Used to be, among practitioners and pundits of contemporary politics, midterm losses for the president's party were virtually a given. In only two elections in the last century did the pattern fray: 1934, during FDR's New Deal, and 1998, in the wake of Clinton's Impeachment Scandal.

Accordingly in 2001, Roll Call reported on "A History Lesson on Midterms: Usually No Picnic for President's Party." The piece qualified that "Although the party out of power traditionally gains seats in the election following a presidential race, historians say that larger forces -- such as recession or political scandal on the scale of Watergate -- are required to spark major change in Congress." However, on average, Roll Call tells us, the White House's party loses 25 House seats and four senate seats.

In the coming two years, however, there may be a temptation to approach the mid-term elections differently, both on the part of the party in power and the press. Both of the last mid-term elections spared the incumbent significant losses, and the memory of the last harsh mid-term is now a decade in the past. Many will point to 1998 and 2002 and predict that the off-year election curse is broken along with the Curse of the Bambino.

So can we thumb our noses at the mid-term losses rule of thumb?

"Surge and Decline"

In 1960, the seminal work The American Voter (1960) was published in a collaboration of scholars from the University of Michigan's school of political science. The work contained the classic "Surge and Decline" model of midterm elections. The model stipulated that the winning president attracted voters of the other party and that independent voters jammed to polls to support the winning candidate. When those forces vanished two years later, so did the president's party's advantage, and midterm losses were the result.

In 1987, James E. Campbell revisited Voter's model. Contrary to the theory of Voter, Campbell found a surge in turnout among partisans and a change in vote choice among independents, instead of the other way around. He agreed with The American Voter's argument that "Midterm elections are, at least in part, repercussions from the previous presidential election." But the centerpiece of this theory was that "Short-term forces manifest themselves in the vote choices of independents and in the turnout of partisans." (emphasis added)

Why? Because, he speculated, Independents are the voters most likely to be influenced by the course of the campaign. By contrast, partisans "contribute to presidential coattails...primarily through their turnout decisions."

Certainly this jives with what we just saw. Leading up to Tuesday, in race after race at the Congressional level, pre-election polling cast a dim view of Republican chances: in Florida, Kentucky, Alaska, and Louisiana, for instance, Senate candidates looked headed for trouble. But the massive turnout of the Republican base banked on by the Bush White House appears to have lifted all boats, and all those Senate candidates won outright. (In the case of Louisiana the question was more whether GOP candidate David Vitter would avoid a runoff.)

The reversal of these coattail forces -- with Independents being persuaded by the winning presidential campaign, while more importantly partisans of that campaign turn out in higher numbers -- in theory would sharpen the impact of the president's actual win on the midterm election. That is, if it is in fact a surge of the president's party that helps bring him to office -- not the surge of Independents that Voter had theorized -- then the loss of that surge two years later would have an extremely strong effect on the success or failure of congressional candidates.

The findings are clear, said Campbell: "A party's congressional midterm losses are proportional to its presidential vote in the previous presidential election." Two other factors must be taken into account, namely the state of the economy and the popularity of the president on election day. Controlling for those externals, though, Campbell developed a crucial statistical rule of thumb:

"A one percentage point increase in the presidential vote sets the party up to sustain a loss of four to five [House] seats in the next midterm."

After Eisenhower's big win in 1954, for instance, the Roll Call piece noted that, "While many Republicans had swept into office on Eisenhower's coattails during the preceding presidential election, they were swept out with the predominant political trends of the time two years later." By contrast, in 1962, after Kennedy's bare victory over Nixon, as we've seen, the Democrats sustained only minor losses. The same was true in 1998 after Clinton's strong 1996 win was held under 50% by H. Ross Perot, and in 2002 after George W. Bush lost the popular vote but won the presidency.

Is there a positive effect during the Surge, that is, "coattails"? Campbell finds that in a mirror image of midterm elections, when controlling for the number of seats already held, in presidential years, "for every additional percentage point of the presidential vote, the party can expect typically to gain...better than three House seats."

Thus the size of the president's win has an extremely important impact on the shape of the Congress, both during the presidential election and two years later.

The Referendum Effect

In 1992, James Campbell updated his model to check for a so-called Referendum Effect, which was proposed as an alternative explanation by other political scientists (especially Tufte in 1987). The Referendum Effect scholars theorized that the success or failure of the president's agenda had a dramatic effect on Congressional losses in the midterms.

