2010: The Wave That Was Visible Early

Republicans gained 63 House seats in the 2010 midterm, the largest change in any House election since 1948. The House Historian's table of majority changes records the shift as 64 seats, from 257 Democrats and 178 Republicans to 193 and 242, while the figure in general use is 63, measured against the 179 seats Republicans held on the eve of the vote. Both numbers are official and they count from different starting points.

The standard account describes a wave that built through 2010. The evidence points at something harder for the losing side to hear. The wave was measurable from 11/03/2009, roughly a year before the vote, and the instruments that caught it were ordinary ones that both parties already owned.

What would prove this wrong is a 2010 in which the early signals pointed the other way or stayed silent until the summer. The record runs against that. What follows is the gap between when the evidence arrived and when the published forecasts moved.

What the 2010 midterm produced

The 2010 midterm produced a Republican House majority of 242 seats, a Republican gain of six Senate seats that left Democrats in control of that chamber, and Republican control of a majority of governorships and state legislatures. The Senate moved from 57 Democrats and 41 Republicans to 51 and 47, with two independents caucusing with the Democrats in the 112th Congress. On the Clerk of the House's official recapitulation, Republicans took 44,593,666 votes to the Democrats' 38,854,459 in House races, a margin of 6.61 points.

House figures from the Office of the Historian of the US House, whose counts are the party divisions as of the initial election results. Senate figures from the United States Senate's party division table; the pre-election Senate line is the chamber as it stood from 02/04/2010. Governorship figures from the National Conference of State Legislatures' partisan composition tables for January 2010 and January 2011.

Chamber or office	Before	After	Change
House Republicans	178	242	+64
House Democrats	257	193	-64
Senate Republicans	41	47	+6
Senate Democrats	57	51	-6
Republican governors	24	29	+5
Democratic governors	26	20	-6

The House before and after the 2010 midterm

Seats by party in the 111th and 112th Congresses, as the House's own historical table records them.

Two corrections belong with those numbers. The first is the comparison: computed from the Clerk's own table of political divisions, the 2010 change of 64 seats is the largest in any House election since 1948, when Democrats gained 75, and the largest in a midterm since 1938. The second concerns turnout. The United States Elections Project puts 2010 turnout at 41.0 percent of the voting-eligible population for the highest office on the ballot, against 40.4 percent in 2006 and 36.0 percent in 2014. The 2010 wave did not happen on unusually low turnout.

The state-level result was larger than the federal one. The National Conference of State Legislatures' partisan composition tables show Republican legislators rising from 3,246 to 3,944 and legislative chamber control moving from 27 Democratic and 14 Republican to 25 Republican and 16 Democratic. That mattered more than the House gain, because the 2010 winners drew the congressional maps used for the following decade.

What the early signals showed and when

The first measurable signals of 2010 arrived on 11/03/2009, a full year before the election. Bob McDonnell won the Virginia governorship 1,163,651 to 818,950, a margin of 17.36 points on the state's official returns, in a state that had elected Democratic governors twice running. Chris Christie beat the sitting governor Jon Corzine in New Jersey 1,174,445 to 1,087,731, a margin of 3.58 points on the state's official tallies. That same week, Gallup's generic congressional ballot showed Republicans at 48 and Democrats at 44, the first Republican lead of the cycle in a series that had started in July 2009 with Democrats ahead by six.

When the 2010 signal was available against when the forecasts moved

A calendar assembled by hand from ten dated sources, running from 11/03/2009 to election day. Measurable signals above, published forecasts below them in sequence.

The second signal came on 01/19/2010. Scott Brown won the Massachusetts Senate special election with 1,168,178 votes to Martha Coakley's 1,060,861, a margin of 4.76 points on the Secretary of the Commonwealth's official returns. The seat had been Edward Kennedy's since 1962 and Democratic since 1952.

Two contrary signals belong in the record, because a thesis that ignores them is not worth much. Bill Owens, a Democrat, won the New York 23rd district special election on 11/03/2009 in a long-held Republican seat. Mark Critz, a Democrat, held John Murtha's Pennsylvania 12th district on 05/18/2010 with 52.6 percent, in the only district in the country that had voted for John Kerry in 2004 and John McCain in 2008, after national Republicans had framed it as a second Massachusetts. Critz then held the seat again in November.

A third special cut the other way under conditions that limit what it proves. Charles Djou, a Republican, won the Hawaii 1st district special on 05/22/2010 with 39.4 percent against two Democrats who took 58.4 percent between them, in an all-mail single-round election with no runoff. He lost the seat in November by 6.46 points.

Specials are noisy and individually weak. The aggregate was not. By the end of 08/2010, Gallup recorded Republicans at 51 and Democrats at 41 on the generic ballot, which it described as its largest midterm generic ballot lead on record, against a previous high of five points measured in June 2002 and July 1994. Obama's Gallup approval had fallen below 50 percent in interviewing conducted 11/17 to 11/19/2009 and stood at 43 percent in the three days through election day.

