1. 1
Joseph White
Dr. Mohr
April 26, 2014
Final Research Paper
INTRODUCTION:
When dealing with politics and elections, researchers have always wondered why voters
to make a choice for one candidate over another. It may be their stance on a certain issue, or it
may be factors of similarity like gender, race, religion, age, economic upbringing, profession of
work, or other specifics.
This research investigation reviewed the statistics of the CBS poll dataset, which was
conducted in April of 2012, and found two variables that may have a relationship between them.
The voters were polled by CBS via a telephone interview, and asked several questions about the
2012 election. Several of the questions pertained to the presidential candidates and how they
would handle various issues in the government and society. One of the questions asked to voters
was whether they had a positive or negative opinion of President Obama. Voters could choose
between a positive, negative, or neutral opinion, as well as, not answer the question.
Another question asked to voters was whether or not President Obama relates to the
voter. There was not a stated definition of “relate to the voter”; however, for research purposes it
will defined as “sharing common goals, interests, qualities, and environments”. A question of
relationship between presidential public opinion and public relations was established for further
investigation.
The research question created was, “Is the voter’s opinion of Obama affected by whether
the voter can relate to him through personal factors?” Researchers believe this question most
accurately covers what they want to find out and will give clarity to a broad and often researched
topic.
This question is important to public administration and policy because administrative
decisions and policies are created daily that are influenced by politically appointed
representatives. Although administrative policies are enacted with the intent to be non-bias and
cover all racial, social, gender, age, and political parties equally; it has been shown throughout
our country’s existence that political science has played a hand, dating all the way back to the
Bill of Rights.
Through political elections, officials are given the authority to make appointments to
administrative agencies, which in time, could possibly reflect the views and interests of the
voting public. It has been determined through several studies in the past, that public opinion and
public policy are related. By placing certain officials in office, these representatives will create
laws and regulations that will affect how agencies conduct their business operations and
processes. For all these reasons stated, public opinion and political relations are incredibly
important to public administration and policy.
2. 2
LITERATURE REVIEW:
To begin the research, investigations were launched into several journal reviews. First
reviewed, was an article written by Seth Goldman. The article stated that the effect of being an
African American candidate played a great impact on the swaying opinions of white voters. It
stated that some Caucasian voters had a negative opinion of African American males, and used
descriptive words such as: lazy, criminal, and fatherless. (Goldman 2012) However, the main
point that the article argued through facts and surveys, it was discovered that President Obama
raised the positive opinion of African American males and himself, in the eyes of conservative
voters. He accomplished this feat through his campaign actions, his anti-negative stereotype
lifestyle, and his ability to build a bridge with conservative beliefs, including religion and family
structure. The author felt that President Obama was able to show through media outlets and
public events that he could relate to people outside of the African American race, thus raising the
public opinion of himself.
The second article that was examined, focused on religious affiliations of voters and how
they affect their attitudes toward politician’s decisions and stances on political policies. The
article written by Zeynep Taydas, Cigdem Kentmen, and Laura R. Olson stated that religion was
always been an underlying factor in the voters’ opinions of politicians which included Barack
Obama and George W. Bush. The article stated that based on surveys taken of voters, Barack
Obama was viewed positively by Catholics and non-evangelical protestants, and viewed
negatively by evangelical affiliated voters. (Taydas, Kentmen, Olson, 2012) The authors
explore the belief that religious variables reflect public opinion.
There has been some discussion and debate over President Obama’s association with the
Catholic faith. However, after reading the article written by Jason Horowitz, the author explains
how the practicing protestant president became associated with Catholicism. The president
worked with several parishes in the local government, to increase funding to the Catholic
Church. President Obama was already well known in the Chicago African American
communities, and became a larger figure by helping out the local parishes, before seeking his
own faith with Rev. Jeremiah Wright at Trinity United Church of Christ. (Horowitz 2014) Thus,
if President Obama can sympathize or relate with the religious views of a voter, his opinion
rating will increase.
