Abstract: Downballot roll-off is a phenomenon where people vote for top-of-ticket candidates but then do not vote for downballot offices like state legislature. There is remarkably little academic or popular literature examining this phenomenon in state legislatures, and even less exploring whether the phenomenon may have a partisan valence. Building on analyses we conducted in 2020 and 2021, we looked at the Presidential, Gubernatorial, US Senate, state senate, and state house votes for 20 battleground chambers across even-year elections between 2012-2022. We found that, while downballot roll-off in terms of raw vote totals is common among both major parties, it occurred much more frequently for Democratic than for Republican candidates. Vote share (the percentage of the total votes cast in the race) for parties follows a similar trend: downballot Republicans captured a larger share of the vote than top-of-ticket Republicans more often than downballot Democratic candidates outperformed top-of-ticket copartisans. The vote total roll-off pattern was magnified when looking at contested state legislative races only. Republicans saw roll-off in vote total in a minority of contested race comparisons, such that, remarkably, Republican state legislative candidates outperformed top-of-ticket candidates in vote total in most race comparisons. A similar pattern was also seen in vote share for contested races as in all races, with downballot Republicans more likely to capture a larger share of the vote than top-of-ticket candidates than downballot Democrats, something we call an enthusiasm gap. These findings help us understand trends in partisan control of state chambers and top-of-ticket races.
Downballot roll-off is a phenomenon in which a person votes for one or more candidates in “top-of-ticket” races and refrains from voting for candidates further down the ballot. It can be observed as a differential in vote totals between co-partisan top and bottom ticket candidates. Roll-off is a potential issue for political parties looking to build power across levels of government. Based on our prior analyses in 2020 and 2021, we believe the phenomenon may affect Democrats more frequently and deeply than Republicans.
Because of secret ballots, we can’t be 100% certain why vote totals differ up and down the ballot. We know, for instance, that some voters choose to split their ticket between partisans at different levels of the ballot. Additionally, down-ballot races sometimes go unchallenged, and some states and chambers have staggered elections, where only half of the chamber’s seats are up for election in a given year. Both can create situations where party loyalists genuinely do not have a candidate on their ballot to vote for, which may result in a gap in aggregate vote totals at the top and bottom of the ballot.
But when large gaps exist between total votes at the top and bottom of a party ticket, when gaps persist even when only analyzing districts with contested races, or when gaps do not exist in the same magnitude on both sides of the aisle despite similarities in the number of contested races, it suggests that some voters simply chose not to vote for offices lower on their ballots.
Few analyses have explored this phenomenon, and even fewer whether it may affect parties differently. Way to Win’s recent reporting suggests that areas with higher Democratic voter registration tended to correlate with higher roll-off than places with higher Republican registration. They also find that state house roll-off was higher than congressional roll-off across 10 critical swing states (explore this data further with the Way to Win (WTW) downballot tool using the roll-off numbers). The current analysis complements and extends the WTW analysis by focusing mostly on vote totals, as opposed to vote shares, and investigating instances in which downballot voters vote more than top-of-ticket voters (‘overages’) by examining the phenomenon in the context of multiple top-of-ticket races, and by investigating the phenomenon among both entire chambers and contested races only.
This analysis is inspired by SDAN’s 2020 and 2021 general election analyses demonstrating that Democrats saw more voter roll-off between top-of-ticket candidates (President in 2020 and Governor in 2021) and state legislature candidates than did Republicans. In fact, Republicans downballot frequently outperformed their top-of-the-ticket party-mates (in both vote totals and vote share), while the opposite was true for Democrats. The results in 2021 mirror the pattern in 2020. Virginia Republican state legislative candidates received a few thousand more votes than the GOP Gubernatorial candidate, while the Democratic gubernatorial candidate received 60,000 more votes than Democratic state legislative candidates.
This analysis is a retroactive extension of the 2020 and 2021 analyses. Looking at data from 2020, 2018, 2016, 2014, and 2012, this study investigates if the observed pattern holds in prior years, how such a pattern may be affected by top-of-ticket election type, and whether it disappears when uncontested races are removed from the analysis (where data permitted). We specifically looked at the upper and lower chambers of 10 battleground states (Arizona, Florida, Georgia, Michigan, Minnesota, Nevada, North Carolina, Pennsylvania, Texas, and Wisconsin).
