This memorandum is a full review of Magellan Strategies surveys from the 2012 election cycle. The memorandum goes into detail about the challenges of survey research for the cycle, and how well Magellan Strategies research held up to the final outcome.

1. December 6, 2012
MEMORANDUM
TO: INTERESTED PARTIES
FR: DAVID FLAHERTY, CEO
MAGELLAN STRATEGIES
RE: 2012 POST-ELECTION INTERNAL SURVEY ANALYSIS
This memorandum is a review of Magellan Strategies' internal surveys from the
2012 election cycle. This document shares insights to adjustments that we made
during the cycle to improve the accuracy of our surveys. The intent of this
memorandum is to provide political professionals and observers with information
about how Magellan Strategies conducted its survey research in the 2012
Presidential election.
THE CHALLENGES OF SURVEY RESEARCH IN 2012
Measuring voter opinion and candidate support was a challenge for many
researchers in the 2012 election cycle because younger and minority voters made
up a larger than expected portion of the overall voting population. To get it right,
researchers had to make correct assumptions about what the final voter turnout
demographics would be, and make sure their phone sample coverage of the final
voting population was sufficient. If one or both of these assumptions is incorrect,
then the survey will be inaccurate.
To understand what we are talking about when discussing demographics, we are
referring to the percentage of a survey's total interviews broken out by gender, age
or age group, race and geographic region (county, media market, etc.). Another
important demographic that is used by some researchers, but not all, is party
identification.
SURVEY SAMPLE COVERAGE CHALLENGES
Survey sample "coverage" refers to the percentage of the voting population that can
be reached by a phone call. For the survey to be accurate, minimum "coverage"
sample requirements need be met to ensure enough voters have an opportunity to
participate in the survey. For example, if I am conducting a survey in a legislative
district that contains 100,000 voters, I want to be sure the sample used sufficiently
covers the total population. If the sample covers 85% of the population, that
would be considered acceptable to most researchers.

2. Page 2 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
But if the 15% of the voting population that I am unable to contact by phone is
predominantly younger or minority voters and those subgroups are contributing
30% to the overall vote, then the survey will be inaccurate.
The decline of the household or "landline" telephone created additional challenges
for researchers in the 2012 election cycle. To maintain a sufficient level of sample
coverage, most researchers will add cell phone sample. A difficult decision that a
researcher needs to make if they are using landline and cell phone sample is what
the percentage each set of sample should be.
MAGELLAN STRATEGIES SURVEY RESEARCH LESSONS LEARNED
During the 2012 election cycle, Magellan Strategies regularly fielded surveys in all
Presidential battleground states starting in early May. All of our surveys at the start
of the cycle were fielded using an automated voice recorded method of
interviewing. Automated surveys have the benefit of a lower cost and a larger
number of interviews, and political professionals like the ability to “drill down”
among voter subgroups with more confidence than a survey with fewer interviews.
In early June we had concerns that the automated interviewing approach was not
accurately measuring minority and younger voter populations. This made us
skeptical of the results, so we began including cell phone and live landline
interviews of minority voters in almost all of our surveys.
MAGELLAN STRATEGIES APPROACH TO SURVEY DEMOGRAPHICS FOR 2012
The 2008 Presidential election produced a much larger than expected voter turnout
among younger voters, aged 18 to 34 in states across the country. Heading into the
2012 Presidential election, Magellan Strategies, like many other political
researchers believed it was unlikely that younger voters aged 18 to 34 would
turnout in the same percentages as they did in the 2008 election. However, we
believed these younger voters would vote in higher numbers than they did in the
2004 election. Turnout demographics from the 2010 election among younger
voters were higher than they were in the 2006 election. Taking this into account,
we set our 2012 turnout demographics closer to 2008 than 2004.

