Which Models, and Which Precedents, Will Appear on Election Day?

As the countdown to the 2006 midterm elections turns from days into hours, there remains considerable uncertainty about the outcome. The uncertainty flows from something more than the fact that polls show a number of key races as toss-ups. There are deeper questions involved, which have the potential to shake up the way political analysts look at both midterm elections and survey research.

One one hand, the broad historical precedent is clear: the president’s party almost always loses seats in the House and usually loses seats in the Senate in midterm years. Between 1894 and 1994, this occurred in 25 of 26 midterm elections in the House and 18 of 26 in the Senate. Few analysts doubt that Republicans will lose at least some seats in both chambers this year.

On the other hand, the last two midterm elections have been exceptions to the rule. And the rule itself has been growing weaker even when it has appeared. Except for the blowout of 1994, every midterm since 1978—a period now spanning nearly three decades—has had lower than average presidential losses, leading to the obvious conclusion that there is a new normal. Indeed, in 1986, 1990, 1998, and 2002, House seat shifts in either direction have been in single digits.

On one hand, past midterm elections have inspired models for the prediction of midterm results. These models are based on national conditions, including presidential approval rating, congressional approval rating, economic conditions, and national “generic poll” results testing respondents’ party preference in congressional elections. These measures, even when mitigated by strong economic conditions, forecast a drubbing of Republicans. By way of comparison, the approval rating of President Bush (around 38 percent) and Congress (around 19 percent) is just about where President Clinton and the 103rd Congress were in the fall of 1994.

On the other hand, a close look at the elections race-by-race does not seem to show a Democratic tide, but rather points to Democratic House gains right on the edge of the 15 they need for control as well as a three or four seat gain in the Senate—useful tonic for the party, but not a victory that sweeps all before it.

Whether Tuesday will see a Democratic tide or a scenario in which the parties fight it out seat by seat will depend on which of two competing precedents—and the rules of thumb that come with them—will obtain.

The first set of precedents comes from long observation of incumbents in congressional elections, which has produced a famous rule of thumb holding that incumbents who are polling below 50 percent before election day will not win. The logic here is that incumbents are a known quantity and that if voters have not been sold on them yet, they are not going to be. The postulated result is that the undecideds break overwhelmingly for the challenger. If this happens, Democrats will win big, perhaps even taking over the Senate.

The second set of precedents comes from close observation of the last two congressional elections (2002 and 2004), which produces a very different rule of thumb. The rule of 50 percent is dead or dying, goes this school of thought, and the new rule is that Republicans are undercounted in pre-election polls by up to five percent regularly. There are at least two possible reasons this has happened in recent years. Perhaps pre-election polls have not accurately captured pro-Republican sentiment that existed all along; or perhaps they have failed to capture the last-minute successes of the Republican get-out-the-vote machinery. In either case, congressional polls in the last two election cycles have consistently over-estimated the Democratic vote percentage. If that is the case in 2006, Republicans will win the races that now seem close, easily holding the Senate and possibly holding the House, as well.

Unfortunately for prognosticators, there is little data to offer a clue. Both parties claim to be doing well among absentee voters, although the Wall Street Journal has reported that Republicans have the better of that dispute. Otherwise, much of the pre-election analysis is really being colored by a methodological prejudice or partisan wishful thinking. In the end, if Democrats wind up winning big gains, the more conventional way of looking at midterm elections will have held up. If they don’t, analysts will have to go back to the drawing board with some interesting questions, such as: Why is the 50-percent rule defunct? Or were the endangered Republicans actually over 50 percent all along? And if so, why didn’t we know?

Andrew E. Busch is a Professor of Government at Claremont McKenna College and an Adjunct Fellow of the Ashbrook Center.