For clues, look to Trump’s polling in the states that will determine control of the chamber.
For clues, look to Trump's polling in the states that will determine control of the chamber.
W e don’t know yet what will happen in the presidential election, but the odds clearly favor a Joe Biden victory, and even more strongly favor the Democrats holding the House. Harry Truman in 1948 was the only president whose party retook the House before the end of his presidency, after losing it. If Biden wins, much of his agenda will thus depend on whether Democrats take the Senate. If Trump pulls out a surprise victory, his ability to keep confirming judges at a brisk clip will likewise depend on the Senate. In some ways, control of the Senate may be even more important over the next two years than control of the White House. Democrats need a net gain of four Senate seats to gain control, or three to create an evenly divided Senate in which the vice president casts deciding votes.
As of now, the polling shows a close, competitive contest in the Senate. The RealClearPolitics averages favor the Democrats gaining the four seats they need, picking up Colorado from Cory Gardner, Maine from Susan Collins, Arizona from Martha McSally, North Carolina from Thom Tillis, and Iowa from Joni Ernst, while losing Doug Jones in Alabama. Polling may be reliable, but it could also be missing something. Even aside from the general issues why polls missed Republican voters in 2016 and 2018, we may be dealing with a particularly high-turnout election, which could do unpredictable things if there are a lot of the kinds of voters who are hard to capture in polls. So, we should be thinking less about what we see as certainties in the polling, and more about ranges of possible outcomes.
Here are the states for which we have Senate polling, ranked by the “Break” metric I have used in past elections — i.e., what percentage of undecideds in the polling average need to vote Republican to get the Republican candidate to 50 percent. The Georgia special election is listed twice here to compare the Democrat, Raphael Warnock, both to the incumbent Kelly Loeffler and to the Republican poll leader, Representative Doug Collins. Unless one candidate clears 50 percent of the vote, however, both Georgia Senate elections would be decided in runoffs on January 5, possibly with control of the chamber still unsettled.
In an election with upside surprises, Republicans hold out hope for John James in Michigan and Jason Lewis in Minnesota (maybe less in the latter now that Lewis is off the trail for emergency hernia surgery). On the downside, a really bad year could take out some or all of the Republican-held seats in Georgia, South Carolina, Kansas, Alaska, and Montana. To complete the picture, here are the races that are unpolled and not expected to be competitive, more of which are currently in Republican hands:
How common is it for the Senate dominoes to fall hard to one side in a presidential-election year — and is it always towards the side that wins the White House? Let’s take a walk through the last 26 presidential years, going back to 1916 (the first presidential election after the 17th Amendment inaugurated the direct popular election of Senators). How has the party losing the presidential election typically fared? The chart below ranks the performance of the losing party by a rough “Score” metric explained in the notes, ranging from the best losing parties (Al Gore’s Democrats in 2000 and Bob Dole’s Republicans in 1996) to the worst (Herbert Hoover’s Republicans in 1932).
There’s a lot to unpack here. Let’s start with the topline: The average losing party in a presidential election loses two seats in the Senate, and loses 21 of 34 Senate races. Losing a net of two seats would be good news for Republicans; but Mitch McConnell and his caucus are defending 23 seats out of 35, so they will either have to win a lot more races than usual, or lose a lot more seats.
Overall, the party losing the presidential election has gained seats ten times, broken even twice, lost two seats four times, and lost four or more seats ten times. So, 16 out of 26 prior outcomes would be consistent with Republicans holding their majority. But five elections led to major shifts of eight to twelve seats in the Senate in favor of the winners: Warren G. Harding in 1920, Franklin D. Roosevelt ousting Herbert Hoover in 1932, Harry Truman’s surprise reelection in 1948, Ronald Reagan ousting Jimmy Carter in 1980, and Barack Obama in 2008. Those winners had coattails.
Republicans need to win 21 Senate races to keep 51 seats. Only one losing party has managed that: Bob Dole’s Republicans in 1996. Dole is remembered as a disastrously bad candidate who won a popular-vote majority in only six states — the largest of which was Alabama — but his frenetic campaign down the home stretch in 1996 against a complacent Democratic Party saved the Republican majorities in both Houses, leaving Republicans with 55 senators. Only in three other elections (Richard Nixon over Hubert Humphrey in 1968, George H.W. Bush over Michael Dukakis in 1988, and George W. Bush over Gore in 2000) did the losing presidential party win more Senate races than it lost. All of those involved voters choosing Republican presidents and Democratic senators.
Note the distinction between national Republican losses in 1976, 1992, and 1996 — all of which saw Republicans hold their ground or gain seats in the Senate — and the last two Republican losses, in 2008 (when Republicans lost eight seats) and 2012 (when they lost 25 out of 33 races). The change for the worse may be attributable solely to Obama’s unique power to turn out his voters, but it also reflects more broadly the shift toward party-line voting and elections decided by turnout.
