On Parameters Influencing Voters’ Preference

Dynamics In The Indian Context

Debanjan Manna1∗, Prabhu Dayal1†, Rishabh Agrawal1‡
12th April 2024

Abstract
In the realm of democratic governance, understanding the intricate dynamics of voter preferences is of
paramount importance for effective policy formulation & governance. As Abraham Lincoln famously
stated, democracy represents a ”government of the people, by the people, for the people”, underscoring
the pivotal role of voters in shaping the trajectory of governance through electoral processes. This study
embarks on a critical investigation into the factors influencing the electoral decision process of voters
within the complex socioeconomic landscape of India.

The paper delves into a multivariate analysis of influential factors, ranging from party affiliation &
candidate characteristics to policy positions, socioeconomic indicators, social influence, & media impact.
Each of these factors assumes a distinctive role in shaping voter preferences, attitudes, & ultimately, the
choices made in the electoral sphere.

By scrutinizing party affiliation, we aim to elucidate how individuals’ enduring allegiance to political
parties influences their voting decisions. A nuanced examination of candidate characteristics seeks to
discern the impact of personal qualities, qualifications, and leadership attributes on voter perceptions
and preferences. This includes analyzing policy positions to shed light on how voters evaluate candidates’
stances on key issues & the resonance of these positions with their own values & aspirations. Our study
delves into the socioeconomic determinants of voting behavior, exploring the influence of economic con-
ditions and demographic characteristics on voter choices. This involves meticulously analyzing the role of
social influence, including the impact of social networks, family, and community. We also investigate the
pervasive influence of media on voter perceptions and preferences, considering the implications of media
coverage, political advertising, and digital platforms in shaping electoral outcomes. Moreover, we explore
the positive externalities of the current voting system under the purview of the Constitution of India,
acknowledging the monumental effort involved in conducting elections for the most populous nation in
the world.

Apart from the empirical analysis, this study proposes a preference survey of voters aimed at gar-
nering quantitative insights into the aforementioned factors. Leveraging Statistical Survey design theory,
the survey will be designed to capture the nuances of voter behavior & preferences, providing valuable
data for analysis.

With the Indian Lok Sabha elections of 2024 on the horizon, our research will provide valuable in-
sights into the economic determinants of voter preferences, thereby contributing to the understanding
of electoral dynamics in the world’s largest democracy. It will offer policymakers & political strategists
actionable insights for effective governance & democratic representation.

Keywords: Party affiliation, Candidate characteristics, Policy positions, Socioeconomic indicators, So-
cial influence, Media impact, Demographic characteristics, Voter preferences, Positive externalities, Con-
stitution Of India, Statistical survey design theory, Electoral dynamics.
∗ Dept. of Aerospace Eng. & Economics, email: [email protected]
† Dept. of Civil Eng. & Economics, email: [email protected]
‡ Dept. of Material Sci. and Eng. & Economics, email: [email protected]

1
 1 The Problem
Elections are the most important avenue for the citizens of a democratic country to exercise their fun-
damental right to vote to ensure their participation and influence in the decision-making process in the
country. Hence, it becomes imperative for voters to make extremely well-informed choices at the voting
booth to ensure that the right candidate comes through who can potentially help the voter’s local con-
stituency and, in turn, help the country prosper the most.

The factors on which a voter in a population bases his/her opinion are paramount in determining who
becomes victorious in the electoral race. Through this research, we aim to understand & analyze
the parameters based on which voters elect a certain candidate.

Through our research, we target the use of mathematical methods to uncover hidden patterns and
analyze the dependence on the data obtained from our survey on various parameters to determine the
potential political leanings of the survey participants. We aim to understand a voter’s mentality and how
various factors, such as their family background, mode of news consumption, etc., affect their candidate
preference and political orientation.

Emphasizing Fundamental Rights, Fundamental Duties, and Dharmaarthashastra in our re-

search highlights the importance of moral and informed voting. This focus aims to guide young Indian
voters to consider personal gains, national interests, and ethical standards. By doing so, we nurture a
generation of voters committed to the principles that bolster India’s democratic fabric and development
goals, supporting the country’s journey toward becoming a developed nation by 2047 through conscien-
tious and value-driven electoral participation.

