Zarkov: Brown has 82 percent chance of winning
Suffolk University published a poll on January 14 showing that Brown leads Coakley in the Massachusetts Senate race. In a sample of 500 likely voters, 50 percent said they would vote for Brown versus 45.6 percent for Coakley with a margin of error (MOE) of 4.38 percent. Despite Brown’s lead of 4.4 percentage points, the pollsters seem to think they can’t predict who will win the election because the lead is close to the MOE. I disagree because in this Senate election we only care who gets the most votes, not how big the margin of victory will be.
Winning by a mere one percent is just as good as winning by (say) 5 percent. I’ve run the numbers, and my results say that Brown has an 82 percent chance of winning the election if Suffolk University really polled people who are extremely likely to vote in the election. To see why this is so, we need to understand a little about opinion polls and how they go wrong.
- end of initial entry -
Public opinion is often a moving target because people can change their attitudes over time. Every opinion poll is simply a snapshot at a point in time. However based on past data, we know attitudes have usually stabilized near the election date, so the January 14th poll should provide a good estimate of how Massachusetts Senate race will come out if the poll is otherwise accurate.
Polls suffer from two fundamentally different kinds of errors: sampling and bias. Sampling error depends on the size of the sample, and in theory we could make it go away completely by sampling everyone. That would be extremely wasteful and unnecessary because sampling error rapidly diminishes, and most opinion polls use only 500 or 1,000 people. That’s about all you need. Pollsters usually express sampling error by calculating the margin of error. They use standard formulas as given here, which involve various assumptions and approximations that are quite valid for American elections. In the Suffolk University poll, Brown got 50 percent with a MOE of 4.38 percent. This means the poll predicts that Brown should receive between 45.62 percent and 54.38 percent in the actual election (50 plus and minus 4.38). If the pollsters had asked 1,000 people instead of 500, the MOE would have shrunk to 3.09 percent. Increasing the sample size above 500 does not buy very much. The problem with the MOE is that it’s not useful in predicting the winner unless one candidate polls one or more MOEs above the other. Thus had Brown gotten 8 percentage points more than Coakley, Suffolk University would have declared him the likely winner.
Polls also have bias error, and this is much more difficult to determine than sampling error. Briefly bias error is the error you can’t make go away by increasing the sample size; it’s inherent in the design and execution of the survey. Pollsters usually ask a lot of other questions and use the answers to determine if a bias exists, and how they might adjust for it. Suffolk University did exactly that, and here is the full questionnaire along with a tabulation of the answers. Note the very important question: “How likely are you to vote in the Special Election for U.S. Senator?” Note also the tabulation—no one said that they were unlikely to vote! That’s a strong indication that the turnout will be big and all that implies. They also ask some questions to see if people are being truthful in their answers, such as “Can you tell me when the Special Election will be held?” If they don’t know that, are they likely to vote? The questionnaire can also reveal if the sample is representative of Massachusetts voters. For example 196 of those polled declared themselves registered Democrats. This should match voter registration data, which is available. It if didn’t there would be reason to suspect that this particular sample was faulty.
I assume that Suffolk University is a professional enough to have designed their poll well enough to keep the amount of bias error low enough so that sampling error is the dominant kind of error in this survey. I can then use the their results in my own mathematical model which uses Bayesian statistics to predict who will win the Special Senate election on January 19th. Here are the poll counts from here.
I then form three categories: (1) Brown, (2) Coakley, (3) Kennedy, plus undecided. Next I use the counts in each category to calibrate my model and run 10,000 simulations of the poll, and tally the number of times Brown’s count exceeds Coakley’s. This tells me the probability that Brown will win based on the polling data. In 10,000 simulations Brown’s count exceeded Coakley’s 8,166 times, from which I conclude that Brown will be the likely winner. It’s not too close to call.
First, your result is quite close to one of the mainstream polling organizations which puts his chances at 3 to 1, i.e. 75 percent.
Second, according to you, Coakley has just under a one in five chance. That means it’s altogehter possible that she will win. Which, given everything that’s been said recently and given all the polls favoring Brown, would probably come as a major surprise if it happened. She would be come-from-behind winner.
