JS

Let us skip the discussion if debates really matter for a moment and focus instead on who won the debate.

The first point is that, just as regular people, columnists saw the debate through their partisan eyes. In general, Republicans found McCain to be brilliant and Democrats agreed that Obama won the debate handily. Below are a few news headlines [from Real Clear Politics]:

McCain Wins Substantive Debate - David Yepsen, Des Moines Register
Obama Wins Debate On Tactics and Strategies - Joe Klein, Time
The Mac is Back - Roger Simon, The Politco
Tie Goes to Obama - John Dickerson, Slate
McCain Won, But That’s Not Enough - Fred Barnes Weekly Standard

The second and probably more interesting point is that Obama did better among Republicans than McCain did among Democrats.  Polling conducted in California by SurveyUSA shows that of Republicans, 56 percent thought that McCain won, 25 percent said Obama and 19 percent considered it a tie. As for Democrats, 72 percent said that Obama won, 13 percent thought McCain got it and 15 said that it was a tie. In other words, Obama did better than McCain because 1 in 4 of the Republicans answered that Obama won whereas 1 in 8 of the Democrats thought that McCain won.

Finally, also the group of uncommitted voters saw Obama as the winner. CBS asked a nationally representative sample “Who won the debate?” and 39 percent of them said that Obama won while 24 percent said McCain. 37 percent concluded that it was a tie.

written by Jacob \\ tags: McCain, Obama, politics, polls

This post about how polling organizations influence results is interesting. Great work by Charles H. Franklin on house effects.

written by Jacob \\ tags: polls

NASA recently released pictures on a vehicle built to maneuver on the moon. The vehicle is part of an effort to build a permanent base there, and eventually on Mars. From the NASA press release:

In a car commercial, it would sound odd: active suspension, six-wheel drive with independent steering for each wheel, no doors, no windows, no seats and the only color it comes in is gold.

Although NASA is enthusiastic, the American public is less so. A majority does not think that permanent settlements in outer space is a good idea (ABC News Poll, Jan 2008).

My guess is that people have not given this much thought. But even if it is just a “doorstep opinion” it seems as the government needs to be more persuasive on space exploration if it wants to go forward with the program. At this moment there is no public discussion on the issue.

written by Jacob \\ tags: NASA, polls

It is likely an understatement to say that Americans in general are politically unsophisticated. After all, only half the population can place South America on a map (Delli Carpini and Keeter 1996) and less than a third of the public understands political ideology (Converse 1964).

Consequently, any time there is talk of strategic political behavior I raise a skeptical eyebrow.

The last example of alleged strategic voting, and this is pretty amazing if it is true, is that the talk show host Rush Limbaugh has been able to persuade fellow Republicans to go out and vote in the Democratic primaries. Limbaugh wants the Democratic “soap opera” to continue because he believes that it will benefit McCain. Some primaries are open to all partisans so it is possible to vote in the opposing party’s primary in certain states.

In the Mississippi Democratic primary 12 percent of the voters identified as Republicans. 12 percent is a high enough number to suggest that people are actually following Limbaugh’s advice. After all, people who vote in primaries are more politically sophisticated than most citizens. At first blush, the data also supports this idea. The data is from the exit poll in the Mississippi Democratic primary. The sample includes 147 self-identified Republicans who voted for Clinton. From Mystery Pollster’s column:

• 85 percent rated John McCain favorably, and 58 percent had a “strongly” favorable opinion of him
• 41 percent said they would be dissatisfied if Clinton were the Democratic nominee.
• 56 percent said Clinton has not “offered clear and detailed plans to solve the country’s problems.”
• 62 percent said Clinton does not inspire them “about the future of the country.”
• 72 percent said Clinton is not “honest and trustworthy.”

The important point, however, that turns the strategic voting part on its head is that these Republicans dislike - hate is probably a more accurate term - Obama so much that they simply think Hillary Clinton is a substantially lesser evil. Again, from Mystery Pollster Mark Blumenthal these are the Republicans from the exit poll:

• 91 percent said Clinton is more qualified to be commander in chief; only 3 percent said Obama is more qualified.
• 94 percent said Obama does not inspire them “about the future of the country.”
• 89 percent would be dissatisfied if Obama were the Democratic nominee.
• 86 percent said Obama is not “honest and trustworthy.”
• 86 percent said Obama has not “offered clear and detailed plans to solve the country’s problems.”
• 82 percent said Clinton should not pick Obama to be her running mate if she is the nominee.

It therefore seems reasonable to conclude, as Blumenthal does, that the Limbaugh effect has been overstated. Republicans are driven more by emotions and less by strategic voting.

written by Jacob \\ tags: Hillary Clinton, Limbaugh, Obama, politics, polls, primaries

Polls indicate that Hillary Clinton and her supporters should be fairly optimistic in today’s primaries. In Ohio she is leading by around 6 percentage points according an average based on 8 polls. An average based on three polls in the State of Rhode Island and Providence Plantations gives Hillary Clinton an advantage by nearly 10 percentage points.

However, it is extremely difficult for pollsters to get their predictions right because of the uncertainty with who actually turns out to vote. If the polls do not reflect the groups that vote, it is likely that they will be wrong. I have written previously about for example young people and blacks turning out more than usually, and they tend to prefer Obama.

It seems that pollsters are busy tweaking their samples because otherwise it is difficult to understand why the predictions should differ that much. In Ohio, there is a range of 14 percentage points between the predictions. One poll gives Obama a lead by 2 percentage points. At the other end of the spectrum is a poll that gives Hillary Clinton a lead by 12 percentage points.

