Election Season 301

Introduction

Traditionally, voting followed a single, uniform process across America. Voting was done at a polling place, which covered a small geographic area, called a precinct. Polls opened at their given time, voters showed up, checked in, filled in the bubbles, and left. Polls closed at their given time (though of course anyone in line by then could still vote) and then started counting ballots. That made the process of interpreting election returns a function of one variable: geography. For instance, if it’s later in the night and most of the Pennsylvania votes left to count are in Philadelphia, which is heavily Democratic, then the final margin will be more Democratic than the current margin. That basic principle has created some very complex models for interpreting results, and informed practically all of the media systems and intuition for interpreting results as they come in.

Unfortunately, basing predictions on geography is no longer nearly enough to correctly interpret election results as they come in. The coronavirus pandemic changed everything. While it’s true that America has been moving away from the traditional voting model for a while now, including many states in the West which had already radically redesigned their voting systems, a shift of this magnitude hasn’t happened before. Thus, it is increasingly important to recognize the impact non-traditional voting has on interpreting incoming election results. Misunderstanding how these new voting systems create different return patterns than the old ones has already caused multiple premature, incorrect race calls in 2018 alone. Understanding election returns in real time this November will require understanding the new methods of voting and the patterns for election returns they create.

What’s Changed?

Due to the Covid-19 pandemic, a large number of people have opted to vote using some method besides the traditional in-person, election day voting. Unfortunately, our state and local reporting systems were not designed to accommodate such a large shift towards alternative methods of voting, so understanding the reporting process and calling races will be much more difficult this year.

Take the “precincts reporting” metric that most official sources use. It used to be that if someone said that 150/200 precincts in a state had reported their results, it was safe to assume that about ¾ of the votes were counted. But this cycle, in most states, a precinct “reporting” could mean it has only reported in-person, election day votes, which won’t be close to all the votes cast– far less than half of them in some states.

The temptation, then, is to assume that election day in-person votes are going to be a sort of sample of the total vote. If the whole vote is about half in-person, half absentee ballots, then it’s okay to track in-person votes and assume that absentee votes are going to be similar.This will fail spectacularly in yielding accurate predictions..

This brings us to the first new rule of interpreting election results: don’t judge how much of the vote is left to count by the number of precincts reporting.

Types of ballots

The new major division in election returns is ballot types, so it’s important to understand the different terms and what they mean. It’s equally important to use clear language when talking about different kinds of ballots, especially since states sometimes use some terms, officially or colloquially, differently from each other.

• Election day ballots, sometimes shortened in writing to “eday” these are the traditional style of ballots described in the introduction. They are cast on election day and counted immediately after polls close. In previous years, the vast majority of ballots cast in most states would be election day ballots, but that’s no longer true. These are sometimes referred to as “in person” votes, but since there are other kinds of in-person voting, that can be misleading.
• Early in-person ballots: In many states, polling places will open for a period of days or weeks before election day for voters to cast their ballot. Depending on the state, these ballots may either be counted with or before election day ballots. For a while, this was the only way most people could vote early, so it will often be called “early voting”. But early voting can now refer to a category of different techniques.
• Absentee ballots: ballots which are not cast in a polling place, but are instead sent directly to voters. In some states, voters need an approved reason to receive an absentee ballot, but most states with that requirement, waived it for this election.
• Mail ballots, aka mail-in, vote-by-mail, and VBM: In some western states, the polling place model of voting had already been replaced by a system where every registered voter was simply mailed a ballot. These are called mail ballots.
• Provisional ballots: special ballots which are given to voters at polling places when it’s not clear whether that voter is allowed to vote. There aren’t many of these and they are not counted on election night. Instead their validity is decided on by the Board of Elections a few days later. Military ballots: Some military members vote using a special kind of universal ballot if their state can’t get them an absentee ballot in time. Like provisional ballots, these are small in number.

