Top Videos
Related Topics
election 2006
2008 Polls NationalIowaNew HampshireGeneral Election
GOP | DemGOP | DemGOP | DemHead-to-Head

Send to a Friend | Print Article


Republicans Are Lucky They Did Not Lose More Seats

By Jay Cost

Election Day 2006 was a bad day for the Republican Party. But it could have been much, much worse.

My estimate of the House vote (calculated by summing up all current totals reported by the media and projecting the vote totals of the Democrats and Republicans who were uncontested) indicates that the two-party vote was approximately 54% D to 46% R. This would mark a 5.4% decrease in the GOP's share of the two-party vote from 2004. The final two-party vote result was almost a perfect inverse of the 1994 result.

This estimate is very similar to the one that Rhodes Cook offers at Pew. The difference between us seems to be that I have estimated the Democratic votes in the uncontested races (which are 90% Democratic), and he has not. Cook comes out with a smaller Democratic margin than I do (his two-party vote is 53% D to 47% R), but he indicates that 5% of the votes remain to be counted, that they are mostly in Democratic strongholds, and that he expects the Democratic margin to increase.

(Note: The two-party House vote is computed by dividing the Democratic vote by the sum of the Democratic and Republican vote. It speaks to the following question: of the voters who registered a major party preference, what percentage supported the Democrats or Republicans? So, for instance, Cook finds that the public voted 52% D, 46% R, 2% Other. The Democratic two-party vote thus is: 52 / (52 + 46) = 53%. It is commonly used as a way to compare partisan electoral strength across time without getting tripped up by small changes in third-party votes.)

Over Election Day weekend, I offered an estimate of 19 seats based upon (a) Gallup's generic ballot prediction of 54-46 (once again - Gallup was closest to the mark), and (b) an OLS regression model of votes to seats for 1996 to 2004. On post-election Wednesday, I indicated that this latter model was incorrect -- that it overestimated the structural protections the Republicans enjoyed going into the election. It seems that, based upon my vote summations, I was overestimating Republican strength. I was using Gallup's accurate 54/46 estimate to predict 19 seats (with a 95% confidence interval of 11 to 27 seats). This prediction was outside the margin of error, I believe, because there had been, before this election, very little post-1994 variation in the two-party vote, with both parties hovering around 50%. Thus, I could not get a confident read on the slope of the post-1994 votes-to-seats line, as I indicated at the time. In other words, while it was clear that the structural changes that occurred in 1994 were ossifying House seat margins to an extent, the extent itself was not clear. I indicated over that weekend that I needed more observations to further test it. And Election Day provided a good observation.

Generally speaking, my overarching hypothesis about this election was correct: the structure of our system muted Democratic gains; however, its causal power was not as great as I thought it would be. The Republicans lost more seats than my model predicted, but they still lost far, far fewer seats than they would have with this kind of Democratic vote in the pre-1994 era.

By my estimate, between 1946 and 1994, a 54% two-party vote for the Democrats would have produced a 258 D to 177 R Congress. This would have been a 55 seat pick-up for the Democrats. So - the structural protections that the GOP had in place mitigated what would otherwise have been a truly disastrous result. It saved the Republicans about 26 seats.

What do I mean by "structural protections?" The following. Between 1946 and 1994, the House exhibited what might be called a modest "Democratic bias" in the translation of votes into seats. That is, the Democrats always had a higher percentage of seats than they did votes. Since 1996, there has been a slight Republican bias to the House. Accordingly, before 1994, a 54% share of the 2-party vote would give the Democrats, on average, 60% of all seats. This would be a 6% over-performance. This year, a 54% share of the 2-party vote has yielded them 53% of all seats. A 1% "under-performance," and therefore a 7% differential. The Democrats have under-performed in votes-to-seats since 1996 by anywhere between 0.1% and 2.5%.

This relates to a point I have made several times this year. Something has happened in the last 12 years to mute Democratic gains. While this "something" did not have quite the effect I thought it would (I thought it would save the GOP about 10 more seats than it actually did) - it was still very effective on Election Day.

