Intersectional Framework

Abstract

The intersectional framework, which the Medical Subject Headings classify under sociology, is an analytic approach holding that social positions such as race, gender, class, and sexuality are not separable, additive sources of advantage and disadvantage but mutually constituting systems whose combination produces experiences no single axis can capture. Coined by legal scholar Kimberlé Crenshaw and rooted in Black feminist thought, it entered psychology as a corrective to research that studied one identity at a time and treated the others as noise to be controlled away. Its central methodological claim is that the effect of occupying several marginalized positions at once is often more, or less, than the sum of their separate effects. This article traces the framework's origin, the three methodological approaches Leslie McCall distinguished, its uptake in psychology, and the quantitative models now used to estimate intersectional effects.

Keywords: intersectionality, social identity, additivity, quantitative methods, health equity

Intersectionality is the proposition that the major axes of social differentiation, race, gender, class, sexuality, disability, and others, operate not as independent strands but as interlocking systems, so that a person's position at their intersection is qualitatively distinct from the arithmetic combination of the parts (Crenshaw, 1991). The term was introduced by the legal scholar Kimberlé Crenshaw to name a problem that antidiscrimination law and single-axis politics kept reproducing: Black women's claims fell between the cracks of race discrimination, framed around Black men, and sex discrimination, framed around white women, because each doctrine assumed a single ground of harm (Crenshaw, 1991). For cognitive and social psychology the framework matters because it challenges a default research habit, the study of one social category with the rest partialled out, and asks instead how membership in several categories jointly structures perception, treatment, and outcome (Cole, 2009).

What the Intersectional Framework Is

The framework began outside psychology, in critical legal studies and Black feminist thought. Crenshaw's founding argument was concrete and doctrinal: courts hearing the claims of Black women plaintiffs insisted on treating race and sex as mutually exclusive grounds of discrimination, with the result that a harm arising specifically from their conjunction, being disadvantaged as Black women, not as Blacks or as women, became legally invisible (Crenshaw, 1991). The metaphor she offered was a traffic intersection: injury at a crossroads can be caused by cars traveling in several directions at once, and assigning it to a single street misdescribes what happened. The intellectual roots run deeper, into a long tradition of Black feminist analysis that Patricia Hill Collins systematized as a matrix of domination, an image of oppression as an interlocking structure of race, class, and gender rather than a hierarchy with one master axis (Collins, 2015).

Two clarifications separate the framework from its frequent caricatures. First, intersectionality is not the claim that identities simply add up; it is the stronger claim that they interact, so that the meaning and effect of one position depend on the others held simultaneously (Bowleg, 2012). Second, it is not merely a theory of disadvantaged individuals but a theory of systems: the units that intersect are structures of power, racism, sexism, classism, not private attributes, even though their effects register in individual lives (Cho, Crenshaw, & McCall, 2013). As the framework diffused across disciplines its meaning also broadened, and scholars have debated whether so widely borrowed a term retains analytic precision or has become a general gesture toward complexity (Nash, 2008).

McCall's Three Approaches

The framework's migration into empirical social science required a decision it had not originally forced: what, methodologically, to do with the social categories themselves. Leslie McCall gave the influential answer by distinguishing three approaches according to their stance toward categories (McCall, 2005). The anticategorical approach is skeptical of fixed categories altogether, deconstructing them as unstable social fictions and resisting analyses that reify race or gender as if they named natural kinds. The intracategorical approach, closest to Crenshaw's original, provisionally accepts a category but focuses on a neglected group at a particular intersection, the Black woman, the working-class lesbian, to expose experience that broader categories obscure. The intercategorical, or categorical, approach provisionally adopts existing categories precisely in order to document changing relationships of inequality among the multiple groups they define, comparing across the full grid of intersections rather than dwelling in one cell.

The three are not rivals so much as tools suited to different questions, and the choice among them carries real consequences for what a study can see (McCall, 2005). An intracategorical design illuminates a single marginalized location in depth but cannot, by construction, say whether that location is worse than additivity would predict; an intercategorical design can make exactly that comparison across many groups but risks treating fluid categories as fixed. The demonstration below lets the reader move a study between an intracategorical focus on one stratum and an intercategorical comparison across all of them, and watch what each stance renders visible or invisible.

