KEY POINTS FROM THIS ARTICLE
— From 1914 to 2016, presidential cycles featured a higher rate of straight-ticket outcomes than midterm elections, with 74% of presidential-Senate results going for the same party in presidential years. Midterm cycles showed more splits, with just 61% of presidential-Senate results won by the same party. In 21 of 25 midterm cycles that followed a presidential election in the 1913-2016 period, the share of split-ticket presidential-Senate results increased compared to the share in the previous presidential cycle.
— In 2018, the Democratic caucus is defending 26 of the 36 seats (72%) that will contested this cycle, the most for any party in the post-World War II period. That figure includes the Alabama special election in 2017 as a seat for the GOP to defend, which it remarkably failed to do. Democrats are defending more seats than any non-presidential party in previous midterm elections. Of those seats, just over half (14) are in states that leaned toward the Democrats in the 2016 presidential election while 12 are in states that leaned toward the GOP in 2016, including 10 in states that Trump carried.
— Conversely, 2018 Republicans are tied with the 1970 GOP for the fewest seats a presidential party has had to defend (10) in a midterm cycle in the era of popular elections. If we exclude the Alabama race from the defense count for the entire cycle, this year’s GOP is actually defending the fewest seats at the time of the regular midterm election of any presidential party. And almost all of those seats are in Republican-leaning states, meaning that there are only a few GOP-held Senate seats that Democrats can realistically target.
— In 39 open-seat contests in seats previously held by the presidential party where the partisan lean favored the non-presidential party, the presidential party has won just nine times (23%). The winning percentage for the presidential party in open seats that lean just toward it is better (59%), but still not far beyond a coin flip. Presidential party Senate incumbents in states that leaned a bit away from their party have won around two-thirds of their contests (64%), which is another reason not to count out Sen. Dean Heller (R-NV). Non-presidential party candidates have evenly split their attempts to retain seats that just leaned against them, though they’ve won most of the few open-seat contests in states where the partisan baseline was even worse for the non-presidential party.
Unstoppable force meets immovable object?
As the Crystal Ball has noted repeatedly since early 2017, the Senate elections in the 2018 cycle feature two contrasting forces: highly polarized, partisan voting in elections running up against a tendency for non-presidential party incumbents to do well in midterms, even in states that backed the other party in the most recent presidential election. With 10 Democratic Senate incumbents running for reelection in states that President Donald Trump carried in 2016, the force that proves to be more unstoppable or immovable will be key to determining the majority in Congress’ upper chamber in November. After all, if voters hypothetically were to align their states’ Senate seats with their 2016 presidential party winner this fall, Democrats would lose 10 seats they currently hold while Republicans would lose just one, handing the GOP a net gain of nine seats. That theoretical result would give Republicans a 60-40 filibuster-proof edge.
Yet we know that such an outcome is highly unlikely at this point. While Republicans are favored to retain control of the chamber, and possibly gain seats in the process, Democrats do have a chance of winning a slim majority this November. To win 51 or 52 seats, Democrats would have to hold most or all of those 10 seats in Trump-won states while also adding one or two other seats in states the president carried — Arizona and Tennessee — while also winning the lone Clinton-won state the GOP is defending this cycle — Nevada. Although this is not a likely outcome, the fact we can credibly discuss it as a possible result is evidence that the historical trend of non-presidential party incumbents winning most midterm races may prove to be the stronger force once again in 2018. It is also evidence of the historical tendency for competitive open-seat contests (e.g. Arizona and Tennessee) and races in seats held by presidential party incumbents in states that lean toward the non-presidential party (e.g. Nevada) to sometimes swing to the out-party in midterm cycles. This article will dig into some of these trends and will place 2018 in the context of past contests, from the first midterm cycle in 1914 on.
Midterms usually feature more split-ticket outcomes than the previous presidential cycle
The 2016 election cycle was highly partisan. Even with two unpopular presidential nominees, around 90% of voters in each party backed their respective standard-bearers. Just as telling, the 2016 election marked the first time in the history of popular elections that 100% of the Senate outcomes matched the presidential party winner in each state with a Senate race. That is, the same party won the Senate and presidential races in the 34 states that held elections for Senate. The previous record had been the 1920 cycle, when 97% (33 of 34) of presidential-Senate party winners matched.