Also, this time, unlike in 1992 or 1996, there is no third-party candidate whose voters might pull the Republican lever downticket. The 1948 election is explained in good part by the fact that the Democrats split in three at the presidential level (with Henry Wallace defecting from the party’s left wing, and Dixiecrat Strom Thurmond defecting from its right) while those voters pulled the lever for House and Senate Democrats.
Losing the Close Ones
Those are the outcomes. But how did we get to them? One thing that jumps out of the data, notwithstanding a few exceptions (such as 2000, 1932, and 1920) is that there is not actually a very strong relationship between the national popular vote (for which, as usual, I use the two-party vote) and a party’s performance in Senate races. Partly, that is because of the uneven role of incumbency in Senate races, but it is also because of the way the Senate map works. Each state counts equally regardless of size, and a third of the states have no Senate race in a typical presidential year. So, the presidential ticket’s performance may look quite different when matched up with the Senate map.
I calculated, for the column marked “Avg2P,” the two-party popular vote each year across only the states with Senate races. I used an unweighted average (the two-party vote for each state, divided by the number of races) rather than add up the votes, to approximate the equal weight of each race in the Senate. In some years, that gave the losing presidential party a significantly better showing. Dole may only have won a majority in six states, but five of them had Senate races, and one (his home state of Kansas) had two. And the same class of Senate seats that was up in 1996 is up again this year. By this measure, Nixon in 1960, Gerald Ford in 1976, and John McCain in 2008 all averaged a majority of the two-party popular vote in states with Senate races. It did not do Republicans much good in 2008, however — with the same map as 1996.
Based on current polling, Trump’s unweighted average is 52 percent of the two-party popular vote across all states with Senate races (counting Georgia twice), and 50 percent across the 23 Senate races in which we have Senate polling. In that sense, Republicans are again playing on home turf in the Senate races. It does not matter that Trump will get massacred in California, New York, Hawaii, Vermont, Maryland, Connecticut, and Washington, because none of those states has a Senate race in 2020.
Averages, however, can also be deceiving. Blowout or uncontested races can significantly skew the picture of what the landscape looks like for contested races. Let’s look at how the outcomes of Senate races in presidential years were distributed:
Unsurprisingly, the best and worst showings by a losing party tend to line up with the years when the losing party won, or lost, a lot of the races decided by single-digit margins. Democrats in 2000 ran the table in close races, even winning one contest with a dead man running against the incumbent, and Republicans in 1996 also won more than half of the single-digit races. In extreme Democratic blowout years such as 1932, 1936, and 1964, by contrast, there are very few close races at all. Republicans also had more blowouts than the Democrats in 1996: twelve races won by double digits, compared with six for the Democrats.
1980 is an extreme example of a sweeping presidential victory carrying a lot of Senate seats with it. Senate Democrats, on average, won 51.2 percent of the vote, running nearly ten points ahead of Jimmy Carter. But winning the Louisiana Senate race by 93 points did not matter nearly as much as the fact that 19 Senate races that year were decided by less than 13 points, and Democrats lost 16 of them. The Reagan Revolution would have looked very different if Reagan’s victory had not carried in all those Republican senators. That had long-term consequences: 1984 would be the first time Republicans won ten or more Senate races by double-digit margins since 1928; safe Republican seats have been a fixture ever since. The 19 such Republican wins in 2016 was the most ever. An outcome like 1980, with Trump in the role of Jimmy Carter, would be the Republican nightmare scenario. On the other hand, one factor that dispirited Democratic voters that year in the Western states was that Reagan’s victory was obvious from the East Coast results announced while the polls were still open. That is not happening in 2020.
One notable item in Senate history that has now all but vanished is the uncontested race. Tom Cotton this year is only the fifth Republican Senate candidate since 1916 to face neither an official nor de facto Democratic opponent (I counted independents where it was clear which side they would caucus with.) By contrast, 49 Democrats have run uncontested, the great majority of them in southern states before 1980. It was once commonplace for Republicans to give up entirely on three or four Senate contests in the South every year.
Other than 2008 and 2012, the states in which the losing party got clobbered in competitive Senate races tended to be years when the losing party did quite poorly, on average, in the presidential vote in states with Senate races, creating an undertow that the party’s Senate candidates could not resist. Likewise, the years when the presidential ticket did well in the Senate states tended to see a much better performance in the competitive Senate races. That bodes well for Republicans if Trump does as well as, or better than, his current polling across the states that control the Senate. As of now, it does not yet appear that the bottom has dropped out of Trump in states such as Texas, Georgia, Arizona, Iowa, or North Carolina. That could matter quite a bit even if Trump himself goes down to defeat.