Abstract

This study investigates how high school students perceive ideological moderation in themselves and in potential political candidates. Individual measures of moderation are collected by self-identification. Candidate measures of moderation are measured by hypothetical election vignettes comparing two candidates, one with a package of extreme views, and the other with a package of moderate views. These issues include immigration, healthcare, foreign policy, gun control, and drug policy. The moderate vs extreme issue package was derived from analyzing a variety of public opinion surveys, and choosing positions that huge majorities support as the moderate position. I find that the majority of my sample is moderate, with significant numbers of extremists. However, I also find that no matter the survey respondent’s individual characteristics, the most important predictor of their support for moderate self-identification or moderate candidates is refusing to identify with one of the two major political parties.

Introduction

Political debate in the United States is often discussed in terms of the extremes: should we abandon capitalism on the one hand, or go back to a 1950s traditional family values on the other? In political science, a large amount of evidence shows that the majority of American voters are actually moderate, and it is the elites of our two parties that occupy the political extremes. Moreover, it seems that Moderates may not just be “in between” partisans but rather repersent a distinct set of preferences that blend pragmatism with incremental reform (Fowler et al. 20231).

If voters are disproportionately moderate, extremists could suffer a penalty when they run for office. This was seen in the study, Ideology and the US Congressional Vote. Which was about ideology vote choice, where researchers found that ideological positioning strongly predicts electoral outcomes and further reinforces the cost of extremism (Shor and Ragowski 20162). In other words, there is an electoral reward for moving towards the median voter, who typically lies in the middle of the American political spectrum.

There is a paradox, however. If there is a reward for electoral moderation, how do extremist candidates make it to the general election in the first place? There are many factors, but one of the most important is the institution of the primary. In primary elections, held months earlier, moderate candidates suffer a penalty since the people who participate (by voting or donating) are a small bloc of unrepresentative extremists. This was seen in the study, Who Punishes Extremist Nominees? Which found the despite the fact that extremist’s might win in the primary by energizing bases they may alienate swing voters in a General election (Hall and Thompson3).

This paper concentrates on the general election. It examines moderation in two respects: the individual and the candidate. I find that moderate candidates are very much advantaged electorally no matter the individual characteristics of the survey respondent. In other words, in my sample, there is very much a reward for moving to the middle.

Analysis

Data

One fine Friday afternoon, I sent out an anonymous survey to the entire Emery High School Student body, asking about their voter demographics and which they favored in two election hypothetical either extremist or moderate in each. Within two days, I collected an impressive 82 survey responses and began analyzing the data.

Party

I asked my respondents to identify themselves in terms of partisanship. Nearly half the sample are Democrats, and the rest are split between Republicans and neither party (Independents). This is important because party identification is known to be one of the most powerful influences in political behavior.

Forms response chart. Question title: Do you think of yourself as closer to the Republican or Democratic party?. Number of responses: 82 responses.

Gender

I asked people to self-identify their gender. Whilst looking at surveys during the research phase of my project, I noticed that recently men and women have diverged politically with men becoming more conservative and women becoming more liberal. A very small number preferred not to say, and only one said non-binary, so I dropped these from the analysis for simplicity.

Forms response chart. Question title: What is your gender?. Number of responses: 82 responses.

Ideology

I asked people to self identify their personal ideology on a 7 point scale, where 1 is an extreme liberal, and 7 is an extreme conservative, and 4 is the midpoint. The histogram below illustrates the responses.

Forms response chart. Question title: Place yourself on an ideology scale.. Number of responses: 82 responses.

I then calculated the ideology of the median survey respondent. The median is a 3, indicating that from my sample the respondent in the middle of the data between the two extremes is a moderately liberal person. I chose to do a median because of a potentially skewed distribution, and also because in politics the median voter determines what is the majority opinion. This is known as the median voter theorem which dictates that in majority-rule elections with voters arranged along an ideological spectrum the preferences of the median voter will win.

median(survey$ideo7)
## [1] 3

Here are the median ideology by party.

survey %>% 
  group_by(party) %>%
  summarize(median_ideo = median(ideo7))
## # A tibble: 3 × 2
##   party      median_ideo
##   <chr>            <dbl>
## 1 Democrat             3
## 2 Neither              4
## 3 Republican           5

As expected, the median Democrat is a moderate liberal (3), the median Republican is a moderate conservative (5), and the median person who doesn’t identify with a party is in the middle between them (4).

survey %>% 
  filter(!is.na(gender)) %>%
  group_by(gender) %>%
  summarize(median_ideo = median(ideo7))
## # A tibble: 2 × 2
##   gender median_ideo
##   <chr>        <dbl>
## 1 Female           3
## 2 Male             4

Interestingly, people who identify as female are slightly more liberal (3) than those who identify as male (4), or equivalently, people who identify as male are slightly more conservative than those who identify as female. This confirms existing findings that women and men have predictably different preferences.