I would point to three items that might explain this phenomenon - three potential "structural features" that might be at play in our electoral system.

A. The South

I think the most important reason that the GOP has gone from under-performing in House votes/seats to over-performing has to do with the fact that - whereas once the South offered the Democrats a solid floor of seats, it now offers the GOP a solid floor of seats. It used to be that, thanks to the South, the Democrats were more competitive nationwide than Republicans. Democrats could compete in most places, while Republicanism in the South was, as the Monty Python boys used to say, "Right out!" Now, it appears as though large portions of the South are non-competitive for the Democrats. This, in turn, has given the Republicans a floor of secure seats that they did not once have.

Of course, the floor is not as high for the GOP as it was for the Democrats. There are still about 20 "blue dog" Democrats in the South; there are also many minority-majority districts as well as genuinely competitive swing congressional districts here and there (mostly in Florida). So today, the GOP controls 59% of seats in the House, whereas the Democrats used to control 100% of seats. But, while this floor is lower - it seems to be solid. GOP seats were largely unaffected in the South this year. By my count, the Republicans lost 11 in the Northeast, 10 seats in the Midwest, 4 in the West and only 4 in the South. The West's numbers are not terribly impressive for the GOP, considering that 20 of the Republican seats in the West were in gerrymandered-into-stasis California, which only saw Abramoff-tainted Richard Pombo go down.

The South's numbers are noteworthy. The GOP still carried states like Georgia, Florida and Texas by comfortable margins despite losing the total popular vote by several million votes (point of comparison, the Democrats winning 54-46 in the two-party vote is a little better than Clinton's victory over Dole in 1996). While their vote margins were reduced in the South, they were still strong enough to mute losses. In point of fact, two of the GOP's southern seats that were lost (TX 22 and FL 16) were probably lost because the GOP candidate was not even on the ballot, one of the seats (FL 22) is an urban district Ft. Lauderdale district, the final seat (NC 11) featured one of the least liked and poorest-performing representatives, and the GOP almost picked up 2 seats (GA 08 and GA 12). While the party was devastated in the rest of the nation, the South held fairly firm for them. Again, the South is not as solidly Republican as it was solidly Democratic - but, it is still very solid.

B. Safe Districts

There are very few genuine swing congressional districts. A major reason is that most districts have come to align their votes for Congress and their votes for President. The net benefit this year seems to have been on the Republican side.

This year, 18 of the 29 Democratic victories came in districts that either (a) gave more votes to Kerry than to Bush, or (b) featured notably weak Republican candidates. The message here is that most Republican incumbents represent districts that are themselves Republican - sufficiently Republican to protect the incumbent despite a significant shift in the public mood. Early data indicates that independent voters who moved from the Republicans to the Democrats - and that this movement was nationwide. However, most Republican incumbents could suffer the loss of some independents - thanks to a strong base of right-leaning independents and Republicans. They were, simply stated, insulated from the swing. Their two-party vote totals declined, but nevertheless stayed above 50%.

The Democrats do not really share this advantage - at least to this extent. Going into 2006, there were 41 Democrat-held districts that Bush carried (compared with just 17 Republican-held districts that Kerry carried). Democrats are thus more susceptible to anti-Democratic moods because many Democratic representatives have relatively few strong Democratic voters upon whom they can rely. Thus, a 5.4% pro-Republican swing in the two-party vote would probably have pushed out many more Democrats than this year's 5.4% anti-Republican swing pushed out Republicans.

Of course, a given safe Democratic district tends to be much safer for Democrats than a given safe Republican district is for Republicans. Why did Bush win 58.6% of all congressional districts in 2004 but only 50.7% of all votes? It is because the districts he lost went more strongly for Kerry than the districts he won went for him. In Kerry's worst congressional district, he still pulled in about 25% of the vote. In Bush's worst congressional district, he pulled in about 10% of the vote. Accordingly, the GOP would probably do worse with a massive anti-GOP shift (say, on the order of 8%) than the Democrats would with a massive anti-Democratic shift.

