I'm continuing to follow polling from Votemaster Andrew Tanenbaum's www.electoral-vote.com, this week I've run another pair of simulations of the general election. Most state polls have a 4% margin of error, so one simulation uses that. Those numbers are a reflection of what would happen if the states followed the most recent polling results in each state, taking into account random sampling error as the only source of variability. In addition, for the past several weeks I've used a wider margin of error to try to capture some possibility of movement beyond just sampling error. Taking the 2004 data I've done a linear regression fit assuming the error of polls declines the closer to the election we come. That regression currently gives an 11.9% margin of error, so that's the other simulation.

4% Margin of Error
Obama wins 98.93%, averages 298.8 EV (low 252, median 297, high 367)
McCain wins 0.48%, averages 239.2 EV (low 171, median 241, high 286)
Electoral tie 0.59%

11.9% Margin of error
Obama wins 87.1%, averages 299.3 EV (low 181, median 299, high 391)
McCain wins 12.1%, averages 238.7 EV (low 147, median 239, high 357)

This week's numbers are not materially different from last week's, but there may perhaps be slight movement in McCain's favor.

In the 4% margin of error simulation I noticed the unusual case that an electoral tie (in which case Obama would likely become president in the ensuing state-by-state vote of the House of Representatives) was more likely than all of the winning McCain scenarios combined.

If Obama wins all the states Kerry won, plus Nevada, New Mexico, Iowa, then he gets to exactly 269 electoral votes. Obama is currently leading in all of these states right now, plus Indiana and Colorado, giving him a cushion of 20 electoral votes.

Compared with last week's numbers, the new state polls do mostly show gains for McCain. A Survey USA poll gives McCain a 6 point lead in Florida, , and a Rasmussen poll has McCain leading by 7 in Missouri. In Wisconsin, where Obama had double-digit leads in June and July, two new polls have Obama ahead by 6 and by 4.

In Oregon, Survey USA finds a 3 point lead, down from a 9 point lead in a Rasmussen survey in July. The two firms have done several polls of Oregon, with Survey USA usually showing a much closer race. Indeed their prior poll, from June, also showed Obama ahead by just 3, so this could simply be a difference in sample weighting between the two polling firms.

The best news for Obama this week is a poll now giving him a 7 point lead in Michigan, one point short of his largest lead this year. The other new polls were in states where one candidate has a large lead.

The state polling had been basically static since Obama got a bounce from clinching the Democratic nomination. The past two weeks may now indicate a small shift starting towards McCain, but Obama is still has a commanding lead in the current electoral map. McCain needs a surge in the polls to be able to win.

It is possible that McCain's attacks are starting to have an effect. Conventional wisdom is that most of the public doesn't really pay attention to the race until after Labor Day anyhow, so it's quite possible to read way to much both in the current state of polling (still quite favorable to Obama) and the recent changes (mostly in McCain's direction).

I'll be on vacation for the next few weeks, likely without internet access, so I this may be my last update before Labor Day.

Yet again I've burned a more CPU cycles simulating the November elections, using data from Votemaster Andrew Tanenbaum's www.electoral-vote.com.

We're now just over 3 months away from the election, and there were a number of new state polls this week, but in terms of frequency of winning simulated elections, essentially nothing has changed since last week. Or, arguably, since mid-June: Obama continues to lead in more states, and McCain needs to improve markedly in state polling to win the race. I'm running two simulations, one using a 4% margin of error that is a snapshot of the most recent polling from each state, and another using a 12% margin of error.

4% Margin of Error
Obama wins 99.05%, averages 309.2 EV (low 255, median 310, high 378)
McCain wins 0.68%, averages 228.8 EV (low 160, median 228, high 283)
Electoral tie 0.27%

12% Margin of error
Obama wins 90.7%, averages 306.0 EV (low 182, median 307, high 404)
McCain wins 8.8%, averages 232.0 EV (low 134, median 231, high 356)
Electoral tie 0.5%

This week I took a closer look at how recent the polling is in each state. All the polls were from June or July except for five states: Wyoming's latest poll was in May, Maryland's was in early March, and Vermont, Deleware, and Hawaii were last polled in late February. All those states have double-digit or larger leads for a candidate, except Delaware, where Obama led by 9, so those states are not in play.

Of the 10 states last polled in June, only Indiana (Obama up 1 in a June 23 poll from Survey USA) had a lead of less than 8 points. So we've got recent polling from the closest states.

