OCT 29 UPDATE! A Republican tsunami! My official election prediction for the House of Representatives.

This prediction has been replaced/updated with this one.

In my first analysis of the election, I forecast a Republican gain of 61 seats in the House.

In my second, I created three more sophisticated models and forecasted a GOP gain of 78 House seats. Each time a new Generic Congressional Poll was released, I updated my forecast in the comment section. I will do so again here.

The following is mostly the same text from Sunday’s post with updated polling data.

First, one must predict the vote totals for each party. Currently, RCP’s average of the “Generic Congressional Vote” shows Republicans winning 48.7% to 42.4%, a decline from Sunday’s 49.3%-41.6% margin. I then remove the polls with highest and lowest spread to eliminate outliers (in this case, I am removing one poll showing Republicans up by 14 and one that shows Democrats up by 3). Excluding those two gives us a much smaller range of +3 to +13, though this is much wider than Sunday’s range of +7 to +11, a small ver 4 point range. Removing the highest and lower outlier, Republicans lead Democrats 49.0% to 41.7%, a narrower spread than last week’s 49.8% to 40.6%. Eliminating the undecideds (if they have not decided by now, they are unlikely to vote) gives a two-party vote total of 54.0% for the Republicans (down from 55.1%) and 46.0% for the Democrats (up from 44.9%).

I now have three models to convert vote totals to House seats.

40-year model: Regression of House seats vs. vote total for every election since 1968.

8-year model: Regression of House seats vs. vote total for every election since 2002. Because of the increase in partisanship and computerized gerrymandering, there are now many more safe seats.

1994 & 2006 model: In these two mid-term elections, control of the House switched sides against an unpopular President. The same will likely occur this year.

The models produce the following results:

40-year model: Republicans win 257 House seats, gain of 79 seats.

8-year model: Republicans win 241 House seats, gain of 63 seats.

1994 & 2006 model: Republicans win 246 House seats, gain of 68 seats.

Taking a simple average of the three, I now predict the Republicans will win 248 House seats, a gain of 70 seats. On Sunday, I had predicted a gain of 78 seats. So the Generic Congressional Polls moved against the Republicans this week, but only barely.

However, the last poll of the week moved significantly in the GOP’s favor. Individual races, which had been moving against Republicans last week and earlier in this week, have also moved in the GOP’s favor toward the end of the week. I suspect that most of the movement we are seeing is statistical noise and not a change in voters’ opinions. So let’s take an average of my averages over the past week. Doing so, I now predict the GOP will gain 72 seats in the US House of Representatives.

UPDATED! A Republican tsunami? My official election prediction for the House of Representatives.

This prediction has been replaced/updated with this one.

I have added new models to my previous prediction of a Republican gain of 61 seats in the House.

First, one must predict the vote totals for each party. Currently, RCP’s average of the “Generic Congressional Vote” shows Republicans winning 49.3% to 41.6%. However, you clearly see two outliers, one to the upside (Gallup LV Lower Turnout) and one to the downside (Newsweek). Excluding those two gives you a pretty tight pack varying from +7 to +11, a small 4 point range versus the huge 20 point range if you include the two extreme polls. Based on these five closely-packed polls, Republicans lead Democrats 49.8% to 40.6%. Eliminating the undecideds (if they have not decided by now, they are unlikely to vote) gives a two-party vote total of 55.1% for the Republicans and 44.9% for the Democrats.

I now have three models to convert vote totals to House seats.

40-year model: Regression of House seats vs. vote total for every election since 1968.

8-year model: Regression of House seats vs. vote total for every election since 2002. Because of the increase in partisanship and computerized gerrymandering, there are now many more safe seats.

1994 & 2006 model: In these two mid-term elections, control of the House switched sides against an unpopular President. The same will likely occur this year.

The models produce the following results:

40-year model: Republicans win 268 House seats, gain of 90 seats.

8-year model: Republicans win 247 House seats, gain of 69 seats.

1994 & 2006 model: Republicans win 253 House seats, gain of 75 seats.

Taking a simple average of the three, I now predict the Republicans will win 256 House seats, a gain of 78 seats.

