Take a good look at that chart.

In 1996, 82.3% of people registered to vote voted.

In 2000, 85.5% of people registered to vote voted.

In 2004, 88.5% of people registered to vote voted.

In 2008, 89.6% of people registered to vote voted.

Based on the trend, it is a good bet that 90% or more of the people registered to vote in this country will vote.

More from the Census Bureau:

Historically, the likelihood that an individual will actually vote once registered has been high, and 2008 was no exception. Of all registered individuals, 90 percent reported voting, up slightly from 89 percent in the 2004 presidential election.

Now, what do you think is more likely to be accurate?

• A mysterious alchemy formula likely voter screen.
  
• Multiplying the registered voter number by 90%.
  

See, what pollsters try to do is look into the demographics of their poll results and match those to historical patterns in census data. This is how they come up with these likely voters. They look at income, race, ethnicity, age, education, geography, etc. Some formulas are better than others. Some suck. Regardless, it is a complex science that just may be too cute by half. Someone with high income may be considered highly likely. But what if that person also has only a high school education? What about someone who has a high education but low income? How do you "screen" a white woman with only a high school education and low income, but also happens to be gay and an environmentalist? This is a crap-shoot "science."

That's not even getting into what or who called whom and where and who responded. It is all so maddeningly crazy, you don't know what to believe now do you?

Do this. Find the best registered voter number you can from a reputable pollster. Multiply each number by 90%. Now watch the trend over time. That's the best way those of us who don't have access to campaign polls can get a general sense of what is going on from the more cheaply done media polls.

For example, kossack Scan points to a Florida poll by CNN. Says,

LV Romney 49   Obama 48
RV Obama 50   Romney 43

So you're better off going with 45 Obama Romney 39. Then just watch the trend.

That doesn't mean the vote is going to come down 45 to 39, obviously. What it does mean is that the likely voter screen is assuming something that doesn't make any fucking sense. Because as we know from census data, 90% of the registered voters are going to show up. If the registered vote has Obama ahead by six points, then it is a good bet that Obama is ahead. If you see that number tighten up, then you might want to get into the likely voter number and watch the trend there.

You're welcome.

P.S. -  Some folks might be tempted to say...well what if Romney's registered voters show up by 95% and Obama's 85%? Here's what the Census Bureau has to say about that:

The differences in turnout among groups is small. Romney will get almost almost an all white 89.4% turnout vote. Most of the people voting for President Obama will be white too. Also 89.4% In fact, a great deal will be White women who turn out more than White men. The lowest performing group, Latinos, is STILL at 84% and I bet it will be better than in the swing states where the campaign is engaged.

Interesting factoid: the most likely person to turn out to vote is a Black woman. Go sisters!