CDI Methodology

Methodological Overview

This page provides technical documentation of the CROWN Discrimination Index (CDI) methodology for academic audiences, potential research collaborators, and institutional review processes. It describes the survey instrument design principles, the calibration approach linking survey data to hardware-verified diagnostics, the validation pathway, and the statistical framework underpinning the index.

The CDI draws on methodological expertise at the Faculty of Psychology and Educational Sciences, University of Geneva. The methodology adheres to established psychometric standards for survey instrument construction and composite index design, while introducing a novel calibration element: the integration of self-reported discrimination data with objective, hardware-verified hair diagnostic data from CROWN’s multi-sensor device.

Survey Instrument Design

Construction Principles

The CDI survey instrument is constructed following classical test theory (CTT) principles supplemented by item response theory (IRT) analysis. The instrument measures three latent constructs corresponding to the CDI’s three measurement pillars: discrimination prevalence and experience, discrimination severity, and economic impact.

Item generation draws on three sources: (1) the existing validated instruments in discrimination research, including the Everyday Discrimination Scale (Williams et al., 1997), the Perceived Racism Scale (McNeilly et al., 1996), and the Schedule of Racist Events (Landrine & Klonoff, 1996); (2) qualitative data from structured interviews with individuals reporting hair-based discrimination in European contexts; and (3) domain expertise from CROWN’s founding team and academic advisors.

Content validity is established through expert review by researchers in discrimination studies, clinical psychology, and psychometrics. Items are evaluated for relevance, clarity, cultural appropriateness across European populations, and absence of acquiescence and social desirability bias.

Response format employs a combination of Likert-type frequency scales (measuring how often specific discrimination experiences occur), categorical items (identifying contexts, perpetrators, and institutional settings), and open-ended economic items (capturing financial impact data). The instrument is designed for completion in 15 to 20 minutes.

Cross-Cultural Validity

A central challenge for any European discrimination index is cross-cultural validity. Discrimination is experienced, interpreted, and reported differently across cultural contexts. An instrument validated in Geneva may not produce equivalent measurements in Paris, Berlin, or London.

The CDI addresses this through several methodological provisions:

Translation and adaptation. Items are developed bilingually (English and French) from inception, not translated post hoc. Subsequent language versions will follow the International Test Commission’s guidelines for test adaptation (ITC, 2017), including forward and back translation, cognitive interviewing in each target language, and differential item functioning (DIF) analysis.

Cultural decentring. Items are written to describe concrete, observable experiences rather than culturally loaded interpretations. “My hair was touched without my permission in a professional setting” is more culturally stable than “I experienced discrimination at work.”

Invariance testing. Measurement invariance will be tested across language groups, countries, and demographic categories using multi-group confirmatory factor analysis (MGCFA). Configural, metric, and scalar invariance are required before cross-group CDI score comparisons are considered valid.

Hardware Calibration

The Calibration Innovation

The CDI’s methodological contribution is the calibration of survey-based measurement against hardware-verified diagnostic data. This is, to our knowledge, the first application of objective biometric calibration in discrimination research.

The calibration works as follows. Participants who complete the CDI survey instrument are invited (with separate informed consent) to undergo a diagnostic assessment using CROWN’s multi-sensor device. The device produces a CROWN Hair DNA profile — an objective, multi-dimensional characterisation of the individual’s hair including fibre diameter, cross-section ellipticity, cuticle condition, curl pattern classification, porosity index, hydration level, protein integrity, and chemical treatment history.

This creates a linked dataset: subjective discrimination experience (survey) paired with objective physical characteristics (hardware). The analytical possibilities are significant.

Controlling for Self-Report Bias

Self-report bias in discrimination research takes multiple forms. Underreporting occurs when individuals normalise discriminatory experiences (“It’s not really discrimination, it’s just how things are”), when they fear social consequences of reporting, or when they lack the conceptual vocabulary to categorise their experiences as discrimination. Overreporting can occur in survey contexts that prime discrimination awareness. Categorisation inconsistency arises when similar experiences are reported differently across cultural and linguistic contexts.

The hardware calibration provides a partial correction. By linking reported discrimination to objectively measured physical characteristics, the CDI can:

• Identify underreporting patterns: if individuals with hair characteristics strongly associated with discrimination in the broader dataset report low discrimination levels, this may indicate normalisation or underreporting rather than absence of discrimination.
• Detect context effects: by comparing survey responses to objective characteristics, the methodology can assess whether survey framing systematically inflates or deflates reported discrimination levels.
• Establish dose-response relationships: correlating specific, measurable hair properties (e.g., fibre diameter distribution, ellipticity index) with reported discrimination severity enables a more precise understanding of what physical characteristics drive discriminatory responses.

Chemical Treatment History as Evidence

The diagnostic device’s near-infrared spectroscopy module detects chemical treatment residues — evidence of straightening, relaxing, or keratin treatments. This data has particular methodological value.

