We look forward to parsing the results of a series of public polls over the next week.
Here’s one to watch: The University of Virginia’s Battleground Tracker.
The tracker, created in 2016, is a very dry public poll with six minor polls — a live operator interview and three automated surveys — per week. A HuffPost/YouGov poll conducted in September suggested a 3.5 percentage point lead for Clinton over Trump. The major public polls were very close, with the average for CBS News, Monmouth and ABC/WaPo being a point behind.
The Battleground Tracker data — which leaves us in possession of actual numbers rather than major poll averages — may be more reliable than the polls it feeds to you, particularly with weighting. Here’s a chart from the tracker’s data on the impact of national debt policy in the 2000 election: A 1-point lead for Gore would have meant an improvement over the previous election — 4 percentage points for Bush in 2000 and a 5-point lead for Gore in 2000 versus 40-point gap between the two in 2000.
The tracker also has been similarly accurate in congressional races: An average of the tracker’s results (including automated and live-interview surveys) would have given the edge to Virginia’s 4th District Democratic Rep. Donald McEachin over his Republican opponent, state Del. Scott Taylor, by 2 percentage points on Nov. 7, 2016. (Taylor won by 7 points.)
The tracker, as you can imagine, can’t tell us everything. And yet it’s been fairly accurate overall: It’s been slightly more accurate than the average of national pollsters, which we assign to “PRX,” on average. The average also includes public polls outside of PRX.
One caveat: The Battleground Tracker doesn’t take race into account when weighting the numbers (as opposed to polls that take race into account and only weight by age). And the tracker provides more custom data — like average internal polls, more detailed demographics and the average of other surveys conducted in a given state.