How Red Or Blue Is Your State? Your Congressional District?

Last fall, Joe Biden became the first Democratic presidential candidate to win Arizona since 1996 and the first to carry Georgia since 1992. But does this mean Arizona and Georgia are now blue states? 

Well, not yet — at least by our definition.

Allow us to introduce (or reintroduce)1 you to FiveThirtyEight’s partisan lean metric — our way of measuring the partisanship of a state or district, similar to the Cook Political Report’s Partisan Voter Index or Inside Elections’s Baseline. We define “partisan lean” as the average margin difference2 between how a state or district votes and how the country votes overall. For example, if a state has a FiveThirtyEight partisan lean of R+5, that means it is 5 percentage points more Republican-leaning than the nation as a whole. Put another way, in an election that’s exactly tied nationally, we would expect Republicans to win that state by 5 points.

By this definition, Arizona and Georgia are still (slightly) red states — R+7.6 and R+7.4, respectively. While they may have voted for Biden in 2020, they did so by margins smaller than his national popular vote win of 4.5 percentage points. (These partisan lean values also factor in the results of other elections, in which Republicans did better in Arizona and Georgia. More on that in a minute.) But here are the new FiveThirtyEight partisan leans for every state for the 2021-22 election cycle, updated with the results of the 2020 elections.

The FiveThirtyEight partisan lean of every state

Average margin difference between how each state votes and how the country votes overall in congressional and gubernatorial elections, according to a blend of presidential and state-legislative election results

State Partisan Lean State Partisan Lean District of Columbia D+68.2 Arizona R+7.6 Massachusetts D+32.6 Florida R+7.6 Hawaii D+31.6 Iowa R+9.7 Vermont D+27.5 Texas R+12.0 Maryland D+25.9 Ohio R+12.4 California D+25.5 Alaska R+14.6 Rhode Island D+24.0 South Carolina R+18.6 New York D+20.0 Indiana R+20.0 Delaware D+13.7 Montana R+20.0 Illinois D+13.4 Mississippi R+20.3 Washington D+12.4 Louisiana R+20.5 Connecticut D+12.1 Kansas R+20.7 New Jersey D+12.0 Missouri R+21.2 Oregon D+10.6 Nebraska R+24.8 New Mexico D+7.0 Utah R+26.3 Colorado D+6.4 Kentucky R+27.1 Virginia D+4.6 Tennessee R+29.4 Maine D+4.0 Alabama R+29.6 Minnesota D+1.9 Arkansas R+31.8 New Hampshire D+0.3 South Dakota R+32.2 Michigan R+1.6 West Virginia R+35.5 Nevada R+2.5 Idaho R+37.0 Pennsylvania R+2.9 Oklahoma R+37.2 Wisconsin R+4.1 North Dakota R+37.2 North Carolina R+4.8 Wyoming R+49.7 Georgia R+7.4

Sources: State election websites, Daily Kos Elections

We’ve also calculated partisan lean for every congressional district, which you can find on our GitHub page. (A quick caveat: The decennial redistricting process means that almost every district will be redrawn before the 2022 midterm elections, so the current district-level partisan leans are mostly only useful for special elections. Don’t worry, though: We’ll calculate the partisan leans of the new congressional districts as soon as they’re finalized.)

So now that you know what our new partisan lean scores are, let’s talk about how — and how not — to use them. First, we want to emphasize again that FiveThirtyEight partisan leans are expressions of relative partisanship; that is, they don’t necessarily tell us how red or blue a place is in absolute terms. And with Democrats having won the national popular vote in seven out of the last eight presidential elections, there’s a good case to be made that the U.S., as a whole, is actually a tad left of center, and that a state with a FiveThirtyEight partisan lean of, say, R+1 may actually vote Democratic more often than it votes Republican.

But the reason we prefer a relative metric to an absolute one is so that we can better understand how a state or district might vote in a given national environment. For example, if polls of the generic congressional ballot are suggesting Republicans will win the House popular vote by 6 points, we might be able to infer that a Democratic congresswoman sitting in a district with a D+4 partisan lean is in danger. (Assuming a uniform national swing, Republicans would carry that district by 2 points.) Orienting partisan lean around a nationally tied election doesn’t imply that a national tie is our default expectation; it just simplifies the math of overlaying it on different national moods (D+6, R+3, etc.).

The other thing to know about our partisan lean score is that it doesn’t simply reflect the results of a single presidential election (i.e., 2020). Rather, this version of partisan lean (meant to be used for congressional and gubernatorial elections) is calculated as 50 percent the state or district’s lean relative to the nation in the 2020 presidential election, 25 percent its relative lean in the 2016 presidential election and 25 percent a custom state-legislative lean based on the statewide popular vote in the last four state House elections.3  

Empirically, we’ve found that this blend of a few different elections yields the most accurate expression of a place’s true partisanship. While the most recent presidential election is the most important input, the previous one has predictive power too, especially for states (like Arizona and Georgia) that shifted suddenly toward one party (they often experience some reversion to the mean). And the state-legislative component accounts for the fact that some states (like West Virginia) vote differently for down-ballot offices than they do for president.

As a reminder, partisan leans for all 50 states (plus Washington, D.C.) and all 435 congressional districts are publicly accessible on our GitHub page. We enthusiastically invite you to use it for all your 2022 congressional- and gubernatorial-race needs; we’ll be using it in our own analyses, but we’d love to see it start to catch on with others too. If you do anything cool with it, don’t forget to tag us or drop us a line.

Mary Radcliffe, Aaron Bycoffe, Nate Silver and Geoffrey Skelley contributed research.