Chemical straightening is, in many cases, a direct behavioural response to discrimination or the anticipation of discrimination. The NIH’s 2022 study linking chemical hair straightening to elevated uterine cancer risk underscored that this is not merely a cosmetic choice but a health-relevant behaviour driven by social pressure. The CDI can correlate chemical treatment history (objectively measured) with reported conformity pressure (self-reported), providing a behavioural validation of survey responses.

Composite Index Construction

Aggregation Framework

The CDI aggregates three sub-indices — prevalence, severity, and economic impact — into a composite score. The aggregation follows the methodology established by the UNDP for the Human Development Index, adapted for discrimination measurement:

1. Normalisation: Raw scores on each sub-index are normalised to a 0-1 scale using minimum-maximum normalisation, with goalposts defined by theoretical minimum and maximum values for each construct.

2. Weighting: Sub-indices are combined using weights derived from both expert judgment (Delphi method with discrimination researchers) and empirical analysis (principal component analysis of the underlying data structure). The weighting scheme will be published upon completion of the validation study.

3. Aggregation: The composite CDI score is computed as a weighted geometric mean of the three sub-indices — a choice that penalises extreme imbalance between dimensions, following the logic that high prevalence with low economic impact, or low prevalence with high severity, should both register as elevated CDI scores.

Disaggregation

The CDI is designed for disaggregation across multiple dimensions:

• Geography: Country, region, city, neighbourhood
• Sector: Workplace, education, healthcare, public services, social
• Demographic group: Age, gender, ethnicity, hair type, socioeconomic status
• Time period: Enabling longitudinal tracking and pre/post-intervention comparison

This disaggregation capacity is what makes the CDI useful for policy impact assessment — a government can measure CDI in a specific sector, implement legislative or institutional changes, and measure CDI again to assess effectiveness.

Validation Pathway

The CDI validation follows a three-phase pathway:

Phase 1: Pilot Study (Current)

The pilot study, developed in consultation with the University of Geneva, serves three purposes: (1) psychometric evaluation of the survey instrument (item analysis, reliability, factor structure), (2) feasibility assessment of the hardware calibration protocol, and (3) preliminary CDI score computation. The pilot targets a sample size sufficient for stable item statistics and initial factor analysis (minimum n = 200, stratified by hair type, ethnicity, gender, and country of residence).

Phase 2: Peer Review and Replication

Following the pilot, the methodology and initial results will be submitted for peer review in a relevant academic journal. Simultaneously, the study will be designed for independent replication — a requirement for any instrument intended for cross-national use. Replication sites will be selected to ensure geographic, linguistic, and demographic diversity within Europe.

Phase 3: Multi-Site Deployment

Upon validation, the CDI will be deployed at scale through CROWN’s network of research partners, corporate partners, and the growing installed base of diagnostic devices. Each deployment generates additional data that further refines the instrument’s psychometric properties and calibration models.

Statistical Methods

The CDI’s statistical framework encompasses:

• Factor analysis (exploratory and confirmatory) for validating the survey instrument’s dimensional structure
• Item response theory for item-level analysis and computer-adaptive testing potential
• Multi-group confirmatory factor analysis for measurement invariance testing
• Structural equation modelling for testing the hypothesised causal relationships between hair characteristics, discrimination experiences, and outcomes
• Mixed-effects models for analysing hierarchically structured data (individuals nested within contexts, contexts nested within countries)
• Missing data treatment via multiple imputation, given the sensitivity of discrimination-related questions and the expected pattern of non-random missingness

Ethical Considerations

All CDI research is conducted under ethical review standards. Participant data is anonymised at collection, stored in compliance with the Swiss Federal Act on Data Protection (nDSG) and the EU General Data Protection Regulation (GDPR), and governed under CROWN’s privacy policy. Participants provide separate informed consent for each data collection component (survey, diagnostic, data linkage).

The CDI is designed to measure discrimination — a sensitive topic that can cause distress in participants who are asked to recall discriminatory experiences. The survey instrument includes wellbeing checkpoints, and participants are provided with referral information for psychological support, including CROWN’s own therapeutic protocol.

Collaboration Opportunities

The CDI methodology is designed for collaborative development. CROWN welcomes inquiries from researchers interested in:

• Instrument validation in additional languages and cultural contexts
• Replication studies using the CDI methodology
• Integration of CDI data into existing discrimination research programmes
• Co-supervision of student projects on CDI-related topics

For academic collaboration inquiries, contact contact@crown.ngo or visit our Research Partners page.

This page provides a summary of the CDI methodology. Technical details sufficient for peer review will be disclosed through the standard academic publication process. CROWN’s proprietary algorithms and weighting coefficients are not published here but will be made available to reviewers and replication partners under appropriate agreements.