Tuesday, August 23, 2016

Dewey Defeats Truman: What If The Polls Are All Wrong?

In my last piece I explained why Clinton would likely win. In this piece I explore why Trump may be doing better than we think.

November 3, 1948 was an all-time kind of bad day for the Chicago Daily Tribune. That is the day they printed the totally false, hugely lettered, front page headline made so famous in our history classes. DEWEY DEFEATS TRUMAN!

History is pretty tough on that headline. To be fair to the Tribune, the polling all pointed to exactly the outcome they printed. The problem for the Tribune is that if you print the news before it happens you could be wrong, and they were wrong. Truman enjoyed the error at their expense, the country moved on and Thomas Dewey became a footnote in American history. So, how the heck did it happen?


American politics is as old as the country itself. Polling is not.

Life was rapidly changing in post World War II America. It's hard to find a more accurate time in history to apply the phrase: "To the victor go the spoils." America entered the war late, fought hard, and ultimately came out on the victorious side. The costs during the war were huge for all involved, but Americans had one very important advantage on the back end, because we didn't have to fight in our own country. Japan, Europe, huge areas in Russia, Great Britain, all were heavily decimated by the war. America had the attack on Pearl Harbor, but otherwise our major cities remained totally untouched. We had infrastructure, materials, and people. Plus, we had factories ready to be converted from full time arms manufacturing. It was like moving into a fully-furnished apartment and then telling everyone that has to move all their furniture that you're faster at moving then they are. It's true, but it's also because of an inherent advantage. America had that advantage and to our credit did a fantastic job utilizing it for its full potential.

The post war boom had as much to do with the economic positioning of the world at that time as it did with the American spirit of hard industrial work. With this came the middle class boom. It also created a new crop of college educated young men (and women, but still mostly men) thanks to GI bills. GDP was up, incomes were up, education levels were up.

This all happens along side the proliferation of in-home technology. Phones. Radios. Eventually televisions. Electrically powered washing machines, fridges, freezers, on and on and on it goes. Everything was the best thing since sliced bread. (Ironically, I find myself really enjoying freshly baked loafs so much more than sliced bread. Cutting bread, it turns out, is not that hard.) It was a revolutionary time. Sure, by today's standards almost EVERY single new electronic marvel of that time is outdated (and a fire hazard) but at the time the idea of putting your clothes into a machine and coming back to them being completely clean, then putting them into ANOTHER machine and coming back to them being completely dry was f^&$ing amazing.

Along with all these new inventions came the advent of modern political polling. Social scientists began asking people for their opinion on things to shape the message and make predictions. The 1948 Presidential election was the first real test of polling. It's also worth noting this was the first time in 16 years polling would have even been useful. FDR won, and then won, and then won, and then won again. 1932, 1936, 1940, 1944, all went to FDR and none were terribly close.

When Truman ran for a “second term” he was really running for the first time as far as his exposure to the American people. Dewey had been there before, running and losing to FDR in 1944. Both had name recognition, and in the early days Truman, an “accidental” President, wasn't polling well. In fact, Dewey held a significant lead in the polls right up to the election, which (in part) prompted the false headline. Dewey did not defeat Truman, but the polls showed that not only would Dewey win, he was supposed to win by a fairly wide margin.


Just like you probably don't want to use a toaster from 1948 you also don't want to use polling methodology from 1948. The new versions are way better, and both do a much better job at creating an even outcome. Seriously, modern toasters rule.

The Dewey polling debacle is the result of an imperfect science. Polling was still new, and methodologies were relatively untested. Today we hear the phrase “random sample” in polling and advertising. There's a reason for that. In 1948 the samples were, well... not as random. Sample sizes were small and homogeneous. It's not hard to imagine that if you hand picked a group of people to answer your questions you will likely have bias in your process, even if it is unconscious bias. It's human nature and it's why random sampling is so important.

The very subject of the 1948 polling errors spawned volumes of research and writing. Ask any political science undergrad about their Research Methods class, I still remember mine. It's a big part of curriculum now for social scientists. 

What's important is this: Polling, for the most part, was fixed after this and has been very good at predicting the Presidential elections since then.


I think the sampling issues of 1948 may be back, but with a modern twist. I'm basing this on what we saw in 2012. For months in 2012 every media outlet was running with polling numbers showing Obama and Romney in a very tight race. There was talk of a tie along side talk of either being able to win, and that was the joint narrative right up until the results came out. As it turned out, Obama won handily, not a landslide per se, but by much more than the polls had predicted.

