It’s the holy grail of presidential election campaigns: knowing which states will be decisive in the Electoral College. We have our guesses: States like Arizona, Michigan, North Carolina and Wisconsin are widely expected to be among the top swing states for 2020. But how confident can we really be in those expectations five months before the election?
To find out, we went back and checked which states election handicapper The Cook Political Report thought would be swing states as of mid-June 2004, 2008, 2012 and 2016. We then compared those early ratings to what the actual swing states turned out to be after the election. (To be clear, Cook does great work, and we’re not trying to pick on them — we’re simply using them as a proxy to see where conventional wisdom stood at the time. It’s actually surprisingly hard to find historical sources for this, and Cook is one of the only outlets we could consistently find with early assessments of the presidential race dating back to 2004. In an email, Charlie Cook, the founder of The Cook Political Report, told us he thinks of its ratings as rough guidelines, particularly at this point in the cycle: “My view has always been that we are not trying to predict outcomes as much as help our readers see which states are safely in a party’s column, which aren’t but could get into play, which ones are competitive but one side seems to have an advantage, and which ones are really, really close.”)
[Related: Our Pollster Ratings]
Importantly, much of how accurate these early ratings are hinges on how you define what is a swing state. For instance, “swing state” could mean simply a state that is closely contested — a toss-up state in Cook’s lingo. But it could also mean a bellwether state — aka a state whose results closely match the national popular vote. This might sound like semantics, but it’s actually an important distinction: In a roughly tied election, toss-up states and bellwether states are more or less the same; however, in an election in which, say, one candidate leads nationally by 10 points, the bellwether states may not actually be very close. So how you define a swing state matters.
And as it turns out, early ratings are often wrong about which states will be toss-ups in November. But, as we will explain, that doesn’t mean they are useless or even bad. In fact, Cook does have a pretty good track record of identifying an election’s bellwether states early on, and arguably that’s more important.
First, though, let’s look at which states Cook expected would be toss-ups at this point in the cycle in the last four presidential elections and which states actually ended up being toss-ups.1 The teal states in the chart below are those that Cook correctly rated as toss-ups; the pink ones are states that Cook rated as toss-ups but did not turn out to be that close; and the gray ones are states that Cook did not rate as toss-ups but ended up close enough to count as one.
As you can see, Cook’s record of calling toss-ups is mixed. In June 2008, Ohio was the only toss-up state it called correctly. In late May 2016, it correctly identified New Hampshire and North Carolina as toss-ups but missed nine other states that turned out to be tight. On the bright side, in April 2004,2 Cook correctly foresaw seven states’ toss-up status, missing only four and incorrectly calling toss-ups in just Florida and Missouri. And in late May 2012, it correctly pegged all four of the eventual toss-up states; while at first glance it appears they cast too wide a net by also rating Colorado, Iowa, Nevada, New Hampshire and Pennsylvania as toss-ups, those states all just barely missed the cutoff for being actual toss-ups (then-President Barack Obama won them each by 5 to 7 points).
But what went wrong in 2008 and 2016 wasn’t that Cook misread the electorate in specific states; it’s that it didn’t realize how competitive (or uncompetitive) the presidential race would be overall. (To be clear, this is not Cook’s fault; there was no way for anyone to anticipate that the economy would collapse in fall 2008, sending Obama to a landslide victory, or that then-FBI Director James Comey’s Oct. 28 letter to Congress seemed to bring the 2016 campaign within a normal-sized polling error of a Donald Trump victory.)
This is clear when you look at the states that Cook’s springtime ratings implied would be bellwethers: Most of them turned out to be actual bellwethers in November.3
In late May 2016, Cook was wrong to think former Secretary of State Hillary Clinton was an overall favorite, but it did correctly anticipate that Colorado, Florida, Michigan, Nevada, Pennsylvania, Virginia and Wisconsin would closely mirror the national popular vote (although three other states also did). Similarly, in June 2008, Cook’s ratings implied a presidential race that was too close to call, but even after the race shifted in Democrats’ favor, four of the six states Cook viewed as the likeliest bellwethers indeed turned out to be bellwethers (although so did four other states Cook missed). And in 2004 and 2012, Cook had even more success pre-identifying bellwether states than they did pre-identifying toss-up states.
So where does that leave us in 2020? Well, as of June 16,4 The Cook Political Report lists six states5 as toss-ups: Arizona, Florida, Michigan, North Carolina, Pennsylvania and Wisconsin. And since Democrats have the advantage in states worth only 232 electoral votes and Republicans have the advantage in states worth only 204, Cook is therefore also expecting these six states to be bellwethers.
[Related: The Latest Political Polls Collected By FiveThirtyEight]
History suggests you can largely trust those bellwether ratings, too — though you should expect the election to still throw us a few curveballs. One or two of those states will probably not be that close to the national popular vote; two or three states not on this list probably will turn out to be good national bellwethers.
The toss-up ratings, on the other hand, should not be taken literally — at least not at this stage. And that’s because, while handicappers are pretty good at assessing how states will vote in relation to one another, it is more challenging to forecast the national political environment so far in advance. But in a way, that’s OK; the bellwethers are really what matter.
It doesn’t really matter if, say, Virginia or Texas is decided by just 1 point, because if either one of those scenarios comes to pass, the election is likely a landslide. Knowing the bellwethers in advance is arguably far more valuable because we can then focus our energy on watching who is leading in them — because whoever wins the bellwethers wins the election.