I guess I'll note this here: The Field Poll is shutting down – after 70 years of charting California’s political highs and lows https://t.co/s2HV8yLijb pic.twitter.com/obTRsIVdd4— The Sacramento Bee (@sacbee_news) December 9, 2016 The end of a genuine California political institution.
http://election.princeton.edu/2016/12/10/the-comey-effect/ "However, the big change does coincide well with the release of the Comey letter. Opinion swung toward Trump by 4 percentage points, and about half of this was a lasting change. This was larger than the victory margin in Michigan, Pennsylvania, Florida, and Wisconsin. Many factors went into this year’s Presidental race, but on the home stretch, Comey’s letter appears to have been a critical factor in the home stretch." I'm not sure whether the consensus is that Comey acted out of animus toward Hillary, or fear of being made to look bad by those rogue FBI agents in New York. I'm not even sure if there is a consensus. Either way, his action, by itself, put Trump in the White House. His UNPRECEDENTED, WARNED AGAINST action. <sigh>
Okay. It is now time for me to discuss what went wrong with polling. Taking advantage of some variables that are publicly available (Dave Wasserman, Nate Silver, census data, Rutgers' Center for American Women in Politics), I have identified some interesting relationships, reasons, and discussion that will help people understand what happened this year. First, the polls. 50 states + DC cast electoral votes. Taking Hillary Clinton's margin and comparing it to the polls-only forecast generated by 538, I present this graph (DC omitted for viewing reasons): The blue line is a simple abline from 0-100, and the red line is the smoothed fitted values of the regression of hillaryshare ~ silverpredict. So clearly Nate Silver was off - mostly, he was conservative in that he thought red states would be more blue, and blue states would be more red. But which states were most off? The following states had less Hillary vote share than the polls forecast by greater than 2.5 percentage points: ID (-8.4% Hillary) ND (-6.9% WV (-6.8%) TN (-6.4%) SD (-5.8%) WY (-5.5%) KY (-5.2%) AK (-4.4%) OK (-4.3%) VT (-3.5%) IA (-3.4%) MO (-3.3%) WI (-3.1%) SC (-3.1%) KS (-2.9%) NE (-2.8%) AR (-2.6%) IN (-2.5%) MN (-2.5%) It should be noted that of the 19 states included here, only three were battleground states (bolded), and only those three of the bunch were regularly polled. In total, thirty states had a lower Hillary vote share than the forecast predicted. One state was right on the money - Virginia - and another 19 had more Hillary vote share than predicted. The following states had higher Hillary vote share than the forecast predicted by greater than 2.5 percentage points: DC (8.0%) HI (4.3%) CA (3.4%) NM (2.6%) ... and that's it. Of the 13 swing states identified by Wasserman in his chart (modified by me to remove Iowa and add Virginia in), Hillary only underperformed the polls by an average of 1.05%. The worst offender on the list was Wisconsin at 3.1%. Pennsylvania was off by 1%, while Michigan was off by 1.1%. Polls were not to blame. If the states that strayed outside the polling margin of error were re-assigned to their "rightful" winner (polls correct and the election wrong), Hillary still loses the election, 296-242. BUT BUT BUT Because I had all this data, I decided to run a few models. And here's the one I find most fascinating. The dependent variable is the vote margin between Hillary and Trump. Hillary's regular vote share, or the drop from Obama '12 to Clinton '16 do NOT find these results: In short, there are three variables that predict the gap between Clinton and Trump: the percent of state citizens with a college degree, Obama's approval rating in 2016 (October, Gallup), and the percentage of female state legislators in the state in 2015. The way to interpret that coefficient is that for each percentage point increase in the proportion of female state legislators, the Clinton margin over Trump increases by 0.42%. For those of you who want to know, Michigan's share of state legislators is 20.3%, Pennsylvania's is 19.0%, and Wisconsin's is 25.8%. The national average is 24.6%. If Virginia and Pennsylvania had female legislators at the national average, Hillary would have won those states by -0.2 + (24.6-20.3)*0.42 = 1.6% and -0.7 + (24.6-19.0)*0.42 = 1.7%, respectively. Even worse is Wisconsin, who as recently as 2004 had 28% of its legislature as women. That number declined to 22% when Democrats took power in the late Bush years, and has not recovered. These tentative findings indicate that there are a sizeable chunk of voters who normally vote for Democrats but are uncomfortable voting for