* Again, I will update these numbers as new polls come in. But if you look at the polls on RCP over the last month, Republicans have been consistently in the lead by 8 or 9 points. Barring some late breaking change in this election, I don’t expect these number to change much. But who know how accurate the polling is this year? We won’t know for certain until November 2.

A Republican tsunami? My official election prediction for the House of Representatives.

This prediction has been replaced/updated with this one.

Many, if not most, political analysts are talking about the Republican tsunami due on November 2. According to current projections at FiveThirtyEight, Republicans are expected to capture 7 Senate seats, 50 House seats, 6 Governorships. Analysts are calling this the biggest year since 1994. Michael Barone says it may be even bigger than that:

For months, people have been asking me if this year looks like ’94. My response is that the poll numbers suggest it looks like 1994, when Republicans gained 52 seats in a House of 435 seats. Or perhaps somewhat better for Republicans and worse for Democrats. The Gallup high turnout and low turnout numbers suggest it looks like 1894, when Republicans gained more than 100 seats in a House of approximately 350 seats.

Maybe the results will be closer to Michael Barone’s high-end prediction than RCP and FiveThirtyEight. To betray my political affiliation, as if you didn’t already know, I hope it is. But currently, the polling suggests it won’t be. FiveThirtyEight’s statistical analysis has the Republicans winning 228.4 House seats. RCP hs the Republicans at 235.5 if we give half the Toss Ups to the Republicans. So based on current projections, the Republicans will win 52.5% or 54.1% of all House seats. Currently, the Democrats have 255 House seats or 58.6% of them. For Republicans to match the Democrats’ current position, Republicans would have to win 77 House seat. (Remember, all House seats are up for grabs each two-year election.)

This certainly puts the “tsunami” into perspective. While this election may be called a tsunami based on the number of seats changing hands, I would not call it an outright tsunami when Republicans are not likely to win as many seats in 2010 as Democrats did in 2008.

This tsunami vs. non-tsunami picture can be seen as either good news or bad news for Republicans (and inversely for Democrats).

From the one perspective, it would be very disappointing that the Republicans can only manage 52.5% (FiveThirtyEight) or 54.1% (RCP) of the House seats in a tsunami year. How can the Republicans “fix” Washington and the country when they cannot win the large majorities that the Democrats have been able to?

From another perspective, this may mean that the polls, RCP, and FiveThirtyEight are way off and that some predictions of even larger Republican gains will come true. The generic Congressional polls currently gives the Republicans a 7.2% advantage. However, if you eliminate the three oldest polls in the average, Republicans have a 9.8% advantage. That would imply a Republican vote total of 54.9%. In 2006, the Democrats won 53.6% of the total vote and captured 53.6% of the House seats in an election that switched party control. So, even with the benefits of incumbency, each party should receive the number of House seats equal to its vote total. Therefore, based on the generic Congressional polls, Republicans should win 54.9% of the House seats, giving them 239 seats, a 61 seat gain.

Let’s look at an alternative scenario. What is Gallup’s most Republican-favorable margin of victory plays out? If Republicans do indeed win by seventeen point, they would win 58.5% of the total vote and about the same number of House seat. That would results in 254 Republican seats, a gain of 76 seats. That would be huge, a tsunami as they are saying, but nowhere near the 100 seats Michael Barone said is possible.

Now what if the low end model plays out; the one where Republicans only win by five points? The Republicans would then capture about 228 seat, a gain of 50 seats.

So here is my official prediction for the House of Representatives:

Republicans gain 61 seats and finish with 239 seat, about 55% of the total. This is more than the 50-seat gain predicted at FiveThirtyEight and the 58-seat gain predicted by RCP, but less than some others are talking about. On the low end, I would expect Republicans to win 50 seats. So my low end projection is equal to FiveThirtyEight’s middle-of-the-road projection. On the high end, I think it unlikely that Republicans win more than 76 seats*, much less than some are saying is possible. (I will update this prediction in the comments section as new polls come in.)

* All that said, this election is probably the most difficult to model in years, possibly ever, and the polls and models they are based on are less accurate than most years. The gap between Gallup’s registered voter and likely voter (lower turnout) models is a huge 12 points. Nobody know which model, if any, is the best and therefore the range of possible results this election is huge. But lacking anything better than the current polls and models, I used those.