Wednesday, May 18, 2016

For the record: #FeelTheDerp

Hey, it's not Latinos that made the mistake!
Wishful thinking is not generally effective thinking. Accepting something you noticed in somebody's blog ("Sanders has a significant lead among Hispanic voters and other races" in a May 5 poll), and then translating it into a more majestic sentence that means something totally different ("Bernie has won most Latino votes to date") does not get you closer to the truth but further away. Not checking the link, or not noticing that there is no link, doesn't help either. It is likely that Clinton has received considerably more total Latino votes than Sanders to date, and on the other hand it is likely that Sanders's current popularity in the Latino community is nearly as high as Clinton's and could be even higher, and it's really hard to judge how true either is, because the exit polls are not perfectly reliable and the samples of ethnic minorities within broader samples are just too small. Storify from this morning below the fold:

That is: the Ipsos poll Beijer was citing didn't have enough Latino respondents to judge from. That wasn't what the poll is about (they used the ethnic data for statistical correction, in weighting the results). "Confidence interval" is the online-poll equivalent of "margin of error", and whatever I said on Twitter it's probably closer to 40% in this case. Beijer imagined the numbers meant something to because he wanted it so bad. That's Dr. Bill Kristol's favorite error, and it's not something we ought to be doing on this side of the net, where we like to think of ourselves as a reality-based community.

But the main thing is if you want to heap scorn on the head of Josh Marshall you really need to make sure you've got your facts straight.

No comments:

Post a Comment