None of the state chambers targeted by Democrats flipped in 2020. As in any complex system, many factors likely contributed to these results. Let’s take a look at three possible groups of explanations that may have contributed to state legislative outcomes: gerrymandering, roll-off/ticket-splitting, and enthusiasm/awareness gap.
This analysis aims to describe what happened, but not why the outcomes occurred. An analysis of the “why” would need to address issues like the resonance of Democratic messaging, the impact of the pandemic on Democratic field programs, the impact of misinformation, voter suppression tactics, and other factors. (For an early analysis of some of the “why”s, check out our early analysis.)
One theory is that Democrats weren’t able to flip chambers because of gerrymandering. It goes like this: Democrats picked up the easiest seats in 2018 (minus MN and some PA Senate seats, which weren’t up in ‘18), so what was left in 2020 was harder and Democratic votes couldn’t crash over the “red wall.”
If this is true, we would expect Democrats running for state legislative seats to have, at the chamber level, received more votes than Republicans, but won fewer seats.
For instance, we often use Wisconsin in 2018 as an example of this “seat share gap:” Democrats running for state assembly received 54% of the votes, but only won 36% of the seats (18% seat share gap).
Bottom line: results are mixed.
In most chambers, we should have picked up more seats than we did, particularly in NC (both chambers), WI Assembly, GA House, and PA Senate. We also should have at least tied the MI House and flipped the MN Senate. But generally speaking, Democrats running for state legislative seats did not get enough votes to have flipped chambers even if the maps had been more fair.
[Note: we generally have not restricted analysis to contested seats because gerrymandering purposefully creates “safe” seats by packing like-minded voters into the same districts, and we need to look at the whole chamber to truly see the effect of gerrymandering. The exception is Wisconsin, where we cannot locate vote totals in uncontested races, so our analysis includes contested races only.]
Given this year’s state legislative chamber performance:
2020 Election State Legislature Chamber Seat Share Gap
Chamber | Total Votes Won by Dems |
Seats Now Held by Dems |
Seat Share Gap |
MN Senate | 50.12% | 47.76% | 2.36% |
MI House | 49.82% | 47.27% | 2.55% |
NC House | 49.06% | 42.50% | 6.56% |
PA Senate* | 48.95% | 36.00% | 12.95% |
NC Senate | 48.50% | 44.00% | 4.50% |
GA House | 48.35% | 42.78% | 5.57% |
WI Assembly** | 48.30% | 38.28% | 10.02% |
PA House | 46.79% | 44.55% | 2.24% |
AZ Senate | 46.48% | 46.66% | -0.18% |
TX House | 43.40% | 44.67% | -1.27% |
FL House | 42.61% | 35.00% | 7.61% |
***As of 11/12/20 for AZ, FL, MN, NC, PA Senate, WI. As of 11/13/20 for GA, MI, TX, PA House.
*Only half of PA Senate seats were up for election but the whole state could vote for President. This may affect findings (i.e., the other 25 seats could lean more Democratic).
**We were unable to find vote totals for uncontested candidates in Wisconsin, so totals only reflect contested races.
Note: Seat Share Gap indicates the difference between the percentage of the votes that Democratic legislative candidates received in the chamber and the percentage of seats the Democrats won in the chamber.
If it’s not just gerrymandering, then what else could explain the results? It could be ballot roll-off and/or ticket-splitting.
If it’s roll-off (where a person votes only for the top of the ticket and not further down the ballot), we would expect there to be significantly fewer total votes for down-ballot candidates in the chamber than the total number of votes received at the top of the ticket in the chamber.
If it’s ticket-splitting (where a person votes for different parties at the top and bottom of their ballots), we would expect to see downballot candidates get higher vote totals or higher vote shares than upballot candidates running in the same party. In the case of this analysis, all non-presidential candidates are considered downballot candidates compared to presidential candidates. However, there are also cases in which the aggregate total for state legislative candidates is larger than the vote total for Senate or gubernatorial candidates from the same party.
Bottom line: results are mixed. It appears that there were a significant number of people who voted for Biden (regardless of the voter’s party registration), who either voted for Republican state legislators (split their ticket) or who did not vote for any state leg candidate (rolled off).
Evidence of Ticket-Splitting and Roll-Off
Chamber | Biden- statewide | Trump - statewide | Dem Fed Senate | GOP Fed Senate | Dem Gov or Senate Special | GOP Gov or Senate Special | Dem Leg | GOP Leg |
---|---|---|---|---|---|---|---|---|
MN Senate | 1,716,207 | 1,483,551 | 1,565,815 | 1,397,606 | N/A | N/A | 1,596,658 | 1,532,446 |
TX House | 5,211,406 | 5,860,096 | 4,844,433 | 5,931,602 | N/A | N/A | 4,478,832 | 5,682,507 |
NC Senate | 2,680,501 | 2,754,103 | 2,566,389 | 2,661,404 | 2,830,501 | 2,582,720 | 2,572,688 | 2,662,458 |
NC House | 2,680,501 | 2,754,103 | 2,566,389 | 2,661,404 | 2,830,501 | 2,582,720 | 2,614,446 | 2,565,882 |
MI House | 2,795,184 | 2,649,063 | 2,725,692 | 2,640,672 | N/A | N/A | 2,657,658 | 2,651,208 |
FL House | 5,295,138 | 5,667,716 | N/A | N/A | N/A | N/A | 3,795,537 | 5,075,082 |
WI Assembly | 1,630,570 | 1,610,030 | N/A | N/A | N/A | N/A | 1,253,361 | 1,325,547 |
AZ Senate | 1,670,260 | 1,659,272 | 1,714,466 | 1,635,352 | N/A | N/A | 1,377,358 | 1,586,063 |
PA House | 3,400,711 | 3,341,535 | N/A | N/A | N/A | N/A | 2,991,838 | 3,356,734 |
GA House | 2,471,918 | 2,457,846 | 2,372,004 | 2,458,626 | 2,375,836 | 2,422,219 | 2,200,949 | 2,350,385 |
***Federal/gubernatorial numbers of 11/13/20; As of 11/12/20 for AZ, FL, MN, NC, PA Senate, WI. As of 11/13/20 for GA, MI, TX, PA House.
