President Clinton won reelection by shaping a new coalition, while Democrats failed in their bid to retake the House in part because they allowed the composition of their support to stagnate or deteriorate in the face of a changing electorate, according to a detailed analysis of exit poll data.

A comparison of data from 1992 -- a winning year for both Clinton and House Democrats -- with last Tuesday's results shows a major divergence between the presidential and congressional wings of the Democratic Party.

This year, Clinton enlarged his advantage among such key constituencies as women, Catholics, Hispanics and young voters by margins more than large enough to counter losses in other groups. But House Democrats saw virtually no change in their support among women, young voters and Hispanics, while losing ground with Catholics.

The ability to build new sources of support was crucial in the election for Clinton and the House congressional candidates because both saw their support slip with men and decline less significantly among married voters and voters from union households. Clinton saw no change in the level of backing among Protestant voters, while House Democrats saw a significant erosion of support in this key group.

"A lot of the coalition Clinton put together fragmented and stuck with Republicans for Congress," said GOP pollster Bill McInturff, adding that if House Democrats had been able to grab on to the same demographic groups that Clinton succeeded with, "they would be a majority today." Instead, McInturff said, "people were saying they {the Democrats} had not been out long enough, they had not really learned the lesson, and if they were put back, they would do the same thing -- raise taxes, be too liberal."

Democratic pollster Alan Secrest did not disagree. "Voters have not fallen back in love with the Democrats," he said, adding that the closing weeks of the campaign demonstrated how vulnerable members of his party were to GOP themes that Democrats are "puppets of union bosses" who will revive "tax and spend" policies, and that "the last thing we need is a liberal in Washington."

While support for House Democratic candidates, especially less well-known challengers, dropped during the closing assault using these themes, Clinton set out to insulate himself from just such liabilities. The result was a Democratic victory at the top of the ticket that was not replicated further down the ballot.

Rahm Emanuel, special projects director at the White House, contended that Clinton's gains among Catholic voters were crucial to his ability to carry the battleground terrain of the industrial Midwest. Clinton won Ohio, Michigan and Illinois, where the Catholic-ethnic vote is substantial. In Illinois, exit polls show that Clinton lost to Dole among Protestants, 43 percent to 46 percent, but overwhelmed Dole among Catholic voters, 54 percent to 38 percent.

Hispanic voters, in turn, were key to Clinton's win not only in California, but in what had been a solid GOP state, Florida, where many of the Hispanic voters are of Cuban descent and much more strongly inclined to vote for Republicans than are Hispanic voters with roots in other countries. In Florida, Clinton held his own with Hispanic voters, losing to Dole by just 2 percentage points, 44 percent to 46 percent.

The differences between the demographic gains and losses of Clinton and House Democrats are striking.

In the crucial area of gender, for example, Clinton made substantial gains among women. He beat Republican nominee Robert J. Dole among female voters by 9 percentage points more than he beat George Bush. House Democrats, in contrast, showed no improvement among women over these four years, while support from men fell by a disastrous 13 percentage points. Clinton's gains among women allowed him to absorb a 4 percent decline in support among men.

Similar patterns emerged with different age groups, with members of different religions, with different income groups, particularly the lower rungs of the middle class, and with the increasingly important block of independent voters.

Clinton, from 1992 to 1996, improved his performance over his Republican opponent by 2 percentage points among independents, while Democrats running for the House saw their level of support fall by 10 percentage points among these voters. Similarly, Clinton gained 8 points among moderate voters, while House Democrats saw virtually no change in their level of support among these voters.

The calculation of the shift in presidential vote is based on comparison of the differences of the Clinton-Bush percentages in 1992 and the Clinton-Dole percentages in 1996 among different demographic groups, a method used by many partisan pollsters and political scientists. For congressional contests, which were not clouded by third-party candidates, the calculations are based on straightforward differences in the votes for House Democrats and Republicans for the two years.

The ability of candidates and parties to adjust the demographic and ideological structure of their support is central in determining long-term success.

The electorate is constantly changing. In the relatively short period from 1992 to 1996, for example, the percentage of voters casting ballots who had completed college grew from 39 percent to 44 percent. Ideologically, the electorate moved 5 points to the right, with the percentage of voters identifying themselves as liberals dropping by 1 point, and the percentage of conservatives growing by 4 points. In the close contests that determine the makeup of the House and Senate, these kinds of shifts are crucial.

At the same time, the parties are developing new core constituencies. Single working women, for example, have become a bedrock of the Democratic Party, supporting Clinton by nearly 40 percentage points, 65 percent to 26 percent. White married men, conversely, are now solidly with the GOP, supporting House Republican candidates by 61 percent to 37 percent, according to exit polls.

McInturff said post-election research by his firm found that some of the groups showing the highest levels of ticket splits -- voting for Clinton and Republican congressional candidates -- were younger voters under the age of 35 and women with college degrees.

Fully 82 percent of voters believed that Clinton would be reelected, McInturff said. The groups that voiced the strongest fears of a Democratic Congress and a Democratic White House, he said, included not only such traditional Republican constituencies as Christian Coalition members, households with National Rifle Association members and opponents of abortion, but also moderate Republicans.

In another reflection of the liabilities of the Democrats, McInturff said, the percentage of people who said Clinton would govern "as a liberal" jumped from 23 percent to 43 percent when voters were asked if Clinton would govern differently with a Democratic Congress.

Stanford University political scientist David Brady, who conducted his own analysis of the House races, said the most striking finding was that the main vulnerability of the GOP lies in keeping in office those who represent moderate districts. The 18 districts where Republicans lost, he said, had electorates that supported Clinton by substantially higher percentages in 1992 than the districts Republicans retained. The average 1992 Clinton vote in the losing districts was 47 percent, 4 points higher than the 43 percent plurality he won with that year, while winning Republican House members represent districts where the average vote for Clinton was 38.9 percent.

Brady said, however, that the failure of the Democrats to retake the House this year is likely to compound the party's difficulties, because senior Democratic members, many representing marginal districts, now have more of an incentive to retire because they will remain in the minority, out of power.

"A lot of the old bulls are going to drift off into retirement," he said, and these retirements will give the GOP new opportunities to pick up seats, as it did in a number of southern districts this year that had been represented by senior Democrats who decided to end their congressional careers.