A Word to the Wise: Y'all Politics Poll Unwittingly Suggests GOP Candidates Are Vulnerable

Y’all Politics released a poll today that should be embarrassing to all those involved, as it fails to even attempt to mirror the demographic profile of the state’s electorate. 

The poll falls short in the typical ways IVR can be problematic. An IVR survey is a survey conducted through landline phones where the respondent uses the keypad to answer the poll questions. Because these automated polls go only to landline phones, respondents tend to be much older than the actual electorate, people of color are under-represented and women are over sampled.  Reputable IVR surveys weight the results to more accurately reflect the likely demographics turnout in the upcoming election. Additionally, each respondent’s self-identification of party affiliation is the norm, rather than an arbitrary constraint at the outset that Mississippi voters identify themselves as Democrats and Republicans in equal numbers and that there is a static 20% of “independent” voters.

In this poll, 71% of respondents are age 65 or older, and 91% of respondents are at least age 55. 60% of respondents are women and only 12.4% are African American.  Even in the 2nd Congressional District, which is nearly two-thirds African-American, only 20% of respondents to this poll were black. 

These errors matter a great deal no matter where one polls and are even more egregious in Mississippi where voting is often racially polarized.

Advisors to the Republican candidates in 2018 who understand how to properly read polls and believe this poll to be accurate will be very alarmed by these results by a close reading. 

  • In this poll Steven Palazzo leads Jeramey Anderson by 54%-37% in an electorate that is, according to the poll’s crosstabs, just 8.4% African-American. Readers should remember that the likely turnout in November will be around 20% African-American in the 4th CD. Without the crosstabs, it is difficult to determine how bad it is for Palazzo.  But in the best-case scenario for Republicans (where every African-American surveyed has already pledged to Anderson and therefore the Democrat is closer to reaching his ceiling) Palazzo may lead 47%-44%, or 48%-43%, depending on the racial mix of November electorate. That’s an incredibly weak showing for an incumbent against an opponent who has had very little money to reach voters.
  • Similarly, in the special election for Senate, Cindy Hyde-Smith is in deep trouble if she can only muster support from four in ten voters in this poll’s electoral mix.  This poll assumes nine in ten voters qualify for an AARP card and African Americans are polled at about one third the rate of likely  turnout in the fall.  When properly weighted this survey suggests that Espy almost certainly leads Cindy Hyde-Smith outside the margin of error. 

Ironically, this sloppy attempt at measuring voter sentiment and bolstering the political fortunes of Republican incumbents has the opposite effect: if accurate, this poll confirms the vulnerabilities of both Congressman Palazzo and Senator Cindy Hyde Smith. GOP insiders will read these results and fret. 

From our perspective, the survey methodology is so flawed that we caution everyone—Republicans AND Democrats—against drawing any solid conclusions about the races discussed therein.



  1. Joshua V's avatar
    Joshua V
    | Permalink
    Any idea what these numbers look like for Baria's campaign?

    I'm so happy to hear that Anderson has a real chance.
  2. Drew's avatar
    | Permalink
    If you assume that 90% of the "left out" African Americans vote for Baria and 5% vote for Wicker (roughly equivalent to national in 2012 and 2016), you get a 53-40 win for Baria. That roughly jives with the MS-04 vote. If MS-04 (Cook PVI of R 21) is R 5%, then MS as a whole (R 8) should be 26% more Democratic (21-8)*2 or just around a 20% win for Baria.

    My guess is that "fixing" the poll makes it way too Democratic, since Democrat party registration and current partisanship probably means that a number of the white voters who identify Democratic would vote D, with a few conservative flake offs that would be accounted for w/ black voters. As a result the races should just be adjusted by 4% or so. (So you would get Wicker 49-36, instead of Wicker 53-32).

    Of course "poll fixing" is a perilous path, and is not legitimate. The poll should just be discarded as "goddawful".
    1. Leave a Comment