“The Ballot is Stronger than the Bullet”
-Abraham Lincoln
I’m writing this newsletter a little over a week after Election Day and a few days after the news networks declared Joe Biden the President Elect. For the purposes of this article I am going to assume that Biden is our next President. While I personally don’t see anything that would over turn the current vote counts, I also don’t get to make those decisions. What is even more unclear, is who is going to be in charge of the Senate come January. Due to the resignation of Senator Johnny Isakson for health reasons, both Senate seats in the state of Georgia were up for election this month. Under Georgia law, if no candidate gets at least 50% of the vote, the top two candidates have a run-off election in early January. That is where we now stand, there are 48 Democrats and 50 Republicans in the Senate and those two run-off elections, in the same state, will determine if the Senate keeps a Republican majority or if the Senate is tied, which assuming once again that Harris is Vice-President means the Democrats would control the White House and both Houses of Congress.
I find this very unlikely to happen. Based on the two senate races and the Presidential numbers, Georgia is a pretty evenly divided State. Even if the vote is split, where one of the Republicans and one of the Democrats win, then the Senate stays in GOP hands. This would mean divided government, and this is what the markets have been celebrating for the last week, along with positive vaccine news. A divided government means we don’t have the instability of Trump (even for Trump’s supporters, you have to admit he says a lot of things that are at best unclear and that often made the markets nervous) and yet you rein in the most extreme regulatory and tax policies that Biden has proposed. Depending on your political view-point this may be a negative on one side or the other, but from a Market standpoint, this is baby bear’s porridge.
But I can’t really talk about that with any major authority right now, because nobody knows what’s actually going to happen. Maybe the Democrats win both Senate seats? Maybe some actual election fraud is uncovered that casts doubt on Biden’s election? Maybe, and of more immediate importance, the Pfizer vaccine turns out to have a problem and doesn’t get approved by the FDA? Any one of these issues would greatly affect the markets and how we should invest in them. That’s why we didn’t make changes based on the polls, because man, were those things wrong! And that’s what I really want to talk about this month since I can’t talk about election clarity as I assumed. I want to talk about polls, surveys and big data and how they affect what we do and how wrong they can be.
First, for the sake of this illustration I’m going to assume the current vote totals are correct. In Nevada the pre-election polls had Biden up by 5%, he won by 3% - that’s within the margin of error. Likewise, in Pennsylvania he was shown up by about 4% and won by about 1%, a little outside the margin of error but not horrible. But then in Ohio and Florida for example, where the pre-election polls had Biden down 1% and up 3% respectively, he ended up losing by 8% and 4% - meaning the polls were off by 7% in both states. This was even worse than the notoriously bad polls in 2016. Biden’s leads were much bigger so the misses didn’t seem to flip the election, but the polls for Senators and Congress People were even worse. Pre-election polls showed Maine Senator Susan Collins getting drubbed, yet she ended up winning by almost 10%! What the heck happened here? Then of course, there is baseball. In the sixth game of the World Series, up 1-0 and facing elimination, the Tampa Bay Rays pulled starting pitcher Blake Snell. Snell had only given up two hits and struck out nine batters through those six innings (for non-sports fans, that’s good) and yet he was pulled. He was taken out because of analytics, the math that pretty much runs how baseball is played now days. The numbers tell you that after a pitcher has faced every batter on the opposing team twice, his effectiveness drops and you’re better off putting in a new pitcher. Finally, there is economics, the profession with which I am technically affiliated. Market projections, inflation assumptions, interest rates, unemployment numbers and the like are all precisely calculated, sometimes down to meaningless hundredths or thousands of a percent. These projections are a combination of surveys, like the way they get polling data, and analytical number crunching, like the type that tell you a very effective pitcher who the other team can’t hit should be pulled out in favor of a much less talented pitcher. (By the way, the Dodgers came back and won the game and the World Series in case you’re wondering where I’m going with all this.)