With all the hand-wringing about national polls and how Obama may be slipping back, I find it interesting  that what movement we see this week in the "big three" swing states is favorable to Obama. Obama led by 2 in a Quinnipiac poll in Florida, after leading by 1 and trailing by 2 in Rasmussen and ARG surveys last week. While it's not statistically significant evidence that Obama is gaining, or even that he's ahead, it is a marked improvement over spring polls where Obama had trailed by as much as 10. Florida looks like it will be close, which means McCain will have to fight hard to win it, and McCain needs Florida *much* more than Obama does.

Similarly Obama is ahead by 2 in a Quinnipiac poll of Ohio, reversing a 6 point lead for McCain in the prior week (the Votemaster's algorithm averages the two polls, since their midpoints were within 7 days of each other). So where last week my model would have given McCain Ohio almost all the time, now it's closer. I'd guess that McCain is still likely ahead in Ohio, but it's tighter than 6 points.

The latest Pennsylvania poll (Quinnipiac again; they've been busy this week!) has Obama ahead by 7, up from 5 and 4 point leads in two previous Rasmussen polls from last week and late June. The differences could be sampling noise, or minor differences in weighting between samples, so perhaps Pennsylvania isn't moving at all.  The most likely outcome as of now appears that Obama will win Pennsylvania and one of Ohio or Florida. But in the current map, Obama needs to win just one of them. And to win all three, McCain needs to see significant movement in his direction.

There is good news for McCain in other close states.  The two previous polls in Missouri in early July agreed on a 5 point gap but disagreed on the leader. This week Survey USA sees McCain up 5, so it now McCain is likely at least ahead there. After an early July poll showing Obama ahead 5 in Montana, McCain is now back ahead by one (both polls from Rasmussen). This is likely movement back in McCain's direction, but it could also be statistical noise.

North Carolina was the only other state closer than 5 points with a new poll this week, and it showed McCain ahead by 4,  no change from the previous leads of 3 and 5.

The state polling landscape hasn't materially changed since Obama's bounce from clinching the Democratic nomination in early June:  Obama is ahead in all the states Kerry won, and is also leading in 6 states Bush carried in 2004: Colorado, Florida, Indiana, Iowa, Nevada and New Mexico.

Virginia is a tossup in current polls, and McCain's leads are less than 5 in previously solid red states North Carolina, Montana, and the Dakotas.

Certainly if opinion shifts towards McCain nationally, this race could be a dogfight, and it's far from assured that Obama will win. But he's still in a far stronger position, and McCain has yet to make sigificant progress in eroding Obama's lead in state-by-state matchups.

Here's another post of my Monte Carlo simulations using state-by-state polling data from Votemaster Andrew Tanenbaum's www.electoral-vote.com. Most state polls have a 4% margin of error, so I use that as input in one simulation. In the other, I'm using an interpolated value based on the accuracy of 2004 polling data in predicting the 2004 election results - that interpolated value gives a 12.1% margin of error now, and it will continue to shrink as we get closer to the election.

4% Margin of Error
Obama wins 99.40%, averages 308.9 EV (low 234, median 305, high 363)
McCain wins 0.58%, averages 229.1 EV (low 175, median 233, high 304)

12.1% Margin of Error
Obama wins 89.6%, averages 303.8 EV (low 174, median 304, high 404)
McCain wins 9.7%, averages 234.2 EV (low 134, median 234, high 364)

The big changes this week affect some of the big, traditional swing states. Obama had been leading slightly in Ohio before, but Rasmussen's latest poll now shows McCain ahead by 6, which helps McCain especially in the 12.1% margin of error. This is partly offset by two new Florida polls, one by Rasmussen showing Obama now ahead by 1, and another by ARG showing McCain up by 2. McCain had led by 7 here in Rasmussen's previous Florida poll, so the Sunshine State moves from being likely McCain to a virtual tossup. These two states show moves of about 8 points, in opposite directions, which suggests real opinion change and not simply noise from sampling in the states.

Some other states also suggest movement in McCain's direction. Obama's lead in Minnesota had been double-digits, but Rasmussen shows him up by 8, and a Quinnipiac
poll has him ahead by just 2. Michigan's two new polls this week show Obama ahead by 4 and 2, while the prior two surveys had given him 6 and 8 point leads. New Hampshire is closer also, as three new polls show Obama up by 6, 2, and 3, after previously leading by 12 in June.

Virginia still looks quite close, with the latest survey showing both candidates tied at 44, while Obama retains a 5 point lead in Pennsylvania. Colorado is close: a Rasmussen survey gives him a 7 point lead, while a Quinnipiac poll showed McCain up by 2. So Obama likely has a small lead here.

