Charlie Cook has a very good column in the National Journal (subscription), pondering whether or not Democrats will be able to take the House and/or Senate this year.
Mr. Cook makes the point that your conclusion to that question will depend on whether or not you adhere to a macro or micro theory of Congressional elections.
If you believe the micro theory, you say things like this: "The truth is that the districts have been so gerrymandered that every year, there are very few targeted ‘swing’ districts. Let’s look at the statistics: Incumbents get re-elected 99 percent of the time.” Additionally, you believe that most citizens hate Congress, but love their own Congressman…"
On the other hand, if you believe in the macro theory, you view Congressional elections as one big national election (rather than 435 small elections). If that’s the case, you believe the zeitgeist determines elections. And that means you probably believe the Abamoff scandal will have a dramatic impact on elections.
As Charlie Cook puts it, the truth is that most elections are decided on the micro level – but you never know when something dramatic will cause sweeping changes:
In any "normal" election year, the micro-political, race-by-race method is clearly superior, since those are the kind of races that late House Speaker Tip O’Neill had in mind when he said "all politics is local." But there are years, most notably 1958, 1964, 1966, 1974, 1980, 1982 (in the House only), 1986 (in the Senate only) and 1994, when politics was anything but local. In those years, the political laws of gravity were effectively suspended, causing large numbers of seats to change hands. In those elections, the inherent advantages of incumbency eroded significantly.