Another opinion survey

Civis Analytics (CA), a firm started by 2012 Obama campaign veterans, conducted a telephone opinion survey February 28 about American policy priorities.  Powerline reported their results here, and Vox here.

Powerline excerpts two charts from the CA survey.

The first chart shows responses of likely Democratic voters - of whom 45% give top priority to “health care”!  Whut!? Because Obamacare didn't work?  Because this time, they’ll get it right fer shur?  

Reminds me of an old Flip Wilson punch line:  “Hell no, you broke yours off already!” 

The second chart shows responses to the same questions, only this time the sample group is all likely voters – in other words, not just Democrats.  Notice how the percentages change from the first chart to the second.  The second chart reveals far less less support to “health care”.

Yet Civis Analytics has this to say:  “Democratic voters, and voters in general, seem very clear in their preference that health care come first.”  Vox opines that “the numbers are strikingly similar, with answers more concentrated around health care and guns”.   Really? Voters in general?  Strikingly similar??

I don’t think so.  I say CA and Vox have it wrong.  I say Powerline has it right: “the results skew when all likely voters—not just Democrats—are reported”.  How much does it skew?  Assuming CA surveyed roughly equal numbers of likely Democratic and non-Democratic voters, the results in the two charts imply about 17% support for “health care” among likely non-Democratic voters. Do 45% and 17% seem strikingly similar to you?   Do 45% and 17% mean underlying agreement?   

Of course not.  CA and Vox both err in looking at the average of the combined surveys as though that average reflects unified public opinion. It’s an error because the responses of the two survey populations show a clear and sharp difference of opinion about “health care”.  Therefore it’s false to claim the overall average represents any general preference.   CA’s conclusion is like claiming that, on average, Americans have one testicle and one ovary.  It’s only “true” when you ignore the reality underneath the average. 

Yet despite Civis Analytics’ (and Vox’s) equivocations, I think the CA survey does reveal two important truths – (1) “health care” remains a divisive issue among Americans and (2) the division still appears to have more to do with politics than with the actual substance of “health care”.