This analysis looks at voter turnout in terms of 1) total votes cast for major state-wide top-of-ticket races in a single year (President, Governor, U.S. Senate); 2) total votes cast for state legislative candidates across all legislative districts, and 3) total votes cast just in contested state legislative districts (where district level top-of-ticket data was available), and 4) vote share, to investigate the enthusiasm gap between the top and bottom of the ticket (see more in results on the relationship between roll-off and enthusiasm gap).
Our hypotheses are: 1) that the pattern of top-of-the-ticket Democrats outperforming state legislative Democrats in terms of vote totals holds acrosstime periods; 2) that this phenomenon is stronger for Democrats than it is for Republicans; 3) that the vote total trends will hold in just contested state legislative races; and 4) the trend will hold when looking at vote share enthusiasm gaps (e.g., when the president gets 51% of the presidential vote and state legislative copartisans get 49% of the state legislative vote), and enthusiasm overages, or times where the downballot copartisans receive a larger vote share than their top-of-ticket party-mates.
We analyzed chambers in 10 battleground states: Arizona, Florida, Georgia, Michigan, Minnesota, Nevada, North Carolina, Pennsylvania, Texas, and Wisconsin. These were the states with the 10 closest 2-way Presidential vote margins in the 2020 election. Further, 8 out of 10 of these states have been in the 10 closest 2-way margins in 2+ of the last 3 Presidential elections (2012, 2016, 2020), indicating that many of these states are consistently competitive. Texas and Georgia, the 2 states that have only been in the top 10 once in the past 3 Presidential elections, have rapidly changing demographics, as demonstrated by narrow victories for Joe Biden in Georgia and ground gained in Texas by popular Democratic candidates like Beto O’Rourke.
State legislative data from elections since 2012 and the accompanying Presidential, Gubernatorial, and U.S. Senate races in the same elections were collected from the Secretaries of State from each state and Wikipedia. The dataset comprised 3 Presidential elections across 19 chambers (2012, 2016, 2020), 21 Gubernatorial elections across (2012, 2014, 2016, 2018, 2020), and 34 US Senate elections (2012, 2014, 2018, 2020). Please note that the higher vote total cast for each party was used in the case of 2 US Senate elections occurring on election day in a single year. Comparisons were made between the aggregate number of votes for Republicans and Democrats in the state house and the number of votes cast for both parties in each top-of-ticket race held in that state in a specific election year, and between the aggregate number of votes for Republicans and Democrats in the state senate and the number of votes cast in each top of ticket race. This yielded a total of 318 comparisons between downballot and top-of-ticket votes cast in the full dataset: 114 presidential data points (57 per party), 78 gubernatorial data points (39 per party), and 126 US Senate data points (63 per party).
Top-of-ticket data in contested state legislative districts only was from DailyKos. Races were contested if two major party candidates were officially nominated by their parties and running in the general election against one another. Only a subset of the chambers analyzed in the full dataset had available data for the contested dataset (see below), which resulted in 204 comparisons in the contested dataset: 94 Presidential data points (47 per party), 50 gubernatorial data points (25 per party), and 60 US Senate data points (30 per party). The data were analyzed descriptively.
1) To determine if downballot roll-off occurred in a chamber, we subtracted the number of top-of-ticket party votes in a state from the number of downballot votes for candidates of the same party.
It should be noted that some differential between top and bottom of ticket raw vote totals may result from ticket-splitting, not roll-off. However, recent reports indicate an all-time low for ticket-splitting, indicating that many individuals who vote for a party’s top-of-ticket candidate but not their state legislative candidate simply did not vote downballot. This conclusion appears to be supported by aggregate data, which often shows thousands more top-of-ticket than downballot voters. While not a definitive indicator of roll-off, it serves as a good approximation.
