Analysis: Here's What Civil Beat’s Political Polls Really Found

By Christine Gralow

​Ige could be leading Hanabusa, Hirono might be more popular than Gabbard, and Schatz’s approval rating might be less than one percentage point below Gabbard’s. That’s what two political polls Civil Beat published last month actually found, given the polls’ margins of error.

Margin of error is a statistical measure of a poll’s reliability. The American Association for Public Opinion Research has cautioned journalists to, “Remember, margin of sampling error applies to the poll’s estimate of EACH candidate, and a candidate usually needs to be ahead by 1.5-2 times the margin of sampling error to be statistically significant.”

Merriman River Group conducted several polls for Civil Beat in early May. Civil Beat published articles about the polls from May 22 through May 24. A large majority of people polled were on landlines vs. cell phones.

One poll asked likely Democratic primary voters, “If you vote in the Democratic primary, which candidate are you most likely to support for Governor – David Ige, Clayton Hee, or Colleen Hanabusa?”

Of the 707 people polled, 37% responded that they would most likely vote for Hanabusa, 31% responded that they would most likely vote for Ige, and 11% responded that they would most likely vote for Hee. The poll had a 3.7 percentage point margin of error.

That means the pollsters could be reasonably confident (assuming there were no significant polling problems) that, in early May, somewhere between 33.3% to 40.7% of Democratic primary voters were likely to vote for Hanabusa, and somewhere between 27.3% to 34.7% of voters were likely to vote for Ige. So while it's possible Hanabusa did actually lead Ige in early May, it’s also possible that Ige led Hanabusa by 1.4 percentage points (34.7% to 33.3%).
​
Nonetheless, Civil Beat chose to publish the oversimplified headline, “Hanabusa Leads Ige By 6 As Primary Draws Closer.”

​In another poll, Merriman River Group asked a sample of 967 registered voters if they "generally" had "a positive or negative opinion of” Trump, Obama, Ige, Caldwell, Schatz, Hirono, Hanabusa, and Gabbard.
​
This poll had a 3.2% margin of error. The pollsters found they could be reasonably confident that approval ratings in early May for the four current Congressional delegates were:

Gabbard: Somewhere between 57.8% - 64.2%
Hirono: Somewhere between 55.8% – 62.2%
Schatz: Somewhere between 50.8% – 57.2%
Hanabusa: Somewhere between 40.8% - 47.2%.
 
Given these intervals, it is possible that Gabbard had the highest actual approval rating in early May. It is also possible that Hirono’s approval rating was 4.4 percentage points above Gabbard’s (62.2% to 57.8%). The difference between Gabbard's and Hirono's numbers was statistically insignificant. The poll also found that Schatz's actual approval rating in early May could have been 1.4 percentage points above Hirono's (57.2% to 55.8%) and just six tenths of one percentage point below Gabbard's (57.2% to 57.8%). 
 
Yet Civil Beat’s editor chose to run with the headline, “Tulsi Gabbard is Hawaii’s Most Popular Politician.” The article featured a cute picture of a happy Gabbard with the caption, “Hawaii’s voters love Tulsi Gabbard.” Half of the article was devoted to Gabbard.

Did Civil Beat choose click baiting over accurate polling analysis? Or is its editor statistically challenged?
-------------------------------------------------

Clarification, June 2, 2018: Merriman River Group conducted three other polls for Civil Beat in early May that are not analyzed here. Those polls focused on the Lt. Governor's race, the First Congressional District race, and a potential constitutional convention.