UNPACKING THE PPH CD2 POLL & WHY IT IS FAR OFF
Last week, Emily Cain released an “internal poll” claiming she was tied with Congressman Bruce Poliquin in the race for Maine's Second Congressional Seat. The last time Ms. Cain released an “internal poll” it was in October of 2014 – and she missed by 13 points, claiming she was up by 8 points, but losing by 5 points just days later.
Then just hours later, the Portland Press Herald released a poll also claiming the race was close, but giving Congressman Poliquin the lead. There are issues with this poll as well.
What we got with the PPH CD2 poll, however, is just bizarre.
1. Out of the 469-609 person sample, the CD2 portion is only reportedly 227 voters.
This sample size is not large enough to get an accurate reflection in ME-02, the largest geographic Congressional District east of the Mississippi River.
2. The geographic cross-tabs show two glaring issues, maybe three.
First is that Congressman Poliquin, who is strong in central Maine, only shows a small 3 point lead over Ms. Cain. Second, Congressman Poliquin shows a 16 point lead in Downeast Maine. Neither of these is accurate in a real geographic sense of Maine. Having seen plenty of data since the start of the 2014 election, it is clear Congressman Poliquin is doing very well in Central Maine. And, while I would not be surprised to see Congressman Poliquin with a lead in downeast Maine, I think this lead is a little bit exaggerated. Third, a possible flaw, what is the definition of northern Maine? Are we talking northern Penobscot and Piscataquis County plus Aroostook county? If so, why is the sample size nearly 40% of the total?
If I use a turf-cutter application, selecting voters from Bangor north statewide on an actual map of voters, and exclude Washington County with the expectation that they are in the “Downeast” segment of the poll, I end up at about 33%. These issues are likely the result of polling that relied too much on specific subsets of geographic regions, or poor definitions of central, Downeast and northern Maine by the polling firm. It's almost as if the pollster ended up with a sample overweight with more liberal voters in “central Maine”, underweight with liberal voters in “Downeast Maine” and, depending on what the pollster defines as “Northern Maine”, I'm betting on an overweight sample in either the Orono/Old Town region or the St. John Valley. 3. Age sample is also a mess.
Without even getting into the issue of the fact that older voters show up much more regularly at the polls, this CD2 poll misses the mark, skewing 6 points toward younger voters as a percentage of the CD2 population, not even accounting for the typical lower turnout.
PPH Poll= 18-49 year old voters 47% vs actual CD2 43.8%
PPH Poll= 50+ year old voters 53% vs actual CD2 56.2%
4. Education demographics way off
In the education level cross-tabs for CD2, the only segment Emily Cain actually wins among is post-graduate level. She wins it by a whopping 24 pts. (Cain 54% vs Poliquin 30%.) However, the CD2 sample is a massive 19% of those polled. U.S. Census Bureau data, however, shows that Maine reports only 8.95% of residents with a post-graduate education level.
In other words, in the only education demographic Emily Cain gains ground on Congressman Bruce Poliquin among, she is relying on a sample that is double the size of the voters who actually exist in that demographic. In the statewide cross-tabs for the Presidential race, it is even more pronounced, with Hillary Clinton's lead built on an oversampling of the post-graduate demographic by nearly 2.5 times the number of voters the U.S. Census bureau reports as actually existing. Among the voters in the post-graduate education demo, Clinton wins 59%-19%, gaining back serious ground she loses in other demographics.
But again, Clinton's big margin comes among 22.2% of those polled claiming post-graduate education vs. an 8.95% rate reported by U.S. Census.
To Conclude, after looking at the age and education demographic skews, I can't help but wonder if the samples were based heavily on certain pockets of the voting population, such as might be found in some college towns, like Orono.
And I stress, while I think this analysis clearly suggests that Republicans are in a stronger position than the poll would have readers believe, I am not suggesting that anything was done to intentionally make the poll reflect the announced results.
It just happens to look like voters in specific demographic groups who favor Cain and Clinton were over-represented in some important places.
It's too hard to get even a vague snapshot of a race through a sub-600 sample on a statewide level, or a sub-400 sample on a Congressional district level.
I'm not a betting man, but if I was, I'd be putting money on Republicans in ME-02.
- Jason Savage, Executive Director, Maine GOP