C. Safe Incumbents

The Republicans, I think it is fair to say, would have lost many more seats than they did if they had suffered more retirements. The GOP lost all but one of its open seats that went for Bush by 55% or less. Thus, any more Republican retirements in slightly conservative districts would probably have resulted in more Democratic pickups. This and (B) were the two reasons that I was so bullish about a GOP retention of the House in the Spring. At the time, the Democrats only had a net advantage in open seat contests of about 8 seats, and more than 90% of the GOP caucus was running for reelection. My thinking was that these incumbents, most of whom were in Bush-leaning districts, would be able to survive the coming anti-Bush vote.

Unfortunately for Republicans - the advantage they had in April by having relatively few Republicans retire was diminished by the ever-increasing number of disadvantaged Republican candidates. By my count, poor candidate quality came to be a contributing factor in 8 of their 29 loses. And when I say, "poor candidate quality," I mean either (a) their candidate was not on the ballot, or (b) serious ethical questions had been raised at some point about the Republican. Furthermore, I would add to this list an additional 9 could-have-been-quality Republican incumbents who did not do everything they could have done to secure reelection (although, when this number of votes shift from one party to another, it is not untypical for some of the losing party's incumbents to be caught unawares and to be defeated unexpectedly). Factor in the 5 other open seats the GOP lost, and there were only about 7 Republican incumbents who lost despite their best efforts.

The lesson to be learned here is that, by and large, incumbents are in charge of their own destiny. If they choose to do everything they can to get elected, which includes first and foremost not making ethically questionable choices in the recent past, and which then includes running a modern campaign that is fully funded and that is duly cautious about the potential threat - the chances are very high that they will be reelected. Not 100%, of course. But very high nevertheless.

All in all, the incumbency advantage probably kept things from being worse for the Republicans. Like I said, any open seats that went anything less than 55% for Bush would probably have been Democrat pickups. Tom Reynolds and the Republican leadership should be credited with minimizing GOP retirements. They should also be credited with making sure that the GOP caucus knew well and good that this was going to be a rotten year for the party, and that everybody was a potential target. The fact that only about 9 Republican incumbents lost because they underestimated their vulnerability is a credit to the party leadership.

What exactly did the incumbency advantage do for the GOP? Whereas (A) and (B) had a pro-GOP effect by distributing votes in a pro-Republican way, so that the 54-46 vote spread kept seat swings to a minimum, the incumbency advantage probably kept the vote spread itself to a minimum. Think of all the pissed off voters who nevertheless voted for their incumbent Republican because he or she spent 6 months and $3 million dollars to convince them to do so. They "count" as Republicans, but in another age, when incumbents were not so informed about voter feelings and so aggressive about changing them - would they not have counted as Democrats? Could Cubin, Wilson, Hayes, Porter, Ferguson, Reichert, Walsh, Schmidt, Pryce, Gerlach, Drake, Doolittle, Shays, Reynolds, Kuhl, and about a dozen other Republicans have held on 50 years ago? I do not think so. 50 years ago, I think most of them would have gone down because they would have lacked sufficient capacity to separate themselves from the national mood. In other words, the GOP's structural barriers did not just diminish the effect of votes upon seats; they directly influenced votes.

Conclusion: What is the Correlative Year?

Republicans should thus count themselves very lucky. With this kind of vote share prior to 1994, the Democrats would have an 81-member majority, as opposed to the 29-member majority they now enjoy. The South, the alignment of the electorate, and the incumbency advantage all probably combined to create a result in which the Democrats - rather than over-perform in their vote-to-seat ratio - under-performed by a slight amount.

I am not entirely sure that, as some have argued, this is an "average" midterm result. From the standpoint of the statistical average, it most certainly is. The Democrats picked up about the average number of seats in the post-WWII era that a sixth-year "itch" election dispenses. However, how much does this average mean? After all, the number of seats you pick up is highly conditioned by how many seats you have to begin with. For instance, if you only have 20 seats, and there is a 2% swing in your favor, you are going to pick up more seats than you would if you were starting with 415 seats.