I studied the article, written by Herb Boyd, in which the author investigates the
relationship that Barack Obama had with the left and right side media outlets, as well as, his
relationship with reporters and journalists of various races. The article expresses the belief that
the unknowledgeable voter of the same race may be drawn to Barack Obama due to his physical
appearance, smooth communication skills, and intelligent wit. However, the author claims the
opposite of knowledgeable voters of the same race, whom the author used examples: journalists,
news anchors, professors, and analysts. The author states that several African American
Publications claimed that “Barack Obama, in their opinion, was nothing more than an
opportunity for the ruling class to present its case and to continue its global dominance in the
guise of a black face” (Boyd, Black Scholar, 2008). The article continued to make claims and
give examples of how other minority races including Hispanics and Asians felt that the
democratic candidate and party did not show them enough time and respect of their issues.
Several Hispanic publications stated that they were not given access to political conferences and
3. 3
not invited to the Democratic National Convention. When Asian reporters were asked about
their opinion of Barack Obama, nothing more than a sly smile would come across their face.
This article continued to argue the idea that Barack Obama was more marketable and likeable to
the white majority voter, and not to the minority voter, especially when minority voter’s
knowledge of the political issues was increased.
In the final article, posted by Frank Newport, which stated that “In 2012, age is a
significant correlate of voting behavior -- as has been well established in previous elections -- but
basically only among non-Hispanic white voters. Nonwhite voters are so strong in their support
for Obama that age makes little difference.” (Newport, 2012). This article explains that through
Gallup Poll surveys, the author can see differences of opinion, based on age, at the age of 40 and
up, with a negative effect on Obama’s approval rating as the voter gets older. The article
explains that this “age factor” only applies to Caucasian voters, where voters of races still hold
President Obama in favorable opinion through every age bracket.
After reviewing the discussed articles, researchers have obtained a mild understanding of
what factors of similarity are affecting President Obama’s public opinion rating. His race and
religion play strong positive roles in his public opinion. President Obama’s gender has a neutral
effect on his ratings, and his age shows a negative effect only with Caucasian voters.
THEORETICAL AND STATISTICAL MODEL:
A research model was created that contains one dependent variable, one explanatory
variable, and three main control variables (which are broken into several dummy variables).
The dependent variable will be the voter’s opinion of Obama, which can be found in the
CBS DATASET category “AR”. Initial research places weight in the claim this dependent
variable is the strongest reason for why a voter chooses to select President Obama as their
president. There are several different ways that a voter or survey taker can answer this question.
They may answer the question by giving a positive opinion, negative opinion, or a no reply.
They may not have enough information to answer the question, or they may have enough, but
still might not be able to give a definitive answer. It will be described how this study will
operationalize this variable further on in my paper.
The explanatory variable that will be used is the category of relate to Obama, which can
be found in the CBS dataset “AX” slot. Like stated above in the hypothesis, researchers believe
that this variable will affect the dependent variable, and believe that as the explanatory variable
increases or decreases, so does the dependent variables percentage. If strong evidence of
correlation is shown between the two categories, political strategists will be able to save or shift
time, money, and manpower; as well as, obtain great success rates in elections.
Several other factors may play into the dependent variable and model. These factors are
labeled control variables, and could be used to explain away by relationship between my
dependent and explanatory variables. When people make decisions on which candidate to vote
for, many ideas go through their heads. Political issues, backgrounds, educational and
professional experience, all possibly come into play during this decision process. However,
researchers wanted to take this decision making process to an even more simplistic level.
4. 4
Similarities including race, religion, and age all could affect the voter’s judgment. These
factors also show similarity in relationship to the president. Many people argue that a voter’s
race or religion is the strongest reason why they vote for someone, so they feel that political
parties should endorse candidates that share similar features. This topic is hard to argue, based
on voter polls and success rates.
The control variables that will be used will be race, age group, political party and
religion. The control variable of race can play a strong factor in questioning the research
validity. President Obama is African American, which could influence the vote. This variable
was chosen because many critics feel that race plays a strong hand in influencing voters. This
factor could hurt the president’s opinion due to unrealistic bias that some voters hold. But at that
same effect, it may garner him positive opinion points that may not necessarily be warranted by
the voter of the same race.
