Smart, Responsible, and Upper Caste Only: Measuring Caste Attitudes through Large-Scale Analysis of Matrimonial Profiles

Ashwin Rajadesingan, Ramaswami Mahalingam, David Jurgens

TL;DR

By analyzing over 313,000 profiles from a major Indian matrimonial platform (Shaadi.com), researchers quantified caste attitudes through individuals' willingness to enter an intercaste marriage. They found that while younger generations are notably more open than their families, openness is heavily shaped by socioeconomic status, and counterintuitively, the Indian diaspora in the United States is substantially less open than demographically matched peers in India.

What Problem Does This Paper Address?

Despite legal protections and social movements, caste-based prejudice continues to shape social mobility and everyday life in India and the South Asian diaspora. Historically, arranged, endogamous marriage has served as the core institution reinforcing caste boundaries (only ~5% of marriages in India are intercaste).

Measuring modern attitudes toward caste is notoriously difficult. Surveys are plagued by social desirability bias (respondents claiming to be more tolerant than they actually are), while counting actual marriage rates conflates willingness with structural opportunity and parental coercion. This study addresses the measurement gap by analyzing self-reported partner preferences in online matrimonial profiles—high-stakes personal decisions where private attitudes directly dictate prospective matches.

Methods

The authors examined whether profiles checked the option: "Not particular about my partner's Caste/Sect (Caste No Bar)", which defaults to "uncheck" on the platform.

  • One-to-One Almost Exact Matching: To isolate generational differences and diaspora effects without demographic confounding, the authors matched self-posted profiles to family-posted profiles (and US profiles to India-based profiles) controlling for age (within 5 years), account age (within 90 days), caste, income, affluence, and education.
  • Multilevel Logistic Regression: Mixed-effects models incorporated state-level random effects to capture geographic variations while estimating the effects of gender, education, caste, affluence, and parental employment. Coefficients across self- and family-posted models were compared using two-sample $Z$-tests.
  • Text Mining & Content Analysis: Extracted desired partner preferences using regex patterns and classified traits into 11 theory-driven categories (e.g., Personality, Family, Profession) with high inter-annotator reliability (Krippendorff’s $\alpha = 0.91$).
  • Spatial Density Estimation: Measured the number of candidate profiles within a 50-mile radius in the US to test whether mate availability correlates with selectivity.

Datasets

  • Shaadi.com Matrimonial Profiles: A scraped dataset of 313,308 profiles created between 2017 and 2018 (298,400 in India and 14,908 in the US; 214,650 self-posted and 98,658 family-posted).
  • IHDS-II Classification: Castes from 352 self-reported designations were categorized using the Indian Human Development Survey into Brahmin, Other Forward Castes (OFC), Other Backward Castes (OBC), Scheduled Castes (SC), and Scheduled Tribes (ST).

Key Insights and Ideas

  • The Education Divergence: While higher education strongly increases openness to intercaste marriage when an individual creates their own profile, an individual's education has no significant effect when family members write the profile. Highly educated individuals still face familial moral policing and infantilization.
  • The Caste Paradox: Higher social status (income, affluence) generally predicts greater openness, but caste status deviates sharply: Brahmins—at the top of the traditional hierarchy—are by far the least open to intercaste marriage, whereas Intercaste, OFC, and SC individuals show substantially higher openness.
  • Collectivism vs. Individualism: Users open to intercaste marriage prioritize personal traits (Personality, Kindness) in their partner descriptions, while caste-conscious users emphasize external social context (Family, Caste).
  • The Diaspora Paradox: Indian immigrants in the US are significantly more conservative regarding caste endogamy than their counterparts in India. This supports bicultural acculturation theory: immigrant minorities often essentialize and cling to traditional cultural identity markers in private life while assimilating into professional spheres.

Key Results

  • Generational Shift: In exactly matched pairs, individuals posting for themselves were significantly more open to intercaste marriage than families posting on their behalf (23.94% vs. 20.82%, $p < 0.001$).
  • Diaspora Gap: US-based individuals were markedly less open than demographically identical peers living in India (13.69% vs. 22.91%, $p < 0.001$). However, US-raised individuals were more open than Indian-born immigrants in the US.
  • Regional Disparities: When controlling for demographics, self-posters in Northern Indian states exhibited significantly higher openness to intercaste marriage than Southern Indian states, contradicting assumptions based on South India's higher literacy and development indices (likely due to Dravidian cross-cousin marriage traditions).
  • Urban Density in the Diaspora: Diaspora openness rose steadily from ~13% to ~17% as the local pool of candidate profiles within 50 miles increased from 0 to over 2,000.

Limitations and Future Work

  • Platform Demographics: Online matrimonial platforms overrepresent educated, affluent, urban, and forward-caste individuals (Brahmins and OFC are overrepresented; SC and ST are underrepresented relative to national census estimates).
  • Selection Bias: Profiles written by individuals versus families may reflect unobserved personality or lifestyle traits beyond age and demographics.
  • Ethics and Data Access: The dataset violates the platform's Terms of Service and contains sensitive personal data. Consequently, the authors did not publicly release the dataset, limiting external reproducibility.
  • Future Work: Further qualitative research is needed to explore why Scheduled Tribes show unexpectedly high resistance to intercaste marriage, and why individuals seeking remarriage in the diaspora exhibit lower openness.

Summary updated

Abstract

Discriminatory caste attitudes currently stigmatize millions of Indians, subjecting individuals to prejudice in all aspects of life. Governmental incentives and societal movements have attempted to counter these attitudes, yet accurate measurements of public opinions on caste are not yet available for understanding whether progress is being made. Here, we introduce a novel approach to measure public attitudes of caste through an indicator variable: openness to intercaste marriage. Using a massive dataset of over 313K profiles from a major Indian matrimonial site, we precisely quantify public attitudes, along with differences between generations and between Indian residents and diaspora. We show that younger generations are more open to intercaste marriage, yet attitudes are based on a complex function of social status beyond their own caste. In examining the desired qualities in a spouse, we find that individuals open to intercaste marriage are more individualistic in the qualities they desire, rather than favoring family-related qualities, which mirrors larger societal trends away from collectivism. Finally, we show that attitudes in diaspora are significantly less open, suggesting a bi-cultural model of integration. Our research provides the first empirical evidence identifying how various intersections of identity shape attitudes toward intercaste marriage in India and among the Indian diaspora in the US.
BibTeX
@inproceedings{rajadesingan2019smart,
  author = {Ashwin Rajadesingan and Ramaswami Mahalingam and David Jurgens},
  title = {{Smart, Responsible, and Upper Caste Only: Measuring Caste Attitudes through Large-Scale Analysis of Matrimonial Profiles}},
  booktitle = {ICWSM},
  year = {2019},
  url = {https://openreview.net/forum?id=cLxAyZw-TxU},
}

Loading related content...

Figure