The Winston Group is a strategy and research firm dedicated to making ideas matter.

Without Proper Context, Polls Are Not Very Helpful

by David Winston

Here are a couple of gee whiz facts that, in this age of razor-thin presidential elections, most political observers and pundits seem to have forgotten. In the 1988 presidential contest, Vice President George H.W. Bush defeated former Massachusetts Gov. Michael Dukakis by just under 8 points, which produced a staggering 426 electoral votes for the winner.

Although the Democrat at one point led the sitting VP by 17 points, Bush rallied to beat Dukakis by 314 electoral votes. (Dukakis lost one when the Electoral College convened.)

In that election, Bush carried states such as Maryland, New Jersey, Connecticut, Illinois, Vermont and California. Yes, California. A short four years later, the tide had turned. Bill Clinton won by a margin of 5.6 points and got 370 electoral votes.

In the end, he trounced Bush by 202 votes in the Electoral College, carrying states such as Georgia, Tennessee, Kentucky, Louisiana and West Virginia.

No, the 1988 and 1992 elections were not held in a parallel universe. In fact, looking at recent history, these kinds of margins in presidential races were more the norm than the exception.

Five months from the 2008 election, political pundits and operatives are trying to use national and state polls to predict a winner in this fall’s presidential sweepstakes.

In the stampede to handicap the odds for the November election, the debate over the value of national polls versus state polls has become an ongoing argument in political circles. What the discussion needs now, as the campaign moves into the general election phase, is context.

In close election battles, like the two most recent presidential races, the resulting margin in the Electoral College has been so close that a slight change in voter preference in just a state or two could have changed the eventual outcome. A small difference in Florida in 2000 or Ohio in 2004, and we would be saying goodbye to President Gore or in the midst of President Kerry’s re-election battle.

But the closeness of those races has made us forget that when the margin at the national level becomes large enough, the margin in the Electoral College becomes not just bigger but exponentially bigger. What changes the dynamics of the presidential contest is the size of the winning national coalition, and that change simply overwhelms state-by-state dynamics.

In other words, when the breadth and depth of a candidate’s national coalition becomes substantial enough, it tends to tip many states in one direction. Historically, if a candidate wins the popular vote by approximately 5 points or more, the result in the Electoral College is far more lopsided than one might expect.

Harry Truman won in 1948 by 4.5 points yet defeated Thomas Dewey in the Electoral College by 115 votes (although one of his electors voted differently when the college convened). On the other hand, a win below 4 points tends to produce the kind of races we have seen in the past two elections, with Electoral College results of 271-266 in 2000 and 286-251 in 2004.

The one oddity came in 1968, when Richard Nixon won by less than 1 point but topped Hubert Humphrey by 110 electoral votes. This was the result of George Wallace’s third-party effort, which hurt Humphrey and skewed polls as predictors.

For a concrete example of how this phenomenon works, look at the 2004 election. Both presidential campaigns were state- focused and not national-coalition-focused. Instead, they relied on base strategies that depended on finding like-minded voters in key states and turning them out.

This strategy was unfortunate for Republicans at the House level because it caused them to lose independents for the first time since the Gingrich revolution and set the table for the one-sided 2006 elections. With both campaigns focused on their respective bases, their reach was minimized; and, as a result, neither candidate had much of an ability to get more than 50 percent.

Each was talking to just enough voters to win in the key states they needed to eke out a national victory while the big middle, hardly a small part of the electorate, got short shrift. This year’s presidential campaign is much different.

Both Sens. John McCain (R-Ariz.) and Barack Obama (D-Ill.) have talked about building large coalitions and running 50-state campaigns. They are competing for voters whom presidential candidates did not pursue in the two previous elections.

Whether it is McCain going after the old Reagan Coalition by reaching out to working- class voters, independents and Catholics, or Obama focusing on independents, young people and upscale voters, both are trying to build a coalition far different and far larger than their immediate predecessors. Regardless of the outcome, this approach is better not only for the election process but, inevitably, for whomever becomes the next president.

Why does it matter whether a president wins by 370 electoral votes or just seven? Close races deny presidents the kind of indisputable mandate that gives him or her the political and popular strength to govern, and that is the purpose of a president just as it is a political party — to govern.

In the next few months, daily national tracking and polling in key states will fuel a running debate on the actual status of the race. While state-by-state polls could be predictive, keep an eye on the national polling margins.

As long as one side or the other stays ahead in these polls by at least 5 points, it will be clear which candidate is succeeding in building a broad-based winning coalition.

Article on rollcall.com