We're now a little more than 3 months from the election, and while there's been small movement in many states, Obama retains his overall advantage by being competitive in many places. While the chances he may lose Ohio have gone up this week, the chances he wins Florida also have risen, keeping his overall chances in my simulations about the same. While the 4% margin of error is probably a fair snapshot of what would happen if the election were today, the 12.1% margin of error still overstates how likely I think Obama is to win in November. But the longer the campaign goes without McCain making significant progress, the better it is for Obama.

I thought about simply cutting and pasting last week's post, or the previous week's
that, or the one before that.... The data aren't changing much, but
what shifts in state-by-state there are tend to favor Obama a little
more. We're at a lull in poll movement. As usual, I'm using composite
polling data from Votemaster Andrew Tanenbaum's www.electoral-vote.com
as input to two simulations, one using a 4% margin of error (the
standard for most state polls), and another using a 12.25% margin of
error (the current value I get from a linear regression of the 2004
state-by state polling data errors compared with the actual 2004
election).

This week's results:
4% Margin of Error
Obama wins 99.94%, averages 317.3 EV
McCain wins 0.04%, averages 210.7 EV
Electoral tie 0.02%

12.25% Margin of Error
Obama wins 96.6% averages 314.0 EV

McCain wins 3.1%, averages 214.0 EV
Electoral tie 0.3%

The
4% margin of error simulation captures what current polls say, and
Obama is simply ahead by too much in too many places to see how he
wouldn't win in an election held now. He has double-digit leads in
states with 211 electoral votes, while McCain has such a lead in states
with just 83, so while several states in the middle might go either
way, McCain would need to win just about all of the marginally close
states to eke out a win.

The most relevant new polls this week
are almost all breaking in Obama's favor: Obama now leads in Michigan
by 8, continuing a trend in his favor. It's his largest ever lead
there, and is the third straight poll there where Obama has improved
(moving from a 4 point deficit in late May). Early i

Iowa,
Rasmussen now shows Obama ahead by 10 (up from a 4 point lead in Survey
USA's mid June poll, and a 7 point lead in Rasmussen's previous poll in
early June), suggesting this one-time swing state may also be moving
beyond the polling margin of error.

Rasmussen now shows Obama
retaking the lead in Nevada, where he's now up by 2 after traling by 3
in their previous poll in June. Obama lead in February and March, then
trailed regularly, and now he's back in front there.

McCain
still leads in North Carolina, but it may be tightening a little more,
as Survey USA shows him up by 5, and Rassmussen now shows him ahead by
3.

Obama is making more of the Mountain West competitive, as he
now trails by just 4 in South Dakota (following last week's gains in
North Dakota and Montana).

The other new polls are in states
clearly in one column or another, with some movement favoring Obama,
and other movement favoring McCain, but nothing which yet suggests any
of these states may be competitive in November.

The
state-by-state polling data certainly do paint an optimistic picture
for Obama, but these simulations assuredly overstate the likelihood of
his winning. While there hasn't been much movement recently in winning
percentages, McCain was leading Obama in mid-May, which shows that
things can change over time.  We're still  over three months from the
election, and most of the country isn't really paying close attention
yet.

As a point of comparison, political futures sites like www.intrade.com and the Iowa Electronic Markets
currently give the GOP about a 30-35% chance of winning in November,
with the Democrats at about 65%-70%. That's surely more accurate than a
95% or better chance for a Democratic win, but my biased personal
opinion is that Obama's chances are better than the markets say; I'd
guess he has about a 75% chance of winning.

I was out of town over the weekend, so this week's simulations are a little later than last week's. As in the past, I'm taking data from Votemaster Andrew Tanenbaum's www.electoral-vote.com and running a simulation of 10,000 elections each, using both the 4% margin of error common to most state polls and a 12.4% margin of error, the latter coming from a linear regression of errors in 2004 polling data in predicting the final outcome.

This week's data:
4% Margin of Error
Obama wins 99.87%, averages 315.6 EV
McCain wins 0.08% averages 222.4 EV
Electoral tie 0.05%

12.4% Margin of Error
Obama wins 94.2%, averages 310.1 EV
McCain wins 5.4%, averages 227.9 EV
Electoral tie 0.4%

The standard caveats apply: we're still just under 4 months away, and many of the most recent polls are rather old.  The 4% margin of error case is at best a snapshot of the current state of the race: that allows only for variation caused by sampling error, not the possibility that opinion might shift during the course of the campaign due to outside factors.

I intended the larger margin of error to allow for that as well, but I strongly suspect it also overstates Obama's chances of winning, because my model is assuming each state's result is a completely independent trial. It is more likely that if general public opininion is shifting, it will tend to shift in the same direction in most states, making it more likely for McCain to come back (or for Obama to win in a larger landslide).

