2002: When the President's Party Gained Seats

Republicans gained eight House seats and two Senate seats in the 2002 midterm, and took control of the Senate in the process. The Office of the Historian of the US House records the House division moving from 221 Republicans and 212 Democrats in the 107th Congress to 229 and 205 in the 108th. That is a small number in absolute terms and a very large one in context, because the president's party almost never gains seats in a midterm at all.

The standard account files 2002 as an anomaly produced by the environment after 09/11/2001. The evidence points at something a campaign can use. The midterm penalty is a conditional rather than a law, it switches off under conditions that are measurable in advance, and 2002 met them. Treating the penalty as automatic is how a party writes off seats it could hold and defends seats nobody is attacking.

Two conditions did the work in 2002: a president well above 50 percent job approval, and a governing party holding close to its natural number of seats rather than an inflated one. What would prove this wrong is a midterm in which a president above 60 percent approval lost heavily anyway. Across the 18 midterms Gallup has measured, there is none.

What happened in the 2002 midterm

The 2002 midterm moved the House, the Senate and the governorships in three different directions. Republicans gained eight House seats. Republicans gained two Senate seats and with them control of the chamber, going from 49 seats to 51. Democrats gained three governorships, reducing the Republican advantage among governors from 27 to 21 down to 26 to 24, by UPI's election-night tally.

House figures from the Office of the Historian of the US House, which states that its counts are the party divisions as of the initial election results. Senate figures from the US Senate's party division table; the pre-election Senate line is the chamber as organized from 06/06/2001. Governorship figures from UPI's election-night report.

Chamber or office	Before	After	Change
House Republicans	221	229	+8
House Democrats	212	205	-7
Senate Republicans	49	51	+2
Senate Democrats	50	48	-2
Republican governors	27	26	-1
Democratic governors	21	24	+3

The Senate arithmetic needs its history attached. The 2000 elections produced a Senate of 50 Republicans and 50 Democrats, organized under a power-sharing agreement adopted on 01/05/2001. Vice President Dick Cheney's tie-breaking vote gave Republicans the majority from 01/20/2001. James Jeffords of Vermont then left the Republican party, announcing the decision on 05/24/2001, and the Senate convened on 06/06/2001 with Democrats as the majority party. So the 2002 election was fought over a chamber Republicans had held, lost without an election, and needed two seats to recover.

The Senate before and after the 2002 midterm

Seats by party, as the chamber was organized from 06/06/2001 and as the 108th Congress opened.

Republicans won Georgia, Minnesota and Missouri; Democrats won Arkansas. Two of those involved deaths. Jim Talent won the 11/05/2002 Missouri special election for an unexpired term and took office on 11/23/2002, replacing Jean Carnahan, who had been appointed after Mel Carnahan died during the 2000 campaign and was elected posthumously. In Minnesota, Paul Wellstone died in a plane crash on 10/25/2002, Walter Mondale replaced him on the ballot, and Norm Coleman won. On the Clerk of the House's official count, Republicans took 37,091,270 votes to the Democrats' 33,642,142 in House races nationally, a margin of about four and a half points.

How often the president's party gains seats in a midterm

The president's party has gained House seats in four of the 41 midterm elections held between 1862 and 2022: 1902, 1934, 1998 and 2002. Brookings tracks every one of them in its Vital Statistics on Congress series, and the pattern is close to a law. The average is a loss. The American Presidency Project puts the average midterm loss between 1934 and 2014 at 27 House seats and four Senate seats, and Gallup puts the average House loss since 1946 at 25 seats.

House seat change for the president's party in every midterm since 1862

Each column is one midterm election, colored by the president's party. Only four of the 41 sit above the line.

Three of the four gain years involved gains in both chambers: 1902, 1934 and 2002. The 1902 case carries a caveat that is worth stating plainly, because it is the kind of thing that gets lost in a seat count. Brookings notes in a footnote to the same table that although Republicans gained nine seats in 1902, they lost ground to Democrats, who gained twenty-five, because the House itself grew after the 1900 census. Reapportionment legislation enacted on 01/16/1901 increased the House to 386 seats from 357. Republicans held more seats and a smaller share of them.

That leaves 1934 and 2002 as the clean cases. Both followed an event that reorganized public opinion around the president rather than around the normal business of government, and in both the president's party was defending an unusually small number of exposed seats.

Why Republicans gained seats in 2002

Republicans gained seats in 2002 because George W. Bush sat at 63 percent job approval, a level at which no president's party has ever lost more than five House seats. Gallup's final pre-election reading, interviewed 10/31 to 11/03/2002, put Bush at 63 percent approval. Gallup's own long-run finding is that presidents below 50 percent approval have seen their party lose an average of 37 House seats, against 14 seats for presidents above 50 percent.

George W. Bush's Gallup approval through 2002

Gallup job approval readings through the midterm year, against the level recorded in the days before 09/11/2001.

Presidential approval against the midterm seat result, 18 midterms

Each point is one midterm between 1946 and 2014, placed by the president's Gallup job approval and by his party's net House seat change.

The level matters more than the trajectory, and the trajectory was downward. Bush was at 51 percent in Gallup interviewing conducted 09/07 to 09/10/2001. He reached 86 percent within days of the attacks and peaked at 90 percent on 09/21 to 09/22/2001, the highest figure Gallup had recorded for any president. He then lost ground for thirteen straight months. A campaign reading the trend line in the summer of 2002 would have seen decline. A campaign reading the level would have seen a president still 12 points above where he sat before the attacks, in a system where 50 percent is the threshold that matters.

