Monday, September 22, 2008

About Those Zogby Interactive Polls...(The McCain Bounce Revisited)

As I alluded to a week ago, I wanted to examine the post-convention bounce McCain enjoyed without the seemingly over-inflated numbers from Zogby's internet polling outfit. Why? Well again, I did this same thing with the post-clinch bounce Obama had throughout June. The first wave of Zogby state polling was seemingly overly supportive of Obama and really skewed the bounce the Illinois senator got in his favor. The underlying message is that the polling that Zogby has been doing through voluntary online surveys seems to drift with the political winds. That may not be a fair characterization since the first wave involved 34 states total, while the second and third waves have been pared down to ten or so battleground states. It may not be fair, but I'll stick with that description for the time being.

Well, what were the results (...of these now week and half old polls)?

Zogby Interactive -- Wave Three (Sept. 9-12)
StatePollMargin
Colorado
Zogby Interactive
+2
Florida
Zogby Interactive
+10.3
Michigan
Zogby Interactive
+5.7
Missouri
Zogby Interactive+6.1
Nevada
Zogby Interactive
+7.6
New Hampshire
Zogby Interactive
+6.3
New Mexico
Zogby Interactive
+1.5
North Carolina
Zogby Interactive+1.5
Ohio
Zogby Interactive
+5.9
Pennsylvania
Zogby Interactive
+4.8
Virginia
Zogby Interactive
+6.5

There is a lot of red on that list and surprisingly that red stretches into a trio of states, Colorado, New Hampshire and Pennsylvania, that are currently among the group of Obama toss ups. North Carolina, once again, has a result contrary to what has been witnessed throughout other polling recently. That has been the mark of Zogby's efforts in the state across the three iterations of surveys, though. Finally, Michigan of all states is apparently immune to the red drift in many of these states, turning in a solidly blue 5.7 point margin for Obama.

That's all well and good, but what effect did these poll have on the electoral college map? Nothing really. Not one of these states changed categories and that can be attributed largely to the number of polls that have been done in all these toss up states and New Mexico over the entire campaign. With an increasing amount of polling activity in the most competitive states, outliers are absorbed into the backend of the weighted average with little, or in this case no, effect. But while the electoral college was left unchanged, the assessment of the bounce the McCain-Palin ticket got out of the GOP convention did not. The picture without these polls from Zogby was a bit mixed. McCain gained, especially in traditionally Republian states that had been closer than history would indicate prior to the convention. However, his momentum in FHQ's toss up states was less pronounced. The Arizona senator had the averages moving in his direction in seven of the 11 toss up state, but Obama still had New Hampshire, Pennsylvania, Missouri and Indiana moving toward him since the pre-convention baseline was set on August 24.

[Click Map to Enlarge]

But New Hampshire and Pennsylvania were among the Zogby polls that showed a McCain lead. They were and both those polls had the effect of shifting the averages in the Arizona senator's direction. In Nevada and Ohio the bounces increased as well, moving from the "barely moved" range into the middle category of movement -- a one to two point change in the average over the convention period and its aftermath. All was not lost for Obama, however. Based on the strength of Zogby Michigan poll, the trend in the Wolverine state began to favor the Illinois senator. And the surprising Obama lead in North Carolina helped mute the effect McCain got out of his convention in North Carolina.

Given that we see such volatile changes based on a series of polls that may not be all that representative, why include them at all? A valid question. One that I'll answer with another question: What happens to the weighted averages when we back out all three waves of Zogby polling? If the effect is minimal, no harm, no foul, right? But if there is a decided shift toward one of the candidates, then the idea of the polls' inclusion may need to be revisited. Since the first two waves were seen as favorable to Obama, it could be hypothesized that McCain would stand to benefit from those polls being omitted.

Here is the electoral map from today:
[Click Map to Enlarge]

And here is how the map would look tomorrow if the Zogby polling were dropped from the averages (Huh? Tomorrow? How does that square with the map for today? This gives you a sneak peek for tomorrow's update. It includes the polling released today, all 18 polls from 15 states.):
[Click Map to Enlarge]

Yeah, that's it. North Carolina is the only state that changes categories based on the three waves of Zogby polling being dropped. But as we have seen, changing the color on one state on a map doesn't really provide us with the full picture.