Jake Jacobsen writes:
I have been making an effort to train myself to not care about this sort of thing, it is instinctive and yet his win will empower and enthuse the supporters of the Party who has been trying to destroy this nation 25% less than the other guys for the last eight years.
What are we getting excited about exactly?
Hmm, could it that we are excited about the prospect of stopping Obamacare and Obama’s other radical measures?
I think your question shows a loss of perspective.
Jake Jacobsen replies to LA:
Let me be clear what I’m saying here, while it’s a fine thing that Obama’s kill America at all costs plans will be thwarted, it is at the cost of empowering the very same party that we have been fighting an existential battle with for the last several years.
How is that a loss of perspective exactly? I think I have gained a healthy perspective.
I’ll grant that you have a perspective. But it’s a perspective that requires you downplay and even dismiss the importance of stopping an imminent, deadly peril, because stopping that imminent, deadly peril does not also stop other deadly but much less imminent perils.
As I see it, this reflects a chronic flaw in paleocon-type thinking going all the way back at least to the America First movement. Because fighting Hitler would lead to an empowerment of the federal government and an advance of liberalism, the Rightists opposed fighting Hitler—and, today, in the person of Patrick Buchanan, they still retroactively oppose fighting him. They didn’t then, and don’t today, grasp the evil of Hitler controlling half the world as compared with the evil of an enlargement of the U.S. federal government. That’s what I call a tragic, even deadly, loss of perspective.
Similarly, because a successful fight against Obamacare may empower the Republicans, you said (in your earlier comment): “I have been making an effort to train myself not to care about this sort of thing” (i.e., to care about defeating Democrats and stopping Obamacare).
That’s a perspective, all right. But it’s a perspective that fails to grasp different degrees of evils and misses the bigger picture.
Re Brown’s 82 percent chance of winning: don’t underestimate the ability of the
Democrats to steal votes and cheat.
James P. writes:
My fear is that even if Brown wins, the vote will be close enough to demand a recount, allowing the Democrats to steal it just as they did in the Coleman-Franken race.
“don’t underestimate the ability of the Democrats to steal votes and cheat.”
In the previous governor’s race in Washington (2004), Republican Dino Rossi won, but the margin was so close that the Secretary of State, by law, ordered a recount. On the recount, Rossi won, but by an even narrower margin. However, it was still close, and a party can ask for a recount, paying the cost. The Democrats asked for a hand recount. This time, the King County election office managed to scrounge up enough new votes for a win for Gregoire. Read this and you’ll see how razor-thin the margin of victory was: 129 votes. A small margin of victory opens the doors for such skulduggery.
The election was held on November 2, 2004, with the initial count showing Gregoire trailing Rossi by 261 votes. However, a legally mandated machine recount reduced that lead to only 42 votes, then a hand count that was requested and funded by the state’s Democratic Party gave Gregoire a 10-vote lead. Following a State Supreme Court ruling that allowed several hundred ballots from King County to be included, her lead was further increased to 130 votes, but when the vote was certified by the state’s Secretary of State, Sam Reed, at the end of December, one vote which had been counted in Thurston County past the deadline was disqualified and her lead was reduced to 129 votes. Washington’s Republican leadership then filed suit, claiming that hundreds of votes, including votes by felons, deceased voters, and double voters, were included in the canvass, but on June 6, 2005, Judge John E. Bridges ruled that the Republican party did not provide enough evidence that the disputed votes were ineligible—or for whom they were cast—to overturn the election. Judge Bridges did note that there was evidence that 1,678 votes had been illegally cast throughout the state, but found that the only evidence submitted to show how those votes had been cast were sworn statements from four felons that they had voted for Rossi. He subtracted those four votes from Rossi’s total and upheld the election.
Rick U. writes:
If the election is close, I agree with Edward that the chances of stealing the election is very high. All we need to do is look back to Minnesota. Dems have perfected the art of post election fraud.
