1. ANALYSISIt was a nice, slow day for state polls yesterday as only twelve state polls were released. Barack Obama is leading in eleven of them. The one exception is Survey USA’s South Carolina poll, but there’s something very interesting about that poll’s crosstabs and a slow day gave me the time to dig down and notice it.
The topline for Survey USA’s South Carolina poll shows John McCain with a 14 point lead there. Looking at the crosstabs, though, McCain’s lead drops to +13.06 points when gender is considered, and it drops to +12.33 points when age is considered. These differences could simply be due to poll noise and rounding.
But when ethnicity crosstabs are considered, McCain’s lead drops to only +6.09 points in South Carolina. I’ve noticed this discrepancy in most of Survey USA’s polls for a few months now: Their crosstabs for ethnicity seem to throw their polls out of whack. Do they know something that other pollsters are unaware of? Or are they terribly off-the-mark with ethnic weighting this year? Most, if not all, of their state polls show more than 10% support for McCain among African American voters, well above the national average of around 5%.
Below are a few more states where Survey USA seems to show major discrepancies with ethnic weighting. The topline result in each poll was closest to the gender crosstab:
Alabama – Gender McCain +27.26, Age McCain +27.28,
Ethnicity McCain +20.84Georgia – Gender McCain +7.22, Age McCain +6.83,
Ethnicity McCain +1.66Illinois – Gender Obama +14.28, Age Obama +14.30,
Ethnicity Obama +19.42Maryland – Gender Obama +10.17, Age Obama +11.00,
Ethnicity Obama +18.22Don’t get me wrong, I believe Survey USA’s polls to be more accurate than most. But based on the information above, I firmly believe Survey USA has a serious weighting problem with their state polls this year in regards to ethnicity.
Meanwhile, two states change columns today. Quinnipiac University is showing Obama leading by 17 points in Wisconsin, moving that state on average into the Strong Obama column today. Quinnipiac is also showing an 11 point lead for Obama in Minnesota, but Minnesota moves one column to the right today due to older polls falling off and the ARG poll from last week with Obama leading by only one point in Minnesota staying in the mix.
The national polls continue to soar for Obama (see Figure 4c below), and it is becoming almost commonplace to see a double-digit lead for Obama in these polls. Also, Obama’s national popular vote lead is closing in fast on the 7 million mark (+6.8 million).
Note: Trend lines in graphs are only accurate up to five days in the future, so we need to view the trend lines with that in mind.2. NEW STATE POLLSScale:
Strong Obama (10% or greater)
Weak Obama (5% to 9%)
Lean Obama (0% to 4%)
Lean McCain (0% to 4%)
Weak McCain (5% to 9%)
Strong McCain (10% or greater)
Colorado
Obama 52, McCain 43 (Quinnipiac, 10/12 +/- 3.0, 1088 LV)
Colorado
Obama 47, McCain 43 (Suffolk University, 10/11 +/- 4.0, 600 LV)
Delaware
Obama 56, McCain 41 (Rasmussen, 10/10 +/- 4.5, 500 LV)
Michigan
Obama 54, McCain 38 (Quinnipiac, 10/12 +/- 3.0, 1043 LV)
Minnesota
Obama 51, McCain 40 (Quinnipiac, 10/12 +/- 3.1, 1019 LV)
Missouri
Obama 48, McCain 46 (Public Policy Polling, 10/12 +/- 3.1, 1009 LV)
North Carolina
Obama 49, McCain 46 (Public Policy Polling, 10/12 +/- 2.8, 1196 LV)
Ohio
Obama 50, McCain 45 (Survey USA, 10/13 +/- 4.2, 575 LV)
Pennsylvania
Obama 55, McCain 40 (Survey USA, 10/13 +/- 4.4, 516 LV)
Pennsylvania
Obama 51, McCain 38 (Muhlenberg College, 10/13 +/- 4.0, 602 LV)
South Carolina
Obama 41, McCain 55 (Survey USA, 10/13 +/- 4.2, 561 LV)
Wisconsin
Obama 54, McCain 37 (Quinnipiac, 10/12 +/- 2.8, 1201 LV)
3. ELECTORAL VOTE SNAPSHOT^ FIGURE 3a. This graph shows the daily tracking of Obama’s projected electoral votes from three sources. 270 is a majority of the electoral votes. The blue line shows daily tracking from 538.com, which projects what the outcome should be on election day. The red line shows daily tracking from Electoral-Vote.com, which shows a current snapshot of Obama’s electoral progress if the undecideds break evenly. The green line is my own projection, and it shows a current snapshot of Obama’s electoral progress if the undecideds break the same way as the swing states. A trend line is used (dashed green line) to show the average progress since May 19 and possible route for the next seven days.