Intersectionality in Psychology

Psychology's formal adoption of the framework is usually dated to a single agenda-setting article. Elizabeth Cole asked what it would mean to actually conduct intersectional research in psychology and answered with three questions a study should pose: who is included within a category, what role inequality plays among categories, and where there are similarities across seemingly different groups (Cole, 2009). The point was to move intersectionality from a critique that psychologists admired but did not use into a set of concrete design choices. Around the same time a cluster of papers built the methodological scaffolding: Stephanie Shields framed gender itself as inescapably intersectional, arguing that gender cannot be understood in isolation from the other identities that inflect it (Shields, 2008), and Leah Warner offered a best-practices guide translating the theory into procedures for sampling, measurement, and analysis (Warner, 2008).

The uptake was not merely programmatic; it also provoked reflection on disciplinary fit. Moin Syed cautioned that importing intersectionality wholesale risked flattening the deep disagreements about method and epistemology that the framework carries with it, and that psychology should engage those tensions rather than adopt a sanitized version (Syed, 2010). Lisa Rosenthal later situated the framework within psychology's stated commitment to social justice, arguing that taking intersectionality seriously reorients research toward the structural determinants of inequity rather than individual deficits (Rosenthal, 2016). The through-line across these treatments is a shift in the default unit of analysis, from the isolated variable to the person situated at a specific, multiply-determined social location (Cole, 2009).

Estimating Intersectional Effects

The framework's defining empirical claim, that combined positions are not simply additive, is a claim about statistical interaction, and translating it into estimation has been the central methodological project of the last decade. The starting difficulty is conceptual as much as technical. Lisa Bowleg put it sharply: the additive question a survey naturally asks, how much of an outcome is due to race plus how much to gender, misframes an intersectional reality in which the two are lived simultaneously and inseparably, so that the very measurement instrument can impose an additivity the theory denies (Bowleg, 2012). Nicole Else-Quest and Janet Shibley Hyde laid out the epistemological and measurement issues this raises for quantitative psychology, distinguishing the framework's demands from the tools conventionally used to meet them (Else-Quest & Hyde, 2016).

Two methodological programs answer the challenge. The first, in population health, was set out by Greta Bauer, who showed how intersectionality theory could be incorporated into research methodology by modeling social positions as jointly constituted rather than as independent covariates, and by attending to which intersections a design can and cannot identify (Bauer, 2014). The second is a specific modeling strategy, multilevel analysis of individual heterogeneity and discriminatory accuracy, abbreviated MAIHDA, in which individuals are nested within intersectional strata defined by the crossing of several categories, and the variance is partitioned to ask how much of an outcome is explained at the stratum level (Evans, Williams, Onnela, & Subramanian, 2018). A systematic review by Bauer and colleagues found the quantitative literature growing rapidly but unevenly, with recurring confusion about what an interaction term does and does not capture (Bauer et al., 2021). The demonstration below builds a set of intersectional strata and lets the reader vary how much of the outcome variation lies between strata rather than within them.

Additivity and Its Failure

The clearest way to see what the framework asserts is to contrast two models of the same data. An additive model predicts the outcome for any group by summing a baseline, a main effect for each disadvantaged position, and nothing more; an intersectional model allows an interaction term that is added when several disadvantaged positions coincide. When the interaction is zero the two models agree and intersectionality reduces to bookkeeping. When it is non-zero, the doubly or triply marginalized group departs from the additive prediction, and the sign of the departure matters: a positive interaction means compounded disadvantage beyond the sum of parts, a negative one means the combination is less severe than additivity would forecast, sometimes because the separate main effects were themselves estimated on non-overlapping groups (Bauer, 2014). Table 1 sets the ingredients side by side, and the first demonstration and the worked example both turn on exactly this comparison.

Table 1. The additive and intersectional models of a two-axis outcome, differing only in whether the interaction term is constrained to zero.

