Sampling isn't done based on party ID. Party ID is one of the questions asked during the sample.
Do you think the turnout will be at 2008 levels? There isn't any chance of that happening, zero chance. If that is the case any poll showing these spreads is based upon faulty data.
I took a look at the polls that list party identification numbers at the top of the RCP Ohio page.
There are 9 polls and the averages work out to be
Democrat 39.3%
Republican 33.1%
Independent 26.3%
The CNN exit poll has the following
D 39%
R 31%
I 30%
So they have .3% more Democrats, on average, in their samples. 2.1% more Republicans and 3.7% fewer independents.
As points of references....
2010 CBS exit poll out of Ohio had
D 36%
R 37%
I 27%
2004 CNN exit poll had
D 35%
R 40%
I 25%
I think one of the main reasons that these polls are still favoring Democrats is because of lax likely voter screens. One poll had 603 likely voters taken from 611 registered voters for instance.
In any case for these polls to be accurate you have to assume that 2012 turnout is going to be real close to 2008.