Chart 1 displays the percentage of straight- and split-ticket presidential-Senate results for every federal election cycle from 1914 to 2016. For midterm cycles, I compared the party winner in a Senate race to the party that won the state in the previous presidential election. The data include special elections held at irregular times during a cycle to give a more complete accounting of Senate results. For example, the 2014 cycle includes Democratic special election wins in Massachusetts and New Jersey in 2013, and 2018 data in this piece will include Alabama’s special election result from 2017. Additionally, Sens. Angus King (I-ME) and Bernie Sanders’ (I-VT) election victories in 2006 (just Sanders) and 2012 (both senators) are included as Democratic wins because they caucus with the Democrats in the Senate and are both up in 2018 as de facto Democrats.
Chart 1: Percentage of straight- and split-ticket party winner outcomes in presidential-Senate elections by cycle, 1914-2016
Notes: Presidential results in midterms are based on the previous presidential election outcome. Any state won by an independent or third-party candidate at the presidential and/or Senate level is excluded from the analysis. The 2006 and 2012 wins for Sen. Bernie Sanders (I-VT) and the 2012 win for Sen. Angus King (I-ME) are treated as Democratic wins because they caucus with the Democrats. Special elections contested prior to the regular Election Day are included in their respective cycles. Concurrent special elections held on the regular Election Day to complete a few months of an expiring term are excluded from the analysis. Click on image to enlarge.
From 1913 — when the 17th Amendment established popular voting for Senate elections — through 2016, there were 1,881 elections for the U.S. Senate. After excluding concurrent elections for expiring terms and cases where a third-party or independent candidate won a state’s Senate race or presidential contest, there were 1,801 major-party presidential-Senate results to examine (i.e. a Democrat or Republican won the presidential and Senate races in question). In total, 67% (1,210) had straight-ticket winners while 33% (591) had split-ticket results. Perhaps unsurprisingly, given the potential power of presidential coattails, presidential cycles featured a higher rate of straight-ticket outcomes, with 74% of presidential-Senate results going for the same party in each. Midterm election outcomes showed more splits, with just 61% of presidential-Senate results won by the same party. This makes sense: The president’s party tends to lose ground in midterm elections — practically always in the U.S. House and often in the Senate as well. Causes of this “presidential penalty,” such as disproportionately high turnout among the non-presidential party’s supporters and the electorate’s desire to check the president and his party, can open up opportunities for a party to succeed in a state where the president’s party may have recently been stronger. These factors can also boost Senate incumbents from the non-presidential party who are battling to win reelection in states that are tough for their party at the presidential level.
In 21 of 25 midterm cycles that followed a presidential election in the 1913-2016 period — excluding the 1914 midterm because the 1912 presidential election falls outside that range — the share of split-ticket results increased compared to the share in the previous presidential cycle. So, in most midterm cycles the percentage of split-ticket outcomes relative to a state’s last presidential party winner increased. The only exceptions were the cycles in 1978, 1990, 2002, and 2014. Since 1990, there has been a relatively steady trend toward greater straight-ticket voting outcomes, culminating in the 2016 presidential cycles’ 100% straight-ticket result. The 2014 result might suggest that potential Republican gains in Trump states could be numerous in 2018, but as things stand, we currently rate four Trump-state seats held by Democrats as Toss-ups while the other six are rated as Leans Democratic or better for the Democrats. For the sake of argument, assume that the Democrats hold the seats we lean toward them. That would result in at least six split-ticket results out of 35 elections held this coming November, and seven out of 36 overall in the 2018 cycle when including Alabama’s special election, which works out to an 81% straight-ticket share of outcomes.