Individual states have bounced around a bit, but really not much has changed since Obama got a bounce from clinching the Democratic nomination: he's been winning nearly 100% of the current snapshot model, and about 95% of the larger margin of error for the past month. This week is no exception, as there are a few new polls which mostly cancel each other out in effect.

Missouri is a tight battleground state, and this week had contrasting polls, with Rasmussen finding McCain up by 5, while Research 2000 gives Obama the same 5 point edge. Following the Votemaster's algorithm, these two polls get averaged to make Missouri a toss-up, an improvement for Obama over last week, where recent polling had McCain slightly ahead. Earlier polls had given Obama a small lead, so this one truly looks like it could go either way. But it's one of an increasing number of "must win" states for McCain if he's going to have any serious chance of getting 270 electoral votes.

Another new data point is the first North Dakota poll since April: Rasmussen shows a 43-43 tie. McCain had a 6 point lead before in a Dakota Wesleyan poll, while the only other North Dakota poll was by Survey USA in February, showing Obama up by 4. This shifts ND from likely McCain to a tossup, another favorable development for Obama.

Rasmussen has McCain trailing by just 5 in its latest poll of New Jersey, which gives McCain a glimmer of hope he could take the Garden State, and their survey of Washington shows Obama's lead falling to 8, the first time he's led by less than 10 since late May. Rhode Island, Illinois, and Maine also have recent polls showing Obama's lead shrinking, but in each case his lead is well above 10%, so Obama wins these states virtually all the time in both simulations. The Rasmussen poll in Illinois is just the second of this matchup, and it gives Obama a safe 13 point lead, but one much smaller than the 29 point lead Survey USA saw in February. The only other new poll released this week is from Alabama, where Capital Survey has McCain up 49-36, his smallest lead of the season.

The overall map remains favorable to Obama: even though Florida is currently quite solidly in the McCain column, Obama wins overall quite comfortably. And there's evidence trickling in from Mountain West states like Montana and North Dakota that the 50-state strategy is chipping away at the GOP base. Without a significant public opinion shift in his favor, McCain will lose.

Each week I've been doing Monte Carlo simulations of the general election based on data from the Votemaster's www.electoral-vote.com. In addition to a simulation using a 4% margin of error for the polls, for the past few weeks I've been using a wider margin of error,  this week about 12.5%, taken from a regression of 2004 polling data errors compared to the final results. For each margin of error, I've run 10,000 simulated elections. This week's results:
4% Margin of Error
Obama wins 99.5%, averages 308.4 EV
McCain wins 0.5%, averages 229.6 EV
No electoral ties

12.5% Margin of Error
Obama wins 93.4%, averages 307.3 EV
McCain wins 6.2%, averages 230.7 EV
Electoral tie 0.4%

McCain is doing a little better than in the last two weeks because he leads by 7 in the most recent Florida poll, a Rasmussen poll from June 26th. It's not clear that this poll represents a genuine shift - the previous Rasmussen poll actually gave McCain a larger lead (by 8), whereas Obama led in the latest Florida polls from ARG and Quinnipiac. The Votemaster's algorithm (which I use) averages the latest poll with any other polls from other pollsters within 7 days of that most recent poll. Since the ARG and Quinnipiac polls were a little older than that, they're ignored in my analysis, which makes Florida swing from a toss-up to quite likely McCain. If McCain is taking a large lead in Florida, it's certainly good news for him.

Obama's good news for the week is taking a 5 point lead in Montana. Sure, it's only 3 electoral votes, but it's yet another formerly solid red state where Obama may well be competitive. This is the advantage of the 50-state strategy: the more states Obama might be able to pick up, the less dependent he is on winning previously key swing states like Florida, Ohio, or Pennsylvania.

Pollsters may well differ on Georgia: Insider Advantage gives McCain just a 2 point lead this week, and their prior poll in mid-June had McCain up 1. By contrast Rasmussen's last poll gives McCain a 10 point lead, the same lead he had in their poll in early June. I suspect that they're differing in how they weight the samples, and that the race is probably more like a 4-7 point difference that has remained rather stable, rather than swinging between a 10-point McCain lead and a virtual dead-heat. Georgia (and Montana) should be solid McCain country if he's going to win; that he's perhaps in trouble in both is certainly good news for Obama.

Louisiana may be in the same situation, but with different pollsters: here Rasmussen's two last polls were showing McCain up 9 and 11, wheras Opinion Research shows a 16 point McCain lead in their last two polls. Unlike Georgia, no pollster shows this a close race, so these fluctuations don't affect the simulations at all.

