Media analysts are slowly coming around to the reality that the poll numbers of Trump and Sanders are inflated, but not necessarily for the reasons public opinion scholars and experts have been explaining.

In this article from the Washington Post Aaron Blake argues that Trump’s and Sanders’ poll numbers are inflated because “the voters” don’t yet care about electability. While this is intuitively plausible and no doubt accurate with regard to some voters (perhaps even a few who readers know personally), it is NOT what a responsible interpretation of polling data suggests, and it falsely props up the predictive capacities of early horse race polls.

To prove his point, Blake cites data from current polls showing that although respondents are stating preferences for Trump and Sanders at high rates, very few claim to care about electability. He then cites exit polls indicating that primary voters are primarily motivated by electability concerns. From this, Blake reasons that primary voters are presently registering incomplete preferences. While it is reasonable to assume that voters, in general, are not yet focused, it is entirely unreasonable to assume that early polls are surveying representative samples of actual primary voters.

Blake is correct in his conclusion that the polls don’t really mean Trump and Sanders are frontrunners, but wrong in assuming that they are measuring the preferences of the same population that will choose the parties’ nominees. Media analysts need intuitively plausible narratives that can be easily followed by unsophisticated readers, but by misusing polling data to prop up these narratives they are exacerbating public misunderstanding and distrust of survey research.