Therefore, it seems likely that split-ticket results will make something of a comeback in 2018, at least compared to the past couple of cycles. Should Democrats manage to retain all the seats they hold — possible but probably unlikely — and pick up Arizona and Nevada to win a 51-49 Senate majority, that would result in 11 split-ticket outcomes relative to the 2016 presidential election (10 Democratic-held Trump states and Arizona, a state Trump carried). Once we account for the Alabama special election, that would add up to 12 split-ticket outcomes out of 36 races in the 2018 cycle as a whole, or one-third. Such a result would be similar to the 2006 cycle, which featured the same class of senators (Class I) and a 31% split-ticket rate.
Parties on the defensive in midterm elections
Of course, Senate elections are not truly national elections the way House contests are. All 435 seats in the House are up for grabs every federal election cycle, whereas only about one-third of the Senate’s 100 seats are contested, depending on the number of special elections. Thus, electoral fortune is partly dependent on what seats happen to be up in a given cycle. The current composition of Senate Class I is such that Democrats and the two independents who caucus with them hold 25 of the 33 regularly-scheduled seats up this November. Meanwhile, partly because of their successes in 2010 and 2014, Republicans hold 21 of the 33 seats in Class II and 22 of the 34 seats in Class III. Were either of those classes up in 2018, Democrats might be solid bets to take control of the Senate because the GOP would be mostly on the defensive. Instead, the most Democratic-heavy map, which also features the most red-state Democrats, is up in 2018, and the Democrats are mostly on the defensive as a result. Although this might seem like a curse for Democrats, there is a silver lining for them: Together, the environment and the Democrats’ incumbency advantage (all 26 are seeking reelection) might help them retain seats in states that Trump carried, including five seats in states that the president won by 18 percentage points or more. Had Hillary Clinton won the 2016 presidential election, a high single-digit to low double-digit net gain might have been in the offing for Senate Republicans in 2018.
To put the level of Democratic exposure — that is, the large number of seats that they are defending — in context, the Crystal Ball examined the history of exposure for parties in midterms back in December 2016 and found that Democrats had one of the largest numbers of seats to defend for any party in a midterm. As this piece is taking into account the entirety of a cycle — thereby including irregular special elections — I have adjusted the numbers to get a look at the full cycle exposure for parties. In 2018, the Democratic caucus is defending 26 of the 36 seats (72%) that will contested this cycle, the most for any party in the post-World War II period. That figure includes the Alabama special election in 2017 as a seat for the GOP to defend, which it remarkably failed to do. Table 1 presents the data for the parties with the highest percentage of seats to defend over the course of a midterm cycle.
Table 1: Most exposed parties based on share of seats to defend over the course of a midterm cycle, 1914-2018
Notes: *The 2018 Democratic number of seats to defend includes Sens. Angus King (I-ME) and Bernie Sanders (I-VT), who both caucus with the Democrats. Based on the entire cycle of 2018, Democrats have gained one net seat so far via their 2017 special election win in Alabama.
The 2018 Democrats rank as the fourth-most exposed party in any midterm, trailing Democrats in 1938 (92%), Republicans in 1926 (81%), and Democrats in 1942 (73%). Of these 10 most-exposed party cycles, only once prior to 2018 did one of these highly-exposed midterm parties not occupy the White House. The preponderance of presidential parties having many seats to defend in these cycles has much to do with election timing. In many cases, a presidential party was stuck defending a large number of seats in a midterm six years after a presidential win produced gains down the ballot. This includes Republicans in 1926 (Warren Harding’s landslide “Return to Normalcy” victory) and Republicans in 1986 (six years after the Reagan Revolution of 1980 featured a 12-seat net gain for the GOP). But in other cases, the high level of exposure came from repeated successes in consecutive Senate election cycles. This helps explain, for example, Democrats’ large number of seats to defend in 1942 after big gains in 1930 and 1936, or Republicans in 1930 after numerous pickups in 1918 and 1924. It also describes Democrats in 1970, when they defended 23 of the 35 seats contested after a 16-seat net gain in the 1958 cycle and some additional wins in 1964. But with Republican Richard Nixon in the White House, Democrats avoided sweeping losses, even though they were defending 13 seats in states that leaned Republican based on the 1968 presidential election result. Lastly, the 2018 cycle is also an example of a party putting together a series of successful Senate election cycles: Democrats gained four net seats in 2000, six net seats in 2006, and two net seats in the 2012 cycle. Now they have to defend many of those seats in 2018.
