Has the New York Times been caught playing with its poll numbers in U.S. Senate midterm races?

Endangered Species: Democrat Senators

I thought this was sort of stinky when I first saw it this morning...

I checked Rasmussen and it had a poll from February showing Tom Cotton five points ahead of Mark Pryor.  So the idea that Pryor is now ten points ahead of Tom Cotton now really does not make a lot of sense.

Jim Geraghty is questioning this.  One of Jim's readers notes that there is shenanigans afoot in the NYT's poll cross tabs:

Patrick • 4 hours ago 

From the poll's crosstabs:  

Of Arkansas poll respondents, 27% claimed they voted for Romney in 2012, 26% for Obama, 8% for someone else(!), and then some don't knows, didn't votes, etc. Actual Arkansas result in 2012: Romney 60.6%, Obama 36.9%. This looks like a very unrealistically Dem-heavy sample. That does seem to be the "All Persons" breakdown (they didn't provide separate stats for Registered voters) but still there's no way that equates to a representative sample. 

The other states in this NYT poll are similarly very skewed toward underrepresenting Romney voters. The Kentucky sample reports that 31% voted for Romney, 28% for Obama. Romney won Kentucky 60.5% to 37.8%. Louisiana poll breakdown is even more skewed, 31% Obama, only 28% Obama. But Obama lost Louisiana 40.6% to 57.8%. North Carolina rounds out the superfecta: poll 38% Obama voters and only 31% Romney ones, though Romney won NC 50.4% to 48.4%. 

I could conduct a poll showing that McCain would beat Obama by ten points in 2008, if I started by suppressing the weight of 2004 Kerry voters by 30% and inflating the weight of 2004 Bush voters by a similar amount.

And Arkansansans do the right thing: Vote for Tom Cotton

Update:
Americans for Prosperity launch ObamaCare attack ads on four vulnerable Democrat Senators