2 Motivation
Our motivation to work on this project stems from trying to gauge voters’ preferences, especially
considering the forthcoming 2024 Lok Sabha elections. Many factors, including socioeconomic status,
cultural background, and personal experiences, can influence political orientation, making it a complex
and multifaceted phenomenon. Setting up relationships between such complicated factors can be a highly
arduous task.

To perform a rigorous analysis of the data that we have obtained from our survey. Further, we
want to look to explore possible relationships between different variables, which might help us determine
a person’s political orientation and also help us potentially unearth how an individual’s responses to
questions asked in our survey would indicate their choice of candidate at the voting booth .

3 Objective
Our objective is three folds:

1. Ascertaining the voting parameters influencing the contemporary electoral behaviors of the
GeneZ students of IIT Kanpur.
2. Finding the political spectrum distribution of the sample (Far Left, Center Left, Center, Center
Right & Far Right)

3. Exploring the applicability of Dharmaarthashastra in the process.

4 Literature Review
The dynamics of voters’ preferences have been a topic of study for a long time, and multiple studies have
been conducted in this domain in different contexts worldwide.

2
Our primary inspiration for this research comes from the work of Kulachai, et. al.[1], which pro-
vides a comprehensive literature review of the factors influencing voting decisions. The authors identify
several key factors, including socioeconomic status, political ideology, and candidate characteristics. They
also highlight the importance of understanding the context-specific factors influencing voting decisions
in different countries. This paper has largely inspired us to explore the potential relationships between
different variables in determining political orientation. We wanted to perform our analysis to verify the
hypothesis put forward by the authors in this article.

We further explored the paper by Stephanie Tawa Lama-Rewal[2], which examines the scientific
& political debates surrounding election studies in India. The author argues that election studies in
India have been shaped by scientific and political considerations, with scholars and politicians seeking to
understand the electoral process and its implications for Indian democracy.

Roberto (2017)[3] & OOSGA (2024)[4] explored the influence of social media on voting decisions
in India, emphasizing the significant role social media plays in shaping public opinion and political dis-
course, and this study provided us with the motivation to add a question requiring participants to fill
in their preferred mode of news consumption. It turns out that most of the Gen-Z population relies
on social media to get their news, and hence, this medium becomes of utmost importance in trying to
analyze what shaped public opinion in this day and age.

5 On Dataset
To reduce the omitted variable biases and make sure that the sample is sufficiently homogeneous, the
data is extracted from IIT Kanpur Gen Z1 (Those born after 1996) students only to gather
valuable insights and draw potential conclusions on the political orientation of each person in our
dataset based on certain assumptions & parameters.
Further our study adopts a gender-agnostic approach, thereby refraining from introducing gender-
based controls.

• Population: GenZ IITK Current Students.

There are total 4089 Undergraduates(UGs), 1897 Postgraduates(PGs) & 2250 Doc-
toral Students a . Number of GenZ students will be atleast (4089 + 1897 + 214) = 6200

So, Population Size ≥ 6200 (This is an approximate calculation to find the lower
bound)
a NIRF 2023

5.1 Survey Design

We designed a survey based on [5] so as to accurately understand the preference set of the target popu-
lation.
We have conducted two surveys call them Survey1 & Survey2 (henceforth : S1 & S2) with two different
goals in mind.

The elements out of our sample (ie Population − Sample) are similar to the elements inside the
sample in all respect. IITK students often tend to have a similar background (same kind of motivations,
aspirational class family background and a hunger to learn and achieve more in life; from our practical
experiences we have observed them to behave in an often predictably identical way)
All IITK students are connected equally (all of them have smart phones and access to stable highspeed
internet connections from the institutes). Provided that we conducted our survey in google forms
propagated through WhatsApp with personal calling and reminders
1 Pew Research Center What We Know About Gen Z So Far

3
 Figure 1: Population vs. Sample

Survey1

• Aim of the S1: Understading voting preferences of the given population

• We consider a sample data: Our sample size is 134.
• Mode of Survey: Google Form filled through WhatsApp often by setting personal re-
minders.