The Electoral College Spectrum*
HI-4
(7)**
MA-12
(157)
CO-9***
(269/278)
ND-3
(160)
KS-6
(64)
VT-3
(10)
WA-11
(168)
NH-4***
(273/269)
WV-5
(157)
NE-5
(58)
RI-4
(14)
IA-7
(175)
NV-5
(278/265)
SC-8
(152)
AR-6
(53)
IL-21
(35)
MN-10
(185)
OH-20
(298/260)
AK-3
(144)
TN-11
(47)
CT-7
(42)
OR-7
(192)
VA-13
(311/240)
TX-34
(141)
KY-8
(36)
NY-31
(73)
NJ-15
(207)
IN-11
(322/227)
GA-15
(107)
AL-9
(28)
CA-55
(128)
WI-10
(217)
FL-27
(349/216)
SD-3
(92)
WY-3
(19)
MD-10
(138)
NM-5
(222)
MT-3
(352/189)
MS-6
(89)
OK-7
(16)
ME-4
(142)
MI-17
(239/316)
MO-11
(363/186)
LA-9
(83)
ID-4
(9)
DE-3
(145)
PA-21
(260/299)
NC-15
(175)
AZ-10
(74)
UT-5
(5)
*Follow the link for a detailed explanation on how to read the Electoral College Spectrum.
**The numbers in the parentheses refer to the number of electoral votes a candidate would have if he won all the states ranked prior to that state. If, for example, McCain won all the states up to and including Pennsylvania (all Obama's toss up states, but Michigan and New Mexico), he would have 299 electoral votes. Both candidates numbers are only totaled through their rival's toss up states. In those cases, Obama's number is on the left and McCain's is on the right in italics.

***
The line between Colorado and New Hampshire is the where Obama crosses (or McCain would cross) the 270 electoral vote threshold to win the presidential election. That line is referred to as the victory line. Both states are currently favoring Obama, thus the blue text in those two cells.

As we look at how the Electoral College Spectrum and Watch List would look with today's polls included, but the Zogby polls omitted, what we see is that there are not that many changes, but what changes there are, are almost wholly within Obama's coalition of states. Maine, Maryland and Massachusetts all saw decreases in their respective averages. Not enough, to warrant any worry from the Democrats or joy from Republicans, but noticeable changes for states that have remained largely unchanged over the summer. The biggest shake up is among the Obama lean states. Of those, only Washington kept its same position, close to being a strong Obama state. Iowa came out stronger without the Zogby polls in the equation, jumping New Jersey, Oregon and Minnesota. Without the June Zogby poll, Minnesota also inches even closer toward being competitive as measured by FHQ's weighted average.

The Watch List*
StateSwitch
Alaskafrom McCain leanto Strong McCain
Delawarefrom Strong Obama
to Obama lean
Massachusettsfrom Strong Obama
to Obama lean
Nevadafrom Toss Up McCainto Toss Up Obama
New Mexicofrom Obama leanto Toss Up Obama
North Carolinafrom McCain lean
to Toss Up McCain
Ohiofrom Toss Up McCain
to Toss Up Obama
South Carolinafrom McCain leanto Strong McCain
Texasfrom Strong McCainto McCain lean
Washingtonfrom Obama lean
to Strong Obama
Wisconsinfrom Obama leanto Toss Up Obama
*Weighted Average within a fraction of a point of changing categories.

And the Watch List? Virginia will be back on the list tomorrow -- with the Zogby polling included -- but without the Zogby date, the Old Dominion would be back off the list. Ohio is very close to sliding off the list without those polls as well. Georgia will come off the list tomorrow with or without these polls and South Carolina is slightly safer for McCain with that 1 point Obama lead in their June poll dropped from consideration. Finally, Massachusetts reenters the list as well after a prolonged absence. Well, the Bay state would be back on if the Zogby poll there weren't propping the state's average up to some degree.

In the end, the biggest surprise was that South Carolina didn't shift into safer McCain territory. North Carolina's shift was expected, but even that move was muted. The lesson here is that, yeah, Zogby adds some noise, but the overall effect is not that severe. Should we keep them, should we drop them? That, my friends, is certainly up for discussion.


Recent Posts:
The Electoral College Map (9/22/08)

Today's Agenda

The Electoral College Map (9/21/08)

2 comments:

SarahLawrenceScott said...

Drop them. Even if noise is negligible, or is a wash, it's still noise. We have both methodological and empirical evidence for distrusting these polls.

Josh Putnam said...

I'm still waiting on John Zogby for the counterpoint.

Honestly, their inclusion was based on a lack of polls during the period immediately following the July 4 holiday. And the inclusion of subsequent iterations was more legacy than anything. I really don't have a problem continuing to included them...in states where we have a significant amount of polling. But the inflation for Obama in states like Oklahoma and Tennessee, where there has been far less polling -- and for good reason -- the shift with/without the polls was significant.

And you can't included the polls in some instances and not in others. This is one black and white scenario we have in this race.

I'm inclined to remove them from the data if only because the reason for inclusion was tenuously based in the first place.

This one really is a no-brainer. But now that we've had these polls incorporated for a while, we can see what and where their effect is felt.