For more than 25 years, the Cook Partisan Voting Index (PVI) has served as a useful metric to compare the partisan leanings of all 435 House seats and 50 states. Today, we're publishing historical PVI values for every set of congressional districts dating back to 1997 — all in one place. Together, the data illustrate how polarization and shifting boundaries have cut swing seats in half, profoundly altering House members's electoral incentives.
The new Cook PVI archives are more than a great academic resource; they allow for a full comparison of every Congress since 1997 and show how voters's increased self-sorting — more than gerrymandering — has made primaries the only race that matters in most seats.
About the Cook Partisan Voting Index (Cook PVI℠)
First introduced in 1997, the Cook PVI measures how each state and district performs at the presidential level compared to the nation as a whole. A score of D+2, for example, means that in the 2016 and 2020 presidential elections, the district performed about two points more Democratic in terms of two-party vote share than the nation did as a whole, while a score of R+4 means the district performed about four points more Republican.
So, for example, if the national popular vote was 50% Democrat and 48% Republican, a district with a PVI of R+4 should expect to see a GOP vote share more like 52%. If a district performed within half a point of the nation in either direction, we assign it a score of EVEN.
For this release, we've updated every historical dataset to be consistent with the formula we used for our 2022 post-redistricting release, which weights the most recent presidential cycle three times more heavily than the preceding one to assess partisanship (rather than the past 50/50 mix). In other words, the newest PVI dataset gives 75% weight to the 2020 presidential election result in each district and 25% weight to the 2016 result.
In most cases, the formula adjustment barely altered the PVI score. But in seats that have shifted rapidly towards one party — such as those in South Florida or the suburbs of Atlanta, recently — the new formula "leans in" to the transformation underway.
The Incredible Shrinking Swing Seat 1997-2023
Today, just 82 of 435 districts sport Cook PVI scores between R+5 and D+5 — our traditional definition of what constitutes a "swing seat." That's down exactly 50% from the high-water mark of 164 back in 1999. Since then, "R+5 or More" seats have jumped from 150 to 189, while "D+5 or More" seats have jumped from 121 to 164. That means Democrats need to win about 66% of swing seats to win the majority today, up from 59% back in 1999.
But the trading range of what's "competitive" has also narrowed over the past few decades as split-ticket voting has declined and fewer candidates have been able to outrun their party. Back in the 105th Congress (1997-1998), there were 26 Democrats sitting in districts with PVIs of R+3 or more and 14 Republicans in seats with PVIs of D+3 or more. Today, there are just five Democrats and four Republicans in seats surpassing those thresholds.
In fact, the number of "hyper-swing" seats — those with Cook PVI scores between D+3 and R+3 — has declined even more precipitously. Today, there are just 45 seats between those benchmarks, a 58% decline from the high water mark of 107 in 1999.
A big driver of the House's dominant "play-to-the-base" psychology: over time, the two parties’s constituencies have drifted even further apart: today, the median House Democrat represents a district with a PVI score of D+13, up from D+8 in 1997. The median House Republican holds a district with a PVI score of R+13, up from R+7 in 1997. In other words, the gap between the two parties’ median districts has grown from 15 to 26 points.
What's Most to Blame for the "Hollowed Out" Middle?
Our 13 unique sets of Cook PVI scores give us a powerful tool to isolate and quantify the impacts of changing voting patterns and redistricting on the makeup of the House. And though both play a role, urban/rural polarization has driven most of the competitive decline.
In many minimally altered districts, the electorate has simply become much more homogeneous than it used to be. For example, the boundaries of rural eastern Kentucky's 5th CD haven't changed much since 1997, but its PVI score has shifted from R+2 to R+32 as its voters have abandoned national Democrats. Likewise, the PVI score of Wisconsin's 2nd CD has ballooned from D+8 to D+19 as Madison has continued to attract like-minded liberal voters.
We've released new PVI scores in odd-numbered years following presidential elections: 1997, 2001, 2005, 2009, 2013, 2017 and 2021. We've also recalculated scores in six even-numbered years following redistricting: 1998 (after court-ordered redistricting in North Carolina and Virginia), 2002 (after the 2000 census), 2006 (after mid-decade redistricting in Georgia and Texas), 2012 (after the 2010 census), 2018 (after new court-ordered maps in Pennsylvania) and 2022 (after the 2020 census).
Of the net 77 swing seats that have vanished since 1997, 58% of the decline (net -45 seats) resulted from areas trending redder or bluer from election to election, while 42% of the decline (net -32 seats) resulted from changes to district boundaries.
The Impact of Elections vs. Redistricting on PVI Scores, 1997-2023
This disparity would be even more lopsided if not for the 2020 election, which produced a decent rebound in swing seats after a hyper-polarized 2016 cycle. The strong anti-Trump trend of diverse and highly college-educated suburbs, particularly in Texas and Georgia, pushed a net 14 seats below the R+5 threshold. Likewise, Trump's strong performance in working-class white and Hispanic seats pushed a net six seats below the D+5 threshold.
However, the 2022 redistricting cycle gave partisan mapmakers the opportunity to retrench and shore up their own vulnerable seats, and the number of R+5 to D+5 seats receded from 90 to 82. Texas' new GOP gerrymander alone accounted for the entire decline.
This realignment and retrenchment fits a longer-term pattern: elections have produced the biggest increase in strongly Democratic seats, while redistricting has swelled the number of strongly Republican seats. Why? Over the last three decades, most populous suburbs have trended blue, especially at the presidential level. But Republican state legislators have held the power to redraw far more congressional districts than Democrats.
Finally, the impact of redistricting hasn't been even. Across the nine states where independent or bipartisan commissions held primary responsibility for redistricting as of 2022 (Arizona, Colorado, California, Hawaii, Idaho, Michigan, Montana, New Jersey and Washington), swing seats fell 39% between 1997 and 2023, from 41 to 25. But across the 19 Republican-controlled states, swing seats fell by a whopping 70%, from 56 to 17.
The modest success of state-based redistricting reform efforts have slowed the decline of competitive seats by offsetting the impacts of gerrymandering elsewhere. For example, in the early 2000s, both California and New York's legislatures passed blatant "incumbent protection" gerrymanders. For 2022, California's commission and a court-appointed expert in New York drew plans that didn't take partisanship into account.
A consequence of this unevenness: the battle for control of the House is increasingly being fought in a handful of states with more “neutrally” drawn maps. The 18 states where courts or independent/bipartisan commissions implemented maps in 2022 (this list includes Iowa, where Republicans approved a map drawn by a nonpartisan state agency) account for 47% of all House seats, but 65% of all swing seats.
Although it's likely maps in North Carolina, Ohio or other states will be revisited prior to 2024, our next nationwide overhaul of Cook PVI values will be for the 119th Congress, following the 2024 presidential cycle. The upcoming election could answer whether we hit "peak polarization" in 2016, or if the slightly depolarizing 2020 election was simply a blip on the path to an even more geographically fractured electorate.
Our subscribers have first access to individual race pages for each House, Senate and Governors race, which will include race ratings (each race is rated on a seven-point scale) and a narrative analysis pertaining to that race.