^ FIGURE 3b. This graph shows Obama’s green line and red line from the previous graph, and compares it to John Kerry’s 2004 progress (orange line) as shown from Electoral-Vote.com. (Both the red line and the orange line show data from Electoral-Vote.com)^ FIGURE 3c. This chart organizes all the state polls into groups categorized by how much a certain candidate is leading. “Strong” is used to indicate a candidate leading by 10 points or greater in a certain state. “Weak” is used to indicate a candidate leading between 5 and 9 points in a certain state. “Lean” is used to indicate a candidate leading between 0 and 4 points in a certain state. A total of electoral votes for each category is given at the bottom of each column.4. POPULAR VOTE AND NATIONAL POLL SNAPSHOT^ FIGURE 4a. I am projecting 136 million votes on election day, which is the total 2004 votes plus population growth and new voter registration totals for each state. A total projected popular vote is calculated for each candidate based on projected votes on election day this year multiplied by each state’s current poll results. This graph shows the daily tracking of the popular vote calculations.^ FIGURE 4b. Obama’s cumulative lead in the national daily tracking polls is illustrated in this graph. Showing it this way helps us see the Obama Campaign’s momentum shifts much easier over time.^ FIGURE 4c. This chart compares each polling agency’s latest polls with their averages for each candidate this year. The top section of the chart compares the Daily Tracking polls and my own projected popular vote for each candidate. The bottom section of the chart compares the National Tracking polls (usually released monthly). Each section is totaled and averaged, and an overall average is shown at the very bottom of the chart. The latest polls released for each polling agency are displayed in the center columns for each candidate, as well as Obama’s current lead in those polls (a negative number denotes Obama trailing McCain in that poll). The columns on the right show each polling agency’s averaged results since May 1.5. OBAMA’S SEVENTEEN^ FIGURE 5a. The “Obama Seventeen” is the group of 17 states where the Obama campaign is concentrating much of its efforts this year. This widget shows how the candidates will do if the undecideds nationwide break the same way as the swing states. Each of the seventeen swing states are listed below the indicator’s current percentage of the total 1700 possible. The closing figures on Intrade are shown next to each state, with a column to show the change from the previous day. The scale for these is 0 to 100, with 50 being a majority. The number of electoral votes for each state is shown in the left margin, and the most recent polling for each state is shown in the far right column.^ FIGURE 5b. This graph shows the daily tracking of “Obama’s Seventeen” swing states total from the Obama’s Seventeen widget above. To show Obama’s progress in the swing states, divide the latest closing figure by 1700.^ FIGURE 5c. This graph plots all of the state polling results for Obama’s Seventeen swing states, and displays a trend line to illustrate each candidate’s progress in these states as a group.6. WIGAND ELECTORAL AVERAGE^ FIGURE 6a. The Wigand Electoral Average is our own indicator, and shows how the candidates will do if the undecideds nationwide break the same way as the swing states. Each of the twelve swing states are listed below the indicator’s current score. The closing figures on Intrade are shown next to each state, with a column to show the change from the previous day. The scale for these is 0 to 100, with 50 being a majority.^ FIGURE 6b. This graph shows the daily tracking of the swing states total from the Wigand Electoral Average widget above. To show Obama’s progress in the swing states, divide the latest closing figure by 1200.^ FIGURE 6c. Wigand vs. Trading and Real Clear Politics Average. The Wigand Electoral Average uses both poll averages and trading (swing states), and this graph compares it to the national daily tracking average from Real Clear Politics, as well as the closing trading figures from Intrade (people buying shares with real money) and Rasmussen Markets (people using fake money).^ FIGURE 6d. This graph shows the daily tracking of each swing state’s closing figure on Intrade. The scale is 0 to 100 with 50 being a majority. I commonly refer to the range from 40.00 to 60.00 as “The Zone” because the states in The Zone are true swing states this year.The Wigand Electoral Average uses public polling results to determine base states for both parties and a short list of swing states, and combines the actual electoral votes from base states with a weighted Intrade calculation for electoral votes from swing states to derive an average on a scale of 0 to 100. On this scale, greater than 50.00 is considered a win for Barack Obama. An average of 50.10 would yield 270 electoral votes, and an average of 100.00 would yield 538 electoral votes. More Information 7. LINKS AND SOURCESYesterday’s WidgetFull Weekly TrackingSaturday Data DumpPast editions of THE MATHSources:
Pollster.com
FiveThirtyEight.com
Electoral-Vote.com
Intrade
Rasmussen Markets
RCP Average
RCP Latest Polls
3 Blue Dudes.