Most of the other changes this week favored Obama, although none had a
significant effect on the simulations. Obama increased already large
leads in New York, New Jersey, Connecticut, Massachusetts and Rhode Island. And while McCain still leads in Alabama, the last poll shows a 15 point lead, down from 20-30 point leads in earlier polls.

I'll add the usual caveats: we're still four months away from the elections, and much can and will change. McCain's chances of winning are probably more like 30% than 5% or less, but the current polling certainly still favors Obama. Despite currently losing Florida, he still projects to win by about 80 EVs.

McCain is shaking up his campaign staff, giving more responsibility to Karl Rove protege Steve Schmidt. Current polling data suggest some sort of change is needed.

Here's another in my series of weekly general election simulations, using data from Andrew Tanenbaum's www.electoral-vote.com.

I do 10,000 trials each with two different margins of error: a 4% margin of error reflects the sampling error of most of the state polls used in the analysis. That is effectively "if the election were held today, what would the results be?". To try to make a better prediction, I also ran a regression analysis of 2004 polling data against the final results, with the number of days to the election as the independent variable. That extrapolated margin of error is now 12.7%, so that's the other value I'm using.

This week's results:
4% Margin of Error
Obama wins 99.8%, averages 322.8 EV
McCain wins 0.2%, averages 215.2 EV
No electoral ties.

12.7% Margin of Error
Obama wins 94.5%, averages 315.2 EV
McCain wins 5.1%, averages 222.8 EV
Electoral tie 0.4%

These results are quite similar to last week's, with the model showing Obama is very likely to win. The usual caveats do apply: these are polls still over 4 months before the election, and a lot can and will change. At best the polls are a snapshot of current opinion. And even when I use a wider margin of error, I'm still treating each state's result as an independent event, which understates the chances of a popular opinion shift in favor of McCain in many states at the same time. Personally, I think McCain has a much better than 5% chance to win the election, but this analysis indicates he's got to make gains in a lot if states simultaneously to become competitive.

Obama is leading solidly now, an observation supported both by the national Newsweek poll that gave him a 15 point lead, and the L.A.Times/Bloomberg poll that gave him a 12 point lead (or 15 also, if you added Bob Barr and Ralph Nader into the mix).

The Votemaster has a lot of new data this week, but there has been relatively little that materially affects the simulations. The latest poll in Missouri showed McCain moving out to a 7 point lead, after trailing slightly beforehand. This is balanced by an Indiana poll that now gives Obama a 1 point lead, where he previously had been trailing by high single digits. Ohio and Pennsylvania might be tightening, as Obama's leads shrunk there, but his lead in Michigan might be widening, and there's another Mississippi poll giving McCain just a 6 point lead, and the two last polls from Texas showed McCain up 9 and up 5, where prior to that he'd led by 13 in two earlier polls.

So while little has changed overall, the optimistic reading of the new data for Obama is that he is making inroads in the deep South, what should be solid McCain territory. If McCain needs to devote resources to holding Texas or Mississippi (!), then Obama is in fine shape indeed.

The Federal Elections Commission now has a quorum, after the Senate approved by voice vote yesterday the nominations of 5 new commissioners, three Republicans and two Democrats. Hans von Spakovsky had previously withdrawn his nomination, but Senate Majority Leader Harry Reid said he was withholding a vote until Wisconsin Senator Russ Feingold had a chance to speak with each nominee.

Now perhaps things can move forward on resolving John McCain's self-financing status for the primaries.

Thomas Friedman's column today provides proof that the print media are spiraling towards irrelevance:

... this from a president who has so neutered the Environmental Protection
Agency that the head of the E.P.A. today seems to be in a
witness-protection program. I bet there aren’t 12 readers of this
newspaper who could tell you his name or identify him in a police lineup.

Regular TPM readers, of course, could both name and recognize Stephen "Stonewall" Johnson. Yet Friedman is convinced the esteemed readership of the Times would be unable to do so. I guess we're not the Times's target audience any more.

Here's another weekly installment of my simulations of the general election results using polling data from Votemaster Andrew Tanenbaum's www.electoral-vote.com.

The usual caveats apply: it's still June, polling is unreliable, and many of the polls are quite old now anyhow. But to paraphrase our former secretary of defense, you analyze the data you have, not the data you'd like to have.

