’s average swing of their four states was 8.125% and the NYT/Siena’s average swing was 10.67% in six states, for the record.)
There is nothing in the state polls that are fundamentally telling a different story than the national ones – and, until we get some more new live caller state polls, we won’t be able to see whether this recent uptick in Biden’s lead in national polls is real or not at a state level. There could be a widening in the state and national level results split that led to a Trump win with a popular vote loss, but if Quinnipiac is right, and the President is losing whites by 6%, that won’t happen. There’s still time between now and November, but the President is in trouble, and according to the national polls in further retreat. The bad news for the President is there’s little solace in the state polling.
Evan Scrimshaw (@EScrimshaw) is Managing Editor and Head Of Content at LeanTossup.ca and a contributor to Decision Desk HQ.