When the published forecasts caught up

The published House forecasts did not reach the eventual result of 63 seats until the last week of 10/2010, eleven months after the first signals. The Rothenberg Political Report projected a Republican gain of 28 to 33 seats on 07/19/2010, noting that larger gains were quite possible. The Cook Political Report revised its outlook to a Republican gain of 32 to 42 seats on 07/29/2010.

Published House forecasts against the 2010 result

Each bar is one published projection range, with the eventual Republican gain of 63 seats marked.

Rothenberg moved to 37 to 45 seats on 10/05/2010. Cook published a final range of a Democratic net loss of 48 to 60 seats on 10/26/2010. Rothenberg published 55 to 65 seats on 10/28/2010 and held that range through election day. Only that last range contained 63, and Cook published its own with the note that higher losses were possible.

This is worth stating carefully, because it is easy to turn into an unfair attack on forecasters. Public raters are measuring seat by seat, they are constrained by what they can defend in print, and they moved steadily in the right direction all year. The point is about the distance between availability and action. A party reading the same public information in 11/2009 had the direction. A party reading it in 06/2010 had the magnitude. The forecasts that matched the result were published eight days before the vote, which is long after the budget, the candidate recruitment and the incumbent retirement decisions were all locked.

Which Democratic seats were exposed and how early that was knowable

Forty-eight House Democrats went into election day 2010 holding districts John McCain had carried in 2008, and 36 of them lost. CBS News, which called the group the McCain 48, reported that Republicans won 75 percent of those districts and that they accounted for more than half of the Republican pickups. Democracy in Action's parallel count of the 49 McCain districts that had elected a Democrat in 2008 finds Republicans winning 37 of them, with Democrats taking none back.

What happened to the Democrats holding McCain districts

The 48 House Democrats who went into election day 2010 in districts John McCain carried in 2008, by what happened to the seat.

The exposure measure is the most useful thing in the 2010 record, because it required no polling at all. Presidential results by congressional district were published in 2009. Any analyst with that table and a list of House members could count the seats sitting in hostile territory, and the count was 48 or 49 depending on where you drew the line. James Campbell's account of the cycle counts 47 Democrats defending districts carried by George W. Bush in 2004 and McCain in 2008, against six Republicans in districts carried by John Kerry and Barack Obama.

That asymmetry is the whole structure of the year in one pair of numbers. One side had four dozen seats standing on ground that had voted the other way for president. The other side had six. No message and no amount of money changes that ratio in a single cycle, and it was countable in 2009.

The enthusiasm measures said the same thing earlier than most people remember. Gallup reported on 06/21/2010 that Republicans were net plus 14 on midterm voting enthusiasm while Democrats were net minus 21, and called the gap the largest relative party advantage it had measured in a single midterm election-year poll. By the final pre-election reading, 63 percent of Republicans and Republican-leaning independents said they were more enthusiastic than usual about voting, and Gallup's later retrospective puts the Democratic figure at 44 percent.

What a party should do with evidence it does not like

The 2010 record shows a losing side holding good evidence for roughly a year and acting on it late. The information was public, cheap and unambiguous by the summer. The response, in retirements, recruitment, money and candidate behavior, came after the ranges moved in late October.

Three practices follow, and all three are about process rather than analysis. The first is to fix the measurement calendar before the year starts, so that off-year and special election results are scored against a prior expectation rather than argued about afterward. A special election is only a signal if somebody wrote down beforehand what a normal result would look like.

The second is to keep an exposure list and update it the week presidential results by district are published. That list changes slowly and it sets the plausible range of outcomes, which is what a forecast is for.

The third is the hard one. Assign somebody to argue the case the leadership does not want to hear, and give that person the same data and standing as everyone else. In 2010 the evidence was plain and it was unwelcome, and unwelcome evidence moves slowly through an organization unless somebody is paid to carry it.

Questions about the 2010 midterm

How many seats did Republicans gain in the 2010 midterm?

Republicans gained 63 House seats in the 2010 midterm by the figure in general use, and 64 by the House Historian's own table of majority changes, which counts from 178 Republicans at the opening of the 111th Congress to 242 at the opening of the 112th. The 63 figure measures against the 179 seats Republicans held on the eve of the election. Republicans also gained six Senate seats.

Was the 2010 wave predictable?

The direction of the 2010 wave was measurable from 11/03/2009, when Republicans won the Virginia and New Jersey governorships and took their first generic congressional ballot lead of the cycle in Gallup polling. The magnitude was measurable by 08/2010, when Gallup recorded its largest midterm generic ballot lead on record. Published seat forecasts did not reach the eventual 63 until the last week of October 2010.

What were the McCain 48?

The McCain 48 were the 48 House Democrats who went into election day 2010 holding districts John McCain had carried in the 2008 presidential election. CBS News reported that Republicans won 36 of the 48, or 75 percent, and that those seats accounted for more than half of the Republican pickups. The exposure was countable from published presidential results by district a year earlier.

Was 2010 a low-turnout election?

Turnout in 2010 was 41.0 percent of the voting-eligible population for the highest office on the ballot, by the United States Elections Project, against 40.4 percent in 2006 and 36.0 percent in 2014. The 2010 midterm had higher turnout than the midterm before it, so the result was not produced by an unusually small electorate.