One bad poll is an outlier, two bad polls is a problem.

Even more worrisome for Trump: Emerson College polling, which now has Clinton up two in Arizona, risks understating Democratic support because it only calls landlines in conducting its surveys, which tends to mean an older, more conservative sample. The “house effect” for Emerson per FiveThirtyEight is a two-point Democratic tilt, meaning that if these numbers are capturing the true state of the race, you’d expect Clinton to win by four if the Arizona election were held today. That’s 11 electoral votes if it flips, nearly as many as North Carolina. If this state starts to get away from him, the whole election is going sideways.

Four new polls released today by Emerson College show a tightening of the presidential race, and for the first time, Hillary Clinton has taken a lead in Arizona, by a margin of 44% to 42%…

Three of the states polled have a sizable Hispanic population: Florida (19%), Nevada (22%) and Arizona (22%). Among Hispanic voters, Clinton has a solid lead over Trump in all three: 59% to 21% in Nevada, 54% to 40% in Arizona, and 59% to 36% in Florida.

The only other post-debate poll of the state, from OH Predictive Insights, had it 42/42, similar to Emerson. A few surprising results from the Emerson crosstabs: Clinton leads Trump by five points among independents and “the genders split fairly evenly” between the candidates. That’s not what you see in most state polling, although a close split among women in Arizona might make sense given the state’s Republican tilt. Women there broke for Romney by seven points in 2012; if, as expected, Clinton’s overperforming with women in Arizona this time, a narrow margin either way seems plausible. I don’t know how to explain a tight race among men, though: Romney won them by 17 points last time and they’re one of Trump’s core constituencies this year. If Clinton’s making a race of it among men in a red state, Trump has major problems.

As for the 54/40 split among Latinos, you can take that as encouraging or discouraging for Trump as you prefer. The encouraging spin is that he’s performing much better with that group than Romney did. Obama won Latinos in Arizona 74/25. Trump has improved 15 points(!) among a demographic that’s supposed to find him toxic. The discouraging spin is that if Emerson’s numbers are overestimating Trump’s support among Latinos, or if there’s a late break towards Clinton by this heavily Democratic group, then he stands a real chance of losing the state and, with it, whatever hope he had of winning the election.

More surprising than any of these numbers, I think, is the fact that Clinton has fewer field offices in Arizona than Trump does, a rare case of him out-organizing her in a state. You can understand the logic of that — Arizona’s a longshot for Democrats, they obviously don’t need it to win, therefore they’re better off devoting the money and manpower to extra offices in Florida or Pennsylvania. The counterargument, though, is that there’s little evidence in the polling that Arizona is all that much of a longshot this year. Here’s the entirety of the surveys of the state tracked by RCP since early August:

Only once since the convention has Trump been ahead by more than four points and not since late August has he cracked 42 percent. In fully half of the eight polls of Arizona over the past two months, he’s either trailed Clinton or been tied. His lead in the RCP average is down to a single point. The FiveThirtyEight “nowcast,” which is based on the most recent polling, actually has Clinton a slight favorite to win, although the longer-view “polls-plus” forecast still gives Trump a better than 60 percent chance. Makes me wonder what’ll happen if Hillary gets a couple of bad polls next week in one of the midwestern states she’s counting on, like Wisconsin or Michigan. Arizona could replace one of those states in her blue wall if Trump flips one of them due to his strength in the Rust Belt. You might see Democrats pay some attention to AZ after all.

By the way: John McCain leads his Democratic Senate challenger by 16 points in Emerson, meaning he’s running nearly 20 points ahead of Trump. That’s apples and oranges insofar as he’s a longtime incumbent and his opponent isn’t nearly as well known as Hillary Clinton, but it gives you a sense of the degree of ticket-splitting that might happen.