The second control variable that will be used is religion. The president of the United
States is Protestant and he carries a strong following of voters that practice this same religious
faith. There are several other religious affiliations in the country, and they may affect the way
voters view the president in a positive or negative light. Since the president is Protestant, some
believe that he would get extra positive opinion points from voters, and this may affect the
dependent variable of Opinion Obama.
The third control variable that will be used is the age group of the voter. The president
falls into the age group of 45-64 years old. From research into the article reviews, it is stated by
authors that his approval rates and positive opinion rates go down as the voter grows in age.
However, he still has a strong following in his own age group.
The final variable that will be include in the analysis is the voter’s political party ID,
which could be either Democratic, Independent, or Republican. The president associates with
the Democratic Party; so research assumes he will have a high positive correlation with this
group, and a negative affect with the Republican Party. The outstanding question is how the
Independent Party relates to the Opinion Obama variable.
To give the reader a better comparison for the testing, analysis was also conducted on
how the survey taker’s opinion of Mitt Romney compared to the same control variables. This
will show the reader how each candidate measures up to each race, religion, and political group.
This method of testing with show the reader, or could show a campaign manager where each
person’s strengths and weaknesses lie, and may shed light on where they need to focus.
The data set that was used to conduct the model and obtain descriptive statistics was the
CBS News New York Times National Poll. This survey was taken on April 1, 2012. The survey
questioned 957 voters on a wide variety of topics. The questions pertained to their personal
characteristics including the factors above, their education, family, careers, income, and views on
the issues of the national government. The survey did not just ask questions about Barack
Obama, but also Mitt Romney, Rick Santorum, and other political figures in the national
spotlight. The company that completed this descriptive research gives a disclaimer on the initial
set of documents stating that the information provided is not representative of the national
population. Each question that was asked of voters gives them the opportunity to not answer the
question, and gives them several options to choose from with their answers. This increases the
5. 5
validity of the answers provided to me, and gives me a stronger chance of making a case for my
hypothesis. The dataset has varying levels of missing values, which go as low as eleven percent,
and as high into the fortieth percentile. These missing values were eliminated from the analysis
to compete the assessment and make a stronger case.
These variables will be used to test the hypothesis.
Opinion
Obama
Relate
to
Obama
Religion Age
group
Race Party ID
DV X1 Catholic=X2 X6 Black=X7 Democrat=X9
Jewish=X3 Other=X8 Independent=X10
Religion
Other=X4
Religion
None=X5
Statistical Model
DV= B0 + B1(X1) + B2(X2) + B3(X3) + B4(X4) + B5(X5) + B6(X6) + B7(X7) + B8(X8) +
B9(X9) + B10(X10)
HYPOTHESIS
If President Obama can relate to the voter as a person, then the voter’s opinion of
President Obama will increase.
NULL HYPOTHESIS
Whether or not Obama can relate to the voter as a person, the voters’ opinion of Obama
will not be affected.
DATA
Now, there were different coding schemes for each variable, and the decision was made
to recode these variables to meet the analysis measures, more appropriately. Each category had
several no replies, which were deleted, and several smaller groups within each major category,
which were separated with the SPSS software.
The data set that was used to conduct the model and obtain descriptive statistics was the
CBS News New York Times National Poll. This survey was taken on April 1, 2012. The survey
questioned 957 voters on a wide variety of topics. The questions pertained to their personal
characteristics including the factors above, their education, family, careers, income, and views on
the issues of the national government. The survey did not just ask questions about Barack
Obama, but also Mitt Romney, Rick Santorum, and other political figures in the national
spotlight. The company that completed this descriptive research gives a disclaimer on the initial
set of documents stating that the information provided is not representative of the national
6. 6
population. Each question that was asked of voters gives them the opportunity to not answer the
question, and gives them several options to choose from with their answers. This increases the
validity of the answers provided to the reader, and gives researchers a stronger chance of making
a case for the hypothesis. The dataset has varying levels of missing values, which go as low as
eleven percent, and as high into the fortieth percentile.