female candidates for office. That's in part why Hillary lost - minute changes in a wide variety of states would have given her the lead in the Electoral College. Ignore the whole argument over white-nonwhite, or education (although it certainly played a role), or Hillary not campaigning in Wisconsin. She lost because there are a number of Democrats who are uncomfortable with female leaders. It ain't the GOP - Trump's vote share was pathetic. It's the "progressive" party. Now here's the rub. I can't prove this relationship using a 50-state model; there's too many assumptions I have to make. Chief among them is that the Hillary vote share is depressed in areas where female candidates cannot win. Fortunately, there's a way to test this assumption. Due to a database of some 308,000 elections of state legislators, I can compare counties where female state legislators ran and barely won, to those who ran and barely lost, in the past 3-4 election cycles. This group would constitute a "regression discontinuity," or a situation where the losers and winners only found themselves there due to dumb luck or random chance or other factors beyond the control of their campaigns. I can then see if Hillary's 2016 vote share was higher by a statistically-significant amount in the "bare winners" group to the "bare losers" group. If it was, then it means having female representatives makes you okay with more female leaders. Anyway, for those of you who read obscure academic journals, check for this paper in 2-3 years.
Not sure if this is the right place. Anyone see a pattern here, in these comparison of Dem candidates to the congressional average? @American Brummie
Is X the vertical one? If so, it's how liberal they are compared to Congress. Sanders was in the 99th percentile as most liberal...or in this case, I guess, the 1% least conservative. The horizontal axis just ranks them from least liberal to most liberal, with winners in red.
I guess I'll put this here. This is a pretty extraordinary chart. Clinton's numbers fell instantaneously. AGAIN: The national media, far left, far right, Trump and GOP pummeled Hillary's image beyond recognition. THAT is why she won't be POTUS. pic.twitter.com/ywbHVArrCv— Peter Daou (@peterdaou) December 17, 2016 Right off a cliff. This really is a remarkable chart and something to keep in mind when critics say Dems nominated a historically unpopular candidate. https://t.co/2fGsIYr0cV— Anthony Zurcher (@awzurcher) December 17, 2016
Look at the year when it fell off, 2015. But is not like we did not know this was going to be the Republican tactic, they had been getting ready for her for over 8 years.
This is another interesting piece of analysis: https://ecri-prod.s3.amazonaws.com/downloads/1611USCO_Essentials_EXCERPT.pdf Now for the details. Whitesmade up 81 percent of the labor force (that includes employed and unemployed workers) in 2007. Overall, the economy has gained net 9 million jobs from November 2007 to November 2016, but whites experienced a net job reduction of more than 700,000 jobs. During the same period, Asian Americans—5 percent of the work force—gained close to 2.5 million jobs, while black workers, who were 11 percent of the work force, gained a little over 2 million, and Latinos, at 14 percent of the work force, gained almost 5 million net jobs. If that 81 percent figure seems high, know that for this data ‘white’ includes those who identify as Hispanic by ethnicity, and white by race—the same goes for black and Asian Hispanics. The overwhelming majority (89 percent according to the Bureau of Labor Statistics) of Hispanic workers identified as white by race, so the job numbers for non-Hispanic whites would appear to be slightly worse given the job gains among Hispanics overall. As per the BLS, the racial demographics of the work force had shifted only by a percentage point or two from 2007 to 2014. The data is even more striking when we take age into account. Looking only at those of prime working age—25 to 54 years old—whites suffered a net job loss of 6.5 million. For Latinos, Asians, and blacks in the same age cohort, the net job gains were 3 million, 1.5 million, and 1 million respectively. In fact, those three groups have seen net job gains in every age category since the pre-recession employment peak nine years ago. Furthermore, taking race and class—measured by educational attainment—into account, ECRI reported in October that “for seven long years, the majority of less-educated non-Hispanic White adults has not been employed.” Given that Democrats’ primary argument to the white working class has long been an economic-centered one, that statistic says a lot about why theyvoted the way they did in 2016.