The second explanation is the national security argument itself. The House passed the Iraq war resolution 296 to 133 on 10/10/2002, with Republicans voting 215 to 6 in favor and Democrats splitting 81 to 126. The Senate passed it 77 to 23 just after midnight on 10/11/2002. The proposed Department of Homeland Security produced a separate fight that ran through the autumn, and the mechanics of that fight are routinely described backwards. Five cloture votes on the homeland security bill failed between 09/19/2002 and 10/01/2002, and in each one Senate Democrats voted overwhelmingly in favor of cloture while Senate Republicans voted overwhelmingly against. The dispute was over personnel rules for the new department. Republicans campaigned on the stalemate successfully, which is a fact about message discipline rather than about who held up a bill.

The third explanation is effort. The Associated Press reported on 10/22/2002 that Bush had headlined 66 fundraisers that year and raised more than $140 million for Republicans, and he closed the campaign with rallies at 17 stops in 15 states between 10/31/2002 and 11/04/2002. Published totals for the cycle range from about $136 million to $145 million.

What the 2002 redistricting cycle did and did not do

The 2002 midterm was the first election run on districts redrawn after the 2000 census, and whether that redraw produced the Republican gain is genuinely contested among people who have studied it. The Census Bureau's apportionment brief records 12 seats shifting among 18 states, with Arizona, Florida, Georgia and Texas each gaining two and New York and Pennsylvania each losing two.

Sam Hirsch, writing in the Election Law Journal in 2003, attributes essentially the whole Democratic loss to the redraw. His article argues that the entire six-seat net loss Democrats suffered in November 2002 can be attributed to the correction of population inequalities that had accrued during the 1990s. Hirsch served as counsel to Democratic redistricting interests, which is worth knowing when weighing the claim, and his own competitiveness figures complicate it: he reports that the 2002 plan contained roughly nine or ten fewer competitive districts than the 2000 plan, and twelve more districts where neither presidential candidate had won by more than two points.

Alan Abramowitz, Brad Alexander and Matthew Gunning, writing in the Journal of Politics in 2006, find no support for the redistricting hypothesis as an explanation for declining competition, and attribute more of it to incumbency and to the sorting of voters. Gary Jacobson's account of the cycle credits three causes together: the rally around Bush after the attacks, redistricting, and stronger Republican turnout.

Midterm turnout in 1998, 2002 and 2006

Share of the voting-eligible population casting a ballot for the highest office on the ballot.

Turnout can be checked quickly, and in aggregate it does not carry much weight. The University of Florida Election Lab puts 2002 turnout at 39.5 percent of the voting-eligible population, against 38.1 percent in 1998 and 40.4 percent in 2006. Whatever happened in 2002 did not show up as an unusual number of Americans voting. If Republican turnout was stronger, it was stronger relative to Democratic turnout inside a normal-sized electorate.

What 2002 tells a campaign about midterm exposure

The structural fact underneath 2002 is that a party cannot lose seats it does not hold. Republicans entered the 2002 midterm holding 221 House seats, which is a bare majority and a small surface area. Brookings records 35 House incumbents retiring in 2002, 398 seeking reelection, and 383 of them winning, a 96.2 percent reelection rate. Eight incumbents lost in the general election and eight in primaries. There was very little in play.

Compare that with a normal midterm loss. A party loses 25 or 30 House seats when it is defending 30 or 40 marginal ones, which happens after a good year in which it picked up seats it had no business holding. The size of a midterm loss is set years earlier, by how many of the party's seats sit in territory that does not naturally belong to it.

That is the transferable lesson, and it is measurable now rather than in November. Count the seats your side holds in places where the underlying partisan lean runs against you. Count the open seats where an incumbent's personal vote is about to disappear. Those two counts, taken together, put a floor and a ceiling on the year before a single advertisement is bought. Bush's approval rating explains why Republicans did not lose in 2002. The small number of exposed Republican seats explains why they could not have lost very much even if he had been unpopular.

The 1902, 1934 and 2002 exceptions all share that second property. Each followed a period in which the president's party was holding close to its natural level rather than an inflated one. Approval ratings and national events decide which direction a midterm moves. Exposure decides how far it can move, and exposure is knowable in advance.

Questions about the 2002 midterm

Did Republicans gain seats in the 2002 midterm?

Republicans gained eight House seats and two Senate seats in the 2002 midterm, and took control of the Senate. The Office of the Historian of the US House records the division moving from 221 Republicans and 212 Democrats to 229 and 205. Democrats gained three governorships in the same election.

How often does the president's party gain seats in a midterm?

The president's party gained House seats in four of the 41 midterm elections between 1862 and 2022: 1902, 1934, 1998 and 2002, by the Brookings Vital Statistics on Congress series. Three of those four also produced Senate gains. The average midterm result is a loss of about 25 to 27 House seats.

What was George W. Bush's approval rating before the 2002 midterm?

George W. Bush stood at 63 percent job approval in Gallup interviewing conducted 10/31 to 11/03/2002, the last reading before the 11/05/2002 election. He had been at 51 percent just before 09/11/2001 and peaked at 90 percent later that month, the highest Gallup had recorded for any president.

Did redistricting cause the Republican gains in 2002?

Scholars disagree. Sam Hirsch argued in the Election Law Journal in 2003 that the entire net Democratic loss in 2002 can be attributed to the post-census redraw, while Abramowitz, Alexander and Gunning found no support for the redistricting hypothesis in the Journal of Politics in 2006. The 2000 census shifted 12 seats among 18 states, which is a real structural change whatever weight it carried.