Earlier this month, JMC Analytics and Polling independently polled both the Alabama GOP Senate primary and runoff (those polls can be seen here and here ). In an article in yesterday’s Washington Examiner, a lengthy memorandum written by Travis Smith from Senator Luther Strange’s campaign was released disputing the findings of two runoff polls conducted since the August 15 primary (JMC’s poll was one of those two polls critiqued, and can be found here). While a rebuttal from Senator Luther Strange’s campaign is an understandable reaction from an incumbent who ran second in his party’s primary (and who, one week into the runoff, trails by 19 points), the campaign’s rebuttal was written with minimal research done to back up its assertions. JMC respectfully wishes to set the facts straight.
Response to “Fake Polls” paragraph
- This paragraph begins with an assertion that polls released both by JMC and DDHQ are “extremely suspect.” This assertion left out a few facts: (1) JMC was sequentially the first of three pollsters in the primary (Cygnal and Trafalgar were the other two) to show Senator Strange in trouble and former Chief Justice Roy Moore’s making a late surge. That poll can be found here and curiously was only indirectly mentioned in his memo; (2) the primary poll was conducted before President Trump endorsed Senator Strange, so poll numbers would not have reflected the impact of late deciders who in the end decided to support the incumbent. Still, that poll showed Moore ahead of Strange 30-22% (19% favored Congressman Brooks, 12% favored one of the other candidates, and 17% were undecided). That’s an eight point spread, and quite similar to the six point spread on primary night. In other words, saying that the poll had a “pro Moore bias” is not grounded in the facts;
- The memo states that the runoff poll does not pass the “smell test” because it showed Moore getting less than the “respectable 33%” he received in the primary. In reality, a 33%/second-place finish for an incumbent (albeit an appointed one) Senator in a party primary is not a show of strength, particularly when taking into consideration Senator Strange’s overwhelming financial advantage. Furthermore, the poll was conducted in the immediate aftermath of the primary and showed the senator with 32% (statistically identical to his primary showing), and another 17% undecided. Had Mr. Smith even bothered to read the second page of the runoff poll release, a separate question was asked after the ballot test showing that those who supported a losing candidate by a 51-26% margin supported Moore over the (appointed) incumbent. In other words, the crossover support Senator Strange desperately needs to win the runoff is not “crossing over”, particularly since Senator Strange has (according to the primary poll) a 50-35% disapproval rating;
- The memo notes a “significant under sampling” and points to Birmingham as “Exhibit A.” Actually, the first and last pages of both polls clearly state that media markets (as defined by Nielsen) were used as the basis of geographic segmentation. Furthermore, the Birmingham media market (which includes smaller cities like Anniston, Cullman, Gadsden, Jasper, Talladega, and Tuscaloosa) also includes five small counties in the Columbus GA, Columbus MS, and Meridian MS media markets. In the primary, Moore defeated Strange 40-38% in that region. This disclosure is on page 3 of the runoff poll;
- The same paragraph notes that Birmingham accounted for one third of the statewide vote. It’s not clear what he is referring to here (i.e., was he referring to Jefferson County, Metro Birmingham, or the Birmingham media market?), but had Mr. Smith read the first page of both of JMC’s released polls, there was a disclosure that the Birmingham media market accounted for 42% of the sample both in the primary and runoff.
Response to “JMC flaws” paragraph
- Weak Screen: Several factual inaccuracies here need to be set straight. On the first paragraph of both the primary and runoff polls, JMC clearly stated that there was pre-screening done based on recent primary participation. Furthermore, since there are people who do vote in Democratic primaries in Alabama (JMC determined on the night of the primary that 72% chose a Republican ballot – that analysis can be found here), the content of the “weak screening question” was actually (in addition to the pre-screening just mentioned) asking the respondent if he/she “plan(s) to vote in the Republican US Senate special election runoff/primary.” Those NOT voting in a Republican primary were not be allowed to take the rest of the poll;
- Wrong sample: The assertion that the sample “captured too many non Republicans” has already been debunked by the previous paragraph, as Democrats were prevented from taking the poll. Furthermore, in the runoff poll, 68% affirmatively said they were an evangelical Christian. In the primary poll, 70% said they were, and 74% described themselves as conservative, 17% moderate, 4% liberal, and 4% didn’t know;
- Robopoll – JMC conducts live, automated, and “hybrid” polls (where landlines are autodialed, and cell phones are manually dialed by a live operator), and therefore can respond to this assertion in a professionally objective manner. Either automated or live polls are more appropriate depending on circumstances specific to that poll. The main critique on automated polling is that only landlines can be called, and that the omission of cell phones excludes respondents who tend to be younger, non-white, and inclined to support Democratic candidates. A careful analysis of the data, however, (you can read that analysis here) shows that the composition of landlines vs. cellphone voters isn’t as “black and white” (so to speak) as conventional wisdom would dictate. Furthermore, in a special election Republican primary in a state that has a fairly stable electorate, polling an older electorate (who are more likely to retain their landlines) is perfectly acceptable and in fact an appropriate representation of Republican primary voters in Alabama.
Mr. Smith’s passionate defense of his candidate is perfectly appropriate (and understandable) in a contentious Senate campaign. However, it’s unfortunate that he made several assertions not supported by facts or a simple reading of the contents of the poll. Furthermore, JMC’s polls have fully disclosed their methodology, the precise wording of the questions, and have provided detailed crosstabs within the poll release. JMC respectfully asks that similar disclosures be done by the Strange campaign.