The emphasis on a wide range of possible outcomes isn’t just a matter of hedging by pollsters and analysts after the shock of 2016, when Mr. Trump beat long odds. The uncertainty reflects the sparse data available and the unusually large number of competitive contests. In dozens of key districts, no more than a few polls have been done, making it hard to be confident whether (and where) Democratic or Republican candidates might hold a lead.
Final House polls have historically been less accurate than polls of statewide contests and presidential races. On average, House polls differ from final election results by around eight percentage points. These polls tend to have relatively small sample sizes, increasing the margin of error, and voters tend to be less familiar with House candidates, meaning more “undecideds” until late in the race.
Democrats hope that these common sources of polling error might break their way this time.
Turnout is always uncertain, but it is more variable in lower-turnout elections, like a midterm, when even modest shifts in enthusiasm can transform the electorate. It is a particularly challenging question this year, in part because the turnout in recent midterm elections has been so low and so Republican.
This year’s early vote tallies already make it clear that the turnout will greatly exceed that of four years ago, but it is far less clear how that will translate to actual votes. It is equally unclear whether pollsters have been assuming the high-turnout, more Democratic electorate implied by early voting and the turnout in the special and general elections since Mr. Trump became president.
Over the last few weeks, the share of voters indicating that they’re “almost certain” to vote has increased by 10 points in Times/Siena polls, from 66 percent to 76 percent of the likely electorate.
In a few cases, the turnout in early voting has shown some polls were off in their projection of who will vote. Early voting in several Texas counties has already surpassed the projected electorate of a Times/Siena poll from early October, and the Democratic Senate candidate Beto O’Rourke seems likely to benefit from the higher turnout. A Monmouth poll of California’s 48th District in coastal Orange County showed the Republican incumbent, Dana Rohrabacher, leading by two points with an electorate in which registered Republicans outnumbered Democrats by 18 points. The actual Republican advantage in early voting is down to 12 points and continues to drop as turnout increases.
The final Times/Siena poll of California’s 48th shows the Democrat Harley Rouda leading Mr. Rohrabacher narrowly with an electorate where registered Republicans outnumber Democrats by 10 points, an indicator of the extent that Republicans are struggling in the wealthy suburban bastions of 20th century conservatism.