Presidential lean and party exposure in midterms
As with the discussion of the 1970 cycle and presidential lean, we can apply that frame of analysis for all midterms. That is, investigating how many seats the parties had to defend that were on favorable or unfavorable turf based on the previous presidential election. After all, in a landslide presidential contest, there have often been many split-ticket results because voters backed the presidential winner all over the country but may have supported the other party’s candidate in the Senate race at the same election or in the succeeding midterm cycle. By looking at the partisan lean of states relative to the national popular vote, we can adjust for the electoral environment and then see which seats each party had to defend and how they performed in a given midterm cycle.
Charts 2, 3, and 4 present data on the non-presidential party in midterm cycles. Chart 2 lays out the overall number of seats defended by the non-presidential party in each midterm cycle, Chart 3 presents the number of seats defended by said party that leaned toward it at the presidential level, and Chart 4 shows the number of seats defended by the non-presidential party that leaned toward the party in the White House.
Chart 2: Seats defended by the non-presidential party in midterm cycles, 1914-2018
Note: Click on image to enlarge.
Chart 3: Seats defended by the non-presidential party that leaned toward it in the previous presidential election, 1914-2018
Notes: Excludes independent and third-party incumbents, except Sen. Bernie Sanders (I-VT) in 2006 and 2018, and Sen. Angus King (I-ME) in 2018. Both were treated as Democrats because they caucus with the party. Excludes seats in states won by a third-party presidential candidate in the previous election (two in 1914 and one in 1970). Based on two-party presidential party marginal lean in previous presidential election. Click on image to enlarge.
Chart 4: Seats defended by the non-presidential party that leaned toward the presidential party in the previous presidential election, 1914-2018
Notes: Excludes seats in states won by a third-party presidential candidate in the previous election (two in 1914). Based on two-party presidential party marginal lean in previous presidential election. Click on image to enlarge.
Among non-presidential parties, the 2018 Democrats are defending more seats overall (26) than the non-presidential party in any previous midterm election. Of those seats, just over half (14) are in states that leaned toward the Democrats in the 2016 presidential election while 12 are in states that leaned toward the GOP in 2016. Of those 12, 10 are in states that Trump carried. The only two Senate seats that lean Republican but Clinton won are in Minnesota, which has elections for both of its Senate seats: a regular election for the seat held by Sen. Amy Klobuchar (D) and a special election for the seat now held by Sen. Tina Smith (D), who replaced Sen. Al Franken (D) after he resigned. Clinton narrowly carried the Land of 10,000 Lakes in 2016 but won it by a smaller margin than she won the national popular vote, thus leaving the state with a slight Republican lean at the presidential level. The only cycles where a non-presidential party had to defend as many or more seats in states that leaned toward the presidential party were the aforementioned 13 Democratic seats in 1970 as well as 12 Democratic seats in 1974.
While Democrats are defending more seats than any previous non-presidential party, and one of the highest numbers of seats that lean toward the White House party, Republicans are defending nine seats this November and 10 overall for the cycle when we include the Alabama special election. Of those, just one — Nevada — leans toward the Democrats. Charts 5, 6, and 7 display data for the presidential party in the same manner as the three previous charts: Chart 5 shows the overall number of seats defended by the presidential party in midterms, Chart 6 presents the number of defended seats that lean toward the presidential party, and Chart 7 lays out the number of defended seats that lean toward the non-presidential party.
Chart 5: Seats defended by the presidential party in midterm cycles, 1914-2018
Note: Click on image to enlarge.
Chart 6: Seats defended by the presidential party that leaned toward it in the previous presidential election, 1914-2018
Notes: Excludes seats in states won by a third-party presidential candidate in the previous election (two in 1926 and two in 1950). Based on two-party presidential party marginal lean in previous presidential election. Click on image to enlarge.