3. Page 3 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
MAGELLAN STRATEGIES SURVEY RESULTS BY OBAMA WIN MARGIN
Two weeks before the 2012 election, we fielded surveys in the battleground states
that merged automated voice recorded, cell phone, and live landline interviews of
likely voters. In the final week before the election, we fielded only automated
voice recorded surveys without any cell phone or landline live interviews. The
following tables show how well both sets of surveys performed against the actual
results. It should be noted the Presidential ballot test for each survey included the
qualified third party candidates in each state. To measure the accuracy of the
surveys, the following two tables show the actual vote percentage "Obama final
win margin" and the survey "Magellan Obama win margin.”
Oct. 23rd
- 24th
ALC* Survey Obama Win Margin Comparison
State
Obama
Actual
Final %
Romney
Actual
Final %
Obama
Final Win
Margin %
Magellan ALC*
Survey Obama
Win Margin %
Obama Final Win
Margin - Magellan
Obama Win
Margin Difference
Florida 50.0% 49.1% 0.9% 0.8% 0.1%
Pennsylvania 52.1% 46.7% 5.4% 4.9% 0.5%
Colorado 51.5% 46.1% 5.4% 4.5% 0.9%
Michigan 54.2% 44.7% 9.5% 10.8% 1.3%
Wisconsin 52.8% 45.9% 6.9% 4.3% 2.6%
Ohio 50.1% 48.1% 2.0% 4.8% 2.8%
Nevada 52.3% 45.6% 6.7% 3.6% 3.1%
Virginia 51.1% 47.2% 3.9% 0.8% 3.1%
North Carolina 48.3% 50.3% -2.0% 1.9% 3.9%
New Hampshire 52.0% 46.4% 5.6% 1.9% 3.7%
Iowa 51.9% 46.1% 5.8% -0.1% 5.9%
*ALC refers to the survey interview method that included automated voice recorded, live landline,
and live cell phone interviews.
Oct. 31st
- Nov. 1st
Automated Only Survey Obama Win Margin Comparison
State
Obama
Final %
Romney
Final %
Obama
Final Win
Margin
Magellan Auto
Survey Obama
Win Margin
Obama Final Win
Margin Magellan
Obama Win
Margin Difference
Colorado 51.5% 46.1% 5.4% 5.6% 0.2%
Ohio 50.1% 48.1% 2.0% 1.3% 0.7%
North Carolina 48.3% 50.3% -2.0% -3.2% 1.2%
New Hampshire 52.0% 46.4% 5.6% 3.5% 2.1%
Florida 50.0% 49.1% 0.9% -1.3% 2.2%
Iowa 51.9% 46.1% 5.8% 3.1% 2.7%
Nevada 52.3% 45.6% 6.7% 2.4% 4.3%
Wisconsin 52.8% 45.9% 6.9% 0.5% 6.4%
Virginia 51.1% 47.2% 3.9% -3.0% 6.9%

4. Page 4 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
SURVEY ACCURACY MEASURING OBAMA AND ROMNEY SUPPORT
The next set of tables looks at how well each survey measured support for Barack
Obama and Mitt Romney. As the tables indicate, the surveys were more accurate
in measuring Mitt Romney's support than Barack Obama's. We believe the “under
measurement” of Barack Obama’s support is due to the challenges of getting
younger and minority voters to participate in a survey. It is also clear surveys that
included cell phone and landline interviews of minority voters did a better job than
just automated interviews alone. We believe measuring Mitt Romney’s strength
was less challenging the measuring Barack Obama’s because Romney supporters
were older and less racially diverse.
Magellan Strategies Survey Obama Support Comparison Table
State
Obama
Final %
Magellan
10/23 ALC*
Survey
Obama %
Magellan
10/23 ALC
Survey
Difference
Magellan
10/31 Auto
Survey
Obama %
Magellan
10/31 Auto
Survey
Difference
Ohio 50.1% 50.1% 0.0% 48.6% 1.5%
North Carolina 48.3% 48.5% 0.2% 46.9% 1.4%
Michigan 54.2% 53.3% 0.9% N/A** N/A**
Florida 50.0% 47.8% 2.2% 47.7% 2.3%
Pennsylvania 52.1% 49.9% 2.2% N/A** N/A**
Colorado 51.5% 48.4% 3.1% 51.1% 0.4%
New Hampshire 52.0% 48.9% 3.1% 50.2% 1.8%
Wisconsin 52.8% 49.7% 3.1% 48.9% 3.9%
Nevada 52.3% 49.1% 3.2% 47.4% 4.9%
Virginia 51.1% 47.9% 3.2% 46.4% 4.7%
Iowa 51.9% 46.5% 5.4% 49.4% 2.5%
*ALC refers to the survey interview method that included automated voice recorded, live landline,
and live cell phone interviews. **“N/A” indicates a survey was not fielded in the final week prior to
the election. We did not think Michigan was competitive, and for Pennsylvania we were concerned
homes without power from Sandy would significantly impact survey response rates.
Magellan Strategies Survey Romney Support Comparison Table
State
Romney
Final %
Magellan
10/23 ALC
Survey
Romney %
Magellan
10/23 ALC
Survey
Difference
Magellan
10/31 Auto
Survey
Romney %
Magellan
10/31 Auto
Survey
Difference
Nevada 45.6% 45.5% 0.1% 45.0% 0.6%
Virginia 47.2% 47.1% 0.1% 51.0% 3.8%
Iowa 46.1% 46.6% 0.5% 46.3% 0.2%
Wisconsin 45.9% 45.4% 0.5% 48.4% 2.5%
New Hampshire 46.4% 47.0% 0.6% 46.7% 0.3%
Pennsylvania 46.7% 45.0% 1.7% N/A N/A
Florida 49.1% 47.0% 2.1% 49.0% 0.1%
Colorado 46.1% 43.9% 2.2% 45.5% 0.6%
Michigan 44.7% 42.5% 2.2% N/A N/A
Ohio 48.1% 45.3% 2.8% 47.3% 0.8%
North Carolina 50.3% 46.6% 3.7% 50.1% 0.2%