2) To determine if downballot roll-off occurred in contested state legislative races only, we used the following available DailyKos data:
- 2020: president: Arizona upper and lower, Florida upper and lower, Georgia upper and lower, Michigan lower; Minnesota upper and lower, Nevada upper and lower;
- 2018 governor: Arizona upper and lower, Georgia upper and lower, Michigan upper and lower, Minnesota lower, Nevada upper and lower, Pennsylvania upper and lower, Texas upper and lower, Wisconsin upper and lower;
- 2018 US Senate: Arizona upper and lower, Michigan upper and lower, Minnesota lower, Nevada upper and lower, Pennsylvania upper and lower, Texas upper and lower, Wisconsin upper and lower;
- 2016 president: all chambers, all states except Michigan upper
- 2014: governor: Arizona upper and lower, Florida upper and lower, Michigan upper and lower, Pennsylvania upper and lower, Wisconsin upper and lower; 2014 US Senate: Michigan upper and lower; 2012 president;
- 2012 president: all states except North Carolina, all chambers except Michigan Senate; 2012 US Senate: Arizona upper and lower, Florida upper and lower, Minnesota upper and lower, Nevada upper and lower, Pennsylvania upper and lower, Texas upper and lower, Wisconsin upper and lower.
We calculated the same measures previously described using the top-of-ticket and data from contested state legislative races (when available).
3) To determine if enthusiasm gaps occurred as evidenced by differences in vote share, we divided the number of votes each party received by the total number of votes cast for that office. For example, if 10,000 votes were cast in a district and 5,400 of them were cast for the Democratic candidate, the vote share for the state legislative Democrat would be 54%. For contested races only, this was narrowed to only include totals from the contested districts. Please note that the Minnesota Senate and the Michigan Senate hold elections every 4 years, so there are only 3 and 2 data points for those chambers in this analysis, respectively. The relationship between roll-off and enthusiasm gaps is expanded on in the results.
H1 – We will observe ballot roll-off in at least one of the two major parties in at least 1 state legislative chamber in all 10 states compared to at least 1 top-of-ticket office (President, Governor, U.S. Senate).
H2 – Roll-off would occur more frequently for Democrats than for Republicans. And conversely, Republicans will be more likely to have vote overages (where downballot candidates receive more votes than their upballot party-mates) than Democrats.
H3 – In at least some states, we still expect to see roll-off when restricting the analysis to contested state legislative races only.
H4 – Chambers will see a vote share (enthusiasm gap) pattern that mirrors roll-off patterns: downballot Democrats will be more likely to garner a lower vote share than their top-of-ticket counterparts, and downballot Republicans will be more likely to garner a higher vote share than their top-of-ticket counterparts.
- The Presence of Roll-off Across 5 Election Years
The first hypothesis was that ballot roll-off would be observed for at least one of the two major parties in at least one of the two chambers analyzed in all 10 states. This hypothesis proved true. Ballot roll-off was observed in the vast majority of comparisons between top- and bottom-of-ticket candidates. This was the case across all types of top-of-ticket offices and all chambers and states.
Of the 318 observed data points, there were 268 instances where a party experienced downballot roll-off.
State legislative candidates underperformed presidents in vote totals in 88.60% of comparisons and underperformed gubernatorial candidates in vote totals in 80.77% of comparisons. In sum, roll-off occurred in over 84% of the state legislative race comparisons included in this analysis.
Table 1. Number of instances of vote roll-off across types of top-of-ticket races 2012-2020
|Type of roll-off||All races|
|Presidential||101 ( / 114* = 88.60%)|
|Gubernatorial||63 ( / 78☨ = 80.77%)|
|US Senate||104 ( / 126*☨ = 82.54%)|
|Total||268 ( / 318 = 84.28%)|
*The Michigan Senate is not up for election in presidential election years, and therefore there is no ability to calculate Presidential or US Senate overages for Michigan in 2012, 2016, and 2020.
☨ The Minnesota Senate is not up for election during congressional midterm years, and therefore there is no ability to calculate Gubernatorial or US Senate overages for Minnesota in 2014 and 2018.
While roll-off is quite common, it is worth mentioning that in our dataset, we observed 50 instances (out of 318 observed data points, or 15.72%) where the candidates in a state legislative chamber received more votes than their top-of-ticket co-partisans (“overages”).