Survey2

• Aim of the S2: Understanding Poltical Leaning of the population

• The sample size has been restricted 134. So as to make it comparable with S1
• Mode of Survey: Same as S1

One important point to note is that we followed a Simple Random Sampling with replacement SR-
SWR. Therefore a strict inference about the population need not be accurate. With Replacement part
has been introduced because of the technical constraints of google form where if you maintain it anony-
mous then people can fill the form multiple times.

Questions Asked In the Survey

• In S1, we asked the following questions to our target group to capture:

1. Impact of Voter’s Family, Candidates’ religion & social identity (like caste, familiy lineage
and other relevant markers)

S1.Q1 How much does your family’s party affiliation influence your vote?
S1.Q6 Does religion of a candidate influence your choice?
S1.Q8 Does social identity of the candidate influence your choice?

2. Impact of traditional media & traditional political campaigning

S1.Q4 How much do you think the media influence other people’s voting preference ?
S1.Q7 How much does media influence your political views?
S1.Q9 What’s your preferred news source?
S1.Q12 How much do political campaigns & rallies influence your opinion of candidates and
parties?

3. Impact on features desired in the candidates by the voters

4
 S1.Q3 Qualities You Look for in Candidates
S1.Q10 How Important for you is a candidate’s track record for delivering on their promise &
ensuring accountability?
S1.Q11 How much does a candidate’s perceived corruption or ethical conduct influence your
vote?
4. Influence of ”issues”
S1.Q13 How much does the Local issues influence your vote ? (Considering you are voting in the
Lok-Sabha Election)
S1.Q14 How much does the representation of issues related to youth influence your vote ?
5. Checking a particular assertion of [1]
S1.Q2 What do you perceive your personality to be like?
S1.Q5 When voting for a candidate, which of these traits do you prioritise?
6. Participation in the upcoming Loksabha Elections
S1.Q1 Will you vote in the next Lok-Sabha elections?

• In S2 we asked the following 10 questions with only one goal in mind :

1. To decide the degree of Right lean of the voter:

S2.Q1 What do you think your political orientation is?

S2.Q2 Do you think populism should focus on egalitarian ideals?
S2.Q3 Should politics be open to progressive reforms?
S2.Q4 Do you think that the government should be secular in its conduct?
S2.Q5 What is your stance on popular sovereignty being essential for national identity?
S2.Q6 Do you think religion should guide government policies in our democracy?
S2.Q7 Should the government emphasize equality and progress over tradition?
S2.Q8 What is your opinion on the Government’s Role in social programs?
S2.Q9 Do you think that the National Identity should be based on shared culture and customs?
S2.Q10 What is your opinion on the government increasing taxes on the wealthy to redistribute wealth?
S2.Q11 What are your thoughts on Economic Policy prioritizing low taxes and less business regula-
tions?

Options for question S2.Q1:

• Center
• Center Left
• Center Right
• Far Left
• Far Right
Options for questions S2.Q2 - S2.Q11:

• Strongly Disagree
• Somewhat Disagree
• Neutral
• Somewhat Agree
• Strongly Agree
A detailed explanations for each questions is provided in the Results section of the paper.

5
 6 Results
6.1 Inferences
In this section we will mention the inferences that we can make from the Dataset of the two surveys
that we conducted (S1 & S2) & will simultaneously propose the dynamics/reasons behind oberving those
inferences, based on the contemporary political scenario and the nature of the cohort from which the data
is collected.

1. Impact of voter’s family, candidates’ religion & candidates’ social identity on the voter:
Here we would refer to responses in S1.Q1, S1.Q6 & S1.Q8 to make the inferences

(a) on the voter (b) on the voter

(c) on the voter

Figure 2: Raw Data: Data collected in the survey

Now we would check the dependence among the above three varaibles considered in pairs. For this
we use χ2 test to decide whether the variables (considered pair wise) are independent or not ; and
if they are note independent then we use the Cramer’s V value value to measure the degree of
associations of those variables.