5. Page 5 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
MAGELLAN SURVEY EXIT POLL COMPARISON
The survey tables that follow compare the Magellan survey results to the 2012 exit
polling data by voter subgroup. To understand how to read the data in the tables,
the exit polling results for each candidate are in the "Exit" row and the Magellan
survey candidate results are in the "Mag" row. The "Difference" row indicates the
difference between the exit polling results and the Magellan survey results for the
voter subgroup. The comparison of the two different survey interview methods
shows the challenges of an automated survey to accurately measure minority voter
opinion.
Colorado Oct. 23rd - 24th Automated, Landline and Cellphone Survey
Total*^ Men Wom
White
Wom Ind Hisp 65+
Difference 3.1% 3% 2% 1% 1% 11% 9%
Obama (Exit Poll) 51.5% 51% 51% 46% 45% 75% 42%
Obama (Mag) 48.4% 48% 49% 45% 44% 64% 51%
Difference 2.2% 6% 1% 1% 11% 7% 10%
Romney (Exit Poll) 46.1% 46% 48% 53% 49% 23% 57%
Romney (Mag) 43.9% 40% 47% 52% 38% 30% 47%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 5.4%, the survey Obama win margin was 4.5% with 3.6% of
respondents undecided.
The Magellan survey interviewed 746n likely voters, for a margin of error of +/- 3.58%.
The exit polling data interviewed 1,029 voters, for a margin of error of +/- 3.05%.
Colorado Oct. 31st - Nov. 1st Automated Only Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind Hisp 65+
Difference 0.4% 4% 4% 10% 10% 24% 8%
Obama (Exit) 51.5% 51% 51% 46% 45% 74% 42%
Obama (Mag) 51.1% 47% 55% 56% 55% 52% 50%
Difference 0.6% 2% 4% 9% 13% 19% 8%
Romney (Exit) 46.1% 46% 48% 53% 49% 23% 57%
Romney (Mag) 45.5% 48% 44% 44% 36% 42% 49%
*The Total (Exit) column shows the final candidate vote percentages as reported on November 29th.
^ The actual Obama win margin was 5.4%, the survey Obama win margin was 5.6%, with 0.6% of
respondents undecided.
The Magellan survey interviewed 1,033n likely voters, for a margin of error of +/- 3.05%.
The exit polling data interviewed 1,029 voters, for a margin of error of +/- 3.05%.