One possible error is the inability of polling to properly sample the modern American voter. Polling is often tied to land-lines. That's a problem. As an example I have a very “traditional” type of life. Middle class family, home, etc. Yet, our home has NEVER had a landline. In fact, the only real landline I've ever had is at my parents home growing up and now at my office. That's it. I've had the same cell phone number for 12 years, but never have I plugged into a jack for anything but internet service and I've long since ditched the DSL for Cable any way. 

I am not odd for never having a landline. In fact, more and more I'm meeting people over 50 who are cutting the cord on their own landlines and going totally wireless. If we know anything about technology is that's when those over 50 begin using something as normal, it's officially here to stay.

Cell phone polling is increasing, but it's still not regular. Plus, when your cell phone rings and you see an unknown number from another area code, how likely are you to pick it up and answer? That leads to both an oversampling and an under-sampling. Those with landlines, and who are home to answer them, are the more likely people to be polled. Those without... not so much. Just like in 1948 we wind up with a skewed group of people providing the answers to the polling question: Hillary or Trump?

So, there is a real chance that the polls, which show Hillary holding an absolutely dominating lead as I explored in my last piece, could actually be wrong.


What I'm saying is: Never say “never” when it comes to the election outcomes. Trump could win. In fact, Trump could actually be winning right now.

That's right. The polling could be totally and completely...

Ok. I give up. I can't do it. There is just no way. I've been thinking about this for weeks because my goal, always, is that I want to examine both sides. It's just that, in this case, that is becoming increasingly harder to do in a meaningful way.

Seriously. Clinton is up by 11 points in Virginia, a state Obama won by 4 in 2012. Even if you take away a 5 point polling error Clinton would still be up on Trump by 6 in Virginia. In Trump's own style “That's Yuge, believe me.” She's up by 10 points in Colorado, 9 in New Hampshire, 9 in Pennsylvania, 7 in Michigan, 9 in Wisconsin and just about 5 in Ohio. If we assume the polling is just totally whacked and subtract 6 points from each of those states from Clinton and give 100% of that difference to Trump, CLINTON WOULD STILL WIN ALL BUT OHIO!!!!

If you've read my other recent pieces you know that those are all pretty important states for Trump to get to 270 electoral votes. In fact, as of today Clinton is leading by a percentage greater than the margin of error in enough states to give her more than 270 electoral votes. That means that if the election was held today and the polls were accurate, Clinton wins. On top of that, her lead has been getting bigger each week since the end of the Democratic convention. Now, Trump's campaign is doing so poorly that even Missouri, once a true bell weather state, then a virtual lock for the Republicans in 2000, 2004, 2008, and 2012, is a genuine battleground state. 


One thing we know from polling is that the vast majority of Americans think Trump and Clinton both lie and that neither are trustworthy. Turns out, two negatives do still make a positive, at least in the polling numbers. Clinton's lead makes it so that Trump would have to nearly morph into another human being to have a chance. The current numbers are the result of almost 450 days of Trump's campaign. As of writing this there are fewer than 80 days remaining until the election. Do we really think that Trump can change 450 days worth of perception in what amounts to about 75 days of “post pivot” Trump?

Add to that the clear fact that Trump and his every changing carrousel of advisors don't seem to understand what a Pivot really is. Pivoting means keeping one foot firm while moving the other around to find a slightly new position without leaving the main spot. It's a circular motion that, if continued, brings you all the way around to the beginning. Trump has pivoted like 28 times already, and this last one doesn't appear much different from those before. Trump can keep pivoting all he likes, but eventually it will become clear that what he's really doing is spinning in a circle.


Can Trump beat Clinton in the debates? Maybe. With some help from Libertarian and Green party candidates... still, maybe. One on one? No way. His base, the “alt-Conservatives” as they have been dubbed, will think Trump wins no matter what he does. That isn't the same as convincing enough voters in Pennsylvania to swing a 9 point loss into even just a .01 point win.

Pennsylvania really is “The Keystone State” to this election. It, along with Ohio, signal the likely outcome of this race. Right now both are squarely in Clinton's camp, and Ohio matters less and less with Clinton's sizable lead in the other bigger states like Virginia. If Trump has another gear, he has to find it right now. If not, it's time to start talking about the Senate and the House of Representatives because those races are now in play in a serious way, at least in 2016.

- Adam Sommer

Friday, August 12, 2016


Up next: What if the polls are wrong?