The election polls require participation in surveys; surveys done online, on the phone and occasionally in person. Have you ever done one? The problem is that people don’t participate in these surveys anymore. It’s not that they call landlines, they call cell phones and they do online polls as well, but our cell phones have gotten smarter about blocking unknown calls and even when people do get through, the average person just won’t answer the survey. Back in the last century, it wasn’t uncommon to get 30% or higher response rates for polling surveys, and those people tended to be more like the electorate at large. Now the response rate is under 1%, and as one pollster put it, the people who actually answer are not reflective of the population at large, the very fact that they are the outlier one percent who will answer already makes them “strange” to use his term. Couple this with the inability to do more in person polling this year due to the pandemic and it’s not hard to see why the polls were so bad. Then there is what you do with those results, this is where the algorithms come in. If you poll 100 people and 56% are male, and 95% are White and 1% are Hispanic, you have to weight those averages based on what you think the make-up of the total voting population is going to be, so you take the 56% males votes and weigh them at 49%, and the 95% of White votes and weight them at 77% and take that 1% of Hispanic voters and weight them up to 18%. But once again, this is a guess, maybe White participation is higher or lower than you are projecting, maybe women who make up 51% of the population end up being 55% of the voters. Or maybe that 1% Hispanic voters, of only 1% of people who answer the phone, don’t reflect how other Hispanic voters will vote, and you’ve just made that wrong assumption 18% of your poll results. Of course, the polls weight for other things as well, such as education level, blue- or white-collar job etc. The more subcategories, the more complicated the algorithm, the more likely a small error in one spot becomes a huge error in the results. Just like the numbers tell you that in the over 300,000 regular season baseball games over the last 100 years, a pitcher is less effective after each batter has seen him twice is a real, provable stat. But that stat applies mostly to insignificant games played in summer, when there’s always tomorrow. In an elimination game of a World Series, in actual reality, some people can handle the extra pressure and some people cannot and what that player's average performance over a career looks like has no bearing on whether or not he can perform at a critical moment. In other words, the math can be correct and and the conclusion can be wrong.
I’ve spoken to you all before about how I feel the formula the Federal Reserve uses to calculate inflation is flawed. Well what if the inputs are also flawed? What if the people answering the Bureau of Labor and Statistics surveys every month about their employment are not being totally honest? Then the unemployment numbers would be wrong. But what if they are honest, but the people they get in touch with are not reflective of the population as a whole, well then, the number could be wrong for that reason also. They try to fix this by that algorithm, taking people’s answers and weighting them against what they think is the working population at large – but what if the algorithm and the surveys are wrong? They often are, that’s why these numbers get revised from previous months when the surveys are replaced by the actual number of people on unemployment or paying into Social Security. This then carries forward to inflation, where the Fed relies on a survey of prices and then their own algorithm, to business sentiment where once again the Small Business Administration is relying on the surveys of the businesses they call, then adjusted by their particular formula. This goes on and on, stock analysts rely on businesses giving them sales and expense projections, the CEO and CFO’s giving those projections rely on their regional managers, their suppliers, their sales force to give them the information to input those numbers. Never before have we been able to process these numbers as quickly as we can today – but what if the inputs themselves, done by good old-fashioned fallible humans, who are more politically divided and suspicious of authority than ever, are not reflective; compounded by algorithms that aren’t accurate anymore? Then we get bad information, faster than ever before. Just like retired comedy musician Tom Lehrer used to say, “Life is like a sewer, you only get out of it what you put into it.”
All these potential problems and sources of misinformation is why we structure investment portfolios around you first and the economy second. Because we do know how old you are, and what your monthly expenses are and who you want to leave your money too, and if you don’t answer my survey, I won’t work with you. All the rest, the changes around the edges we will make if Biden is in fact the next President, the tweaks to those changes we will make if the Senate is controlled by Chuck Schumer or Mitch McConnell, what we use for an inflation rate on our retirement plan projections, what interest rate the bond and cash side of your portfolio will generate – we will of course try to project all these things. But just like with the election, the output is only as good as the data that is inputted and while the calculations are better than ever, it seems like the inputs may be getting worse.
I hope everyone is well and is happy the election is behind them – even as it’s not really behind them. Please as always let me know if you need anything or if you know anyone who might benefit from our help.
