For all the reasons I have stated before, Nate Cohn agrees that if the polls are off in any direction, it is far more likely that they are overstating the Republicans’ advantages than the Democrats’. Unfortunately, the kind of errors that pollsters are likely to make are minimized in this election for several reasons. Most of the competitive races are taking place in states with small Latino populations (Colorado being the most glaring exception). The hardest to reach voters are precisely the ones least likely to vote, particularly in a midterm election. This includes highly mobile people, exclusive cell phone users, urban dwellers, college students, young people, and people on the margins of society. It also includes people for whom English is a second language, which is a group larger than just Latinos. These voters are heavily Democratic and they are likely to be underrepresented in polls, but the fewer of them that there are, the less of a skew will result.
In other words, the same factors which led pollsters to underestimate the size of President Obama’s reelection victory are still largely in play despite legitimate and concerted efforts to address them, but the size of the error is going to be less even if the problems have not been corrected.
This means that Democrats hoping to see an across the board result far better than predicted are probably going to be disappointed. Some states, like Colorado, still have the potential to greatly defy expectations, both because it has a large Latino population and because they’ve changed their voting process to a mail-in format. But most other states don’t have this same potential, or have it to a much smaller degree.
Races polling within the margin of error are going to be competitive, but don’t expect too many results this time like Harry Reid’s four years ago.