Research: Following the Data Where It Leads

My PhD journey didn't begin with crime. It began with insolvency, bank loans, and NPAs, until my supervisor informed me he didn't work in those areas. Then it pivoted to inequality and taxation then I discovered that economic data in India is essentially a national treasure we guard jealously. So I did what any pragmatic researcher would do: I looked for where the data actually existed, and stumbled upon the NCRB (National Crime Records Bureau), a treasure trove of state-level data spanning decades.

Three years into my PhD, I had datasets. I had statistical tools. I had results. And none of it made sense. But that's when the real research began, not from reading papers and identifying gaps, but from observing patterns in the data itself. What I found challenged my assumptions, revealed hidden inequalities, and exposed how institutions actually work versus how we think they work.

A Pattern Hiding in Plain Sight

The Question: I began examining prison demographics, expecting to see what everyone was talking about: the rising prison population in India. Progressive channels were sounding alarms, and I had my own biases about what I'd find.

The Reality: The data told a different story. While Muslim populations fluctuated, it was the Scheduled Caste (SC) prison population that remained stubbornly static, ranging between 22 25% across states for two decades. SCs constitute 16.6% of India's population but 23% of the prison population.

What We Found: Analysing 20 years of data across 20 Indian states revealed something critical: SC atrocity cases showed higher police disposal rates but substantially elevated court acquittal rates compared to other legal categories. The system appears efficient on paper, but it systematically fails in practice.

The breakthrough finding: SC representation in police forces emerged as the single most significant deterrent to criminal behaviour. After the 2014 POA Amendment, states responded in dramatically different ways; some viewed representation and welfare resources as protective factors, while others considered justice system variables to be more significant. The pattern revealed that discrimination doesn't just create economic disadvantage; it creates persistent, hereditary strain that institutions either amplify or mitigate.

[Work in progress; currently analysing this dataset through multiple theoretical lenses; "Institutional Effectiveness, Political Representation, and Scheduled Caste Incarceration: A Multi-State Analysis of India's Criminal Justice System"]

When Legal Reform Meets Reality

The Catalyst: December 2012. The Nirbhaya case. A 23-year-old medical student was brutally assaulted on a Delhi bus. Her death sparked nationwide protests and led to the Criminal Law (Amendment) Act of 2013; comprehensive legal reforms meant to transform how India handles crimes against women.

The Numbers: In 2012, India recorded 244,270 cases of crimes against women. By 2022, that number reached 445,256, translating to 51 FIRs filed every hour.

The Question: Did the reforms work? Was it through reporting or deterrence?

What We Found: Using Regression Discontinuity in Time across 29 states and 604 districts (2007;2019), we discovered something more nuanced than simple success or failure. The reforms caused an immediate 13-19% surge in reported crimes against women, but this wasn't evidence of failure. It was evidence that the reforms successfully dismantled reporting barriers.

The pattern was revealing: medium-crime states showed the strongest sustained impact (a 39% increase that persisted), while high-crime states showed minimal effects. Rural states responded most strongly; urban states showed the weakest responses. However, the critical finding is that after an initial surge (2013-2014) followed by a dip (2015), crime numbers resumed rising after 2015. This temporal pattern reveals that reforms mobilised the disclosure of historically unreported cases, but actual incidents continued to increase underneath, meaning the reforms transformed the reporting infrastructure without deterring the underlying violence. Female political representation, women's employment, unemployment rates, recidivism, and inequality all emerged as significant determinants of these outcomes. The reforms worked through economic mechanisms, empowering victims to report, but couldn't address the socioeconomic roots of violence itself.

[Under revision: "Tragedy as a Catalyst: Evaluating the Long-term Impact of Event-Driven Legal Reforms on Gender Violence in India"]

The Borders Crime Doesn't Respect

The Assumption: Crime is a local problem requiring local solutions. Better policing in State A should lead to a reduction in crime in State A.

The Reality: Crime doesn't respect administrative boundaries, and institutional effectiveness in one state creates ripple effects across borders.

What We Found: Using spatial dynamic panel methods across 30 Indian states (2002;2021), we discovered something counterintuitive: enhanced police capacity in neighbouring states creates asymmetric displacement, reducing murder while increasing crimes against women in adjacent jurisdictions. Correctional policies displace crimes against children and property crimes across boundaries. Economic development in neighbouring states creates strain effects that increase murder, crimes against children, and property crimes in adjacent jurisdictions.