- Roll-off Trends: Differences by Party
Hypothesis 2 was that the trend of roll-off would be stronger for Democrats than for Republicans and, conversely, that downballot overages would be more common for Republicans. We found supporting evidence for both trends.
Out of 57 Presidential data points, Democrats saw 55 instances of ballot roll-off and 2 instances of overages. In the same elections, Republicans had 46 instances of ballot roll-off and 11 instances of overages.
Similarly, of 39 Gubernatorial data points, downballot Democrats saw roll-off in 34 instances and overages in 5. In the same race comparisons, Republicans saw roll-off in 29 instances and overages in 10.
Out of 63 U.S. Senate data points, Democrats saw ballot roll-off in 59 instances and 4 overages. Republicans saw roll-off in 45 instances and 18 overages in the same race comparisons.
While roll-off is far more common than downballot overages, Republicans saw roll-off less frequently than Democrats AND had more overages.
Table 2. Instances of vote roll-off from top-of-ticket to state legislature 2012-2020 by party
|Type of roll-off||All races|
|Presidential||55 (96.49%)||46 (80.70%)|
|Gubernatorial||34 (87.18%)||29 (74.36%)|
|US Senate||59 (93.65%)||45 (71.43%)|
|Total||148 (93.08%)||120 (75.47%)|
There were a total of 57 Presidential contests, 39 Gubernatorial contests, and 63 U.S. Senate contests in the dataset.
As mentioned above, we observed 50 instances of downballot overages out of 318 data points. When overages are broken down by party, Republicans saw far more overages than Democrats.
Democrats saw overages only 11 times across all data points, while Republicans saw 39. This indicates that not only are Republicans not rolling off as much as Democrats, but they are also voting more downballot than upballot with much more frequency than Democrats. Of the total overages observed, 78% were experienced by downballot Republicans, compared to only 22% by downballot Democrats. This equated to Republicans seeing overages in almost a quarter of their total race comparisons (24.53%) compared to Democrats, who saw overages in less than 10% of their total race comparisons (6.92%).
Table 3. Partisan distribution of vote overages across types of top-of-ticket races 2012-2020 by party
|Type of overage||All races|
|Presidential||2 (3.51%)||11 (19.30%)|
|Gubernatorial||5 (12.82%)||10 (25.64%)|
|US Senate||4 (6.35%)||18 (28.57%)|
|Total||11 (6.92%)||39 (24.53%)|
There were a total of 57 Presidential contests, 39 Gubernatorial contests, and 63 U.S. Senate contests in the dataset.
- Restricting Analysis to Contested State Legislative Races Only
The previous findings are aggregated at the chamber level, which may complicate interpretations due to uncontested races and races that were not on the ballot in a given election year. When races are uncontested or not held, it means that some party loyalists do not have anyone on the ballot to vote for, which may artificially bias aggregate roll-off results towards the party with fewer uncontested races.
Hypothesis 3 was that roll-off trends (both in terms of raw presence, H1, and partisan lean of the phenomenon, H2) would hold when the analysis was restricted only to contested state legislative districts.
This analysis included a subset of the chambers in each election year based on which top-of-ticket vote totals by state legislative district were available (see method section for states and chambers included).
From the available data, H3 was supported: roll-off was present in at least one election year for at least one party in every chamber in every state investigated in the contested races only analysis. And the partisan lean of this phenomenon was even more pronounced in the restricted analysis. While instances of roll-off decrease in these contested race comparisons, they decrease much more dramatically for Republicans than for Democrats.
Democrats saw their roll-off decrease by 13.97%, moving from roll-off in 93.08% of all race comparisons to 79.41% of contested race comparisons, which is still the vast majority of comparisons. At the same time, Republicans saw their roll-off decrease by 38.22%, moving from roll-off in 75.47% of all race comparisons to just 37.25% of contested race comparisons. Their decrease is so dramatic that a minority of downballot Republicans with a challenger saw roll-off compared to their top-of-ticket party-mates.
Essentially, fewer than 2 out of 5 downballot Republicans with a Democratic challenger saw roll-off. This compares to contested downballot Democrats seeing roll-off in almost 4 out of 5 comparisons. Put simply, narrowing the analysis to contested races amplifies the partisan differences in roll-off dramatically.
