Deconstructing The Sun Senate Poll

-512Days -12Hours -9Minuts -53Seconds

I’ve spent most of the day buried in Girl Scout cookies and data from the six Senate primary polls that have been taken since November. In chronological order, an October Washington Post poll (Edwards led Van Hollen, 38-28), a November Baltimore Sun poll (Van Hollen led Edwards, 45-31), a January Gonzales poll (Van Hollen led Edwards, 38-36), a February Goucher College poll (Edwards led Van Hollen, 39-37), a March Gonzales poll (Van Hollen led Edwards, 42-41), and finally, last week’s Sun poll, in which Edwards has a 45-35 lead). I have geographical, racial, and gender cross-tabs for only the two Sun polls and the two Gonzales polls (no cross tabs are available for the Goucher poll). The good news is that with two polls from each pollster, some good comparisons can be made.

There’s all kinds of interesting things in these polls, but not wanting to overwhelm you (it took me all day to get this far, and understand it well enough to explain it) I’m going to focus on geography.

There’s certain things we know about the Senate primary. Chris Van Hollen is going to win his home jurisdiction of Montgomery County. Donna Edwards is going to win her home base in Prince George’s County. In the rural areas of the state, Van Hollen is generally favored. Where the election will be decided, it has been felt, is in the Baltimore region, specifically Baltimore City and Baltimore County. Neither candidate being from Baltimore, this is the likely battleground on which the election will be decided.

Early polling bore out these basic assumptions. The first poll, done by Opinion Works for the Baltimore Sun in November, gave Van Hollen a 14 point lead. In hindsight, much of that lead was attributable to a major ad blitz by Van Hollen, which gave him a lead both in Baltimore City and with women. But other than in the City, the basic geographical pattern was set as I’ve described it above. There are nine geographical breakdowns in the November poll cross-tabs, some overlapping.

Baltimore City: Van Hollen 44, Edwards 29
Baltimore County: Van Hollen 46, Edwards 24
Greater Baltimore (City, County, Anne Arundel, Carroll, Harford, Howard): Van Hollen 49, Edwards 25
Anne Arundel/Carroll/Harford/Howard extrapolation: Van Hollen 55, Edwards 23
Montgomery County: Van Hollen 75, Edwards 12
Prince George’s County: Edwards 63, Van Hollen 23
Greater Washington (Montgomery, Prince George’s, Charles, Howard, Frederick): Van Hollen 48, Edwards 37
Charles/Howard/Frederick extrapolation: Van Hollen 55, Edwards 26
Southern Maryland (Charles, Calvert, St. Mary’s): Van Hollen 49, Edwards 22
Eastern Shore: Van Hollen 25, Edwards 24
Western Maryland (Allegany, Garrett, Washington): Van Hollen 29, Edwards 16

Among white voters, Van Hollen led 57-21. Edwards led with black voters, 49-28. Women favored Van Hollen 42-34, while men supported Van Hollen by 51-25.

Other than Baltimore City and with women, the basic pattern was set.

The next poll was a Gonzales poll in January. Gonzales uses slightly different geographical terminology, but it’s essentially comparable for present purposes. There are five categories: Baltimore City, Montgomery County, Prince George’s County, Baltimore suburbs (Anne Arundel, Baltimore, Harford, and Howard counties), and the rest of the state.

Baltimore suburbs: Van Hollen 39, Edwards 31
Baltimore City: Edwards 50, Van Hollen 24
Prince George’s County: Edwards 68, Van Hollen 18
Montgomery County: Van Hollen 60, Edwards 24
Rest of the state: Van Hollen 49, Edwards 24
White voters: Van Hollen 55, Edwards 15
Black voters: Edwards 65, Van Hollen 15
Men: Van Hollen 45, Edwards 30
Women: Edwards 40, Van Hollen 33

Baltimore City moves toward Edwards, as do women. Otherwise, the pattern continues.

The next poll was another Gonzales poll, this one in March. Both sides have continued to advertise heavily in Baltimore, Edwards more so given the expenditures by Emily’s List. But as it’s the same pollster, we can make some direct comparisons, changes from the January to the March Gonzales polls in parentheses).

