The loan game. Harnessing big data to compare naturalistic mental health expressions on student loan debts

Sinha, Gaurav Ranjan

Permalink

https://hdl.handle.net/2142/115649 Copy

Description

Title

The loan game. Harnessing big data to compare naturalistic mental health expressions on student loan debts

Author(s)

Sinha, Gaurav Ranjan

Issue Date

2022-04-21

Director of Research (if dissertation) or Advisor (if thesis)

Larrison, Christopher R.

Doctoral Committee Chair(s)

Larrison, Christopher R.

Committee Member(s)

• Anderson, Steven G.
• Viswanathan, Madhubalan
• Brooks, Ian
• Friedline, Terri L.

Department of Study

School of Social Work

Discipline

Social Work

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• mental illness
• mental health
• student loan
• student debt
• education loan
• education debt
• financial vulnerability
• big data
• social media data

Abstract

Background and Objectives. Growing demands for higher education have forced families to borrow student loan debts at an unprecedented rate globally. Even in countries where college is tuition free, people borrow huge amount of money to cover associated costs, such as living expenses, to complete their degrees. Although not considered “bad debt”, student loan debt has both economic and psychological consequences. While there is plenty of scholarly literature on economic consequences of student loan debt, its effect on mental health is understudied in different ways. First, the studies are predominantly cross-section in nature that do not account for changes. Second, there is a paucity of research on examining people’s naturalistic response which can paint more nuanced picture of the relationship between student loan debt and mental health. Finally, there is a need for studying sentiments and emotions of people with mental illness on student loan debt and how these sentiments and emotions different from people without a mental illness. However, recent developments in research methods and availability of new tools have made it possible to study people’s unfiltered opinions, beliefs, sentiments, and emotions expressed on social media platforms. Thus, the primary objective of this study was to identify patterns in people’s naturalistic expressions on student loan debts on two social media platforms. The secondary objective was to examine how these patterns, sentiments, and emotions associated with student loan debts differ in people who might have mental illness. Methods. Data for this study were collected from Reddit and Twitter (between 2009 – 2020) using four key terms – “student loan”, “student debt”, “education loan” and “education debt” along with first-person pronouns such as I, me, and my as a triangulating measure of postings by individuals – through an artificial-intelligence (AI) analytics company, Brandwatch. Replies and retweets were removed. Only posts in English were used because of the language limitation of the researcher (n=85,664). Data from Reddit were collected from two different communities including many subcommunities in each – reddit finance and reddit mental health subcommunities. Two types of analyses were used to achieve both the objectives. First, Latent Dirichlet Allocation (LDA), a form of unsupervised topic modeling, was used to analyze the underlying patterns in discussions on both Reddit and Twitter. LDA represents distribution over a fixed set of vocabulary for the collection of text to identify topic areas and identifies patterns in the use of words and hidden semantic structures in the body of the text. Next, a supervised machine learning algorithm was applied to classify the posts into expressing mental illness and non-mental illness. In this step, a training model was developed by manually annotating two percent of posts among users who expressed mental illness (UEMIs) and non-UEMIs using four standard instruments – PHQ9, GAD7, K10, and perceived stress scale. The training model was used to classify the remaining 98% posts as UMIs or non-UMIs. Then, AI-assigned emotions and sentiments for each post were used for further analyses. Finally, bivariate analyses were conducted to examine the group differences based on mental illness status (UEMI vs non-UEMIs). Results. Fifty topics in reddit finance and 40 each in reddit mental health communities and twitter posts were found. Major topics in finance communities included school and university expenses, acquiring assets, budgeting and repayment plans, and desire to know. Reddit mental health communities included outlook towards life, desire to know, self-harm thoughts, anxiety and depression, and academic experience/hardships. Twitter users shared similar topics of discussion and include feeling depressed and worrying thoughts, mental health care and treatment issues, interest rates and taxes, stressful feelings, and academic experience/hardships. Statistically significant associations (p < .01) were found between mental illness status and sentiments. UEMIs had more negative sentiments as compared to non-UEMIs (84.8% vs 71.8%) and were more likely to express sadness (43.8% vs 18.0%) and fear (29.9% vs 10.6%) about their student loans. A surprising finding was higher anger among non-UMIs compared to UMIs (30.9% vs 16.0%). Conclusions. The present study demonstrates that cognitive burden created by student loan debts manifest itself on social media. Patterns in these discussions indicate both academic and non-academic consequences of having student loan debt, including users’ desire to know more about their debts. Higher volumes of negative sentiments and emotions of sadness, fear, and anger, especially among the UEMIs warrant immediate attention to reduce the cognitive burden of student loan debts. While student loan debt is not bad, its negative effect on mental health has implications for social work policy, practice, and education. This study also shows that social media can play an important role as a marker to develop a nuanced understanding of people’s expressions on socioeconomic issues.

Graduation Semester

2022-05

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Gaurav Ranjan Sinha

Owning Collections