Table 4. Instances of voter roll-off between top-of-ticket and state legislative races 2012-2020 by party and contested race status
|Type of roll-off||All races||Contested races only|
|Presidential||55 (96.49%)||46 (80.70%)||41 (87.23%)||21 (44.68%)|
|Gubernatorial||34 (87.18%)||29 (74.36%)||16 (64.00%)||9 (36.00%)|
|US Senate||59 (93.65%)||45 (71.43%)||24 (80.00%)||8 (26.67%)|
|Total||148 (93.08%)||120 (75.47%)||81 (79.41%)||38 (37.25%)|
We found a remarkably similar pattern in terms of overages (where state legislative Republicans exceeded top-of-ticket vote totals). Surprisingly, it was actually more common to observe downballot voting overages among contested districts only compared to the chamber-wide analysis.
As described in Section 1 above, using all the comparisons for all races in each state, we found 50 instances of downballot overages. But, in the same period, restricting to just contested downballot races (204 data points compared to 318), we found 85 instances of downballot overages.
This suggests that not having a downballot candidate to vote for does not explain ballot roll-off and definitely doesn’t explain the partisan pattern of ballot roll-off, as both parties ran candidates in all of the contested races. This result suggests that the partisan lopsidedness of downballot roll-off is even more extreme in contested races. Since contested races are the races that decide majorities (given that these are typically competitive districts where partisans on both sides feel an opportunity to compete), this is particularly alarming for downballot Democrats in competitive races.
Table 5. Number of vote overages across types of top-of-ticket races 2012-2020 by party and contested race status
|Type of overage||All races||Contested races only|
|Presidential||2 (3.51%)||11 (19.30%)||6 (12.77%)||26 (55.32%)|
|Gubernatorial||5 (12.82%)||10 (25.64%)||9 (36.00%)||16 (64.00%)|
|US Senate||4 (6.35%)||18 (28.57%)||6 (20.00%)||22 (73.33%)|
|Total||11 (6.92%)||39 (24.53%)||21 (20.59%)||64 (62.75%)|
- Vote Share Pattern Will Match Roll-off Pattern
The last hypothesis focused on the relationship between roll-off/overages and enthusiasm/awareness gap (where downballot candidates see a lower vote share than top-of-ticket candidates) between the two major parties.
The roll-off analyses in Sections 1-3 looked at the number of votes cast, but races are won based on vote share, or the percentage of the total votes cast. It is not uncommon for a party to see roll-off between copartisans up and down the ballot and still win the majority vote share in the downballot races. In these cases, even though roll-off was observed, it didn’t hurt the downballot candidate because they still won, just with fewer votes. However, it is also possible that roll-off directly contributes to situations where top-of-ticket Democrats are winning, but downballot Democrats lose due to partisan lopsidedness in roll-off/overages.
For instance, in 3 of the 4 majority-making contested races in the Arizona and Minnesota Senates in 2020, Democrats won the Presidential vote in the district, but the Democratic state Senate candidate lost. Some ticket-splitting may have happened, where people voted for Biden for President but voted for the Republican state Senate candidate. But also note that over 1,000 voters who voted for President simply did not cast a vote for state legislature in each district. It appears that some combination of ticket-splitting and Democratic roll-off accounts for the entire Democratic loss margin in these 3 critical districts. The result was winning vote shares for Biden and losing vote shares for downballot Democrats in all 3 districts. It is clear that roll-off may directly impact who wins the majority of the vote share downballot.