Absentee and mail ballots sound similar

They are similar. The difference between the two is mostly about the overall system of a particular state If a state has a vote-by-mail system, such as California, then the default voting method involves mailing a voter a ballot called, appropriately, a “mail ballot.” Absentee ballots, on the other hand, are part of a system where voting by mail is not the default, and where sending out ballots by mail is an exception. The process usually involves an application to receive ballots by mail because, otherwise, the state will assume the voter will vote in person. While the difference is often just in terminology, it’s important to realize that voting systems in vote-by-mail states are designed to process large volumes of mail ballots, and usually do so as efficiently and quickly as possible. Unfortunately, some in-person voting states have obscure, unnecessarily difficult ways to count absentee ballots, processes that can start days after the election, take weeks, or even be done in secret until the final results, depending on the state. New York is notoriously bad for these.

Who’s voting how?

It would be nice and simple if the different voting methods were used equally by all types of voters. Sadly, we don’t live in a simple world. Owing to a variety of factors in-person ballots–both early and election day–are expected to be much more Republican than absentee and mail-in ballots. This is an estimate, based on surveys asking voters how they intend to vote, and statistics about ballots requested so far.

It would also simplify things if we knew a nice, stable fraction of each party voting by each method. That way, we could do a little math and find a rough estimate of what absentee votes should look like, given what in-person votes look like, or vice-versa. Unfortunately, even if we had a definitive number nationwide we could pin down from polling, it’s going to vary so much state by state so much that national proportions would be useless anyway. Moreover, different states have different rules for who can request an absentee ballot. These range from states which haven’t expanded their restrictive absentee ballot access at all since the pandemic began, to states which have essentially created a backdoor vote-by-mail system using their already existing absentee ballot system.

Early in-person and election day in-person votes do usually differ from each other in a given state, but how each party performs in each method, and by what margins, varies so much from state to state that there’s no larger trend to speak of, even before the pandemic.

How to watch

There are three things to keep in mind while watching election results come in:

1. Results that seem extreme, even absurd, probably are.

2. Do not assume a race is over before knowing what ballots are left to count.

3. It’s never a bad idea to withhold judgement while waiting to see how things roll out.

Generally, it’s not difficult to tell how most races are going to go. Not exactly of course - that’s why we have elections. Still, results are broadly going to follow two predictors: past results and polling.

Considering polling first, it’s true that polling can be off. In fact, most polling averages will be off by a few points, and every year there’s a race or two where polling misses by 10 or even 15 points. But if the results in a race are really, really off from the polling average, it’s likely because the results that are in are skewed in favor of one candidate or party, and will normalize as more ballots are counted. Similarly, while large shifts on the presidential level do occur, a swing of 20 or 30 points in a single state is unlikely to stick as more results come in. In short: an early result that seems too crazy to be true probably is.

The more polling that has gone into a race, the more polling averages can be trusted to be in the right neighborhood. The much remarked-upon 7% Wisconsin presidential polling miss in 2016 was from an aggregation of only five polls conducted in the last couple weeks of the race. So a polling average of, for instance, a Congressional race where there aren’t any recent polls, or if all available polls are published by the same pollster, should provide less guidance.

Past results in a state or district are a fairly straightforward way to know what to expect from this year. Simply, things are going to change, but not too much. If the most recent previous election for a state or district was in 2018, that’s probably the best baseline to use. If not, the 2016 presidential results have shown themselves to be remarkably predictive for all sorts of races. Ideally, of course, checking whether it’s reasonable to make any assumptions on election night about downballot races would be done by checking whether the election night presidential numbers seem reasonable, since of course ballots are going to be cast for both.

What is and isn’t reasonable is of course eyeballing to a certain extent. A more homogenous district or state will be prone to larger swings on the presidential level. It’s also worth keeping in mind that this year looks like it’s going to be much better for Biden than Clinton. Holding that baseline assumption and focusing on relative shifts is a good way to think about things. For example, If Biden is doing five points better than Clinton nationally, then a particular state being 15 points more Democratic in 2020 than in 2016 is a large but realistic change relative to national changes. On the other hand, given that same assumption, a state being 15 points more Republican than in 2016 is a truly radical shift, and warrants more skepticism.

What’s out?