This huge spread of predictions is also visible at the national level. Predictions range from a lead for Clinton by 1 percentage point (Rasmussen Tracking) to an Obama lead with 16 percentage points (CBS/NY Times). That is a range of 17 percentage points!

written by Jacob \\ tags: politics, polls, primaries

An optimistic sign for Democrats is the huge turnout in primaries and caucuses. If this pattern can be sustained, the party will have a better chance in the general election.

In media we often hear of three groups that are turning out in larger numbers than normally (if 2004 was normal): women, blacks and young people. Tom Holbrook has looked closer at exit polls from the Democratic primaries and caucuses that have been held so far. The figure below reflects exit poll data for all states combined. The figure shows an increase of 4 percentage points for people between 18 and 29 compared to 2004 in the Democratic primaries and caucuses. At Holbrook’s blog, you can see this by state as well. I found it after reading the Monkey Cage.

As I have discussed earlier, one problem for pollsters is to predict who will turn out. They usually adjust the sample to reflect the population that they think will vote. The data from Holbrook indicates that this might be even more difficult this time. That is, if these groups continue to turn out in large numbers. The increased turnout among blacks, women and young people is relatively small, so it might not influence the results tremendously, but it could prove important if the race is close.

written by Jacob \\ tags: politics, polls, primaries

People in some countries really dislike the U.S.

But that does not mean that people dislike Americans. For instance, the poll shows that while only 30 percent of the German population has a favorable view U.S., about 2/3 of the German people like Americans. There is a big difference there. My guess is that people think of George W. Bush when they hear the U.S. but that is not necessarily the first thing that comes to mind when asked about Americans.

And how can anyone dislike a people that create these Old Spice commercials?

With the hero from Evil Dead, Bruce Campbell

[youtube]http://www.youtube.com/watch?v=Af1OxkFOK18[/youtube]

A new one

[youtube]http://www.youtube.com/watch?v=jPYWTkP9NDM[/youtube]

written by Jacob \\ tags: ads, polls, products, USA

After Super Tuesday it seems clear that Democrats are way more happy and energized by their candidates than Republican primary voters. Just as in New Hampshire, there were many more individuals that voted in the Democratic primaries. 73 percent of the votes yesterday were Democratic. This is a huge difference between the parties.

Some American politics 101: Party identification is usually the strongest determinant for vote choice in the general election; people who identify as Democrats vote for Democratic candidates and Republicans vote for Republicans. Crossing the party line happens but it is rare. About 2/3 of the American electorate identify as either Republicans or Democrats and it is almost equally split between the parties. This means that we know what the majority of the American electorate will vote.

Democrats are much more satisfied with their candidates and their primaries draw huge crowds. If this trend is sustained over the general election, it means that more Democrats will turn out than Republicans. Polls often correct/weigh their raw data by a series of factors. In plain terms this means that samples should represent the overall population on for example age and gender. Since party id is so important, some pollsters weigh their results based on this factor as well. For instance, Zogby and Rasmussen do it. Those who do not weigh their sample based on this factor are sometimes criticized for misleading results.

But what happens if many more Democrats than Republicans turn out in the general election? The consequence is obvious. The polls will be wrong if Democrats are down-weighted.

written by Jacob \\ tags: politics, polls

Let us look at how people in California are voting since it is such an important state. The results below are from an exit poll based on 1811 individuals. As always in the US, race is important, but this time the results are somewhat unexpected.

Several important factors stand out. First, most blacks vote for Obama. This should not come as a surprise since this trend has been visible in previous primaries. 81 percent of blacks vote for Obama whereas only 16 percent prefer Clinton. Second, the white vote is evenly divided among Obama and Clinton. Unlike what some thought, Clinton does not have an advantage among white voters. This is important for future elections. At least among white Californian Democrats, race does not seem to matter. Third, Latinos and Asians strongly prefer Clinton over Obama. It is somewhat surprising that 66% and 73% of Latinos and Asians prefer Clinton, respectively. The support of these two groups for Clinton seems critical.

As always with cross-tabulated data, there are other factors that are not controlled for that might drive these results. For instance, income and education are not controlled for. Moreover, while exit polls usually are quite accurate, sometimes they are flawed.

Update: There are more interviews in the current exit poll (see link above) and the percentages are slightly different. However, the general pattern remains the same. I am still waiting for a satisfying explanation to why most Asians vote for Clinton.

written by Jacob \\ tags: politics, polls, primaries, race

I teach an undergraduate class about polls and survey design. In one class I spent some time on arguably the most famous polling fiasco, the Literary Digest poll from 1936.

Polls had just started to become popular back then. The Literary Digest, a magazine, had been able to predict the winner in the presidential election from 1916 up until that point. They gathered their data by sending out postcards to readers and potential readers. The sample was made up of subscribers, car owners and people with telephones. Based on the postcards they got back, the Literary Digest stated who would win. In 1936 they sent out around 10 million postcards and got back more than 2 million - an enormous sample obviously.

One week before the election they presented their results. The Republican Alf Landon, a Kansas governor, would win over Roosevelt, the incumbent president, the Literary Digests predicted.

This prediction made some sense because the saying was that “as Maine goes, so goes the nation”. In other words, Maine was considered a bellwether state. Republican candidates had had success in Maine that year. The Literary Digest poll probably reflected this, people might have reasoned.

In contrast to the poll, Roosevelt won all states except Maine and Vermont; the election was the biggest victory in over 100 years. Based on this, someone later joked “as Maine goes, so goes Vermont”.

The reason for the polling fiasco was that the Literary Digest sample did not represent the nation. In the years following the Great Depression, the readers and potential readers that they had polled were mostly Republican. Their sample included too many affluent individuals, people who tended to vote Republican.

Besides emphasizing the importance of representative polling, the 1936 results also removed Maine’s status as a bellwether state. Hence, just because Romney won in Maine does not mean that he will win the Republican nomination.

Below is a cover from 1922.

written by Jacob \\ tags: politics, polls