Political Knowledge

I asked my sample about about their knowledge of political news. They were presented with a 1 to 7 scale, with 7 indicating the most political knowledge, and 1 indicating the least.

Likelihood of Voting

I asked my sample how likely they are to vote in a given election. They were given four choices, two indicating low likelihood of voting, and two indicating higher likelihood.

Moderates vs Extremists

Who’s a moderate?

I define a moderate as someone who identifies as 3,4, or 5 on the seven point scale. Extremists are those who identify as 1 or 2 (liberal extremist), or 6 or 7 (conservative extremist). This is a binary scale, where 1 is a moderate, and 0 is an extremist.

I start by calculating the number of moderates overall in my sample. In my sample I found that about two-thirds of respondents were moderates and a third were extremist according to my definition.

survey %>%
  tabyl(moderate) %>%
  adorn_pct_formatting(digits = 1)
##  moderate  n percent
##         0 27   34.2%
##         1 52   65.8%

By Party

Next I calculated the number of moderates within each party category. In this survey I found that those who identified with neither party had the highest percentages of moderates with Republican aligned respondents trailing behind in second and little after Democrats trailing behind in last.

# moderate by party
party.sum <- survey %>%
  group_by(party) %>%
  summarize(moderate=mean(moderate))

ggplot(party.sum, aes(x=party, y=moderate)) +
  geom_col() +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Moderates by Party", x="Party",y="Percent Moderates")

Gender

I then broke down moderates and extremist by Gender to see if certain demographics were more moderate or extreme. In this particular study I found that the men were more likely to identify as moderates than women.

gender.sum <- survey %>%
  filter(!is.na(gender)) %>%
  group_by(gender) %>%
  summarize(moderate=mean(moderate))

ggplot(gender.sum, aes(x=gender, y=moderate)) +
  geom_col() +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Moderates by Gender", x="Gender",y="Percent Moderates")

Vote intention

Next I took a look at the respondents likelihood of voting and how that affected if they were moderate or extremist. From this graph I summarized several key observations: The highest percent of moderates were found to be those who were “unlikely to vote,” the next highest being “likely to vote” and the two lowest being those who were “Certain not to vote” and “highly likely to vote.” I found it especially fitting that the extremist were most likely to be found in the two extreme vote intention practices. From this I conclude that people who identify as moderate may be less polarized and thereby less energized which can affect their ambivalence towards voting. It is important to not that moderates are completely disillusioned with politics to say certain not to vote but they are definitely not passionate enough to be “Highly likely” to vote. Extremists on the other hand are more likely to fall under two camps: “Highly” Likely and “Certain not to vote.” Political science literature does suggest strong partisanship increase with political participation suggesting that those who selected “Highly Likely” are probably more ideologically driven either to the left or right. On the flip side extremist may feel certain not to vote because they think that they’ll have little to no political influence contributing to their low voter efficacy.

voter.sum <- survey %>%
  group_by(vote.intend) %>%
  summarize(moderate=mean(moderate))

ggplot(voter.sum, aes(x=vote.intend, y=moderate)) +
  geom_col() +
  scale_x_discrete(limits = c("Certain not to vote", "Unlikely", 
                              "Lilkely", "Highly likely"),
                   labels = c("Certain not to vote", "Unlikely", 
                              "Lilkely"="Likely", "Highly likely")) +
  scale_y_continuous(label = percent_format(), limits = c(0, 1)) +
  labs(title="Moderates by Vote Intention", 
       x="Vote Intention", y="Percent Moderates")

Political knowledge

Political knowledge was measured on a 7 point scale, which I converted to a 3 point scale for simplicity (low, medium, high). The graph below suggests that those who reported they had high political knowledge were substantially more likely to be an extremist than those who reported they had low political knowledge, who were more likely to be moderate. Something that might explain this is that those who follow politics more closely are more likely to be exposed to partisan elites, media, and ideological arguments while moderates who are less informed may remain in the middle because they don’t know enough to take strong particular stances on any specific issues. This is what is sometimes known as the “knowledge-polarization effect” which says that political engagement amplify partisan divides.

know.sum <- survey %>%
  group_by(pol.know) %>%
  summarize(moderate=mean(moderate))

ggplot(know.sum, aes(x=pol.know, y=moderate)) +
  geom_col() +
  scale_x_discrete(labels = c("Low", "Medium","High"),
                   limits = c("Low", "Medium","High")) +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Moderates by Political Knowledge", 
       x="Political Knowledge",y="Percent Moderates")

Putting it Together: Regression Model of Individual Moderation

Next, I wanted to see which of the factors I discussed earlier was most important. I used a linear regression model to predict individual moderation as my dependent variable. My independent variables are a variety of individual characteristics: party, gender, vote intention, and political knowledge.