The Critical Pattern: Federal systems create natural experiments in governance, but they also create vulnerability corridors. When one state strengthens its institutional capacity, criminals don't disappear; they adjust their spatial operations. Educational disparities, unemployment, and economic prosperity in neighbouring states all create cross-border vulnerability patterns through strain mechanisms.

This finding fundamentally challenges traditional deterrence frameworks: an isolated state-level crime control creates unintended consequences that extend across borders. Federal systems require coordinated regional approaches rather than isolated strategies.

[Accepted for publication in Millennial Asia: "Cross-Border Crime Dynamics in Federal Systems: Spatial Analysis of Institutional Effectiveness and Criminal Activity Across Indian States"]

The Digital Court of Public Opinion

The Spark: Two cases, two months apart in 2024. In Kolkata, a junior doctor was raped and murdered during duty at RG Kar Medical College, exposing failures in medical workplace safety. In Pune, a 17-year-old from a wealthy builder family crashed his Porsche into a motorcycle, killing two IT professionals instantly, revealing how elite privilege undermines judicial accountability. What followed wasn't just news coverage; it was over 150,000 YouTube comments, creating parallel trials in the court of public opinion.

The Question: How do Indians actually perceive different types of crime through their digital discourse, and what do their unfiltered responses reveal about victim identity, institutional trust, and demands for justice?

What We Did and Found: Using computational text analysis (BERTopic modelling) on 150,536 YouTube comments, we compared public discourse on gender-based violence (RG Kar Medical College rape-murder) versus elite impunity (Pune Porsche vehicular homicide). This allowed us to analyse how victim identity, gender versus class, shapes the structure, intensity, and framing of public responses at scale. 

Gender-based violence (RG Kar) generated 67% more discourse and framed the crime emotionally- 52.7% of comments discussed women's safety collectively. Institutional critique was concentrated across medical systems, the police, and the state government.

Elite impunity (Pune Porsche) produced technical discourse focused on juvenile justice, bail failures, and procedural loopholes. Comments centred on individual perpetrator accountability rather than systemic issues. Institutional critique fragmented across disconnected targets, and discourse devoted 7x more attention to content creators than to the crime itself.

The Critical Finding: Victim identity shapes discourse structure. Gender-based violence activates emotional, collectively gendered frameworks with concentrated institutional critique. Elite impunity activates technical, individualised frameworks with fragmented critique. The same institutional failures generate completely different public responses depending on the victim's identity.

[Work in progress: "Public Perceptions of Violence and Justice in India: A Computational Text Analysis of Indian Digital Discourse"]

The Pattern Across Patterns

Four different research questions. Four different datasets. One underlying truth: institutions don't work in isolation, and their effectiveness depends as much on systemic integration as on individual capacity.

Whether examining discrimination against Scheduled Castes, legal reforms for gender violence, digital public discourse, cross; or border crime dynamics, the data consistently revealed that:

1. Institutional factors often matter more than individual-level factors in determining outcomes

2. Integrated approaches across multiple institutions work better than isolated interventions

3. Marginalised populations face compounding disadvantages that require systematic analysis, not anecdotal evidence

4. Spatial and temporal dynamics matter; crime patterns shift across borders and over time in response to institutional changes

This isn't abstract theory. These are patterns visible in decades of data across hundreds of districts and millions of lives, highlighting how institutions actually function versus how we assume they work.

This research was conducted as part of my PhD in Economics of Crime at IIT Jodhpur, where I blend Development Economics, Sociology, and Criminology to understand how institutions, discrimination, and economic factors shape criminal behaviour in India.

Beyond the Thesis: Collaborative Explorations

Beyond my core thesis work, I've collaborated on projects exploring emerging frontiers in criminology and social science:

• AI Companion Apps Research: Examining how digital relationships and virtual companionship intersect with social isolation and behavioural patterns

• Institutional Effectiveness and Criminal Rehabilitation: We employed machine learning and dynamic panel methods across 29 states (2002-2021) to demonstrate that successful rehabilitation necessitates a systematic integration of economic development, legal aid systems, and rehabilitation programs, rather than isolated institutional improvements.

• Metaverse and Social Behaviour: Exploring how virtual spaces create new contexts for understanding social norms and deviant behaviour

These collaborative projects extend the core insight that institutions, whether physical or digital, shape behaviour in ways that require empirical investigation rather than assumption.