H0 : The two variables are independent

We select a significance level of α = 0.05.

If: p-value ≤ α =⇒ Reject H0 =⇒ Variable are Dependent
else: Fail to reject H0 =⇒ V

6
(a) Family’s influence vs. the influence of a (b) Family’s influence vs. the influence of a
candidate’s religion on the voter candidate’s social identity on the voter

(c) Influence of candidate’s social identity vs

candidate’s religion on the voters

Figure 3: Frequency Table: understanding the relationships between the three variables taken two at a
time

7
 If we Reject H0 =⇒ There are not enough evidence to suggest that the variables are independent
=⇒ Calculate Cramer’s V value to evaluate the degree of association between the two variables.

if: Cramer’s V ≤ 0.1 =⇒ Small association

else if: 0.1 < Cramer’s V ≤ 0.3 =⇒ Moderate association
else 0.3 < Cramer’s V =⇒ Strong association
Pair of Variables χ2 stat P-value Hypothesis Testing Cramer’s V
Influence of Family 25.526 0.061 Fail to Reject H0 (p-val > α) −
vs Candidate’s
Religion
Influence of Family 28.801 0.025 Reject H0 (p-val ≤ α) 0.156 (moderate association)
vs Candidate’s
Social Identity
Influence of 50.561 0.0 Reject H0 (p-val ≤ α) 0.257 (moderate association)
Candidate’s
Social Identity
vs Candidate’s
Religion
So we can infer that
(a) Family of the voter has NO impact on voter’s preference on candidate’s religion:
This could be explained by the fact that families from wheere our cohort belongs are fairly
secular in their approach towards electoral candidacy. And family doesnot, ingeneral enforce
a strong bias on the voter to elect a candidate from a particular religion.
(b) Family has a CONSIDERABLE impact on voter’s preference towards the candi-
date’s social identity:
This result is quite counterintuitive. Even though families don’t tend to enforce any religion
bias on the candidate on the voter’s behalf; but the they impart quite a strong bias on
the voter’s preference towards a candidate from a particular social background
(caste, familiy lineage and other relvant markers).
It can be explained from the fact that a large respondents from or sample cohort are from Uttar
Pradesh, Bihar, Madhya Pradesh, Rajasthan and other such Hindi heartland states; where,
historically speaking, there has been an overwhelming presence of caste based politics2 .
Our findings reveal that caste politics continues to exert its influence within our
cohort.
(c) A voter’s preference towards religion of the candidate & towards the candidate’s
social identity is SIGNIFICANTLY correlated
Its very natural for those with strong religious bias to have an explainable preference for social
markers of the candidates . Therefore this result is as per our expectations.
2. Impact of Media and traditional political campaigning

To make relevant inference we refer to response of questions S1.Q4, S1.Q7, S1.Q9 & S1.Q12
we conduct χ2 test to check the dependence of these two variables
Pair of Variables χ2 stat P-value Hypothesis Testing Cramer’s V
Influence of Media 41.627 0.0 Reject H0 (p-val ≤ α) 0.222 (moderate association)
on the Other Vot-
ers vs the Voter
itself

(a) There is a SIGNIFICANT association between the influence of media on the voter
& the perceived influence on other voters:
As is visible from the plot5a & 5b people tend to underweigh their own media influence and
expect the media to have a strong impact on other population, which is quite unique.
2 wiki: Caste Politics In India

8
(a) Media Influence on the respondent itself (b) Perceived media influence on other voters

Figure 4: Raw Data: Media’s influence on the respondent vs the influence on other voters as
perceived by the respondent

(a) Frequency plot 1 (b) Frequency plot 2

Figure 5: Frequency Plot: Media’s influence on the respondent vs the influence on other voters
as perceived by the respondent

(b) Increased consumption of Internet based Media:

From 6 we infer that an overwhelming number(67.9%) of folks in our cohort consumes Social
media (exp. Facebook, Twitter, Youtube). Followed by Newpapers & News websites.
Only 32.8% refers to regular TV based news channels when consuming news. This suggest
an increased appetite for internet based media among GenZ; which is quite expected as GenZs
are the first truly Digital Natives
(c) Almost no impact of ”Traditional” political campaigning:
From figure 14 we can decipher that ≈ 90% of our cohort had moderate to NO effect of
”traditional campagining”. By traditional campaigning we mean political rallies, speeches &
processions.