The only significant independent variable (indicated by a single star, which means 95% confidence that the null of no effect can be rejected) is identification with neither party, or being an Independent. Independents who identify with neither party are 36% more likely to be moderate. No other characteristic is significant when controlling for all of them at the same time.

indiv <- lm(moderate ~ party + gender + vote.intend + pol.know, data = survey)

modelsummary(dvnames(indiv),stars=T,
             title="Regression Model of Individual Moderation")
Regression Model of Individual Moderation
moderate
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
(Intercept) 1.054*
(0.472)
partyNeither 0.363*
(0.138)
partyRepublican 0.022
(0.144)
genderMale 0.053
(0.115)
vote.intendHighly likely -0.559
(0.466)
vote.intendLilkely -0.372
(0.476)
vote.intendUnlikely -0.314
(0.548)
pol.knowLow -0.017
(0.194)
pol.knowMedium -0.054
(0.121)
Num.Obs. 76
R2 0.185
R2 Adj. 0.087
AIC 106.8
BIC 130.1
Log.Lik. -43.412
RMSE 0.43

Candidate Hypotheticals

I then gave the respondents two trials of candidate hypothetical: one with a left wing extremist and the other with a right wing extremist.

I ask them who they would vote for. The first hypothetical scenario paired a left wing extremist vs a moderate and the second a right wing extremist vs a moderate.

Hypothetical #1

Below is hypothetical 1 Candidate A (moderate) and B (extremist) policy platforms.

Candidate A * Stance on immigration policy:

Enhanced border security, technology, staffing, and deportation of individuals with serious criminal records but also wants to expand visas for workers in industries facing shortages. Additionally want to create protection for Dreamers (undocumented immigrant youth who came to the U.S. as a minor). * Stance on Health Policy: Keep a mixed system of public and private insurance. Expand subsidies to make private insurance more affordable.  Incrementally expand Medicaid in states that have not yet adopted it. * Stance on Energy policy: Net zero emissions by 2050. Creation of diverse alternatives to fossil fuels such as solar, wind, hydro, and nuclear. Support workers in fossil industries. Create tax incentives for clean energy.

Candidate B * Stance on Immigration Policy: Eliminate most or all border enforcement. Allow for free movement across national borders. Grant amnesty and full citizenship to all undocumented immigrants. Dismantle ICE and other border patrol agencies. Establish immigration as a human right without conditions. * Stance on Health policy: Abolish private insurance entirely and transition to a single government run system within a very short timeline. Guaranteed healthcare for all residents regardless of citizenship or legal status. Ban profit-making in healthcare, including pharmaceutical companies and hospitals. * Stance on Energy Policy Ban all new drilling, fracking and coal mining right away. Shutdown fossil fuel plants within a very short timeline. Bring utilities, oil companies under direct public ownership. Reject transitional fuels like natural gas or nuclear power.

Hypothetical #2

Below is hypothetical 2 Candidate A (extremist) and B (moderate) policy platforms.

Candidate A: * Stance on Gun control: Oppose all background checks, waiting periods, or licensing. Allow concealed and open carry in most public spaces. Allow large capacity magazines and assault rifles. Calls for the dismantlement of gun regulatory agencies such as the ATF. * Stance on Drug policy:  All drugs including marijuana are illegal. Long and severe prison sentences for trafficking. Oppose all medicinal uses of controlled substances including marijuana. Militarized border enforcement to stop drug smuggling. * Stance on foreign policy: Instantaneous withdrawal from multilateral organizations such as the NATO, the UN, and other treaties. End most if not all foreign aid programs. Willing to use military force without negotiation or diplomacy. Make the U.S more isolationist and focus on domestic issues. 

Candidate B: Stance on Gun control: Recognize the 2nd amendment right to bear arms. Oppose total bans but also implement regulation on sales to individuals with mental illness and criminal records. Create safe storage requirements to prevent accidents and unauthorized access.  Stance on drug policy: Treat Marijuana and low risk drugs differently than opioids and methamphetamines. Support legalization of cannabis for medical purposes. Targeted law enforcement of drug trafficking. Tax and regulate low risk substances to fund health initiatives. Stance on Foreign policy: Support NATO and other long-standing alliances. Make sure partners pay their fair share in alliances. Prioritize diplomatic solutions, sanctions, and alliance pressure before military intervention. Support open trade but protects fair labor practices and environmental standards.