6. Page 6 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
Florida Oct. 23rd - 24th Automated, Landline and Cellphone Survey
Total*^ Men Wom
White
Wom Ind Hisp Black 65+
Difference 2.2% 0% 3% 0% 0% 6% 9% 3%
Obama (Exit) 50.0% 46% 53% 41% 50% 60% 95% 41%
Obama (Mag) 47.8% 46% 50% 41% 50% 54% 86% 38%
Difference 2.1% 4% 0% 2% 9% 0% 5% 1%
Romney (Exit) 49.1% 52% 46% 58% 47% 39% 5% 58%
Romney (Mag) 47.0% 48% 46% 56% 38% 39% 10% 57%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 0.9%, the survey Obama win margin was 0.8% with 3.8% of
respondents undecided.
The Magellan survey interviewed 774n likely voters, for a margin of error of +/- 3.52%.
The exit polling data interviewed 4,225 voters, for a margin of error of +/- 1.51%.
Florida Oct. 31st - Nov. 1st Automated Only Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind Hisp Black 65+
Difference 2.3% 2% 5% 3% 6% 15% 25% 6%
Obama (Exit) 50.0% 46% 53% 41% 50% 60% 95% 41%
Obama (Mag) 47.7% 48% 48% 44% 44% 45% 70% 47%
Difference 0.1% 1% 2% 3% 6% 13% 20% 6%
Romney (Exit) 49.1% 52% 46% 58% 47% 39% 5% 58%
Romney (Mag) 49.0% 51% 48% 55% 41% 52% 25% 52%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 0.9%, the survey Romney win margin was 1.3% with 0.4% of
respondents undecided.
The Magellan survey interviewed 888n likely voters, for a margin of error of +/- 3.29%.
The exit polling data interviewed 4,225 voters, for a margin of error of +/- 1.51%.
The Florida October 24th
survey that included cellphone interviews and landline
interviews of African American and Hispanic voters predicted the overall outcome
very well with a win margin of 0.1%, but the final automated survey missed the
mark, predicting a close Romney win. We think this is a good example of how an
automated interview approach can have coverage and penetration problems
among younger and minority voters, thereby under-measuring Democrat candidate
strength.

7. Page 7 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
Ohio Oct 24th - 25th Automated, Landline and Cellphone Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind Black 65+
Difference 0% 5% 4% 0% 1% 11% 4%
Obama (Exit) 50.1% 45% 55% 46% 43% 96% 44%
Obama (Mag) 50.1% 50% 51% 46% 44% 85% 48%
Difference 2.8% 6% 1% 2% 6% 8% 8%
Romney (Exit) 48.1% 52% 44% 53% 53% 4% 55%
Romney (Mag) 45.3% 46% 45% 51% 47% 12% 47%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 2.0%, the survey Obama win margin was 1.6% with 4.4% of
respondents undecided.
The Magellan survey interviewed 713n likely voters, for a margin of error of +/- 3.67%.
The exit polling data interviewed 3,992 voters, for a margin of error of +/- 1.55%.
Ohio Oct. 31st – Nov. 1st Automated Only Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind Black 65+
Difference 1.5% 2% 5% 3% 7% 28% 5%
Obama (Exit) 50.1% 45% 55% 46% 43% 96% 44%
Obama (Mag) 48.6% 47% 50% 49% 36% 68% 49%
Difference 0.8% 3% 2% 5% 3% 23% 7%
Romney (Exit) 48.1% 52% 44% 53% 53% 4% 55%
Romney (Mag) 47.3% 49% 46% 48% 56% 27% 48%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 2.0%, the survey Obama win margin was 2.5% with 4.5% of
respondents undecided.
The Magellan survey interviewed 1,002n likely voters, for a margin of error of +/- 3.10%.
The exit polling data interviewed 3,992 voters, for a margin of error of +/- 1.55%.
Both Ohio surveys predicted the final outcome fairly well, with both being within
the margin of error. As in other surveys with larger minority populations, you can
see how the automated interview approach struggles to get a solid, accurate
measurement of African American voter opinion.