After deleting all of the no replies, a population of 261 surveyors was discovered. A
descriptive statistical analysis of the variables, are listed below.
Variable N Minimum Maximum Mean Std. Deviation
RelatetoO 261 0 1 .49 .501
Catholic 261 0 1 .21 .411
Jewish 261 0 1 .03 .183
Religion Other 261 0 1 .08 .267
Religion NONE 261 0 1 .14 .345
Protestant 261 0 1 .54 .500
Race WHITE 261 0 1 .90 .300
Race BLACK 261 0 1 .08 .267
Republican 261 0 1 .33 .472
Independent 261 0 1 .32 .468
Democrat 261 0 1 .34 .476
Age Group 261 1 4 3.0 .857
OPINIONROM 261 0 1 .49 .500
OPINIONOB 261 0 1 .56 .496
From what the reader can see, the vast majority of the variables range from 0 through 1
and have differing means and standard deviations. The dependent variable is OpinionOB with a
mean of .56. My key explanatory variable is RelatetoO with a 0-1 range, a mean of .49.
AGEGROUP has a range of 1-4, with a Mean of 3.00. The religions were broken into
Protestant, Catholic, Jewish, Religion Other, and none. Each mean reflects the percent of the
voters that answered this question, with a total of 100 percent. The races are broken up into
White (including Hispanic), Black, and Other. The Political Party IDs are shown as Republican
with a mean of .33, Independent with .32, and Democrat with .34.
7. 7
METHOD
Due to the categories that were chosen to use in this research investigation, a binary
regression analysis will need to be completed. This type of regression was chosen because
researchers cannot ethically place a rank order on any of the parties, races, or religions. Thus,
variables were dummy coded dichotomously, with the exception of age range.
ANALYSIS
The reader can see that for the OPINIONOBvariable, of the 261 total cases, 113 voters
had negative opinion of President Obama, as shown in the .00 y row. This equates to 43.2% of
the voters. Listed below those statistics, 148 voters had a positive opinion of President Obama in
the 1.00 column, for a percent total of 56.7%.
Unweighted Cases N Percent
Selected Cases
Included in
Analysis
261 100.0
Missing Cases 0 .0
Total 261 100.0
Unselected Cases 0 .0
Total 261 100.0
Observed Predicted
OPINIONOB Percentage
Correct.00 1.00
OPINIONOB
.00 0 113 .0
1.00 0 148 100.0
Overall Percentage 56.7
It is shown below that the intercept for the dependent variable is .336. This can be
interpreted as a starting point on the Y axis. Exp(B) shows a reading of 1.399, which gives an
odds ratio of the dependent variable. By dividing 148 cases by 113 cases, the reader can get a
ratio of 1.3 to 1.
B S.E. Wald df Sig. Exp(B)
Constant .336 .098 11.695 1 .001 1.399
Upon reviewing this chart listed below, the data shows that that there are differing beta
coefficients with each variable, and there are differing odds between each variable which have a
8. 8
positive opinion of Obama. The data shows that the explanatory variable has a positive
relationship at .212. There is a low significance between the dependent variable and the
explanatory variable at a .500 rating.
The only variable in this equation that is showing significance is the Catholic Variable
which is showing a .040 rating. The beta is negative at -.672 which shows that they thought
negatively of him, possibly due to backlash over legislation that the Vatican looked down on.
However, there is not any current time series data to help determine when the negative rating
began, or if there ever was a neutral/positive rating to start with.
Variable B S.E. Wald df Sig. Exp(B)
Age Group -.028 .152 .033 1 .856 .973
RelatetoO .212 .315 .454 1 .500 1.236
Catholic -.672 .327 4.221 1 .040 .511
Jewish -.109 .716 .023 1 .879 .896
Religion Other .215 .526 .166 1 .683 1.239
Religion NONE -.201 .396 .258 1 .612 .818
Race BLACK .376 .515 .532 1 .466 1.456
Independent -.119 .340 .122 1 .727 .888
Democrat -.347 .389 .797 1 .372 .707
Constant .533 .533 .999 1 .318 1.704
Chi-square df Sig.