Again I'm running 10,000 trials of the election using two different margins of error:
4% Margin of Error
Obama wins 99.87%, averages 327 EV
McCain wins 0.10%, averages 211 EV
Electoral tie 0.03%

12.8% Margin of Error:
Obama wins 94.4%, averages 314.2 EV
McCain wins 5.3%, averages 223.8 EV
Electoral tie 0.3%

The 4% margin of error that reflects the statistical margin of
error of most polls. 12.8% comes from a
linear regression of polling errors in 2004 using days until the
election as the independent variable. The former captures only sampling
error for polls, and does not allow for movement of opinion in the
electorate itself. The latter tries to allow for such movement by using
a wider margin of error, although because I'm still treating each
state's as an entirely independent random event, this doesn't reflect
the likelihood that improving your vote in one state also improves it
in others. So both of these simulations overstate the likelihood of
Obama actually winning right now.

Obama is now trailing by just 4 in Alaska, he's closed to a single-digit deficit in Arkansas, and he's taken a 1 point lead in Virginia. Two recent New Hampshire polls both give Obama double-digit leads, he's now just 2 or 3 back in Nevada, and as Josh noted earlier this week, he trails by only 12 in Kentucky despite his lopsided primary loss. Not all the movement is in his favor: Minnesota has tightened to a 1 point race (this may be in part pollster differences: Survey USA's last previous poll had Obama up just 5, but in between two Rasmussen polls gave Obama double-digit leads); Obama is up just 2 in Colorado; and McCain has closed to 4 back in Iowa. Bob Barr gets enough support in Georgia that Obama was just down 1 in the last poll.

The huge change here is, of course, Florida. Rasmussen had given McCain regular double-digit leads earlier, and their latest poll still says he's up by 8. Quinnipiac had smaller McCain leads before of between 2 and 9 points, but now they show Obama up 4. ARG's first Florida poll, released this week, has Obama ahead by 5. So while Florida seemed perhaps a likely safe McCain state a week ago, it's now quite clearly in play, which greatly changes things. It's hard to envision McCain winning without Florida, so that's the biggest cause of his dropoff in my simulations.

Personally, I think expecting a 95% probability of Obama winning greatly overstates things, but this exercise does capture that he's showing increasing strength in more of the country. It validates the 50-state strategy: the more of the country you put in play, the more ways you have to win. Obama already was able to win without Florida and Ohio, but it would have been quite hard. With states like Virginia, Colorado, Nevada, and North Carolina potentially in play, he has much room to compensate if he's weaker in some areas.

This may not have been the best of weeks for the fourth amendment, but it's been a good one for Obama's polling numbers.

I was certainly disappointed that Senator Obama did not forcefully oppose the FISA surveillance bill that passed the House yesterday. His lack of action here helps keep a bad bill alive, one which, should it become law, would set the awful precedent of condoning illegal behavior simply because the President asked companies to violate the law. Certainly there is risk that the new surveillance powers, due to the lack of any meaningful judicial oversight in using them, will be abused by the present, and future, administrations. This is a case of codifying that Presidential request is a valid excuse for breaking the law.

Senator Obama has shown outstanding political instincts, however, and it is quite likely that his judgment of the politics here is far better than mine.
1. Vocally denouncing the compromise, which has significant Democratic and virtually unanimous Republican support (only one Republican, Timothy Johnson of Illinois, voted no), would work against his attempts to portray himself as a practical politician who can work across party lines.
2. As Congressman Rush Holt noted, the bill sunsets in four years. Obama may rationalize his tacit support by his belief that he will win in November and will not abuse the power this bill would grant, mostly to him. And if he, as president, doesn't try to renew or extend it, the law will expire quietly, lessening the negative consequences of passage.
3. He has stated (along with Harry Reid) that he will try to get the retroactive immunity pulled from the bill. It's still an open question how hard he would try, or how effective such attempts would be, but the bill has not become law yet, and it may yet be amended further.
4. There's little political gain from opposing it strongly, but larger risk. Those who most strongly oppose FISA are overwhelmingly likely to support Obama for many other reasons. McCain gives unqualified support to the bill, so people voting on this issue have nowhere to turn. Now that the primary is over, Obama is understandably shifting more to the center to try to win independents and moderate Republicans, some of whom might believe this bill is "vital" in our war on terror.

If Obama opposed the bill, it would certainly push the issue of anti-terrorism tactics more to the foreground, and there's also a chance that having a more open, overt debate about that could wind up helping Obama. Undoubtedly it would be a big risk, and with his current strong standing in the polls, it's reasonable to think the risk isn't worth the potential gain. If he were trailing, and seemed likely to lose in November, a risky strategy would be more appealing.

Obama's handling of this reminds me of his handling of the Florida/Michigan revote proposals: had he actively and strongly supoprted revotes, they likely would have happened, and he may well have gained  politically from the new elections. But they also risked a large, public defeat that might have sunk his candidacy. When you're leading, don't risk the big mistake.

