It should be no shock to learn that the media is not always honest in what it publishes, especially when elections are concerned. And so we come to an odd conflict in poll reporting this week for the US Senate race between Ted Cruz and Robert ‘Beto’ O’Rourke. One poll claims O’Rourke has taken his first lead in the polls, while another says Cruz has increased his lead to 9 points. So what’s going on? To understand, it’s necessary to look at the methodology and weighting of the two polls.
The poll saying O’Rourke has a 2 point lead comes to us by way of Reuters/Ipsos.
The poll saying Cruz has a 9 point lead is from Quinnipiac.
The Ipsos poll surveyed “roughly 2,000 adults”, and the poll was conducted online. The Quinnipac poll surveyed 807 “self-identified likely voters”, and was conducted though traditional random telephone calling. Further, the Ipsos poll drew their respondents from “Ipsos’ online panel”, meaning individuals known to Ipsos through prior deliberate registration with Ipsos. This means that despite Ipsos’ claim to a ‘random draw’, that is not actually true, nor is it true that Ipsos’ online polls have only a 3.5% “credibility interval”, since no objective science has yet established the veracity of online polling.
With that said, the Ipsos online poll claims to have sampled 423 Democrats, 463 Republicans, and 90 Independents, for a percentage breakdown of 43% D, 47% R, and 9% I. The Quinnipiac poll sampled “35% Republican, 26% Democrat, and 33% Other”. The Ipsos poll reports that O’Rourke enjoys 91% support among Democrats, 8% among Republicans, and 44% among Independents, while Cruz enjoys 3% among Democrats, 87% among Republicans, and 35% among Independents. The Quinnipiac poll reports that O’Rourke enjoys 94% support among Democrats, 6% among Republicans, and 51% among Independents, while Cruz enjoys 94% among Republicans, 4% among Democrats, and 47% among Independents.
I could go on about demographics, gender and age preferences, rural versus urban and so on, but the big point here is that the two polls differ sharply in voter turnout. The Ipsos poll counts 43% Democrat participation, while the Quinnipiac poll only counts 26% Democrats. The Ipsos poll counts 47% Republican participation while the Quinnipac poll counts only 35% Republican. The Ipsos poll only counts 9% Independent participation, while the Quinnipac poll counts 33% Independent participation.
This brings up questions of voter participation in the actual election. The Texas Secretary of State office provides very useful information on voter turnout, both in primaries and in general elections.
The primary turnout tells us to expect Republicans to outshow Democrats in a 2-party comparison, 56% to 43%. Applying that metric to the two polls, weighted to match primary turnout, projects Cruz will claim between 50.0 and 54.4% of the vote and O’Rourke will claim between 43.6 and 43.8% of the vote.
Therefore the Quinnipac poll is far closer to the most likely actual result, with party affiliation weighting results. Make of the media spin what you will.