Setting aside states with vote-by-mail systems and the small number of provisional and military ballots that may decide a race that’s coming down to the wire, the three categories of votes that could be counted separately will be election day in-person, early in-person, and absentee.

Some states will tabulate early in-person votes before election day, and then release them immediately or soon after polls close. If a state suddenly goes from no votes counted or only a few hundred to tens of thousands, it’s likely that it released all of the early votes at once. In other cases, early in-person votes will be tabulated with election day in-person votes throughout the night.

As for absentee ballots, it may take days or even weeks after election day before they’re fully counted in some states, for reasons ranging from logistical difficulties in the counting process to rules, in states like North Carolina, that allow officials to accept ballots post-election day if they’re post-marked by a certain date. In such states, there’s nothing to do but embrace the fact that a race may take longer than usual to call. Moreover, while in previous elections, given accurate information about the ballots yet to be counted, such as their composition and how large a party’s margin in them is, we could make assumptions about the likely outcomes in some races. However, since voting and reporting processes are different this election from previous cycles, good information about ballots will be harder to come by, meaning any assumptions we make are likely to be incorrect and ineffectual. Thereofere, all there would be to do is wait, which, realistically, is not a bad strategy at all compared to jumping the gun and calling a race with insufficient information.

That said, some states will announce how many absentee ballots they have, and some will even announce the absentee ballot number in each county. Using county by county totals, a crude guess about the partisanship of the absentee ballots can be modeled, but this can’t be done on the fly, so if that estimate is important to you, set up a spreadsheet beforehand and double-check your formulas.

Ultimately, there’s no point in the night where waiting is a bad strategy to be using. You may not know how much, if any, of each kind of ballot has been counted in a particular race, and if that’s the case, there’s just not enough of the vote in to tell much.

Vote-by-Mail States Are Better...And Worse

The good news about vote-by-mail results is that they’re simpler to understand. There may be a few in-person votes that need to be counted, but then all that’s left are the mail-in ballots, and following those results is as simple as waiting for them to be counted. But simple doesn’t mean fast. Vote by mail states count the mail ballots over a period of days, or even weeks.

You can get a pretty good picture of the race based on what gets counted on election night. The general rule of thumb is that the later the ballots are counted, the more Democratic they are. If a Democrat’s winning in a vote-by-mail state on election night, they’ve probably won. If a Republican is up by a lot on election night, they’ve probably won. If it’s close, there’s nothing to do but wait. What constitutes a Republican being too far ahead on election night for the remaining ballots to flip the race in favor of the Democrat? It varies by state, but a flip in any race with an election night margin of more than 10% is rare.

States will often release some measure of “ballots left to count” on election night after they stop counting. In general, the more votes left to count relative to votes already cast, the more the race can change. The most dramatic reversals have occurred in races where around half of the vote was counted after election day. This is something to be careful about, because California, Washington, and Alaska all accept ballots that arrive after election day (as long as they’re postmarked by then), so the vote tally will grow.

A Race Just Got Called

A race being called just means that someone, somewhere, is sure one candidate has won. It has no weight or real meaning beyond whoever cares about that person’s opinion. Generally news organizations, most prominently the AP, have staff devoted to analyzing election results, and don’t announce that one candidate has won until they’re convinced it’s not possible for the other candidate(s) to win. The AP has been wrong before, but generally, when they declare a winner, that’s as official as it gets until the government certifies the vote count a few weeks later. Other news organizations beyond the AP will also call races, although their smaller budgets mean they have a less sophisticated operation, and their need to make headlines can lead to premature calls. The only official sources for election winners are secretaries of state when they certify results weeks later.

There are also people on various online platforms, ranging from pundits and local political professionals to teenagers with Twitter accounts, who will call races. It’s best not to put too much stock in this. Some of them are going to be right, but plenty of them are inexperienced, misinformed, or just trigger-happy. Besides, you can just wait for a more trusted source to be sure. They’ll be looking at the same data.

