How to figure out a Latino poll is BS from quite a long way away

Ear­lier this week I got an E-​mail from the great Fausta.

I gave her a call and we talked about a poll she wanted my opin­ion on. It is the Latino Deci­sions polls and she thought the num­bers were kind of weird.

I took a look at the poll and it didn’t take long to notice some .…inter­est­ing issues to wit:

This is the sixth release of an 11-​week track­ing poll of Latino reg­is­tered vot­ers. Each week impre­Me­dia and Latino Deci­sions will release a new rolling cross-​section of 300 com­pleted inter­views with Latino reg­is­tered vot­ers across all 50 states. Bat­tle­ground inter­view­ers are com­bined across all six weeks and are 267 com­pleted inter­views, with Florida account­ing for the largest share of bat­tle­ground states.

So the bat­tle ground fig­ures in this poll are actu­ally mini sam­ples taken over six weeks while opin­ions might be chang­ing. That’s bad enough but con­sider one more thing:

Bat­tle­ground states: FL, NV, CO, AZ, OH, NH, NC, VA, IA, MO (all 7 points or smaller mar­gin in cur­rent pres­i­den­tial polls)

For­get the insan­ity of try­ing to extrap­o­late the latino vote over fifty states based on a 300 per­son sam­ple. Lets look at the 10 bat­tle­ground states 267 answers over 10 states.

I would just say the num­bers but I think I can explain this with a visual much more effectively:

And yes peo­ple are actu­ally falling for this stuff.

Update: Wel­come Green Room & Faustina read­ers. Don’t for­get there are still free tick­ets avail­able for my 100^th show extrav­a­ganza (we’ll feed you) My analy­sis of what has really been going on over­all (quoted by Rush) is avail­able here and don’t for­get for as lit­tle as the price of a sin­gle Star­bucks cof­fee you can get access to my subscriber/​tip jar hit­ter only com­men­taries. A teaser here.