Presidential elections are odd things. They happen regularly and unfold in familiar ways, but because they’re spaced out by several years, we tend to forget some of the specifics. Combine that with the seemingly insatiable (albeit certainly not universal) interest in figuring out how the thing will unfold, and you have a period of two years or so in which Americans grab onto particular ideas or numbers and shake them under peoples’ noses as though they’re meaningful, when they aren’t. They do so even when they should know better.
It is our mandate here at The Washington Post to ensure our readers are as well-informed as they can be. Often that means presenting new information that our reporters have uncovered. Sometimes, though, it means preparing readers to fend off unimportant or misleading information.
With that latter purpose in mind, we have put together this set of guidelines to bear in mind when presented with information about the election. We’re operating somewhat above the elementary level here; we won’t be getting into the fact that the election is decided by the electoral college rather than the popular vote, for example. (Well, except right there.) Instead, we will explore common pitfalls into which people drop when presented with a seemingly interesting bit of information.
The standard map of election results presents the winner of the presidential contest in a state by coloring that state red (if the Republican earned more votes) or blue (if the Democrat did). This convention is less than 25 years old, but it has been standardized. So be it.
For the 2020 contest, that yields a map that looks like this.
But it’s bizarre to present, say, Florida and West Virginia as equivalent. Not only because the former has about six times as many presidential electors, but because Florida was also far less overwhelmingly supportive of Donald Trump.
If we use gradations of support, the map looks like this — with fewer brightly colored states.
We can look at this another way, presenting the 2020 results from the most heavily supportive of Joe Biden to the most heavily supportive of Trump. Did you know, for example, that Texas gave Trump his third-narrowest margin of victory?
The scale is admittedly challenging here, given that it includes D.C., where the results were off the scale. But the point remains: There are relatively few states where the election was a lopsided blowout for one candidate or the other.
Biden and Trump got at least 40 percent of the vote in 38 and 40 states, respectively. Biden got at least 45 percent of the vote in 30 states; Trump in 33.
Over the past nine presidential contests, most states have cast more votes for a Democrat and a Republican at least once. Twenty states (and D.C.) have consistently voted for the candidate from only one party; the other 30 have flipped at least once. The average number of times a state has voted for the Republican in the past nine elections is 4.7 — just under half (since the scale runs from zero to nine, inclusive).
(Each state below includes a nine-square grid, corresponding to those nine elections. They’re identified in the key at the top left next to Alaska.)
There have at times been a lot of emphasis placed on certain geographic locations being “bellwethers” for the outcome of a presidential race. Ohio is a typical one: The candidate it backed from 1988 to 2016 won the presidency. (And before that, too, but you get the point.) Then in 2020, it didn’t.
Some Trump supporters tried to elevate this divergence as evidence that something weird happened. But bellwethers are bellwethers — not because they drive the result, but because they happen to reflect it. And when they don’t, it means they aren’t bellwethers anymore. Give up a hit in the ninth inning and it’s not that the hit is suspicious and didn’t count; it’s that you no longer have a perfect game.
Swing states operate in the same way. We look at certain states as “swing” states if we expect that the results in those states will be close. Usually, this is because the results were close in the past and — importantly — the results in those states were determinative.
As you can see, though, there have been lots of times over the past nine cycles when the results in a state were close (here defined as a two-party margin of less than five points). Over the past nine cycles, there have been 92 state-level results that count as close under that definition. Whether that makes a “swing” state is largely a function of whether we collectively agree to talk about it that way.
Before 2020, for example, there wasn’t a lot of discussion about Georgia as a swing state. Then it became a swing state. So now it’s a swing state. Until it isn’t.
Perhaps the most important thing to remember about the election is that we often try to make polls do things polls aren’t designed to do.
For example, Fox News on Thursday released new polling from four swing states (see above) showing narrow leads for Trump in three. But there is no difference between a “narrow lead” and “a tie” and “trailing narrowly” when we’re talking about a gap of a point or two in a poll with a margin of error of three points.
Fox hyped its poll results by comparing the new numbers with the 2020 results in each state and its most recent previous polls. Here’s what those comparisons look like when we remove the labels and include the margins of error.
Take Wisconsin. Biden won there by one percentage point (after rounding) and the two candidates were tied in January in the new poll. So who’s going to win? No one knows.
Okay, so how about Michigan? Trump was up two points in February and is now up three points. To which the proper response is: So? Does this show traction for Trump? No, since the difference is attributable to the vagaries of how and when the poll was conducted. Not to mention that there are months for actual traction to occur, meaning that Trump’s lead very well could extend significantly or that he could at some point fall behind.
It’s easy to absorb all of this information and collapse into a state of observational inertia. Eyes glazed over, you look at another close poll and think, “Well, this doesn’t really matter.” And you’re pretty much right.
What, you thought I was going to try to disabuse you of that idea?
There’s a lot to learn from polls and observations about the election, certainly. But very little of what we’ll hear about over the next few months will answer the most popular question: Who will win? For that, we’ll probably need to rely on actual votes, not some sort of augury using past vote margins in purple states.