8. Page 8 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
Nevada Oct. 23rd - 24th Automated, Landline and Cellphone Survey
Total*^ Men Wom
White
Wom Ind 65+ Hisp Black
Difference 3.2% 5% 3% 2% 0% 3% 7% 33%
Obama (Exit) 52.3% 48% 57% 46% 43% 44% 71% 92%
Obama (Mag) 49.1% 43% 54% 48% 43% 47% 64% 59%
Difference 0.1% 1% 1% 3% 8% 6% 1% 25%
Romney (Exit) 45.6% 49% 41% 52% 50% 55% 24% 8%
Romney (Mag) 45.5% 50% 42% 49% 42% 49% 25% 33%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 6.7%, the survey Obama win margin was 3.6% with 2.8% of
respondents undecided.
The Magellan survey interviewed 754n likely voters, for a margin of error of +/- 3.57%.
The exit polling data interviewed 4,388 voters, for a margin of error of +/- 1.63%.
Nevada Oct. 31st - Nov. 1st Automated Only Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind 65+ Hisp Black
Difference 4.9% 5% 6% 2% 3% 7% 12% 17%
Obama (Exit) 52.3% 48% 57% 46% 43% 44% 71% 92%
Obama (Mag) 47.4% 43% 51% 44% 40% 51% 59% 75%
Difference 0.6% 4% 2% 1% 2% 7% 1% 5%
Romney (Exit) 45.6% 49% 41% 52% 50% 55% 24% 8%
Romney (Mag) 45.0% 53% 39% 53% 40% 48% 25% 13%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Obama win margin was 6.7%, the survey Obama win margin was 2.4% with 0.7% of
respondents undecided.
The Magellan survey interviewed 844n likely voters, for a margin of error of +/- 3.37%.
The exit polling data interviewed 4,388 voters, for a margin of error of +/- 1.63%.

9. Page 9 of 9
Magellan Strategies | 1685 Boxelder Street, Suite 300 | Louisville, CO 80027
www.MagellanStrategies.com | (303) 861-8585
North Carolina Oct. 22nd - 23rd Automated, Landline and Cellphone Survey
Total*^ Men Wom
White
Wom Ind 65+ Black
Difference 3.7% 2% 10% 14% 6% 11% 2%
Romney (Exit) 50.3% 54% 49% 67% 57% 64% 4%
Romney (Mag) 46.6% 56% 39% 53% 51% 53% 6%
Difference 0.2% 6% 5% 9% 5% 5% 5%
Obama (Exit) 48.3% 45% 51% 33% 42% 35% 96%
Obama (Mag) 48.5% 39% 56% 42% 37% 40% 91%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Romney win margin was 2.0%, the survey Obama win margin was 1.9% with 4.9% of
respondents undecided.
The Magellan survey interviewed 769n likely voters, for a margin of error of +/- 3.53%.
The exit polling data interviewed 4,341 voters, for a margin of error of +/- 1.49%.
North Carolina Oct. 31st - Nov. 1st Automated Only Survey vs. Exit Polls
Total*^ Men Wom
White
Wom Ind 65+ Black
Difference 0.2% 0% 2% 5% 3% 8% 8%
Romney (Exit) 50.3% 54% 49% 67% 57% 64% 4%
Romney (Mag) 50.1% 54% 47% 62% 54% 56% 12%
Difference 1.4% 2% 1% 4% 3% 6% 11%
Obama (Exit) 48.3% 45% 51% 33% 42% 35% 96%
Obama (Mag) 46.9% 43% 50% 37% 39% 41% 85%
*The Total (Exit) column shows the final candidate vote percentage as reported on November 29th.
^ The actual Romney win margin was 2.0%, the survey Romney win margin was 3.2% with 1.1% of
respondents undecided.
The Magellan survey interviewed 1,006n likely voters, for a margin of error of +/- 3.09%.
The exit polling data interviewed 4,341 voters, for a margin of error of +/- 1.49%.
CONCLUSION
We believe the 2012 election cycle was challenging for survey researchers because
of assumptions made that it was unlikely that younger voters would turnout in the
same percentages as they did in the 2008 election cycle. We also believe that the
difficulty of getting younger voters on the phone, even with significant cell phone
sample, contributed to lower sample coverage rates. We see this in our own
surveys that are detailed in this memorandum.
CONTACT MAGELLAN STRATEGIES
Any questions regarding this memorandum should be directed to David Flaherty at
303-861-8585 or by email at Dflaherty@magellanstrategies.com. This
memorandum can be downloaded on the front page of the Magellan Strategies
website at www.MagellanStrategies.com.