3. Impact on Qualities desired in candidates by the Voters:

We refer to questions S1.Q3, S1.Q10, S1.11 and S1.Q5 & S1.Q2
From the figures 8, 9a & 9b we can make the following inferences:
(a) Integrity & honesty, Ability to lead, Future vision are the top 3 qualities that are being
preferred by our cohort. While Ability to work across the party lines is the least preffered
quality. This result is in line with [1] .
A point to note is that ≈ 40% of the respondents do care about environment.

9
 Figure 6: Preferred media of the respondent (voter)

Figure 7: Impact on Political Campaign on the voter

From this we can argue that our cohort seeks a candidate with honesty, strong leadership
, future vision, transparency, accountability, and good academic background.
(b) We have observed that Integrity & honesty are strongly preffered by the electorate. S1.Q11
answers the degree upto which people prefers Candidate’s ethics.
As one can observe from fig. 9b that a whooping ≈ 77% respondents gives a score of 4 → 5;
thereby underscoring the importance of ethics in the candidates.
(c) One can observe in 8 that 63.4% prefers Transparency & Accountability and 49.3% finds
Track records & experience to be important.
S1.Q10 intended to measure the same. From 9a we found out that ≈ 80% rated it 4 → 5

4. Checking an assertion of Kulachai, et. al [1]

The paper[1] suggested that an ”Introvert” person prefers ”Reserved & Thougthful” candidate
and an ”Extrovert” person prefers ”Charismatic & Assertive” candidate.
To address this particular question we use the response of S1.Q2 & S1.Q5

10
 Figure 8: Qualities in Candidates that are preferred by the voters

(a) Influence of Candidates’ track record on (b) Influence of candidate’s Ethical conduct &
delivery & accountability on the voter ccorruptions

Figure 9: Raw Data: As collected in the Survey for these three variables

11
 We have mapped voter’s personality: Introvert→ 0, Extrovert→ 1 & Ambivert→ 2
and Candidates’ nature: Charismatic & Assertive → 0 and Reserved & Thoughtful → 1

(a) Personality of Voters (b) ’Nature’ of Candidates

Figure 10: Raw Data: Voter’s Personality and the Candidates’ Nature

Figure 11: Voter’s Personality vs Candidates’ Nature

From the fig. 10 and 11 we infer the following:

n(”Reserved & Thoughtful”)
n(”Charisma & Assertive”)
vs Introvert Extrovert Ambivert
Personality of Voters
2.273 1.5 2.609

Table 1: Calculated ratios

5. Impact of particular type of issues

We measured the impact of ”Local issues” in a national level election (Lok Sabha election in
India). Further we also measured how the issues related to youth impact our cohort’s voting
preference.
By ”local issues” we meant issues pertaining to a particular constituency such as building a hospital,
providing proper public services & amenities, providing of local schooling etc.
By ”Youth related issues” we meant political empowerment of youths and meeting their aspirations
(proper job market, transparency, higher education facilities etc).
For this we refer to responses of S1.Q13 & S1.Q14

12
 (a) Impact of ”Local Issues” (b) Impact of ”Youth Related Issues”

Figure 12: Impact of Various issues

Figure 13: Impact of ”Youth Issues” vs ”Local Issues”

From fig. 13 & 14 we can infer the following:

(a) Local Issues do matter in a national level election

This may sound quite contrarian; but even among our folks (where a good number of them
come from a economic middle class background) we observe a large priority for ”Local issues”
while voting.
(b) From figure 12b, as expected, we infer that issues related youth do worries a lot of GenZ.
≈ 80% folks have rated it from 3 → 5

