[Fresh Wind]

Not to be a nitpicker, but the oversampling of women was 54% to 46%, or 8%. The 2010 census showed women outnumbering men by about 1.6%, so the PPP oversample of women is actually 6.4%, and that should definitely skew results in Obama’s favor since he always seems to poll stronger with women (Rush’s arousal gap).

The question is whether this oversampling of women is deliberate or a side effect of discarding republican responses to reach the desired R/D mix, which are more likely to have come from men.

But, I agree, any time someone has to decide whether to keep or toss a given sample, there is the potential for mischief.

[dfwgator]

By women, do you mean single women, or married women? Big difference.

[muawiyah]

It isn't that they oversampled women ~ rather, they dropped some sample cells in a multi-characteristic polling scheme, and there were a lot of guys in those cells BUT the numbers reported weren't statistically significant.

This is called an "artifact of the design" and doesn't really reflect opinion ~ just that you are hearing results where they had so many women in that line, but they had too few men in that line to tell you about it.

In short, the guy paying for the poll didn't want to pay enough for a poll that would tell him everything he'd asked about.