Our daily column will return on Tuesday, July 7.
By Jim Ellis
July 2, 2020 — We saw a polling bonanza released yesterday in the swing battleground state of Arizona and, no matter who you support or what you believe, there is survey data for you.Three pollsters released results that tested the presidential race and the Arizona Senate contest between retired astronaut Mark Kelly (D) and appointed Sen. Martha McSally (R). The three pollsters, all conducting their surveys within the June 26-29 period, gave us starkly different ballot test conclusions.
Two of the pollsters are from out of state, while one is based in Phoenix and specializes in researching the Arizona electorate. One of the pollsters tested six battleground states as part of their national polling series, including Arizona, but did not disclose the size of the Grand Canyon State respondent cell. This makes analyzing very difficult. The other two firms revealed similar sized Arizona likely voter sampling universes.
The three pollsters were Change Research, Data Orbital, and Gravis Marketing. All have done national work and are published regularly in political blogs and websites as well as being frequently quoted in national news stories.
Despite testing the same electorate during the same time period, we see an 11-point swing in the presidential race, and a 13-point spread among the three survey results for the Senate race.
Change Research was the most bullish for the Democratic candidates, posting former vice president Joe Biden to a 51-44 percent advantage, and an even larger 53-44 percent margin in Kelly’s race.
Conversely, Gravis Marketing produced the most Republican results. They find President Trump holding a 49-45 percent lead and Sen. McSally up by the same point margin, 46-42 percent.
In between is Data Orbital, the local pollster. They find Biden ahead 47-45 percent, and Kelly up by a similar, but not as pronounced, margin as Change Research found, 50-43 percent.
Why the vast difference among the three? Much has to do with their sampling methodologies. See the summary below, each with a passage taken from the individual organizations’ polling synopses:
• Change Research
(June 26-28; 3,739 US likely voters; No Arizona numbers)
“Change Research reaches voters via targeted online ads that point people to an online survey instrument.”
(Biden +7; Kelly +9)
• Data Orbital
(June 27-29; 600 AZ likely voters)
“This poll of 600 likely General Election voters was conducted through a live survey that collected 60 percent of the results from land lines and 40 percent from cell phones.”
(Biden +2; Kelly +7)
• Gravis Marketing
(June 27; 527 AZ “registered likely” voters)
“The survey was collected from respondents using interactive voice responses and an online panel of cell phone users.”
(Trump +4; McSally +4)
Obviously, the three are profoundly different in developing their sampling sectors, so arriving at different final conclusions should not be viewed as surprising. And, we also see the major sampling options involved in modern survey research: the online poll, interactive voice response, and traditional live interview via telephone or mobile unit.
Traditionally, online surveying has been the most unreliable data sourcing, with live interviews still rating as yielding the most accurate responses. The automated response device option tends to post reliability factors in between the online and live interview methods.
Of these three polls, it appears that Data Orbital is giving us the best read on the Arizona electorate. They have the local knowledge that helps in defining an authentic sampling universe, their methodology is the soundest of the three, and the FiveThirtyEight organization that rates pollsters awards Data Orbital an A/B grade versus a C for Gravis and C- for Change Research.
Additionally, when looking at the large number of polls being conducted in Arizona, the Data Orbital results seem the most reasonable, and certainly among those within our current narrow focus. In terms of weighting the responses, voter history is an important factor in helping to calculate the final response and is also best utilized in the DO poll. Therefore, it is consistent to conclude that Data Orbital appears to provide the most concrete reasoning base for their weighting formula.
Arizona has historically been a reliably Republican state; however, it is now moving to the center, largely because of strong population growth. Such is underscored when seeing that the state is likely to be awarded a 10th congressional seat in the next reapportionment. To illustrate the strong long-term growth trends found here, Arizona had only five districts in the 1980 census.
We can expect to see much more polling coming from this domain in the 120-plus days that remain in the election cycle. Seeing a wide range of results will likely become a familiar pattern.