6. Expected participation in the upcoming Lok Sabaha 2024 election:

≈ 80% of the respondents are intending to vote this Lok-Sabha election 2024. Refer fig. 14

7. Ideology Preference of Individual (we will be analyzing S2 responses)

Exploring ideologies using left-right leaning offers an understanding of the diverse political land-
scape globally. The ideas of left-leaning and right-leaning ideologies come from the French Revo-
lution, which encapsulates the fundamental ideology divide between hierarchy and social equality.
There is a segregation of ideology between left-wing and right-wing, where left-wing properties
are briefly defined as emphasizing more social equity and government intervention. Conversely,
the right wing prioritizes individual liberty and economic freedom, providing a framework
for analyzing political behavior and party systems.

13
 Figure 14: Number of respondents who are going to vote

Considering the case of India, as we all know, the BJP(Bhartiya Janta Party) and the Indian
National Congress are the modern embodiments of these ideological divides. The traits of the BJP
align more with right-wing principles, mainly the focus of the BJP on the Hindu religion, and the
strong central government has resonated with a large portion of the Indian electorate, as evidenced
by its electoral success over the past decades. Opposite to the BJP, the Congress party’s left-leaning
ideology centered around secularism and social welfare has seen declining support from the 2014
general election.

We have floated a survey form S2 among the IITK GenZ students to capture the political spectrum
distribution3 . The study’s classified respondents into five distinct political orientations.
(a) Far Right
(b) Center Right
(c) Center
(d) Center-Left
(e) Far left
The classification is done based on responses to a set of questions in S2. The explanation concerns
the right wing—only the left-wing exhibits opposite characteristics. Classification of Political
Orientation Based on Responses to Survey Section S2
This summary outlines the classification of political orientation based on individual responses to
various statements regarding government role, economic policy, national identity, and other socio-
political beliefs:

(a) Government Role in Redistribution:

• Disagree: Right-wing (Prefers lower taxes, less redistribution)
• Agree: Left-wing (Supports increased taxes on the wealthy for redistribution)
(b) Economic Policy Preferences:
• Agree: Right-wing (Advocates for low taxes and minimal business regulation)
• Disagree: Left-wing (Favors more regulatory oversight and tax policies targeting inequal-
ity)
(c) Expanded Government Role in Social Programs:
• Disagree: Right-wing (Supports limited government involvement)
• Agree: Left-wing (Advocates for expanded government role)
3 Inspired from Participatory ideology: From exclusion to involvement; Peter Beresford

14
(d) National Identity Based on Shared Culture and Customs:
• Agree: Right-wing (Values a unified national identity based on tradition)
• Disagree: Left-wing (Supports a more inclusive or diverse national identity)
(e) Influence of Religion on Government:
• Somewhat agree: Right-wing (More open to religious influence on policies)
• Disagree/Somewhat disagree: Left-wing (Prefers separation of church and state)
(f) Emphasis on Equality and Progress Over Tradition:
• Disagree: Right-wing (Prioritizes tradition over progressive changes)
• Agree: Left-wing (Focuses on equality and progressive reforms)
(g) Secular Governance:
• Disagree: Right-wing (Less emphasis on strict separation of religion and state)
• Agree: Left-wing (Strong advocate for secular governance)
(h) Popular Sovereignty as Essential for National Identity:
• Agree: Right-wing (Emphasizes national sovereignty)
• Disagree: Left-wing (May prioritize global or communal approaches over national sovereignty)
(i) Focus of Populism:
• Neutral : Right-wing (Focuses on national sovereignty and traditional values)
• Strongly agree: Left-wing (Emphasizes empowerment and socio-economic reforms)
(j) Openness to Progressive Reforms:
• Disagree: Right-wing (Cautious about progressive changes)
• Agree: Left-wing (Supports comprehensive progressive reforms)

• Applying some ML models to predict the poilitical Orientation:

