The methodology of two recent Georgia surveys by a Democratic robopoller that found Republican support sliding has drawn withering criticism for tweaking the sample’s racial composition to allegedly tilt the results.

Public Policy Polling, the high-volume automated polling firm, released a pair of surveys over the last two weeks that showed Democratic contenders for U.S. Senate and the governor’s mansion in contention with their GOP rivals.

In the Senate contest, PPP found Democrat Michelle Nunn tied at 42 percentage points with a generic Republican opponent. When matched specifically against Reps. Paul Broun, Phil Gingrey or Jack Kingston, who the pollster identified as supporting the recent 16-day government shutdown, she carved out a shocking 6-point lead.

The pollster also found Gov. Nathan Deal’s approval rating upside down and managing only a mere 4-point lead over hypothetical Democratic rival state Sen. Jason Carter.

Those numbers, whose results were commissioned by a pair of progressive advocacy groups, were far afield anything produced by other respected firms, and numbers cruncher Nate Cohn of The New Republic says he knows why: the pollster radically adjusted the racial composition of Georgia’s electorate.

Whereas the assumed white share of the electorate registered at 71 percent in an August survey of the state by the same firm, it was only 63 and 62 percent this month.

That the electorate grew precipitously more diverse in a matter months–a proposition the pollster argued would prove unlikely in the 2014 election season–speaks to the polls’ results and the pollsters’ business interests as a Democratic firm, Cohn said.

“If PPP had used the August weights for its October poll, Nunn would have trailed by 6 points. Deal would have led by 10 points if PPP used the August weights for its October poll,” Cohn wrote. “And obviously, it was in PPP’s interest to avoid both of those outcomes.”

– James Richardson