Baltimore suburbs: Van Hollen 50 (+11), Edwards 33 (+2)
Baltimore City: Edwards 54 (+4), Van Hollen 27 (+3)
Prince George’s County: Edwards 69 (+1), Van Hollen 12 (-6)
Montgomery County: Van Hollen 65 (+5), Edwards 22 (-2)
Rest of the state: Van Hollen 50(+1), Edwards 30 (+6)
White voters: Van Hollen 64 (+9), Edwards 17 (+2)
Black voters: Edwards 70 (+5), Van Hollen 16 (+1)
Men: Van Hollen 52 (+7), Edwards 32 (+2)
Women: Edwards 47 (+7), Van Hollen 36 (+3)

The hardening of the electorate into clearly delineated geographical, gender and racial camps continues.

The second Gonzales poll was in the field from February 29-March 4. The final poll, the second Sun poll, was in the field March 4-8. There was no significant event in this very brief time period that should have created any major difference between the two polls. But wow, were there differences. Change from November poll in parentheses. Comparisons with Gonzales at the end.

Baltimore City: Edwards 54(+25), Van Hollen 17 (-27) (Edwards unchanged, Van Hollen -10 from Gonzales)
Baltimore County: Edwards 48 (+24), Van Hollen 20 (-26)
Greater Baltimore: Edwards 48 (+23), Van Hollen 23 (-26) (Edwards +15, Van Hollen -27 from Gonzales)
Anne Arundel/Carroll/Harford/Howard extrapolation: Edwards 43 (+20) Van Hollen 32 (-23)

Montgomery County: Van Hollen 87 (+12), Edwards 11 (-1) (Edwards -11, Van Hollen +22 from Gonzales)
Prince George’s County: Edwards 66 (+3), Van Hollen 21 (-2) (Edwards -3, Van Hollen +9 from Gonzales)
Greater Washington: Van Hollen 47 (-1), Edwards 39 (+2)
Charles/Howard/Frederick extrapolation: Van Hollen 35 (-20), Edwards 33 (+7)
Southern Maryland: Edwards 49 (+27), Van Hollen 26 (-23)
Eastern Shore: Edwards 54 (+30), Van Hollen 17 (-8)
Western Maryland: Edwards 59 (+43), Van Hollen 8 (-21)

Overall average of SM/ES/WM: Edwards 54, Van Hollen 17 (Edwards +29, Van Hollen -33 from Gonzales)
White voters: Van Hollen 50 (-7), Edwards 29 (+8) (Van Hollen -14, Edwards +12 from Gonzales)
Black voters: Edwards 67 (+18), Van Hollen 16 (-12) (Edwards -3, Van Hollen unchanged from Gonzales)
Men: Van Hollen 42 (-9), Edwards 42 (+17) (Van Hollen -10, Edwards +10 from Gonzales)
Women: Edwards 46 (+12), Van Hollen 30 (-16) (Edwards -1, Van Hollen -6 from Gonzales)

So after three polls showing a very steady contest, suddenly, from one day to the next, there are major swings of white men in suburban Baltimore, Southern Maryland, the Eastern Shore and Western Maryland from Chris Van Hollen to Donna Edwards. Not little changes, but big, enormous 40-60 point net geographical swings, and 20-30 point swings among white and male voters in a matter of days.

Does that make any sense to anyone? Crazy as it sounds, it looks to me like some of the numbers got turned upside down, primarily in the Baltimore suburbs and rural areas of the state. Look at the bolded cross tabs, imagine them reversed, and ask yourself: which set of numbers makes the most sense?

I’ll leave the bottom line to Bill Murray. Who better to bring a 2:30 a.m. blog post home?

Good night, boys and girls. I stayed up to finish this for y’all and now it’s time for me to get some sleep.

4 thoughts on “Deconstructing The Sun Senate Poll

  1. groenemanblog

    It’s hard to figure out what’s going on without far more details about the survey methods (type of data collection, sample size, % of cell phones in the sample, number of callbacks, etc.). Another critical detail left out is whether the sample is registered voters, likely voters (and how determined?), or all adults. Certainly the segment numbers can change a great deal from poll to poll since the subsamples the percentages are based on are typically very small (except possibly for gender). Finally, I wouldn’t give much credence to the early poll results, as early polls in primary elections are notoriously unreliable. – Sid Groeneman

    Reply
  2. Cl

    You missed the forest for the trees. Undecided for the Sun poll is 31%. The highest out of any of the other polls.

    Reply

Leave a Reply