Accordingly, the average 6th year seat swing will have a high standard deviation, which means that it does not carry as much information as it could. There were big swings to the out-party in 1946, 1958, 1966, 1974. Small (to negative) swings in 1950, 1986 and 1998. The average is 30-ish, but how much does that tell us?

What is more, the average does not mean much when relevant variables have shifted during your observation period. If the Steelers go 1 for 8 on the road and 6 for 8 at home - their winning percentage will be .4375. But how much information does this winning percentage, or average, tell us? After all, the likelihood of a victory was greatly conditioned by where the game was played. In other words, the statistical average, when you are not careful, can induce you to obscure the true state of the world.

Regarding the argument I am offering here, I mean the following: your expected seat gain from a given vote gain depends upon how many seats you are starting with, how many votes you are starting with, and the electoral structures that condition the translation of votes into seats. And, historically speaking, the Democrats were on the low end of their average coming into this election in terms of seats and votes. They were also in an era in which they tend to under-perform by a slight amount, rather than over-perform by a modest amount. With a shift in the two-party vote from 48.6% to 54.0% and a base of 203 seats, not controlling for the post-1994 environment, we "should" have seen about 26 more seats flip to the Democrats. In other words - were it not for the unique structural protections the GOP has in place, this year's results would have been well above the average sixth year result. It would have been more along the lines of the more dramatic sixth-year post-war midterms: 1946, 1958, and 1974.

Not only that, but the GOP's advantages probably kept the vote spread to a minimum. A 5.6% shift 30 years ago might very well have been a 6.6% shift. Did those extra millions of dollars that Republicans spent on advertising and voter mobilization shift 1% of the electorate? My intuition is that they did - and thus, the Democratic share of the vote would have been higher if the same election was held 30 years ago. Simply stated - after 12 years of only, on average, 1% variation in the two-party vote, the political significance of one party losing 5.4% of it should not be understated.

All in all, to which election should we compare this one? I think that what we saw this year was similar to what occurred at the midpoint of Roosevelt's fourth term, which is one of the few times in the post-war era that the wrath of the public has been acutely felt by the party in power. The size of the 1946 vote swing and seat swing - controlling for the post-1994 era - correlates well with this election. The tenor of that election also correlates well. 1946 was a year when the public came to reject the course the government had charted, and responded by shifting the balance of power. In both 2006 and 1946, the public was exasperated with the course of the government and the results it had produced. They voted for change.

Republicans, accordingly, should take note. While the seat swing, in itself, seems average - the seat swing was diminished by structural factors that actually have nothing to do with voter sentiment. A proper historical comparison indicates that we just witnessed a very dramatic election - an exasperation/repudiation election. In 1946, the public swept the Republicans into office in reaction to the maintenance of wartime rationing and the increasing prevalence of union-induced work stoppages. In that election, the public rejected the current state of national affairs and the governance of the Truman Administration. So they have in this election rejected the state of affairs and the governance of the Bush Administration.

This does not mean that the public has necessarily endorsed Democratic solutions to our problems. The Republicans, of course, were promptly tossed out of power in 1948 for having moved too far to the right on domestic issues; the public was angry at union work stoppages, but this did not mean that they wanted Taft-Hartley. The nature of our democratic system does not allow for much sophistication in the message of the public. They voted for change, but what type of change? The results do not - cannot - tell us. Votes are really nothing more than sets of "1's" and "0's" registered by every voter. We must be careful not to draw too many inferences from them. Democrats, similarly, must be careful not to read their own policy preferences into the public mandate. The 33rd President once remarked that, "The luckiest thing that ever happened to me was the 80th Congress." Democrats should take care that a version of this sentence does not find its way into the 43rd President's memoirs, or else Harry Reid will go the way of Wallace White.

Nevertheless, it is clear that exasperation is what the public feels and that change is what it wants. Today, there are enough "0s" where there used to be "1's" to indicate as much. Like 1946, this year's vote was a vote of exasperation with and a rejection of the status quo as set by the President and his party.

© 2000-2006 RealClearPolitics.com All Rights Reserved


Email Friend | Print | RSS | Add to Del.icio.us | Add to Digg
Sponsored Links

Jay Cost
Author Archive