Vote Intention

I then asked the respondents whom of the candidates they would rather vote for. My hypothesis proved to be correct and in both hypotheticals the moderate candidate had overwhelming support from both moderates and extremists alike. From this I conclude that moderation is not just a middle ground but a statistically proven representation of public will and that being moderate confers a meanigful reward electorally.

# hypothetical candidate vote (moderate vs left extremist) by personal moderation
hyp.sum <- survey %>%
  group_by(moderate) %>%
  summarize(left.hyp.modvote=mean(left.hyp.modvote),
            right.hyp.modvote=mean(right.hyp.modvote))

ggplot(hyp.sum, aes(x=factor(moderate), y=left.hyp.modvote)) +
  geom_col() +
  scale_x_discrete(labels = c(`0` = "Extremist", `1` = "Moderate")) +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Hypothetical Candidate Vote (Moderate Candidate vs Right Extremist Candidate)",
       x="Individual",y="Hypothetical Vote Intention for Moderate Candidate")

ggplot(hyp.sum, aes(x=factor(moderate), y=right.hyp.modvote)) +
  geom_col() +
  scale_x_discrete(labels = c(`0` = "Extremist", `1` = "Moderate")) +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Hypothetical Candidate Vote (Moderate Candidate vs Right Extremist Candidate)",
       x="Individual",y="Hypothetical Vote Intention for Moderates Candidate")

Public Support

I then asked respondents which candidate they thought best represented the interests of the public overall. Yet again my hypothesis proved to be correct where the moderate candidate in both hypothetical were deemed by both moderates and extremist alike to represent the public overall. These finding suggest that the moderate candidates were perceived not just as compromise figures but also representative of the public overall. Both moderates and extremist are able to perceive the factual reality that moderate candidates represent the majority of people.

hyp.pub.sum <- survey %>%
  group_by(moderate) %>%
  summarize(left.hyp.public=mean(left.hyp.public),
            right.hyp.public=mean(right.hyp.public))

ggplot(hyp.pub.sum, aes(x=factor(moderate), y=left.hyp.public)) +
  geom_col() +
  scale_x_discrete(labels = c(`0` = "Extremist", `1` = "Moderate")) +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Moderate vs Left Extremist Candidate",
       x="Individual",y="Hypothetical Prediction of Public Support")

ggplot(hyp.pub.sum, aes(x=factor(moderate), y=right.hyp.public)) +
  geom_col() +
  scale_x_discrete(labels = c(`0` = "Extremist", `1` = "Moderate")) +
  scale_y_continuous(label = percent_format(),limits = c(0, 1)) +
  labs(title="Moderate vs Right Extremist Candidate",
       x="Individual",y="Hypothetical Prediction of Public Support")

Limitations

The sample I had access to is composed entirely of high school students at a single school in Houston. It is not representative of the entire United States. Since most of the students are not voting age, students may not be realistic about how they feel when they actually can vote. In addition, even though 82 is a large number, it is probably only a quarter of the whole school. The people who chose to answer may not be representative of the school as a whole.

Discussion

I asked 82 Emery High School students about their political attitudes. What I discovered is that two thirds of them identify personally as moderates. The most important factor predicting moderation was not identifying with a major political party.

However, even those who identify as extremist still favor voting for moderate candidates. Similarly, both moderate and extremist individuals also think (correctly) that the public favors moderate candidates. What we can conclude from this is that perhaps polarization is overstated, since even extremists can see the reality of the American political system is that it rewards moderates. Extremists can have their own point of view, but since they live in a large constitutional democratic republic, they have to compromise towards the center.

  1. Fowler, Anthony, Seth J. Hill, Jeffrey B. Lewis, Chris Tausanovitch, Lynn Vavreck, and Christopher Warshaw. 2023. “Moderates.” American Political Science Review 117 (2): 597–613. doi:10.1017/S0003055422000170.↩︎

  2. Shor, Boris, and Jon C. Rogowski. 2016. “Ideology and the US Congressional Vote.” Political Science Research and Methods. doi:10.1017/psrm.2016.23.↩︎

  3. Hall, Andrew B., and Daniel M. Thompson. 2018. “Who Punishes Extremist Nominees? Candidate Ideology and Turning Out the Base in US Elections.” Quarterly Journal of Political Science 13 (3): 279–318. doi:10.1561/100.000↩︎