Chart 7: Seats defended by the presidential party that leaned toward the non-presidential party in the previous presidential election, 1914-2018
Notes: Excludes seats in states won by a third-party presidential candidate in the previous election (one in 1950 and one in 1962). Based on two-party presidential party marginal lean in previous presidential election. Click on image to enlarge.
As Chart 5 indicates, 2018 Republicans are tied with the 1970 GOP for the fewest seats a presidential party has had to defend (10) in a midterm cycle in the era of popular elections. If we exclude the Alabama race from the defense count, this year’s GOP is actually defending the fewest seats at the time of the regular midterm election of any presidential party. And as Chart 6 shows, almost all of those seats are in Republican-leaning states, meaning that there are only a few GOP-held Senate seats that Democrats can realistically target. While Republicans would have likely benefited from a Clinton presidency in terms of potential Senate gains, they still have a decent chance of gaining some ground in the Senate even while potentially losing the House. Still, as the Alabama special election proved — under a very particular set of conditions — seats that seem quite safe can become competitive, which also seems to be happening in Tennessee. The Volunteer State — which holds its primaries today — is now quite Republican leaning at the presidential level, but former Gov. Phil Bredesen (D) has made it a real race, leading much of the polling against Rep. Marsha Blackburn (R, TN-7). Bad candidates can taketh away — see: Roy Moore (R) in Alabama — but good candidates can giveth, as Bredesen is proving by unexpectedly putting the Tennessee race in the mix for Democrats.
Chart 7 shows this year’s GOP defending just one Democratic-leaning seat, a mark only equaled by Republicans in 2002, when Sen. Susan Collins (R-ME) was the lone Republican defending a Democratic-leaning seat. From day one of the 2018 cycle, Sen. Dean Heller (R-NV) has been the most endangered GOP incumbent up for reelection, and his race with first-term Rep. Jacky Rosen (D, NV-3) looks to be very close.
Presidential lean, incumbent wins and losses, and open seats
The extent to which a seat leans in one direction or the other can play into how vulnerable it is for the defending party, though incumbency has proved to be helpful in overcoming challenging partisan fundamentals. I looked at midterm Senate elections and the presidential lean in each seat based on the previous presidential election. To do this, I divided all eligible midterm contests from the 1914 cycle to the 2014 cycle into four groupings: seats with incumbents from the presidential party, seats with incumbents from the non-presidential party, open seats previously held by the presidential party, and open seats previously held by the non-presidential party. Then I sorted the seats by the partisan lean of their respective states based on how much a state leaned toward or against the presidential or non-presidential party. Table 2 displays the data for incumbent-held and open seats and the performance of the presidential and non-presidential parties in midterm election cycles (remember, this includes special elections that did not take place on the normal Election Day).
Table 2: Incumbent versus open-seat performance for the presidential and non-presidential parties in midterm elections, 1914-2014
Notes: Includes irregular special elections. Excludes independent and third-party incumbents, except Sen. Bernie Sanders (I-VT) in 2006, who is treated as a Democrat because he caucuses with the party. Excludes Senate incumbents who sought reelection in the general election after losing renomination. Excludes seats in states won by a third-party presidential candidate in the previous election. Presidential lean based on two-party presidential party marginal lean in previous presidential election relative to the national popular vote.
As the Crystal Ball has previously noted, non-presidential party incumbents have won 91% of midterm contests since the 1914 cycle, and this remains true when adding in irregular special elections as well. Presidential party incumbents have won about three-fourths of the time (74%). While each category lacks a large sample size, there are still some results worth highlighting. Regardless of partisan lean, presidential party incumbents have lost at a higher rate than non-presidential party incumbents. Open-seat races in seats previously held by the presidential party have proved to be quite vulnerable when they were not relatively safe seats. In 39 open-seat contests in seats previously held by the presidential party where the partisan lean favored the non-presidential party, the presidential party has won just nine times (23%). The winning percentage for the presidential party in open seats that lean just toward it (zero to +9.9 points) is better (59%), but still not far beyond a coin flip. In 2018, this shows why Arizona very quickly came on the board despite its Republican lean once Sen. Jeff Flake (R) opted to retire after becoming persona non grata among many GOP voters. Presidential party Senate incumbents in states that leaned a bit away from their party (zero to -9.9 points) have won around two-thirds of their contests (64%), which is another reason not to count out Heller in Nevada. Interestingly, non-presidential party candidates have evenly split their attempts to retain open seats that just leaned against them (zero to -9.9 points), though they’ve won most of the few open-seat contests in states where the partisan baseline was even worse (if the 2017 Alabama special election were included here, it too would prove to be a non-presidential party win in a state leaning at least 20 points toward the presidential party).