A. Zarkov writes:,
John Fund at the Wall Street Journal reports that 25 percent of those surveyed in Massachusetts think ACORN will try to steal the election for Coakley including one out of six Democrats! The lack of exit polls in this election make that even more likely because exit polls provide some (however imperfect) information about the integrity of the election. If the election is close then we can look forward to the usual post election shenanigans, where votes turn up in the trunks of cars, and waste dumps. Seymour Hersh in his book, The Dark Side of Camelot, documents the 1960 presidential election fraud in his chapter called “The Stolen Election.” Even the normally liberal Hersh couldn’t stomach the blatant fraud that put JFK in the White House.
I am becoming involved in the 2010 election for the 11th District in California, and I want to put the Democrats on notice that I expect them to try their usual tricks. As such I will do my best to make sure Republican candidate has the appropriate diagnostics in place to detect fraud and intimidation. The details must of course remain secret.
Jake Jacobsen replies to LA:
But it’s a perspective that requires you to downplay and even dismiss the importance of stopping an imminent, deadly peril, because stopping that imminent, deadly peril does not also stop other deadly but much less imminent perils.
I disagree, I find this to be comparable to those who are now saying that since I didn’t vote for McCain it is my fault that Obama is a socialist nightmare. I don’t have a problem with this because it doesn’t prevent another peril, but because it simply replaces the current larger existential problem (Democrats) with a slightly smaller one tomorrow (Republicans).
I should be excited about this because?
Your comparison with the 2008 election was off-base. There’s no equivalence between stopping Obama’s ruinous legislation and electing McCain. Stopping the legislation STOPS THE LEGISLATION. It doesn’t do anything BAD. Electing McCain is something BAD. But you are so hung up on the “existential threat” posed by Republicans that you don’t want to stop, or you don’t want to put too much energy into stopping, the Democrats’ legislation out of fear that it will help Republicans in any way at all. I call this a dangerous loss of a correct perspective, and the parent of no end of mischief. It’s the same type of thinking that drove Richard Hoste to argue that we shouldn’t bother protecting America from terrorist attacks, because doing so is a “neocon” thing to do and distracts us from stopping immigration.
I’ve said it many times: we need to deal with each issue on its own objective merits, not relativize or deny one issue because of some effect we think it will have on some other issue. For example, not liking the policy effects of focusing on terrorism, and so trying to downgrade the importance of terrorism by claiming that a 9/11 attack on America once a year would be no more damaging to the country than routine traffic accidents. For example, not liking the effect on American government of fighting Hitler, and therefore claiming that Hitler permanently in control of Eurasia would be just fine and dandy. For example, not liking the Repubican party, and therefore undercutting opposition to Obamacare because such opposition might help strengthen the Republicans.
It is this kind of relativistic thinking, not guided by truth, that has destroyed paleoconservatism as an intellectual and political force.
A. Zarkov writes (1/20/10, 12:22 a.m.):
With 99 percent of precincts reporting, Brown received 1,153,808 votes, while Coakley received 1,052,391, and the ratio is 1.1 to 1. I went back to my simulations and took the ratio for each run, and then took the mean of the 10,000 simulated ratios, and got 1.1 to 1. The polling data with my mathematical model worked perfectly, correctly predicting the winner and the vote ratio five days in advance of the election.
With a victory margin of over 100,000 votes, the Democrats won’t be able to steal this election, so it’s a done deal. The big question now is: how long it will take to seat Brown? Perhaps the Democrats will now have finally developed “Obama fatigue,” and they will seat Brown without a delay and a fuss. [LA replies: That’s clearly indicated by Sen. Webb’s statement saying that there should no further votes on the bill in the Senate until Brown is seated.] If they do this then look for them to start working on Olympia Snowe to help them pass the health bill. In my opinion this would be a wiser course because a delay in seating Brown would generate tremendous anger. If they can turn Snowe, then she takes the heat, not them. Never underestimate how stupid Republicans can be. Look at Michael Steele. He says they are not ready to govern!
That’s true, there’s always Snowe and Collins. Thus the bill may be the Undead.
Posted by Lawrence Auster at January 19, 2010 02:27 PM | Send