While the polling nerds are eating crow after the MI DEM primary, they assure us that it can’t happen again.  Citing the busted MI ’08 DEM primary for why they got it so wrong this time is fair enough (and something that little old me pointed out).   However, that only explains half the margin of their error.  And the pollsters are burying all their other ’16 flubs.

Iowa.  The aggregate polling to actual winning margin for HRC was off by four points.   But exclude the one poll (Q) that had Sanders winning by three, and the aggregate was off by more the four points and they were generally way short on Sanders’ percentage of the vote.

New Hampshire aggegate had Sanders winning by 13.3 points.  Actual 22.4 points.

Nevada.  One right before the caucuses nailed it — only 0.5 too high for HRC.  A month and a half earlier HRC was winning by 23 points.  What’s interesting about the other two polls taken a few days before the caucus is that both came up with “all tied up,” and those polls were closer to the exit polls during the day that had Sanders slightly ahead.

In South Carolina the error was in the opposite direction.  The aggregate twenty points short for HRC.  Whether that was because those supporting Sanders felt it was hopeless and chose not to vote,  at the last minute there was regression to voting for a winner, and/or the pollsters got bad samples isn’t known.  When they correctly project the winner, they get kudos and everybody moves on without asking for any explanation for the error.  A few factors from exit polls all suggest why the polls were off.  1) Age 18-29 was low in comparison to some other states at only 15%.  2)  The gender breakdown at 39% male to 61% female was more disproportionate than other states.  3) The race breakdown POC 65% to 35% is at variance with prior SC DEM participation rates.
Massachusetts.    Not many polls and most of them were in the last two weeks.  Clinton’s estimated wining margin ranged from three to eleven points.  Actual 1.4 points.   The average did peg HRC to get 50%.  However, the average only had Sanders’ at 43.3% — 5.4 points short of the actual.  
Oklahoma not until the last week did one pollster have Sanders’ ahead and then by less than half of the actual.  The aggregate swing 12.4 points.  
Texas the pollsters nailed this one.   Not a particularly impressive feat in light of the ’08 results, the population of the state, and physical size of the state.   Turnout may have dropped like a rock from ’08, but the combination of HRC and Obama (excluding the “not HRC ’08 vote) alone would have projected somewhere near 65% for HRC.  What’s notable in the TX results is that the DEM electorate is different from that of Deep South states.  Not as many New Deal Democrats have been killed off in the past fifty years.

Virginia polling and results look a lot like TX.  Both 65% for HRC.  Several commonalities between these two (former slave holding) states.  Will mention but one — dependence on MIC in-state expenditures.

AL, GA, MS — pfft.  (Not to be disrespectful of the good people in these states, but they aren’t much into “change” unless one of their own tribe is standing tall for it.)

FL polling presents difficulties similar to that in MI.  The ’08 primary was a bust.  OTOH, there isn’t a fourteen year gap between competitive DEM POTUS primaries in FL.  This was another state that HRC won in ’08 (with 51% in a three way with Obama and Edwards).  As HRC received 25% of the AA vote in ’08, this time 70% is her floor and it could go as high as over 90%. However, doubt that turnout will be as high.  Expect INDs (17% in ’08) to increase and be less favorable to her (40% in ’08).  INDs MA were 33% and MI 27%.  The polling looks dreadful for Sanders, but it’s not quite as bad as it was in MI — which  suggests to me that the pollsters may be better samples.

Yet, there appears to be something going on that the pollsters are having difficulty cracking.  And therefore can’t track.  I’ll present the numerical oddities in Part II.