I applied the percentages in Table 2 to the 2018 numbers in Table 3, which lists the partisan details for each of the 36 seats that have been or will be contested in the 2018 cycle.
Table 3: 2018 Senate seats by party, incumbency, and presidential lean
Notes: *Special elections for Senate, including the special election in Alabama that took place in 2017. ^Sens. Bernie Sanders (I-VT) and Angus King (I-ME) are included as Democrats because they caucus with the party. Presidential lean based on two-party presidential party marginal lean in previous presidential election relative to the national popular vote. Some totals may not add up due to rounding.
For example, I multiplied the Democrats’ five seats that were at least 20 points more Republican than the national popular vote margin in 2016 by the historical success rate of 85% for the non-presidential party in similar seats. The result was 4.2, which suggests that Democrats might be in line to retain four of the five seats in question, a result that would probably precipitate some cheers in Democratic circles. Having multiplied out the results for each category based on the historical success percentages for the presidential and non-presidential parties, the arithmetic results in a Republican gain of about one net seat. However, because Alabama is included as a part of the 2018 cycle, Democrats already have a one-seat gain coming into November. That would suggest a two-seat gain for Republicans in the elections to come. That may or may not happen, but it would be an unsurprising result. On the one hand, Democrats are battling to hold onto many tough seats. On the other hand, the Democrats’ challenging map is tempered by a Democratic-leaning environment via having a Republican president in office, one who also has a mediocre approval rating.
Historical midterm results suggest that we are in for a bounce-back year for split-ticket outcomes in Senate elections as they relate to the previous presidential election. This is good news for Democrats because they are defending one of the largest number of seats on record for any party in a midterm. Conversely, the GOP has few vulnerable seats to defend but may find itself limited by the electoral environment as the presidential party. Looking back at the historical success or failure in midterm contests based on incumbency and partisan lean, we have good reason to expect only a small change to the partisan makeup of the Senate, but those shifts could slightly favor the Republicans at the end. The GOP remains favored to retain the Senate, but the fact that Democrats have any chance at all at capturing a majority speaks to the benefit of being the non-presidential party, having so many incumbents seeking reelection, and of having a relatively friendly electoral environment.
1. Sanders won Vermont’s Democratic primary for Senate in 2006 and 2012, but declined the party’s nomination in both elections. Because of this, Democrats officially fielded no Senate candidate in either election. Sanders is expected to follow the same course in 2018. Sanders also pursued the 2016 Democratic presidential nomination and may do so again in 2020. Unlike Sanders, King faced a Democrat in 2012, but Cynthia Dill (D) finished a distant third with 13%. King faces another Democrat, Zak Ringelstein, in 2018, but the incumbent is heavily favored to win and is expected to continue caucusing with the Democrats going forward.
2. In the 1968 presidential election, George Wallace won five states in the South running as a segregationist third-party candidate. Only one of those, Mississippi, held a Senate contest in 1970. Wallace won 63.5% of the vote in the Magnolia State while Hubert Humphrey (D) won 23.0% and Richard Nixon (R) won 13.5%.
3. Seven if you discount Connecticut as a Democratic loss, where Sen. Joe Lieberman (D) won reelection as an independent after losing the Democratic nomination. He continued to caucus with the Democrats afterwards.
4. Three if you count Connecticut as a Democratic gain. Now-Sen. Chris Murphy (D) won the seat of outgoing Sen. Joe Lieberman, an independent Democrat.