Donald Trump is going to lose the election. He will not be President. Those sentences are very likely true. To call it now may be a bit bold (probably isn't) but the Presidential election is just the tip of the iceberg that is the fall 2016 election cycle.

Congress may be up for grabs. The Supreme Court has at least one and very likely two, three, maybe even four vacancies to fill, before 2024. Underlying it all is the Republican party's hold on any redistricting efforts to keep the House as gerrymandered as possible. All of that is on the table in 2016.

With that in mind, the Republicans went out and waded through a group of 347 candidates (give or take 5 either way, I'm not sure the actual total) and managed to pick Donald Trump to carry their party banner. To be fair, the Democrats had a surprisingly good option in Bernie Sanders and still decided to go with Clinton, but this is about Trump and the GOP (never fear, I'll bash the other side next) so let's focus on them.


In looking at the absolute worst possible electoral outcome for Trump I've come to 391 – 147. The map (click here to view ) is really blue. So blue. You can't believe this blue, the best blue... believe me. And it's an absolute embarrassment of a loss for the GOP. For Trump a loss like this would be the “sign of a rigged system” and “proof it was stolen from him” and “not his fault” and whatever other excuse will be used to cover up his four alarm dumpster fire of a campaign.

Trump can try to spin the loss into even greater traction with his main base of support, which could even be a solid launch to a television show, or even a whole channel, devoted to his slogan to “make America great again.” Trump, much like a cat (or like this cat that looks like Trump) will land on his feet. Or, I'm sure he's got a guy he pays for that, any way.

Trump loses to Clinton, he whines about it, 10-15% of he electorate actually thinks the system is rigged (they already do so what's new?) and they spend 4 years treating Clinton the same way they've treated Obama. Not with respect as the President or Commander In Chief that they demanded George W. Bush be given, but as a fraudulent leader that can't be their President. We've seen this movie already, it is what it is.

For the Republican party on the whole, it may not be so simple.


An electoral outcome with Clinton going over 380 electoral votes would mean almost certain doom for down-ballot races. That's because For Clinton to get to 380+ in the electoral vote one of two things would be true: Either GOP voters abandoned the party, or the Libertarian candidate pulled enough from Trump to create a plurality similar to 1992 when Bill Clinton was elected with well under 50% of the popular vote. That would require Gary Johnson to pull Perot numbers, which seems unlikely at this point, but that could change.

More likely is that usual GOP voters either don't vote for Trump, or do vote for Clinton. The ones that don't vote for Trump are the problem for the Republicans. In plenty of cases there will be voters that simply cannot, will not, and flat out refuse to vote for Trump but will still thoughtfully consider the lower races. Historically turnout is much higher for congressional battles in Presidential years, which is why the ones that don't vote are the issue. If a group large enough to represent a 3% swing in the polls favoring Democrats chooses to simply stay home on election day, then the Republican's congressional seats will be very hot indeed.

It could lead to a trifecta of trouble (alliteration!) for the Republicans. Riddle me this: How do you encourage voters to turn out and vote when they don't want to vote for the Presidential race?

It's a trick question. You don't. Republicans have to hope that voter turnout is low overall. (Yelling about the election being rigged is a pretty good way to keep people at home.) That will allow the Republicans in congress to maintain control. If voter turnout is high it's much more likely that people will vote for #nevertrump or just skip the Presidential portion of the ballot and vote down ticket. Wait... wouldn't that be good for the Republicans? Yeah... no, not really.


If voter turn out is high that's going to favor the Democrats in general. More voters usually means that the under 35 crowd managed to get to the polls and statistically that group drastically favors the Democrats. It breaks down like this:
  • Low voter turnout occurs when younger voters and minority group voters don't go vote.
  • High voter turnout occurs when younger voters and minority group voters make it to the polls.
  • Older voters already turn out in higher numbers in all elections, and are more likely (statistically) to vote Trump and the Republicans in general.

So, if we take those ideas as generally applicable, that means that higher voter turnout means greater chance for Democratic victory in all races. That means not only do the Republicans need the voter turn out to be low, but they need it to be low because the Democrats stay home. The problem for Republicans is that lots of Republicans don't like Trump, and Democrats really don't like Trump. And it's enough to get them to the polls.

Another real problem for Republicans is that the married white female voter, a group they've actually done quite well with and won in 2012 (52% of that group voted for Romney), could easily stay home to avoid this election entirely. Sitting it out is a fairly common response in 2016 accross all demographics.