Machine learning is slowly changing various industries and Aspects of daily life worldwide.
Machine learning is everywhere, including healthcare, finance, retail, manufacturing, Agricul-
ture, Entertainment, security, customer service, etc. Data science and machine learning are
slowly replacing humans in almost every aspect of life. Analysts in the political landscape
also utilize machine learning by classifying individual preferences into broader political spec-
trums. This classification aids political entities in understanding constituency dynamics and
formulating targeted campaigns. To investigate the efficacy of machine learning in this area,
we conducted a survey incorporating ten carefully curated questions that implicitly categorize
individuals as ’far-left,’ ’center-left,’ ’center,’ ’center-right,’ and ’far-right.’ Lastly, we asked
them about their ideology for getting a supervised dataset.
The survey yielded a labeled dataset essential for supervised machine-learning techniques. We
implemented four primary classification algorithms: Logistic Regression, Decision Tree,
Random Forest, and Support Vector Machine (SVM). The performance of these classi-
fiers was evaluated based on their prediction accuracy for the political leaning of individuals.
The Logistic Regression model demonstrated a prediction accuracy of 74%. The Decision
Tree classifier outperformed Logistic Regression with an accuracy of 81%. This increase
could be attributed to the Decision Tree’s capability to map non-linear relationships common
in opinion-based data. Interestingly, the Random Forest classifier, a more robust version
of Decision Trees through ensemble learning, did not show an improvement and matched the
Logistic Regression accuracy at 74. The SVM’s performance was similar to the aforementioned
models.
One critical observation was that the dataset comprised only 134 responses, a relatively small
sample size that may hinder the model’s ability to generalize; this is evident in the moderate
accuracy scores and potential underfitting issues. A small dataset limits the complexity of
the patterns that the model can learn, hence potentially failing to capture the general trend
of the population’s political leanings.
We suggest two potential methods to mitigate this limitation: data augmentation and
oversampling. Data augmentation artificially increases the size of the dataset by generating

15
(a) showing proportion of Different ideologies in our
Survey (b)

(c) Correlation Heatmap

Figure 15: Depiction of number of responses for each political orientation type and their correlations

new samples from existing ones, while oversampling involves replicating samples from the
minority class to balance the dataset. Both techniques could enhance the model’s learning
capacity, enabling it to generalize better to unseen data.
Integrating machine learning to predict individual preferences can significantly benefit political
parties and analysts. It can streamline understanding demographic trends, refine policy com-
munication, and strategize electoral approaches tailored to diverse voter segments. Though
constrained by data volume, the preliminary results lay a promising foundation for how ma-
chine learning could revolutionize data-driven political strategy in the digital era.
• How Our Model works
This short Paragraph describes how our model predicts new individuals’ preferences. This
time, we asked individuals only those 10 survey questions and not his/her ideology preference
since we already trained our model using label data. The response filled by new individuals
acts as test data we call predict function with the obtained response and our model will give
ideology preference of an individual.

6.2 Study specific suggestions

The first page of our constitution, The Preamble, lays down the core values and ethics that a citizen of
our country should abide by in their conduct, representing the nation’s ideals. These principles are meant
to guide us daily, especially when we vote. It’s about choosing leaders who embody these principles and
ensuring our decisions at the ballot box mirror the high standards our nation strives for. One of the
keywords in our preamble is that we are a ”Secular” republic, and hence, definitely, our voting decisions
should not be driven by the religion of the candidate. But as we have observed from our survey, a large
chunk of responders voted in the affirmative of a candidate’s religion being a major factor while deciding

16
 (a) Logistic Regression Accuracy (b) Decision Tree Accuracy

(c) Random Forest Accuracy (d) Support Vector Machine Accuracy

Figure 16: The accuracy of machine learning models in predicting an individual’s preferences.

to hand over the reins of their constituency for the next 5 years. This is where the self-consciousness
or the Dharmaarthashastra of each individual towards our great nation should become paramount so
that they are ready to perform their Fundamental Duties to the fullest, and if we follow our Dharma only
then can we think about achieving our goals of becoming a developed nation by 2047.

