News from the Votemaster

Survey Results

Yesterday we had a little survey. If you missed it, click on "Previous report" above and do it. There were 105,034 responses (out of 990,000 visitors yesterday, which is an enormous percentage). Most readers are Democrats, despite a serious attempt to attract people from all over the spectrum including thousands of dollars spent on ads on numerous conservative Websites (as well as progressive Websites). That didn't bring in so many new conservative visitors as hoped for in part because there is nothing in the conservative blogosphere with the volume of Daily Kos or TPM. Below are the raw numbers. Crosstabs will follow in a few days. If you want to see the data for the 2004 survey, it is here.

Some Senators Are Calling for Stevens to Resign

A small but growing group of senators are making one of those tough calls politicians love to talk about--they are saying the Sen. Ted Stevens (R-AK) should resign. The group includes John McCain and Barack Obama, who don't want to appear soft on felons, but also includes Sen. Gordon Smith (R-OR), Sen. Norm Coleman (R-MN), and Sen. John Sununu (R-NH). Perhaps they are unaware that there is an election in a week and if the people of Alaska want to get rid of Stevens they can handle the job themselves. Why are they calling for his head now? For Democrats, it makes them look good and emphasizes that Republicans are corrupt. For Republicans, they want to put as much distance as they can between themselves and Stevens. Smith, Coleman, and Sununu are all in difficult reelection battles and all appear to be losing. All of them fully understand the consequences of appearing to condone the actions of a corrupt senator. Also, should Stevens be reelected and the Democrats move to expel him from the Senate in January, Republican senators would have to vote on the expulsion motion, something few of them particularly want to do as a vote either way will offend many Republicans. For them, the best-case scenario is that Stevens wins reelection, and then resigns, forcing a special election to fill the seat. This election would probably pit Sarah Palin against Mark Begich, something Palin has a decent chance of winning. Of course, if Begich wins on Tuesday, it doesn't make a huge difference if Stevens leaves the Senate now or on January 5, other than a couple of votes if the Senate is called back in session after the election.

McCain's Pollster Foresees Tight Race

John McCain's pollster, Bill McInturff, foresees a close election. Whether he really has data supporting this or this is just brave talk remains to be seen. But elections often get closer in the last week as the undecideds finally get off the fence.

Weekly Reader Calls It for Obama

In contrast to McInturff, the Weekly Reader's quadrennial poll of students from kindergarten to 12th grade predicts that Obama will win big time, with 55% of the vote to McCain's 43%. This survey has been surprisingly accurate in the past, getting 12 of the past 13 presidential elections right, missing only Bill Clinton's win in a 3-way race in 1992. The survey's accuracy may be due to children getting most of their political views from their parents and the children's views may more accurately reflect what their parents are really thinking than what the parents are telling the pollsters. Thanks to David Richardson for the pointer.

Comparison of 2004 and 2008

It is instructive to compare the state of polling in 2004 as of the Tuesday before the election with state of polling yesterday. The table below shows the the Kerry - Bush score on Oct. 26, 2004 in column 2. Hawaii is an outlier because there was one freaky poll, but, on Oct. 26, 2004, the map showed Kerry ahead of Bush by 9 points in California, by 8 points in Illinois, and behind Bush by 8 points in New Mexico. The third column is the Kerry - Bush election result. For example, the prediction of a 9-point lead in California and an 8-point lead in Illinois were pretty good as Kerry won those states by 10.3 and 10.0 points respectively. The mean of the second column was a Bush lead of 4.4 points whereas the final results was a Bush lead of 6.0 points. These averages are not weighted by population, so they are not predictors of the popular vote, but they do give an idea of how good the 51 polls were a week before the election.

Now consider column 4. It shows the difference between Obama and McCain a week before the 2008 election (taken from yesterday's spreadsheet). As of yesterday, Obama was leading McCain in Hawaii by 41 points, was tied in Montana and North Dakota, and leading in California by 27 points, and so on.

Finally we come to column 5, which is column 4 minus column 2, in other words, how much better the Democrat is doing in 2008 compared to 2004. Ignoring Hawaii on account of one outlier in 2004, we see Obama is doing 21 points better than Kerry was in Montana because yesterday's map showed Obama and McCain tied whereas the Oct. 26, 2004 map showed Kerry 21 points behind Bush. Thus Obama is outperforming Kerry in Montana by 21 points. He is outperforming Kerry in Indiana and Georgia by 16 points and 14 points, respectively. On the other hand, he is underperforming Kerry in the lines colored red. For example, he is underperforming Kerry in Arkansas by 11 points and in Alabama by 8 points. Nevertheless, he is doing better than Kerry was in 41 states plus D.C. and underperforming Kerry in 9 states, only one of which he is likely to win (New Hampshire).

In other words, an alternative way to look at the election is how well is Obama doing compared to Kerry and the clear answer is: "a lot better." Also interesting is to see how many rows have different signs in columns 2 and 3. For example, the two entries for Wisconsin are -2.0 and 0.4 meaning the polls a week out showed Kerry losing Wisconsin by 2 points but he ultimately won the state by 0.4 points. There are four states where the signs don't agree: Hawaii, Wisconsin, Colorado, and Arkansas. As mentioned above, there was one weird poll that completely messed up Hawaii in 2004 and the pollsters were saying Arkansas was too close to call (when it wasn't at all). But in 46 states plus D.C. the candidate leading in the polls a week before the election did, in fact, carry the state.

If you want to play with the numbers yourself, an expanded version of the table above is available in Excel format and in .csv format.

Today's Polls

We have 31 presidential polls today. Here are some of the highlights. A poll in Arizona makes it a statistical tie, with McCain leading by a mere 2 points, 46% to 44% in his home state. Colorado looks like a done deal for Obama, as two polls have him ahead by 8 points and 9 points, respectively. Obama also leads in Florida by about 2 points, but this lead has been consistent for a while now. Georgia is surprisingly close, with McCain ahead by only 1 point. North Carolina is still essentially tied. Nevada, in contrast, is looking more and more like an Obama win as one poll today has him 4 points ahead and another has him 10 points ahead. Likewise, Obama's lead in Ohio looks serious, with four polls today putting him ahead, by margins of 4, 7, 9, and 9 points, respectively. As is well known, no Republican in history has ever won the presidency without Ohio. Finally, despite McCain's last-minute attempt to win Pennsylvania, six new polls there show Obama with an insurmountable lead there, with leads of 7, 9, 12, 12, 12, and 13 points, respectively.

We also have two Senate polls. Mary Landrieu is cruising to an easy reelection over Democrat-turned-Republican John Kennedy in Louisiana, 53% to 34%. Karl Rove was the one who convinced Kennedy to switch parties and challenge Landrieu. The plan appears to have failed. In Mississippi-B, Sen. Roger Wicker (R-MS) has taken a strong lead over Ronnie Musgrove (D), 54% to 43%.

We also have four House polls, excluding a new tracking poll in New Hampshire. We will count that one only when the data samples are independent (e.g., after the Oct. 24-26 poll, the next one that will count is Oct. 27-29).