Retroactive immunity itself isn't what I (and, by projection, I believe without evidence most immunity opponents) care most about. What has me more upset is the disregard for the law shown by the Bush administration: its manipulating intelligence to whip up support for war; its twisted justifications of torture; and its arrogance in asserting it is beyond oversight of any other branch of government. The telecommunications companies who collaborated with the government here were "only following orders", which at least somewhat does mitigate their culpability. If allowing retroactive immunity is the price one has to pay for a more direct, public repudiation of the Bush administration's conduct, I'd take that trade. Winning only on retroactive immunity itself would be like convicting Al Capone for tax evasion: it's better than nothing, but it's not really the point.

So yes, I'm disappointed that Obama didn't strongly oppose this bill. But I can understand why, and his judgment is probably tactically correct here.

Senator Obama -

I appreciate and have been moved by your calls for change, and I have supported your campaign both with my vote and with my money. I know it is important to change the direction our country has been on, especially the last seven years.

As a constitutional scholar, I'm sure you understand the erosion of civil liberties and the disregard for law shown by the current administration, and by promising to review its executive orders to repeal those that are not constitutional, you have shown that you intend to undo its excesses. For this I, and true lovers of our limited government Republic, applaud you.

But I have another particular request: please publicly denounce the FISA compromise bill announced yesterday. You had previously supported Senator Dodd when he filibustered an earlier bill. Yesterday's "compromise" bill may in fact be worse than the one Senator Dodd killed with your help.

This bill would dismiss any lawsuits against telecommunications companies that can produce a letter indicating that the President authorized them to engage in the wiretapping program. While it may not result in quite as broad immunity as the earlier bill (perhaps some companies cannot produce such a letter), this "justification" for granting immunity would set a terrible precedent: it would codify, in a law passed by our legislative branch, that the executive can authorize lawbreaking. De facto it means that the President is above the law, that breaking the law is fine when the President asks you to do so. We should not pass a bill that affirms that terrible precedent.

Opposing this bill is not without risk; certainly the Republicans will claim we must pass this bill to ensure our security, and such claims may resonate with many in the general public.

Yet this is a case where you should put principle before politics, and speak out strongly against this "compromise". It is not a compromise about the mundane details of intelligence gathering; it is a compromise of our fundamental value that no one, not even the President, is above the law.

I urge you to oppose it, by filibuster if necessary.

Respectfully,

A Committed Supporter

It's time for another in my weekly series of election simulations, using data from Votemaster Andrew Tanenbaum's www.electoral-vote.com.

The standard disclaimers apply: the polling data are still thin, with the most recent poll in some states still being from February; and we're several months away from the election, so things will certainly change.

As I started to do last week, in addition to running a simulation using the 4% margin of error common to most polls, I'm running another simulation with about a 13% margin of error. I've chosen this number by doing a regression against the errors in 2004 polling data compared to the actual election result, as made available from www.electoral-vote.com. This 13% number will slowly decrease to about 10% just before the election in November - the closer we are to the election, the more accurate the polls tended to be in 2004.

This week's numbers, from 10,000 trials:
4% Margin of Error
Obama wins 98.6%, averages 301.1 EV
McCain wins 1.3%, averages 236.9 EV
Electoral tie 0.2%

13% Margin of Error
Obama wins 77.8%, averages 287.6 EV
McCain wins 20.9%, averages 250.4 EV
Electoral tie 1.3%

Obama is enjoying his post-election bounce, and this is his strongest showing yet. He took the lead in Michigan, widened his lead in Wisconsin, and another Rasmussen poll now has him trailing by just 2 in North Carolina. Not all the news for Obama is good: a new poll in Indiana has McCain on top (the previous two polls, from April, agreed on the 8 point margin, but disagreed on the leader, so Tanenbaum and I were calling it a tie).

One interesting tidbit: Obama is just 8 points down in West Virginia now, polling much more strongly in a general election matchup against McCain than he did in the primary against Clinton. This reinforces the fact that losing a primary doesn't inherently mean that you lose the general election (or the opposite). Democrats will still tend to support the nominee, even if their preferred candidate doesn't get the nomination.

The 4% margin of error numbers are, at best, a snapshot summarizing the recent polling, and Obama is now ahead in states which total over 300 electoral votes. Since my model treats each state as an entirely independent election, the only way McCain wins with a small margin of error is by getting lucky in pretty much all close states: Ohio, Missouri, Virginia, and Michigan.