Campbell agreed, but he brought to light a crucial distinction. "Midterm elections are also in part referenda on the performance of the president," he acknowledged -- but, he insisted, a midterm is "a muted referendum." The referendum effect helps determine the size of the loss, Campbell said, almost never counterbalancing it. He describes the effect this way:

"The president's party consistently loses in the midterm because of the prior presidential surge; its losses vary a good deal because the public's midterm judgments vary, but also, at least as importantly, because the magnitude of the prior presidential surge varies."

According to Campbell, the size of the surge is "a result of the withdrawal of the strong and positive short-term forces of the prior presidential election year and the public's midterm judgment of the administration's performance." That of course includes the economy's state at the time of the election.

However, the overall message remains clear: the margin of the president's victory is the source of the fact of midterm losses; that margin combined with other factors determine its size. Combining three effects -- the size of the presidential victory two years before, the popularity of the president leading up to election day, and the state of the economy -- explains 90% of the variation in seat change from 1946-1990, a whopping number that would make most political scientists drool in envy.

2000 vs. 2004 -- From Parity to Surge

On December 13, 2000, George W. Bush became the first son of a president since 1825 to gain the White House. He lost the popular vote by three tenths of one percent -- 49.8 million to 50.2 million votes. The press' reaction was to predict a terrible hangover after the 36-day post-election party. The Boston Globe, for instance, noted that Bush had inherited with his election "a sharply divided electorate, a whisper-thin Republican majority in the House, and a Senate split 50-50 for the first time in its history."

The implications for Congress? The recent Roll Call piece noted that "if history is a guide, the midterm elections following the installation of George W. Bush as president should be a boon for House and Senate Democrats, who stand a hair's breadth away from capturing majorities in both chambers."

Yet, Campbell's study proved accurate, and Bush gained seats in Congress.

This year, by contrast, the Republicans have seen a clear surge. Karl Rove's much-maligned strategy of appealing to the base jackpotted the President into the White House and a wealth of new legislators onto Capitol Hill.

While votes are still being counted, today the President leads Sen. Kerry 59,117,523 to 55,587,524, or 51%-48% (http://news.yahoo.com/elections). President Bush's share of the two-party vote -- the crucial variable in Campbell's calculation, is higher -- about 51.5%. In 2000, the President won 50,456,002 votes, while Gore captured 50,999,897, for a baseline of 49.7%. The difference is about 2 percentage points. By Campbell Math, that two percentage-point difference translates into a 5-10 seat loss in the House of Representatives. Not an enormous, 1994-style rout, but a set-back nonetheless.

What is to be Done?

Why rain on the Bush parade? Three reasons, if you're a Republican.

First, Republicans on the Hill should understand they face a relatively tough fight in 2006, at least in theory. Under no circumstances should 1998 and 2002 lure the GOP into a false sense of security, that mid-terms have lost their bite, if they know what's good for them. One would expect an uphill race, so Republicans should prepare for it starting right now, today.

Second, Republicans everywhere should prepare the public and the media. Both audiences tend to view the mid-term elections as the Referendum Effect scholars view them: As an up-or-down vote on the president's governing. Getting the word out that a large win like Bush's leads to challenges down the road will help manage expectations.

Third, following the first two, Republicans in the White House have History to consider. To the extent that the focus of the GOP turns now to "legacy" issues, it's important both that the Hill be prepared to fend off a Democratic tide, and that the press understand that tide exists. Otherwise, if mid-term losses do materialize, any accomplishments of the next two years will be muted by a sense the public rejected it.

Obviously, the sense that crusading Democrats may come crashing through the GOP defenses in the House and Senate is not an encouraging message to spread in the days and weeks to come. Such things have an impact on vulnerable incumbents' decision of whether or not to run. But conservatives must be forewarned, and so forearmed.

Finally, note for the record something utterly intuitive. Campbell's model controls for two things: the president's popularity and the economy. In other words, increases in presidential popularity and improvements in the economy can slacken the force of post-surge declines.

Thus managing expectations and preparations for a hard-fought off-year election battle should be coupled with a conscious campaign to improve the economy and the president's standing, taking care to avoid the overreaching that put Clinton's approval ratings into a tailspin and generated the 1994 Gingrich Wave. The economy and the president's approval cannot eliminate a mid-year decline, Campbell teaches, but it can modulate its size.

As Campbell put it, "there is a presidential pulse to congressional elections."

The GOP should be working today to see that the Democrats' 2006 pulse skips a beat.

See: James E. Campbell. 1993. Surge and Decline: The National Evidence. In Controversies in Voting Behavior. Washington, D.C.: Congressional Quarterly, pp. 222-240