Trump is turning off would be Republican voters. They want to vote for a candidate they like. They don't like Trump. Gary Johnson? Too socially liberal. Hillary? No explanation needed, a hard no on that one. If there is no one on the top of the ballot you want to vote for and if you don't like the other candidates enough to even bother then you won't show up at all. Trump is a wet blanket for the smaller races and will suppress the vote. His vitriol has smothered the electorate with a feeling that makes them want no part of the election and covered the rest of the GOP in Trumpiness. It's hard to wash off.

The best hope the Republicans have is to make Hillary enough of a villain to increase the people coming out to vote against her (similar to what the Democrats already have with Trump) and then hope those people are also voting Republican down-ballot. That's putting a lot of eggs in a very flimsy basket.


The scenario I laid out earlier has clinton in the 380-400 electoral vote range. High numbers for sure, but not rarified air. There is an extreme version of this race that makes Clinton's electoral map look less like Barack Obama and much more like Ronald Reagan. Yup, that's right all those Reagan worshiping, wall smashing, REO Speedwagon loving 80's products: Hillary Clinton could be rightfully compared to Ronald Reagan. (I'll wait while that washes over you.)


Ok, so what?

I'm not saying Hillary and Ronald (a much funnier sounding comparison without their last names) are going to be compared policy wise, although it's getting harder to tell the current Democratic platform apart from Reagan, in a lot of ways. I'm saying that when it comes to the electoral map Hillary Clinton could be poised to match and even make history, for an entirely different reason than being the first female POTUS.


In 1984, Reagan won big. No, seriously. He won by a LOT. Reagan received 525 electoral votes in 1984. He lost Minnesota... that is all. Minnesota is one of the single most Democratic leaning states in the Union. In 1972 Nixon won with 520 electoral votes. (Whoops!) In 1936 F.D.R won with 523 electoral votes. (Yahoo!) Those are your 500+ club members. Roosevelt, Nixon, Reagan.

Hillary Clinton in 2016 may very well join them. To do it she needs Texas. Meaning she would, in fact, have to mess with Texas. (As we know, that is generally ill-advised.) Typically I wouldn't entertain this as possible for a Democrat... but this year... with Trump... I am. There are three reasons: 1) Ted Cruz. 2) The Latino vote. 3) Donald Trump, ya'll.

If Clinton wins Texas it means she may win states like Missouri, Indiana, Arkansas, Kentucky, on and on... and could easily climb over 500 electoral votes.


In 1912 Teddy Roosevelt ran for his third term in office. He ran as a third party. Teddy's goal was to win. In part, T.R. had some regrets, like not running a for his “second” term that would have been his third. He wasn't elected the first time, so he could still run again and it would have been ok. This is also pre-22nd amendment, but the two-terms rule was generally understood. Any way, T.R. thought President Taft had taken the Republican party astray and wanted to get rid of him. (That's the most simplified version of that ever written.) Roosevelt lost. So did Taft. Enter Woodrow Wilson, a Democrat that on his own was likely to loose to Taft. Thanks to Roosevelt and the later named “Bull Moose” party, Taft was easily defeated and Wilson becomes the President.

1992 had a very similar dynamic. Bill Clinton won with less than 45% of the popular vote thanks to business man and hilarious SNL parody character Ross Perot. Perot sucked away a lot of Republican voters dissatisfied with Bush Sr.'s administration and handed a relatively unknown Governor from Arkansas the Presidency.

2016 is shaping up to have a similar feel. Hillary Clinton is struggling to get to 50% in the polls, and depending on just how many voters break third party she could languish at 44%-48% in the popular vote. That would be a problem, but it appears the third party impact creates a much greater disparity and loss for Trump, who likely tops out near 42% (maybe lower) with a third party candidate. There is very little chance either of the third party candidates (Libertarian or Green, or whatever the others are) win a single electoral vote (unlike Roosevelt who did win some) but they will keep Trump from winning as many as he could win in a head to head race.

Hence, the Bull Moose effect. Republicans, dissatisfied with their candidate, have split to a third party in greater numbers than the Democrats, thereby paving the way for Clinton's victory.

It means a bad day for the Republicans in the Presidential election and, if things continue as they have, could lead to a loss of the House, the Senate, and maybe both. If all of that happens it won't just be “Obama's third term” as Republicans like to say, it will be a solidification of Obama's policy legacy for the foreseeable future, and possible a schism too deep to heal for the Republican party. It would be their terrible, horrible, no good, very bad day.

- Adam Sommer