Table 6. 2020 critical district examples
|State||District (Dem SL candidate)||Votes for state leg Dem (vote share)*||Votes for Biden (vote share)☨||Difference SL – Biden votes (diff in vote share)||Votes for state leg R (vote share)*||Votes for Trump (vote share)☨||DifferenceSL – Trump votes
(diff in vote share)
|Presidential voters who did not vote for SL (% of total)||SL Dem loss Margin
(diff in vote share)
|AZ||SD 17 (Kurdoglu)||61,363 (47.48%)||69,581 (51.10%)||-8,218 (-3.62%)||67,889 (52.52%)||64,536 (47.39%)||+3,353 (+5.13%)||6,916 (5.08%)||6,526 (5.05%)|
|MN||SD 26 (Borrud)||23,831 (48.99%)||26,742 (53.71%)||-2,911 (-4.72%)||24,740 (50.86%)||22,044 (44.28%)||+2,696 (+6.58%)||1,143 (2.30%)||909
|MN||SD 34 (Westlin)||28,443 (49.18%)||31,335 (52.83%)||-2,892 (-3.65%)||29,347 (50.74%)||26,726 (45.06%)||+2,621 (+5.68%)||1,477 (2.49%)||904
The table above provides examples of 3 legislative districts in which vote total roll-off and enthusiasm gaps (vote share) appear to interact and result in a loss down-ballot but a win up-ballot.
*These columns display the raw vote totals received and the total vote share won by each of the state Senate candidates who ran in these races
☨These columns display the raw vote totals received and the total vote share won by the 2020 presidential candidates from each party in these state Senate districts
-The column highlighted in blue displays the difference between the number of votes received and vote share for the Democratic state Senate candidate and Biden
-The column highlighted in red displays the difference between the number of votes received and vote share for the Republican state Senate candidate and Trump
-The column highlighted in purple displays the number of people who voted for President but did not vote for state legislature and the percentage of total presidential voters than they represent
-The column highlighted in yellow displays how many votes the state Senate Democratic candidate lost by, and the difference in vote share between the two state Senate candidates
When looking at vote share in all races in the dataset, every single instance of a vote total overage overage was an instance in which the state legislative candidates received a larger vote share than the top-of-ticket candidate of the same party.
However, a downballot candidate can still receive a higher voter share than a top-of-ticket candidate even if they see voter roll-off. This may be the case when both downballot candidates see roll-off compared to top-of-ticket co-partisans, but one party sees less than the other. Due to this, there are far more instances in which downballot candidates receive a higher vote share than top-of-ticket candidates than in which downballot candidates receive more total votes: a total of 183 (compared to a total of 50). The partisan breakdown was not quite as stark as for the vote total data, but Republicans continued to have far more downballot vote share overages than Democrats (86.79% for Republicans vs. 28.30% for Democrats), mirroring the vote total findings.
These conclusions were largely echoed when looking at contested races only. Every case of a voting overage for a party resulted in a vote share that was greater than their party mate’s vote share in the top-of-ticket race. This suggests that, when downballot candidates receive more votes than top-of-ticket party-mates, they will almost certainly capture a larger share of the total vote than their top-of-ticket party-mates.
For instance, in the contested state assembly (known as the state house in most other states) races in Nevada in 2020, there were 30,892 more Democratic presidential voters than Democratic Nevada state house voters and 6,351 more Republican presidential voters than Republican Nevada state house voters in contested races. Even though Joe Biden took 51.42% of the presidential vote to Donald Trump’s 46.35% in these contested state legislative districts, this partisan differential in potential roll-off seems to have contributed to the fact that downballot Republicans captured a larger share of the vote (48.24%) than Trump (46.35%). In comparison, the downballot Democrats captured 0.52% less of the vote share (50.90%) than Biden (51.42%). When Democrats roll off more than Republicans in vote total, that appears to give a direct path for downballot Republicans to capture a larger share of the vote than their top-of-ticket copartisans.
It is notable that vote share overages are more common for Democrats in contested races compared to all races (see Table 7).
Table 7. Instances of vote share overages across types of top-of-ticket races 2012-2020 by party and contested race status
|Type of overage||All races||Contested races only|
|Presidential||15 (26.32%)||51 (89.47%)||21 (44.68%)||40 (85.11%)|
|Gubernatorial||14 (35.90%)||28 (71.79%)||9 (36.00%)||19 (76.00%)|
|US Senate||16 (25.40%)||59 (93.65%)||9 (30.00%)||27 (90.00%)|
|Total||45 (28.30%)||138 (86.79%)||39 (38.24%)||86 (84.31%)|
There were a total of 57 Presidential contests, 39 Gubernatorial contests, and 63 U.S. Senate contests in the dataset.
