Skip to content
    Flow-alerta.cab

    Demographic Parity

    Application & enablementStatistical parityIndependence
    Editorially reviewed on: 2 sources
    Your reading view

    The short answer · 15 words

    In short

    Demographic parity compares whether positive model decisions occur equally or sufficiently similarly in protected groups.

    Flow-alerta.cab (2026): Demographic Parity. https://flow-alerta.cab/en/glossar/demografische-paritaet. Editorially reviewed on: 10 October 2026.

    How much detail would you like?

    Practical

    The metric can reveal differences in allocation decisions, but guarantees neither fairness for individuals nor overall legal or societal fairness.

    In an application model, selection rates are compared by group; in addition, the team checks error rates, data quality, process rules and legal requirements.

    Businesses can use the metric as one diagnostic element in documented high-risk processes. Subject-matter, legal and affected people's perspectives remain necessary.

    At a glance

    Core
    A binary classifier satisfies demographic parity when its positive decision rate does not depend on group membership. The criterion does not take ground truth or qualifications into account.
    Use
    The metric can reveal differences in allocation decisions, but guarantees neither fairness for individuals nor overall legal or societal fairness.
    Example
    In an application model, selection rates are compared by group; in addition, the team checks error rates, data quality, process rules and legal requirements.
    Limits
    Treat parity as a certificate of fairness; ignore small groups and confidence intervals; unlawfully process sensitive characteristics; overlook goal conflicts with equalized odds.

    In context

    Equalized Odds

    In binary classification, equalized odds requires equal true-positive and false-positive rates between considered groups.

    Disparate Impact

    Disparate Impact in U.S. law refers to adverse effects of seemingly neutral practices; it does not equate to a single fairness measure.

    Bias in AI and Machine Learning

    Systematic deviation, which can arise from data, assumptions, modeling or application and affects results repeatedly in a particular direction.

    Questions about this term

    Does the same rate mean fair individual cases?

    No.

    Can demographic parity be measured without sensitive characteristics?

    Group comparisons require appropriate data, the collection and use of which must be legally verified.

    All details and background

    A binary classifier satisfies demographic parity when its positive decision rate does not depend on group membership. The criterion does not take ground truth or qualifications into account.

    ELI5 – Explained simply

    Approximately the same proportion of each group considered should receive a positive decision.

    Why it matters in practice

    The metric can reveal differences in allocation decisions, but guarantees neither fairness for individuals nor overall legal or societal fairness.

    Example

    In an application model, selection rates are compared by group; in addition, the team checks error rates, data quality, process rules and legal requirements.

    Pitfalls and limits

    • Treat parity as a certificate of fairness; ignore small groups and confidence intervals; unlawfully process sensitive characteristics; overlook goal conflicts with equalized odds.

    Origin & history

    Statistical parity concepts come from fairness and anti-discrimination research. Current machine learning tools implement them as a difference or ratio between selection rates.

    Use in mid-sized companies

    • Businesses can use the metric as one diagnostic element in documented high-risk processes. Subject-matter, legal and affected people's perspectives remain necessary.

    Related terms

    Matching Flow-alerta service

    TrainingAI governance, fairness and quality testing

    Matching tool / use case

    View use casesTool/Use Case: Fairlearn for analysis of selection rates, supplemented by other metrics and human evaluation

    Sources

    Quellenprüfung: 10.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.

    Editorially reviewed on:

    Next step

    Put knowledge into practice

    Discuss your projectView use cases