We have also explored how family affiliations to a political party influence the decision-making of the
youth when choosing which candidate to vote for. Many of our survey participants are highly influenced
by the choices of their family members while voting. This should not be the case. A family should always
educate their children or young adults about how they need to make informed choices rather than blindly
follow a family’s affiliation. This is highly imperative because, as seen from our survey, for different
individuals, different qualities in candidates are more important than the others.

Suppose a person at their core values is a right-leaning person, but their family has historically voted
for the left-leaning parties. If the person blindly follows their family’s footsteps, then their actual char-
acteristics won’t be reflected in the person that comes to power, and this could lead to the weakening of
the opposition as well, which can never be a good thing for democracy in the long run.

Our survey gathered that most people in the survey demographic rely on social media to get their
daily news. With every other medium, this mode of communication offers several caveats regarding the
authenticity of the information. It is therefore necessary for users to be mindful and should perform
thorough research and fact-checking of whether the news that they have been exposed to is legitimate
or not in the first place before forming their opinions on certain issues and, in the case of elections, the
candidates.

7 Summary & Conclusion

In this study, we investigated the factors that influence the choices made by Gen Z voters in India,
specifically focusing on how their political affiliations, the characteristics of candidates, the socioeco-
nomic environment, and the media landscape affect their preferences. We used our surveys to try and
disentangle the intricate web of influences that shape the voting patterns and opinions of this important
group of young people. Our results provide a window into the complex and multifaceted realm of Indian
electoral behavior.

17
From our research, while trying to find out relationships between different variables, we can conclude
the following:

1. Family Influence vs. Social Identity and Religion: According to our study, family influence
significantly impacts voters’ preferences towards candidates’ social identity, such as caste and family
lineage, especially in regions with a strong tradition of caste-based politics. However, family influ-
ence does not significantly affect the voter’s preference based on the candidate’s religion, indicating
a more secular approach in electoral decisions within the surveyed population.

2. Media’s Role: The results show that Gen Z voters increasingly consume internet-based media,
with social media as their primary source of news and information. Traditional media and political
campaigning have less impact, showing the growing importance of digital platforms in molding
political ideas and decisions.
3. Candidate Characteristics: Voters prioritize three traits in candidates: honesty, integrity, and
leadership abilities. This choice highlights the public’s need for accountable and honest leadership.
Furthermore, a sizable percentage of respondents highly emphasize a candidate’s moral character
and record, demonstrating the need for integrity in politics.
4. Importance of Local and Youth-Related Issues: Voting decisions are heavily influenced by
local issues and youth-related concerns, indicating that voters are interested in addressing issues
that impact the younger population and urgent and community-specific requirements. This focus
on topics pertaining to youth and the local community represents a more practical and forward-
thinking approach to voting among Generation Z voters.
5. Voting Participation and Political Orientation: Strong political participation among the
group polled is indicated by the high intent to vote in the future elections. In addition, the study
looks at the political spectrum from left to right, indicating that the Gen Z population has a wide
range of ideologies.

References
[1] Waiphot Kulachai, Unisa Lerdtomornsakul, and Patipol Homyamyen. Factors influencing voting
decision: A comprehensive literature review. Social Sciences, 12(9):469, 2023.
[2] Stéphanie Tawa Lama-Rewal. Studying elections in india: Scientific and political debates. Retrieved
from http://journals.openedition.org/samaj/2784. Accessed on 05-03-2024.
[3] Roberto J. González. Hacking the citizenry? personality profiling, ‘big data’ and the election of
donald trump. Anthropology Today, 33(3):3–7, 2017.
[4] Social Media in India - 2023 Stats & Platform Trends - OOSGA — oosga.com. https://oosga.com/
social-media/ind/. [Accessed 05-03-2024].
[5] Constantine Boussalis. Basic survey theory and design. Harvard Law School, 2012.

[6] Belle Wong. Top Social Media Statistics And Trends — forbes.com. https://www.forbes.com/
advisor/in/business/social-media-statistics/. [Accessed 05-03-2024].

18