This table indicates how many votes candidates from the major parties received in their state or chamber in the case of state legislative races (all state legislative candidates are included in one vote total).
Key: Green cells indicate candidates/chambers that outperformed their party’s Presidential candidate. Red cells indicate candidates who received at least 200,000 fewer votes than their party’s Presidential candidate OR chambers where a party received at least 400,000 fewer votes than their party’s Presidential candidate.
In the table above, green boxes provide evidence of ticket-splitting and red boxes provide evidence of roll-off. Green boxes are where downballot candidates received a higher number of votes than their party’s presidential candidate.
For instance, Gov. Cooper (D) received the most votes of any candidate on the NC ballot. This means that some voters who voted for Republicans at another level of the ballot also voted for Cooper. Evidence of ticket-splitting may also reflect some number of people who chose to vote downballot but not higher on the ticket.
Red boxes are where downballot candidates received at least 400,000 votes less than their party’s presidential candidate. For instance, Dems running for the Texas House received 733,000 less votes than Biden. This means that many Biden voters chose not to vote downballot.
We also looked at the magnitude of roll-off in 2020 vs 2016. Because we do not yet have presidential vote totals by state legislative race, we had to exclude chambers we worked in with more than 1 uncontested race. This left us with the completely contested chambers of the Minnesota Senate and the North Carolina Senate. The Michigan House was also included since only one seat was uncontested. That candidate’s vote total was deleted from that party’s presidential candidate total to simulate the removal of that district.
Generally, we don’t see a huge change in roll-off between 2016 and 2020 in the Minnesota Senate, North Carolina Senate, and Michigan House. There was a bit more roll-off this year in the North Carolina Senate, almost identical roll-off in the Michigan House, but a bit less in the Minnesota Senate.
2016 vs 2020 Roll-Off in Contested Races
Chamber | 2016 Roll-off |
2020 Roll-off |
Increase |
Minnesota Senate | 104,726 (3.59%) | 90,002 (2.75%) | -14,724 (-0.84%) |
North Carolina Senate | 118,912 (3.56%) | 211,312 (3.83%) | 92,400 (0.27%) |
Michigan House | 189,070 (3.95%) | 221,853 (4.01%) | 32,783 (0.06%) |
***As of 11/13/20.
Another explanation for the results may be an enthusiasm or awareness gap between voters for the top and bottom of the ticket. This gap can also be expressed as a lack of coattails for downballot candidates from upballot candidates with higher profile races.
The results indicate that this is likely the case.
Evidence of Enthusiasm/Awareness Gap
Chamber | % of total pres votes for Biden | % of total chamber votes for leg Dems | Dem Enthusiasm Gap | % of total pres votes for Trump | % of total chamber votes for leg GOP | GOP Enthusiasm Gap |
---|---|---|---|---|---|---|
MN Senate | 52.39% | 50.12% | -2.27% | 45.29% | 48.10% | 2.81% |
TX House | 46.40% | 43.40% | -3.00% | 52.17% | 55.06% | 2.88% |
NC Senate | 48.59% | 48.50% | -0.10% | 49.93% | 50.19% | 0.26% |
NC House | 48.59% | 49.06% | 0.47% | 49.93% | 49.99% | 0.06% |
MI House | 50.56% | 49.82% | -0.74% | 47.92% | 49.70% | 1.78% |
FL House | 47.86% | 42.61% | -5.25% | 51.23% | 56.97% | 5.74% |
WI Assembly | 49.57% | 48.30% | -1.27% | 48.94% | 51.08% | 2.14% |
AZ Senate | 49.40% | 45.60% | -3.80% | 49.08% | 53.52% | 4.44% |
PA Senate | 49.86% | 48.95% | -0.91% | 48.99% | 51.05% | 2.06% |
PA House | 49.86% | 46.79% | -3.07% | 48.99% | 52.50% | 3.51% |
GA House | 49.52% | 48.35% | -1.17% | 49.24% | 51.63% | 2.39% |
***As of 11/13/20.
It is clear that gerrymandering, roll-off, ticket splitting, and enthusiasm/awareness for candidates influenced races in many of the chambers we targeted in 2020. Voting behavior and electoral outcomes are multivariate, and the results we see have numerous causes. Some confluence of these and other factors resulted in Democrats not flipping legislative chambers and not picking up as many seats as they should have in numerous chambers across the country.
For transparency, we are linking to our dataset, which we constructed based upon publicly-available voting information.