Asians are >5% of the population and they got 30% of the gain, not bad at all for them. Also not bad for Latinos, 16% of the population and around 55% of the gains.
It is a function of location and the sectors where the job growth happened. It happened in cities and service industries, not in the countryside, in manufacturing.
Lots of interesting bits and pieces from the Cook Report in here: 56 Interesting Facts About the 2016 Election Both are surprising, actually, but the Texas numbers more so.
If the Democrats want to win in 2020, they better start by picking up the governorships across the Southwest.
I got this wrong also. Probably not a huge surge. Latinos may have a 25-30 % base that will vote Republican regardless, and if we compare the exit polls with the Latino reflections polls, you see the divide between new immigrants and English as the Primary language Latinos. http://www.npr.org/2016/12/22/50634...lican-and-other-myths-about-hispanics-from-20
Polling was even more wrong than we thought, because Trump tells us that 3 to 5 million illegal immigrants voted for Hillary.
When one looks at the makeup of the Congress, one can only conclude that those folks must've been splitting their ballots.
Something I just realized from the CNN exit polls. Income Clinton Trump Other/No Answer Under $50K 36% $50K or more 64% 24558 respondents 64% of the people that answered the exit polls, were "wealthy" people. Making over 50K. Now the average income is about 59K (or in the 50's). But if I remember correctly, about 50% of all working Americans make less than 30K. So people making over 50K are way over represented in the polling, now that could be correct because perhaps people making over 50K are the people that actually vote, so the exit polling has the same/similar breakdown as the voting population. So basically, poor people do not vote (or are underrepresented in the polling). http://www.cnn.com/election/results/exit-polls
Trump's victory was largely coincidental, in that he managed to win the Electoral College while losing the popular vote due to a late surge in the Rust Belt which allowed him several important, but narrow, victories by switching a group of working class workers who felt insecure and responded to Trump's message. A message that told them that their jobs were either being shipped away to foreign lands due to bad trade deals, or was being taken up by immigrants/foreign workers due to lax enforcement of immigration laws. All this, according to Trump, because these policies benefit some 'large corporations' and an establishment beholden to these 'large corporations' and special interest groups. The coincidental nature of Trump's win notwithstanding, however, the Democrats IMO are lowering the bar with respect to Trump to such an extent that unless he turns to be an unmitigated disaster, and makes things worse for those who voted for him, he might end up winning again. That really shouldn't happen, but it might. This is particularly true if the next election is turned into a referendum on some rather weak attempts to claim Trump is a Russian tool or in attempts to paint him as more of a racist than he is. (Trump carries racial prejudices and holds racist sentiments that are largely in line with those who voted for him). In that dynamic, all Trump would have to do to find himself reelected is to start reading out a teleprompter more (as he did in his address to the Congress), and lay off twitter a bit more than he has shown himself capable of so far. Appearing even semi-sane, those who previously voted for him (including those who aren't really his fans) could feel vindicated and even some who didn't vote for him might do so, being turned off by what they will perceive as the establishment ganging up on him. Of course, regardless, Trump will almost certainly lose the next election if the economy is doing badly. Or if the US is involved in an unpopular war that is going badly. But absent either, the best way for the Democrats to win the next elections is to act more like adults and not lower themselves to Trump's level. If the Democrats do that, Trump -- on his own and without the obvious attempts to get him -- will manage to alienate and shave enough from the votes cast for him last time (around 5% of the electorate would would be up for grabs) to be certain to lose the next time around.
@Boloni86 @American Brummie Even WITH the Comey effect, Hillary outperformed every winner. OK, that's more Trump than her...but the idea she ran an epically bad campaign is afactual. If her campaign was THAT awful, why was she on track to outperform the fundamentals by, I'm guessing, 8 points when Comey put his thumb on the scale?
The thing to add to Dave's post about Abramowitz is that his model is calculated in August of the election year. http://www.washingtonexaminer.com/a...odel-predicts-trump-win-51-48/article/2599203