This Is Election Content
Hello Friends,
I know I promised an issue on the Federal Debt, but I wanted to write a quick issue on the election. As always, I’ll still include the performance of my stock picks. Track Performance.
The Federal Debt issue is coming soon. If this email was forwarded to you and you would like to receive future issues, please subscribe!
What’s In This Issue
I’ll explain the nuts & bolts of election polling and why this is a particularly hard year for accurate polling.
Stocks Recommended: None.
Stocks Mentioned: None.
Polling Musings
I know what you are thinking:
Hey! Wait a minute!! This is a newsletter about stocks. There isn’t going to be anything about stocks in it? I want my money back! You have a lot of explaining to do!!! - You (reader)
I understand your frustration! I’ve refunded everyone who has ever paid for this newsletter. Please check your Venmo.
I use this newsletter as a forum to explain complicated concepts in a simple way. At least, that is my hope. Polling reminds me a lot of stock picking. You have some data today and you analyze that data to predict an event in the future. The present-day data will never perfectly predict the future. However, there is some good news. Even if the final results are unknown, most of the risks are known.
Dewey Beats Truman
I’ve been consuming a crazy amount of polling content and now I’m an expert! A month ago, I knew nothing!
An election poll probably calls 1,000 people at random and asks them who they are going to vote for. If 600 people say Candidate A, then the poll will report that Candidate A is leading in the poll 60% to 40% - Me (exactly one month ago)
Whoa, was I wrong! That is a terrible way to poll. It’s how you get something like this
"Dewey Defeats Truman" was an incorrect banner headline on the front page of the Chicago Daily Tribune (later Chicago Tribune) on November 3, 1948, the day after incumbent United States president Harry S. Truman, won an upset victory over Republican challenger and New York governor Thomas E. Dewey, in the 1948 presidential election. It was famously held up by Truman at a public appearance following his successful election, smiling triumphantly at the error.
A lot had to go wrong to create that incredible photo, but polling error played a big role. Due to a worker’s strike, the newspaper had to go to the printing press several hours earlier than usual. The paper needed a headline even before states had finished reporting results. In turn, the paper relied on polling data which showed Dewey as the runaway favorite. Those polls were conducted on home telephones which were only owned by wealthy families. Truman outperformed dramatically with lower income voters and won the election.
In order to get an accurate read on the election, pollsters need to take a different approach. Today, high-quality pollsters split the electorate into different demographic buckets based on age, gender, race, income & other attributes. They determine the polling averages of the individual buckets and then make estimates to the size of each bucket in the voting population.
What about 2016?
But if pollsters are so smart now, how did they mess up 2016 so bad? Remember 2016? Pollsters said Clinton would win, but then she didn’t. Seems like they haven’t learned much in 68 years. - Literally Everyone
There are a lot of theories for why polling was off in 2016, but let’s discuss some of the popular ones.
National vs State Polling - Most of the high-quality polling was done at the national level. It was assumed that if Clinton had a big enough lead in the popular vote, she would win the electoral college. The final national polls were off by 3 percentage points. Clinton finished with a 2-point win in the popular vote. We can empirically say that a 2-point win in the popular vote is not enough to guarantee a win in the electoral college.
Incorrect Education Weighting - Voters with a college degree tend to be more responsive to pollsters. In past elections, this wasn’t a big source of bias. But in 2016, it was a huge source of bias. In otherwise homogenous states (ex: Wisconsin), college attainment was a key differentiator. If your sample didn’t include enough non-college educated white voters, you severely underestimated Trump’s share of votes.
The next two are more speculative. Hard to prove, but many people believe these theories.
Undecided Voters - Voters deciding in the week leading up to the election tended to vote for Trump. The “Comey Letter” could have played a role.
“Shy Trump” Voters - This is the most talked about theory. Some percentage of Trump voters didn’t want to tell the pollster they were voting for Trump.
That’s a lot of stuff to go wrong for the pollsters! I know we want accurate predictions, but it’s a pretty tough gig. Polling organizations have changed their practices to address reasons 1 & 2. Top pollsters are now polling more frequently in battleground states and appropriately weighting college attainment.
Now that the pollsters have addressed some of the gaps, should we expect more accurate polls?
What about 2020?
Each election brings new polling challenges and 2020 will be no different. Tonight, I’ll make some predictions on the polling stories from this election.
Record Turnout
As I write this newsletter, 98.5+ million people have already voted. In 2016, a total of 137 million people voted. Election experts are predicting 160+ million votes in total. If those predictions come true, voter participation will be the highest since 1908! As I mentioned earlier, pollsters attempt to determine turnout rates for each demographic to weight their model. Normally, they can rely on previous election demographic distributions. But in Texas there may be 50% more votes than in 2016! There is no way the distribution will be the same. The laws have changed, voting is easier. Companies are giving employees time off to vote. Facebook won’t shut up about voting. Pollsters use questions to determine if a voter is a “likely voter”. Questions include “when was the last time you voted?” or “do you know the location of your polling place?”. The questions that helped determine past voter turnout may not be as predictive this time.
Look out for big polling misses in TX, GA, FL and/or NC.
Do You Speak English?
Most live telephone polls are conducted in English. As the country becomes more diverse, pollsters will miss more pockets of voters. This will be particularly challenging in states with big Latino populations. In Texas, third-generation Mexican-Americans tend to be more Republican than first-generation. In Florida, Cuban-Americans tend to be Republican and Puerto Ricans tend to be Democrats. If a pollster lumps Latino voters together and only sample voters who respond to an English speaking pollster, they will introduce a lot of polling error.
Look out for big polling misses in TX, FL & NV.
The Popular Vote Is Useful
In a surprise twist, the popular vote is relevant again. With Texas & Georgia becoming surprising swing states, 7 of the 10 most populous states are now swing states (TX, FL, PA, OH, GA, NC, MI).
The winner of the popular vote wins the election. The magnitude of the electoral college win reflects the popular vote.
Hot Take
Election night will go smoothly. A winner will be announced in the early hours of Wednesday, Nov 4th. Both candidates will accept the results graciously. Control of the Senate is determined by two run-off races in Georgia in January. Control of the House does not change.
The End
Thank you very much for reading. Next time, I will cover the Federal Debt. If you were forwarded this email, be sure to subscribe to get future issues.