Step 8.550 11 .663
Block 8.550 11 .663
Model 8.550 11 .663
Upon reviewing the model summary, it is displayed that the Cox and Snell R^2 and the
Nagelkerke R^2 show very low correlations, .03 and .040, that the created model is weak and
does not explain the variance between the variables to the degree that researchers would like to
see.
Step -2 Log
likelihood
Cox & Snell
R Square
Nagelkerke
R Square
9. 9
1 349.152a .030 .040
In addition to the hypothesis testing, a decision was made to run the same analysis for
the opposing party, Candidate Mitt Romney, to show the reader how these same control variables
measure up to a candidate of differing race, religion, and political ID.
As shown below, the reader can see that the OPINIONROM variable has a separation of
132 voters that viewed him negatively, and 129 voters that viewed him positively. This
separates the two categories very closely at 50.5% to 49.5%.
Observed Predicted
OPINIONROM Percentage
Correct.00 1.00
OPINIONROM
.00 132 0 100.0
1.00 129 0 .0
Overall Percentage 50.6
Listed below, the intercept for OPINIONROM is -.023 which would put him at a lower
starting point on the Y axis, in comparison to Barack Obama. The Exp(B) is shown to be .977,
which is very close to 1, for a decent odds ratio.
B S.E. Wald df Sig. Exp(B)
Constant -.023 .124 .034 1 .853 .977
10. 10
After analyzing the statistics listed above, findings show that Mitt Romney started off in
the hole already with a -.023 beta intercept. Looking how this variable correlates with the
variables in the above chart, the reader can see that the voters tend to have a more positive
opinion of the candidate as they age, with a beta of .015. Romney has a positive correlation
with Catholics, and Jewish religions, and negative correlation with the Religion other. Finally,
the reader can see that Romney is not doing well with Independent voters at a -.296 coefficient.
The Democrat variable of -.798 should be observed, however this is not surprising. The
significance is strong with a score of .040.
For research purposes, it is great to compare differing and opposing candidates because it
shows the reader how several groups do not favor either candidate, and that survey takers’
political opinion of the candidates is negative regardless of their groups. This comparison
negates the belief that if a group dislikes one candidate, they must favor the other candidate. It
brings up the mindset and analogy of “picking the lesser of two evils”, that several voters must
decide between.
CONCLUSION
Based on the information that was obtained from the testing, research was able to show
that the relationship between the dependent variable OPINIONOB, and the explanatory variable
RELATEtoO, have a positive relationship at .212 correlation. But this rating is not significantly
different from zero.
Research will fail to reject the null hypothesis of “Whether or not Obama can relate to the
voter as a person, the voters’ opinion of Obama will not be affected”. It can be seen that there is
a positive relationship between the dependent and explanatory variable, and the significance
rating is low at .500, which is not high enough to support the initial hypothesis of “If President
Obama can relate to the voter as a person, then the voter’s opinion of President Obama will
increase.”
Variable B S.E. Wald df Sig. Exp(B)
Age Group .015 .150 .010 1 .919 1.015
RelatetoO .383 .315 1.480 1 .224 1.466
Catholic .059 .325 .033 1 .855 1.061
Jewish .984 .754 1.704 1 .192 2.676
Religion Other -.110 .509 .047 1 .829 .896
Religion NONE .158 .395 .160 1 .689 1.171
Race BLACK -.087 .499 .031 1 .861 .916
Independent -.296 .336 .774 1 .379 .744
Democrat -.798 .389 4.200 1 .040 .450
Constant .053 .528 .010 1 .920 1.054
11. 11
Party Affiliation and Age are two control variables that showed strong possibilities of
affecting the DV and EV relationship, however, the results failed to show significance during the
President Obama comparison. The explanatory variable of relate ability to Obama may
encompass these control variables, as well as, the others listed. But at this time, research is
unable to provide evidence of this.
