House of Castes

Policy Context

Caste-related violence remains common in India, with caste tensions running high even among school-aged children. The social psychological theory of recategorization suggests that one type of group-based identity (like caste) can be sidelined by introducing a new, superordinate identity. Researchers are evaluating a naturalistic policy along these lines: the introduction of the House System to schools, in which children are randomly assigned to one of four Houses, and school activities and rewards are organized around the new House identities.

Research Design

The research team are working with 176 secondary schools in southern India. They randomly assigned half of these schools to introduce the House System for the 2024/25 school year. Among the treated schools, researchers randomly assigned each student to one of four caste-diverse Houses, creating random individual-level variation in a student’s new cross-cutting group identity. In addition, in a random subset of classrooms, the research team randomized the caste shares across the Houses, creating random individual-level in cross-caste collaborative contact, conditional on participating in the House System. The researchers will field an endline survey at the end of the 2025/26 school year.

Results

Findings to date come from a pilot survey conducted across 43 schools, with responses from 636 students—242 in Ramanathapuram district and 394 in Tirunelveli district. While these results are preliminary and do not speak to the efficacy of the House System, they offer important baseline descriptive findings that informed both the intervention design and measurement strategy. One of the most striking findings is the high salience of caste identity among students. Nearly half of all respondents reported wearing colored wristbands—both inside and outside school—that indicate their caste, underscoring the everyday visibility of caste markers in these school environments. In a dictator game designed to elicit social preferences, researchers found little evidence of direct caste bias in monetary allocations: average giving centered around parity regardless of whether the peer was perceived as same-caste or different-caste. However, friendship networks revealed clear caste-based homophily. SC/ST students predominantly named close friends from their own caste, while non-SC/ST students reported very few friendships with SC/ST peers. These patterns suggest that while students may not show overt discriminatory behavior in abstract settings, such as the dictator game, their social relationships remain strongly shaped by caste boundaries. These findings confirmed both the relevance and feasibility of the research design. They highlighted the need for interventions that promote cross-caste collaboration in everyday settings and helped refine the structure of the House randomization and survey instruments.