- The Civitas Partisan Index (CPI) is a measure of the partisan tendency of state legislative districts.
- By the most stringent measure, the CPI was accurate in over 94% of legislative races.
- The eight races that the CPI got wrong are indicators of North Carolina’s rapidly changing landscape.
With North Carolina’s 2020 election finally (almost) behind us, it is time to take stock.
For over a decade, the Civitas Institute has presented the Civitas Partisan Index (CPI) before state legislative races to give readers an idea of which districts are likely to have competitive races and which races are likely safe for one party or the other.
The 2020 CPI proved to be highly accurate. Even when election results varied from the expectations produced by the CPI, those variations informed us of shifts in voting trends in our state.
What is the Civitas Partisan Index?
The CPI is modeled after national indexes such as the Cook Political Report’s Partisan Voting Index. It is a measure of the base partisan leanings of a North Carolina state legislative district relative to the state as a whole. It is made by finding the average results of several statewide races and comparing it to the average results for those races in each district. Based on that variation, we rank districts as safe, likely, or lean Democratic, safe, likely, or lean Republican, or toss-up. For example, if District X has a Council of State average vote total favoring Republicans by 10 percentage points, compared to a statewide COS vote total favoring Republicans by 3 percentage points, District X would have a score of R +7 (likely Republican).
There are two major differences between the 2020 CPI and the 2018 CPI. First, while we used the 10 2016 NC Council of State races (governor, auditor, commissioner of labor, etc.) to construct the 2018 CPI, the 2020 CPI incorporated those races plus four 2018 statewide judicial races to account for the ongoing shift in urban and rural voting patterns.
Another change came from court-ordered redistricting, which significantly shifted the CPI ranking of 17 legislative districts.
How well did the CPI do?
While the Civitas Partisan Index is not a predictive model (since it does not include factors like incumbency and candidate quality), it does provide a baseline of expected major party distribution of votes in legislative races. A well-constructed index should accurately reflect voting behavior.
So, how did the 2020 CPI do?
Pretty well.
As seen in Table 1, a person using nothing except the CPI would have correctly predicted the result in every one of the 108 NC House districts the CPI assigned a partisan tendency.
Table 1: Winning NC House races by Civitas Partisan Index rating.
The CPI was not perfect on NC Senate races but was still highly accurate. Of the 47 senate races the CPI assigned a partisan tendency, it was correct in all but one, an accuracy rate of 97.9% (see Table 2). The only district that went against expectations was NC Senate district 13. The reason for this exception will be discussed below.
Table 2: Winning NC Senate races by Civitas Partisan Index rating.
One thing that helped with the CPI’s accuracy rate is that 2020 was a max turnout election with both sides getting strong support from their base voters. When one side dominates turnout or persuasion, that variation can make some districts go against their normal tendencies. That happened in the “blue wave” election of 2018, when Democrats won in eight “lean” or “likely” Republican districts (the aforementioned NC Senate 13 went against expectations to go Republican that year also). With both sides getting their voters out this year, the results more closely reflected how those districts “should” go based on their general voting tendencies.
What about those toss-up districts, and what about that “lean Democratic” race the CPI got wrong?
Fifteen districts were categorized as toss-ups in the CPI, meaning that they had a one percentage point or less (technically 1.49 or less) partisan tendency. There were even three districts listed as “D+0” or “R+0,” meaning that their rating was nearly even. However, each of those districts has a small partisan tendency.
How well did this CPI do in those toss-up districts? As seen in Table 3, those districts lived up to their rating, with the CPI being correct in only eight of those 15 districts.
Table 3: Results in fifteen CPI “toss-up” districts.
Most of the eight toss-up races the CPI missed this year are not just cases of randomness in close elections. When the CPI got toss-up races wrong, it failed in directions that reflect changing voting patterns.
The two toss-up districts that the CPI rated as slightly favoring Republicans, but which voted for Democrats (House districts 103 and 104), are both located in southern Mecklenburg County. They are located in an area of inner Charlotte suburbs that was formally known as the “Republican wedge,” an area that had regularly sent Republicans to the General Assembly until 2018. That area is shifting towards Democrats faster than the CPI, which is based on recent election results, can keep up.
Similarly, several districts where Republicans won despite a slight tilt in the CPI in favor of Democrats were in rural parts of the state. House District 1 is in the northeastern corner of North Carolina while House districts 43 and 45 are mostly located in the suburban and rural parts of Cumberland County.
Sen. Danny Earl Britt Jr’s reelection in NC Senate District 13 was the only race in which the Civitas Partisan Index assigned a partisan lean that it missed in 2020.
Two districts deserve special mention. House District 46 occupies parts of Columbus and Robeson County. The CPI rated it as D+1, but Republican incumbent Brenden H. Jones sailed to victory with over 60 percent of the vote. An even more extreme result was in Senate District 13, also in Columbus and Robeson counties. The CPI rated as D+2 (lean Democratic), but incumbent Republican Danny Earl Britt, Jr. cruised to a 64-36 win.
Some of their winning margins can be attributed to their incumbency status and relatively weak opponents. However, a possibly bigger factor is the rapid political transformation of Robeson County. Civitas’ Dallas Woodhouse wrote of that transformation, calling Robeson County “a once Democratic stronghold that has gone red.” You should read Woodhouse’s account to see why Robeson has changed so quickly, but I will note one fact: in 2004, the average share of the vote for Republican Council of State (the ten statewide elected executive offices such as governor, attorney general, and treasurer) candidates in Robeson County was 26.6 percent. By 2020, the average for Republican Council of State candidates in Robeson County was 53.9 percent.
Depending on what happens with redistricting next year, both Columbus-Robeson districts are likely going to be rated as “lean Republican” or even “likely Republican” in the next CPI.
The high turnout of the 2020 election and regional voting trends helped make the CPI even more accurate than usual this year. With redistricting and a presidential midterm election, 2022 will likely be a much tougher test for the CPI.
