Then Wednesday happened. A new poll from Reuters/Ipsos dropped, showing O'Rourke with a 2-point lead over Cruz among likely voters, 47-45 percent. The Reuters/Ipsos survey is the first to show O'Rourke with a lead over Cruz, although an Emerson poll released late last month showed the incumbent with just a 1-point lead.
The key difference in the results appears to be the percentage of Democrats relative to Republicans the firms behind the polls believe are going to show up on election day. Quinnipiac reports that its sample was 35 percent Republican, 26 percent Democrat and 33 percent independent. Reuters/Ipsos' sample, on the other hand, was made up 47 percent Republicans, 43 percent Democrats and 9 percent independents.
Ipsos Vice President Chris Jackson told the Texas Tribune Wednesday that his survey elected to include a greater number of Democrats in its likely voter pool because of high enthusiasm among those respondents who identify as being part of the party.
"More Democrats are registering at the highest part of the scale, at the 10, than the Republicans,” Jackson said, pointing to a question Reuters/Ipsos asks about how likely someone is to vote, on a scale from 1 to 10.
Reuters/Ipsos is anticipating a changed electorate, one that looks a lot more like the one that turns out in presidential years in Texas, rather than the one that turns out for midterms. Midterm Texas electorates are much older, much whiter and much more Republican than their presidential counterparts, typically.
"A lot comes down to what you believe the electorate is going to look like. Depending on what you believe the electorate is going to look like, you can potentially get very different results," Rice University political science professor Mark Jones says.
If a poll wants to anticipate a vastly changed electorate, Jones says, it might be better off analyzing its data and releasing multiple results based on different assumptions being made about the electorate.
"If they're going to do that, I would prefer that they give me different scenarios," Jones says. "That is, 'This is our scenario under a sea change election where turnout patterns change dramatically in Texas.'"
Nate Silver, the polling guru behind FiveThirtyEight, responded to a question on Twitter about the seemingly contradictory polls from Dallas Morning News editor and former FiveThirtyEight Managing Editor Mike Wilson with his own thoughts about the difficulty of polling in Texas.
"I think Texas is a tough state to poll (lots of new residents, low turnout among certain voting groups, may be hard to reach Spanish-speaking voters) and it's probably a healthy sign that we're seeing some disagreement," Silver said.I think Texas is a tough state to poll (lots of new residents, low turnout among certain voting groups, may be hard to reach Spanish-speaking voters) and it's probably a healthy sign that we're seeing some disagreement.
— Nate Silver (@NateSilver538) September 19, 2018
For O'Rourke to do the unexpected and make his race with Cruz truly competitive, he needs to turn out voters who don't typically vote in midterms to make election day turnout look more like the Reuters/Ipsos sample than the Quinnipiac one.
"If we see a dramatic change in turnout by millennial and younger Latinos and, at the same time, see a reduction or [smaller] increase by older Anglos, then we could see a dramatically different electorate," Jones says. "If we see an electorate in 2018 that looks anything like it did in 2016 or 2014 or 2012 or 2010, then Ted Cruz is going to win.""If we see an electorate in 2018 that looks anything like it did in 2016 or 2014 or 2012 or 2010, then Ted Cruz is going to win." — Mark Jones
tweet this