Presidential elections are odd things. They happen regularly and unfold in familiar ways, but because they’re spaced out by several years, we tend to forget some of the specifics. Combine that with the seemingly insatiable (albeit certainly not universal) interest in figuring out how the thing will unfold, and you have a period of two years or so in which Americans grab onto particular ideas or numbers and shake them under peoples’ noses as though they’re meaningful, when they aren’t. They do so even when they should know better.
It is our mandate here at The Washington Post to ensure our readers are as well-informed as they can be. Often that means presenting new information that our reporters have uncovered. Sometimes, though, it means preparing readers to fend off unimportant or misleading information.
With that latter purpose in mind, we have put together this set of guidelines to bear in mind when presented with information about the election. We’re operating somewhat above the elementary level here; we won’t be getting into the fact that the election is decided by the electoral college rather than the popular vote, for example. (Well, except right there.) Instead, we will explore common pitfalls into which people drop when presented with a seemingly interesting bit of information.
The standard map of election results presents the winner of the presidential contest in a state by coloring that state red (if the Republican earned more votes) or blue (if the Democrat did). This convention is less than 25 years old, but it has been standardized. So be it.
For the 2020 contest, that yields a map that looks like this.
But it’s bizarre to present, say, Florida and West Virginia as equivalent. Not only because the former has about six times as many presidential electors, but because Florida was also far less overwhelmingly supportive of Donald Trump.
If we use gradations of support, the map looks like this — with fewer brightly colored states.
We can look at this another way, presenting the 2020 results from the most heavily supportive of Joe Biden to the most heavily supportive of Trump. Did you know, for example, that Texas gave Trump his third-narrowest margin of victory?
The scale is admittedly challenging here, given that it includes D.C., where the results were off the scale. But the point remains: There are relatively few states where the election was a lopsided blowout for one candidate or the other.
Biden and Trump got at least 40 percent of the vote in 38 and 40 states, respectively. Biden got at least 45 percent of the vote in 30 states; Trump in 33.
Over the past nine presidential contests, most states have cast more votes for a Democrat and a Republican at least once. Twenty states (and D.C.) have consistently voted for the candidate from only one party; the other 30 have flipped at least once. The average number of times a state has voted for the Republican in the past nine elections is 4.7 — just under half (since the scale runs from zero to nine, inclusive).
(Each state below includes a nine-square grid, corresponding to those nine elections. They’re identified in the key at the top left next to Alaska.)
There have at times been a lot of emphasis placed on certain geographic locations being “bellwethers” for the outcome of a presidential race. Ohio is a typical one: The candidate it backed from 1988 to 2016 won the presidency. (And before that, too, but you get the point.) Then in 2020, it didn’t.
Some Trump supporters tried to elevate this divergence as evidence that something weird happened. But bellwethers are bellwethers — not because they drive the result, but because they happen to reflect it. And when they don’t, it means they aren’t bellwethers anymore. Give up a hit in the ninth inning and it’s not that the hit is suspicious and didn’t count; it’s that you no longer have a perfect game.
Swing states operate in the same way. We look at certain states as “swing” states if we expect that the results in those states will be close. Usually, this is because the results were close in the past and — importantly — the results in those states were determinative.
As you can see, though, there have been lots of times over the past nine cycles when the results in a state were close (here defined as a two-party margin of less than five points). Over the past nine cycles, there have been 92 state-level results that count as close under that definition. Whether that makes a “swing” state is largely a function of whether we collectively agree to talk about it that way.
Before 2020, for example, there wasn’t a lot of discussion about Georgia as a swing state. Then it became a swing state. So now it’s a swing state. Until it isn’t.
Perhaps the most important thing to remember about the election is that we often try to make polls do things polls aren’t designed to do.
For example, Fox News on Thursday released new polling from four swing states (see above) showing narrow leads for Trump in three. But there is no difference between a “narrow lead” and “a tie” and “trailing narrowly” when we’re talking about a gap of a point or two in a poll with a margin of error of three points.
Fox hyped its poll results by comparing the new numbers with the 2020 results in each state and its most recent previous polls. Here’s what those comparisons look like when we remove the labels and include the margins of error.
Take Wisconsin. Biden won there by one percentage point (after rounding) and the two candidates were tied in January in the new poll. So who’s going to win? No one knows.
Okay, so how about Michigan? Trump was up two points in February and is now up three points. To which the proper response is: So? Does this show traction for Trump? No, since the difference is attributable to the vagaries of how and when the poll was conducted. Not to mention that there are months for actual traction to occur, meaning that Trump’s lead very well could extend significantly or that he could at some point fall behind.
It’s easy to absorb all of this information and collapse into a state of observational inertia. Eyes glazed over, you look at another close poll and think, “Well, this doesn’t really matter.” And you’re pretty much right.
What, you thought I was going to try to disabuse you of that idea?
There’s a lot to learn from polls and observations about the election, certainly. But very little of what we’ll hear about over the next few months will answer the most popular question: Who will win? For that, we’ll probably need to rely on actual votes, not some sort of augury using past vote margins in purple states.