The 13% margin of error allows for more variability, trying to capture possible movement from the electorate changing its mind between now and November. This still assumes each state is an independent trial, though, which isn't ideal: it's more likely that shifts in opinion in one state correlate with shifts in opinion in other states moving in the same direction. But simply increasing the margin of error gives McCain a plausible, if still unlikely, path to victory, as he is given some chance to win states like Pennsylvania, New Jersey, and Oregon.

When I use 4% margin of error, only 13 states didn't always go in one column, and in just 6 of them did the person trailing win at least 5% of the time: North Carolina, Virginia, Missouri, Ohio, Connecticut, and Michigan.

By contrast, with a 13% margin of error, just 8 "states" were always in one column: Obama always won Illinois, Vermont, Hawaii, and D.C., while McCain always won Alabama, Arkansas, Tennessee, and Utah. Each candidate won at least 5% of the time in 26 states.

That Obama's electoral vote averages are lower than the total of the states he currently leads reflects that he's taken small leads in several states, and in the simulations he doesn't always win them. For now McCain has more potentially close pickups than Obama has. But the current polling is certainly good for Obama, and unless McCain can swing opinion back his way, he's headed for defeat.

And Obama's plans to continue the 50-state strategy will put more pressure on McCain, because he'll have to spend in "safe" GOP states to keep them that way. It's beginning to look like a best case for McCain would be to replicate the Bush map against Kerry. That he's running for Bush's third term may be true in more ways than we first thought.

Each weekend I've been running a Monte Carlo simulation of the general election using data from Andrew Tanenbaum's www.electoral-vote.com. The Votemaster has stopped updating data for a possible  McCain/Clinton matching, so I'm no longer running simulations of that, either.

My previous simulations have used just the 4% margin of error common in state polls as the only source of variability, which implies that the only source of error in state polls is sampling error. At best this method gives a snapshot of the current state of the election, using the most recent polls. But we're still five months away from the election, so voter preferences will shift, and the model I've been using doesn't allow for any other changes.

As  a first cut, to adjust for this, I've estimated a wider margin of error for the polls, using data from www.electoral-vote.com from the 2004 election. Taking the polling available on a given date (the site has polling data and dates from September 1, 2004 until the election), I've computed the standard deviation of the polling errors from that date, and then I've done a linear regression of that data using the number of days until the election as the independent variable. From this I can extrapolate a margin of error to use for data today. This model gives me a margin of error of about 13%, over 3 times the 4% typically reported from the polls.

So I've run two version of the simulation, one using 4%, and the other using 13%, doing 10,000 trials of each. Results:
4% Margin of Error
Obama wins 88.0%, averages 289.4 EV
McCain wins 10.3%, averages 248.6 EV
Electoral tie 1.7%

13% Margin of Error
Obama wins72.2%, averages 284.0 EV
McCain wins 26.4%, averages 254.0 EV
Electoral tie 1.5%

This is Obama's best showing since I've been doing this, but the current results haven't changed much since last week. Obama now has a slim edge Missouri, where he previously had trailed by varying degrees, and while there have been other new polls, they just confirmed fairly large leads for the candidate leading in a given state, and thus had little impact on the simulations.

The 13% margin of error shows greater variability in the outcome, as it makes it more likely that states will swing to the other column. Intuitively, I find it easier to believe McCain has about a 25% chance of winning than that he has just a 10% chance. But that may still understate the current closeness of the race. In both these models, I'm assuming treating state's result as an independent random variable. While this is reasonable when you're only trying to reflect sampling bias, it's not such a good assumption when modeling opinion shifts. What causes a candidate to improve in one state is more likely to cause improvement in other states also. A different model which accounts for that would likely provide better projections.

In any event, current polling shows Obama ahead, and he may stretch that lead further if the end of the nomination race provides an extra bounce. The results using a 4% margin of error are, at best, a snapshot of the most recent polling data, which is often quite old. The 13% margin of error allows for changing opinions, but in a way that each state's shifts are completely independent of each other state's, which would understate the chances of the trailing candidate to make a comeback (or the leading candidate to win in a landslide).

The political futures market www.intrade.com has shown an interesting, and unusual reversal. The contract for Hillary Clinton becoming president is now trading above the contract for her winning the Democratic nomination.
2008.PRES.CLINTON(H) bid 6.0, ask 6.1, last 6.0
2008DEOM.NOM.CLINTON bid 4.9, ask 5.0, last 4.9

Is the market perhaps pricing in the possibility of an independent run? It's more likely this is simply caused by market inefficiencies (transactions costs of 0.3-0.5 per contract would make it impossible to profit from arbitrage here), but it the idea that the market gives some weight to a possible third party candidacy could also explain this oddity.