If research was continued, an attempt to perform a time series analysis would be
conducted, where researchers would continuously poll the same voters in the survey, month to
month. This may show clues to why certain control variables had positive, negative, neutral
correlations. It would also be interesting to see if the significance of the variable sets would
change by the month, and if so, by how much. There were several variables that were surprising,
such as the Catholic Religion and the Democratic Party. There are several possibilities as to why
they may be showing these results, however, without a longer and more in depth analysis, a
definite reason cannot be defined.
APPENDIX
Below is a listing of the new recoded variables
Opinion Obama
1=Favorable
0=Unfavorable
The study kept all the voters who had a favorable opinion of Obama the same with the
number 1. A decision to move the unfavorable opinion voters to the opposite end of the
spectrum with the number 0 was then made. This meant taking the original dataset voters who
gave a 3 and changing them to a 0. For the survey takers that did not answer, they were
eliminated from the data. Uncertain voters were also removed from the equations, because it
was felt that their uncertainness could affect the end results and skew the data in an untrue
direction.
Relate to Obama
1=Yes
0=No
An observation of this coding was made and a decision was made to make a few simple
adjustments that was believed would help prove the hypothesis. Researchers decided to make a
dichotomous relationship and kept the positive relations ability to stay as 1 and the lack of relate
ability as 0. No answer surveyors were eliminated. By performing this set up, the reader can
see how many of each side reported their views, and the percentage of people that actually feel
they relate to the president on a personal level.
Religion, and Dummy Recode
Protestant, 1/0
12. 12
Catholic, 1/0
Jewish, 1/0
Other, 1/0
None, 1/0
By performing this recode, a stronger weight is being put on the Protestant religion, for
reasons that were discussed above. No answers were eliminated, and the other religious choices
were merged into dichotomous relationships, to see how they relate to the dependent variable.
Age Range
1=18 to 29
2=29 to 44
3=45 to 64
4=Over 64
As for the Age Range grouping, an ordinal ranking group was created with each age from
18 to 29, 29 to 44, 45 to 64, and Over 64 as the final choice. Barack Obama falls into the 45 to
64 category, so it would thought that he would be viewed more positively in this group, however,
the readings show the opposite results, which are that as people age, they tend to favor their
conservative officials.
Race, and Dummy Recode
White 1/0
Black 1/0
Other 1/0
The original data set for the race variable had the category broken into several races,
including White/Hispanic, Black, Asian, Other, and Refused. It was decided to identify Barack
Obama as African American, and have recoded the grouping as listed above. Due to lack of
Asians voters in this study (0), their category was eliminated.
Party ID, with Dummy Recode
Republican 1/0
Independent 1/0
Democrat 1/0
For political identification, a coding for the variables was set up dichotomously as 1 or 0,
depending on what the voter chose for their answer. The assumptions, as stated earlier are that
President Obama will have a higher rating with Democrats versus Republicans. However, the
interest lies with the Independents that have given their answers.
14. 14
Work Cited
Goldman, Seth. (2012). Effects of the 2008 Obama Presidential Campaign on White Racial
Prejudice. Public Opinion Quarterly, Dec 2012, Volume 76, Issue 4, Pg 663.
Taydas, Z., Kentman, C., Olson, L. (2012) Faith matters: Religious Affiliation and public
Opinion about Barack Obama’s Foreign Police in the Greater Middle East. Volume 93,
Issue 5, Pg. 1218-1242.
Boyd, H. (2008) Obama and the Media. Black Scholar. Volume 38 Issue 4, Pg 11-18
Trumbull, M. (2012) With an eye on female vote, Obama, tots progress for women. Christian
Science Monitor. Pg 6
Newport, F. (2012) In presidential election, age is a factor among whites. Gallup Poll Briefing.
Pg 1-1
Harowitz, J. (2014) The Catholic Roots of Obama’s Activism. Retrieved April 11, 2014 from
http://www.nytimes.com/2014/03/23/us/the-catholic-roots-of-obamas-activism.html?_r=0