However, it is important to note that Republican vote share was higher downballot than for Donald Trump in every chamber we had contested data for (see method for states and chambers) in 2020, while Democrats did not exceed vote share over Joe Biden in the contested races in any of these chambers. Four of the chambers (MN Senate, MI House, and AZ Senate and House) were extremely close to tying or flipping partisan control before the 2020 election. All 3 of these states voted for Joe Biden and Democratic U.S. Senators at the top of the ticket. This analysis suggests that downballot roll-off could partially explain why Democrats failed to make more headway in these chambers even while they won the top of the ticket.
Overall, all of the hypotheses had supporting data. While downballot roll-off is common, it is much more common for downballot Democrats to experience it than downballot Republicans. Across 318 comparisons encompassing all top-of-ticket elections analyzed in 5 election cycles, we observed downballot Democratic roll-off 93.08% of the time compared to Republican roll-off 75.47% of the time.
Further, this is especially true of contested downballot races, where Democrats and Republicans saw a dramatically different level of roll-off. Democrats downballot saw roll off in contested race comparisons 79.41% of the time, while Republicans saw roll off in contested race comparisons just 37.25% of the time.
Roll-off may be seriously affecting the fates of downballot Democratic candidates and, in turn, potentially the control of state legislative chambers. It appears clear that Democrats need to do far more than they have been in the past decade to turn out voters downballot, as the effects observed here cannot be fully explained by ticket-splitting or the impact of uncontested races.
Not only was roll-off observed for the vast majority of comparisons analyzed, but there was also a sizeable lopsided gap between the parties in the form of downballot vote share (a measure of an enthusiasm gap between downballot and top-of-ticket candidates). Downballot Republicans in these data often capture a larger share of the vote compared to their top-of-ticket counterparts. Democrats downballot saw enthusiasm gaps 71.70% of the time, while Republicans saw enthusiasm gaps just 13.21% of the time.
This may suggest that downballot roll-off may be directly related to situations like the 2020 Minnesota Senate elections: enough people rolled off from voting for the Democratic presidential nominee, and from voting for president in general, to cover the loss margins in the state legislative races that would have flipped partisan control to Democrats. It may be the case that Democrat downballot roll-off is a major culprit for why Democrats failed to flip close chambers where Biden won the top-of-ticket vote in Arizona, Michigan, and Minnesota.
Further, Republicans are actually voting more downballot than upballot in a sizable number of elections and far more often than Democrats are. We observed overages in vote totals for Republicans in 24.53% of all comparisons, whereas Democrats had overages in only 6.92%.
Finally, this partisan lopsidedness in roll-off translates to vote share, such that downballot Republicans often capture a larger share of the vote (the percentage of the total votes cast) compared to their top-of-ticket counterparts.
It is unclear from this analysis why downballot Republicans experience roll-off so much less frequently than downballot Democrats. The ‘contested races only’ analysis herein focuses on districts that were competitive enough to field candidates from both major parties, and strips away extremely safe partisan districts that contribute to top-of-ticket wins on both sides. As described above, we observe an even sharper partisan skew for downballot roll-off in this subset analysis. It is hard to offer potential explanations for why this is occurring. It is possible that Democratic top-of-ticket voters do not vote downballot if they do not feel adequately educated about the candidates.
In contrast, Republicans may feel more educated about downballot candidates or may not let their lack of education hold them back from voting at that level of the ballot. It may also relate to differences in national party prioritization, such that the Democratic party and Democratic voters focus more on federal races, while the Republican party and Republican voters allocate more attention to downballot races. More research into individual voter thoughts around state legislative races compared to other types of races is necessary to truly understand the partisan trends in these data.
Regardless of the reasons, this analysis suggests these partisan differences in roll-off are not new, and they must be urgently addressed for Democrats to have more downballot success. In addition to investigating individual voter beliefs and behavior in regards to voting for state legislative candidates, future investigations should expand to more partisan states to determine if these trends hold in areas with extreme partisan supermajorities, like California or Wyoming, or if the trends identified here are specific to competitive battleground chambers. Further, data were limited in the current contested races analysis due to data availability. Hopefully, future contested race datasets can include a more complete subset of the chambers across elections years studied.
If you have questions about these findings or would like a copy of the datasets, please email Mallory.