In races that are too local for the AP to call, local news organizations usually fill in the gap. Some will have a good grasp of the way election returns come back in their state, but if they rely on their intuition built up from previous years, they could make some grave errors. The usual pressure to be the first source with a piece of news only grows stronger every year as the internet disseminates race calls faster and faster. Generally, if a news organization (or pundit) is the only one to call a race, caution is warranted. If they’ve called it and the number of votes reported so far is much less than this race had in 2016 or 2018, outright skepticism is warranted.

From the precinct to your screen

Election administration is not standardized in this country. It’s often not even standardized within a given state. As a result, there is no one way that ballots are counted and counts are reported.

Ballots can either be counted by hand or by machine. Most states have machines, either for voters to cast their vote on directly, or to scan ballots that voters fill out. The places that report votes immediately or very soon after polls close will have counted the votes ahead of time and released vote totals once they were allowed to. Generally, this is just for early voting, but some states do it with absentees, too.

For election day voting, the precincts that report earlier will be the ones without lines when polls close, and which can count ballots with machines. Some places will have collected paper ballots throughout the day and will then have to count them after polls close, which takes much longer.

Once a precinct, or in some cases a county, has a vote total that they want to report, the reporting process can take many forms. In some states, officials can upload the results more or less directly to a state reporting website. In others, the process can be as time-consuming as physically driving a flash drive to the county clerk and standing in line while the clerk uploads each flash drive one by one. This creates a large variance in reporting times. In addition, if something goes wrong during a particular counting, results will not be reported from that precinct until the problem or discrepancy has been fixed.

Each state has its own way of reporting the results once they’ve finished the counting process. In some cases it’s directly to an easily accessible state website and you’ll know them as soon as the earliest news organization to find out. In other states the county websites update first and then the state websites grab the results from the counties.

It can get even more decentralized in ways that make it nearly impossible for individuals to track on election night. Fortunately, this is where the AP comes in; they have spent decades as the pre-eminent source for compiling election returns. As a news cooperative, the AP, instead of displaying these results directly to consumers, sells a live feed of the data to other news organizations, where most people see it. A competing service, Decision Desk HQ, has recently gained some attention and earned a reputation for often being faster, but not always as accurate. Almost every election result displayed comes from a state, AP, or Decision Desk HQ feed.

The AP uses a variety of methods to obtain their results. They’ll look at county, city, town, and state websites, call just about every clerk and administrator they can find the phone number of, and use stringers, people who physically go to precincts to read results posted in their windows. Stringers, an old tradition of election night reporting, have been falling out of favor for a while, and this election might be the death blow to that practice.

Warning Sign (of things to come)

There is no actual legal requirement to report votes on election night, and sometimes, especially in small towns in New England, a town clerk will tally up the results, record them, and go to bed without reporting them to the state or media. This is annoying, but ultimately there’s nothing to be done about this. Occasionally, an entire county will do this, which is even more annoying, but is what it is. Places simply not reporting results, or making other silly mistakes, is generally a sign of a system that relies on a large number of humans, not a sign of anything nefarious.

Doing the Math

Projecting the winner of an election with partial results in is tricky business under the best of circumstances. Basically, the procedure is to determine how much of the remaining vote the currently losing candidate would have to win to win overall, and then make a judgement about how realistic that is. The addition of multiple ballot types to track makes this more complicated.

The different ballot types can be viewed separately. So if early voting ballots create a lead for candidate A of 20,000 votes, and election night ballots add another 180,000 votes, then the question becomes whether candidate B can make up 200,000 votes in the absentees. The next step is determining how much candidate B would have to win the outstanding absentees by. So if there are 600,000 outstanding, a 200,000 win margin would be 67% to 33% (400,000 to 200,000). Then compare that margin to absentees that are coming in.

More complicated is the situation when all the methods of voting are simultaneously partially reported. Because the different voting methods will have both turnout differences and differences in partisan composition, making an estimate of the partisan composition of what’s out based on the partisan composition what’s in isn’t really possible without making some assumptions that are going to make the estimate that results too inexact to be used anyway. (Essentially this is because you’re trying to solve an equation for two variables). At this point, the best way forward is to wait until counties have entirely reported, so that they can be compared to 2016, or wait until voting methods have been entirely reported, so you can follow the process described above.