Why PPP's latest poll is just not good

careful... this is mostly a sample size issue

h/t Blue Virginia

Sample size for rational sub-groups is often a challenge ... when n=600... when when looking at cross-tabs... Margin of Error (MOE) can be quite large in small sample sizes. RE: AA's, women, Hispanic, etc. Overall MOE = +/- 4%.

Quoting from Blue Virginia

UPDATE: Tom Jensen of PPP emailed me to say, "The Hispanic numbers are based on something like 12 interviews so that's why they seem erratic...I'm sure McDonnell isn't really going to win the Hispanic vote but when you have a sample size of around 600 for the overall poll and then a crosstab for a small voting bloc there's high potential for that number to be skewed."

Still the basic finding is correct -- this is a close race at this very early data point -- the true effect of intense campaign activity is yet to be measured.

---

by: Riley Murray @ Tue Jul 07, 2009 at 21:33:07 PM CDT

Some corrections to Bharat......

I think Bharat's general point is valid that the sample skews more Republican than the actual electorate will be, but Bharat exaggerates the extent of it.  It's dubious to predict a 7-point Democratic turnout margin over the GOP, as it wasn't quite that big in Virginia last year and it will be less this year with minorities making a smaller share of the total vote.  People of color were 30% of last year's voters, but only 22% of 2006 Senate voters, and this fall's electorate will be closer to 2006 than 2008 (there are no exit polls for odd-year Virginia state races, so federal races have to be used as approximations).  All that said, I do think Democrats will have more than just the one-point advantage in the PPP poll, and independents won't be one-third of the total--they're never that much in Virginia, they're always in the mid-to-high 20s.

Regarding the female and Hispanic vote, the small subsamples do, indeed, explain it.

Ultimately one thing I look for in crosstabs is the Democrat's share of the white vote.  If the Democrat gets 40% or close to it at this stage, then we're in good shape, because 40% is the magic number to eek out victory.

---

by: DCCyclone @ Sun Jul 12, 2009 at 08:37:00 AM CDT

Some corrections to Bharat......

I think Bharat's general point is valid that the sample skews more Republican than the actual electorate will be, but Bharat exaggerates the extent of it.  It's dubious to predict a 7-point Democratic turnout margin over the GOP, as it wasn't quite that big in Virginia last year and it will be less this year with minorities making a smaller share of the total vote.  People of color were 30% of last year's voters, but only 22% of 2006 Senate voters, and this fall's electorate will be closer to 2006 than 2008 (there are no exit polls for odd-year Virginia state races, so federal races have to be used as approximations).  All that said, I do think Democrats will have more than just the one-point advantage in the PPP poll, and independents won't be one-third of the total--they're never that much in Virginia, they're always in the mid-to-high 20s.

Regarding the female and Hispanic vote, the small subsamples do, indeed, explain it.

Ultimately one thing I look for in crosstabs is the Democrat's share of the white vote.  If the Democrat gets 40% or close to it at this stage, then we're in good shape, because 40% is the magic number to eek out victory.

---

by: DCCyclone @ Sun Jul 12, 2009 at 08:37:01 AM CDT