Worked Example

Consider a health-risk score, ranging from 0 to 100, measured across groups defined by two binary axes of advantage, and follow the additive-versus-intersectional contrast with concrete numbers (Bowleg, 2012; Bauer, 2014). The reference stratum, advantaged on both axes, has a mean risk of 20. Being disadvantaged on the first axis alone raises the mean by 12, to 32; being disadvantaged on the second axis alone raises it by 18, to 38. An additive model, which assumes the two disadvantages simply stack, therefore predicts the doubly-disadvantaged stratum at 20 + 12 + 18 = 50. That number is the additive forecast, and it embeds the assumption the framework contests.

Suppose the doubly-disadvantaged stratum is in fact observed at a mean risk of 68. The interaction term is the gap between observation and additive prediction, 68 − 50 = 18, a positive interaction indicating that the two disadvantages compound rather than merely sum. Relative to the additive forecast, the intersectional excess inflates the predicted risk by 18 / 50 = 36 percent, and relative to the reference group the doubly-disadvantaged stratum is elevated not by the 30 points the main effects imply but by 48. Had the observed value instead been 44, the interaction would be 44 − 50 = −6, a sub-additive combination in which the joint position is less severe than the stacked main effects predict. The single observed number for the intersection thus decides the framework's central empirical question, and no amount of information about the main effects alone can supply it, which is precisely why a design that never populates the doubly-disadvantaged cell cannot test intersectionality at all (Bauer et al., 2021).

Discussion

Intersectionality has become one of the most widely cited frameworks in the social sciences, and its very success has sharpened debate about what it is for. A recurring worry, voiced from within, is definitional: as the term spread from Black feminist legal theory into psychology, health, and beyond, it accumulated meanings, an identity, a population, a type of relationship, a methodology, that are not obviously the same thing, and Patricia Hill Collins catalogued these definitional dilemmas as a sign of a field still consolidating (Collins, 2015). Jennifer Nash pressed a related critique, questioning whether intersectionality's reliance on Black women as its paradigmatic subject and its uneasy mix of descriptive and normative claims leave its core ambiguous (Nash, 2008). These are not hostile objections but the ordinary growing pains of a framework asked to do more work than any single formulation can bear.

For empirical psychology the productive response has been to make the choices explicit. Whether a study is anticategorical, intracategorical, or intercategorical is a decision with visible consequences, and naming it disciplines the analysis (McCall, 2005). Whether an interaction is modeled or assumed away is likewise a decision, not a default, and the quantitative literature's main lesson is that the framework's signature claim is testable only when the design populates the intersections it cares about (Bauer et al., 2021; Evans et al., 2018). The framework's contribution to social cognition, and to psychology's account of stereotyping and prejudice more broadly, is to insist that the target of a social judgment is never a single category but a person at an intersection, and that perceivers, institutions, and researchers alike err when they collapse that intersection to its most salient axis (Cole, 2009; Rosenthal, 2016). The cognitive form of that error has a name: because people who occupy two or more subordinate positions fit no group's prototype, they are perceived non-prototypically and rendered relatively invisible, so their distinctive experiences are overlooked and their treatment is poorly predicted by the sum of the single-axis stereotypes applied to them (Purdie-Vaughns & Eibach, 2008).

Current Directions

The liveliest current work is quantitative and methodological. The multilevel modeling of intersectional strata has moved from proposal to active program: by nesting individuals within strata defined by crossing several categories and partitioning the variance, analysts can ask how much of an outcome's variation is genuinely intersectional, that is, located between strata, versus individual, and can do so without estimating an unwieldy tangle of interaction terms (Evans et al., 2018). The systematic review of the field found this approach spreading across epidemiology and health services research while also documenting persistent errors, additive models described as intersectional, interactions reported without adequate cell sizes, that the framework's own logic should have flagged (Bauer et al., 2021). In parallel, the health-equity literature has pushed intersectionality from description toward intervention, asking not only where compounded disadvantage lies but which structural levers, in policy and institutions, could relieve it (Rosenthal, 2016). The open questions are as much conceptual as statistical: how many axes a design can meaningfully cross before strata empty, how to model categories that are themselves fluid, and how to keep the quantitative apparatus faithful to a theory that began as a critique of exactly such apparatus (Bowleg, 2012; Else-Quest & Hyde, 2016).

Common Misconceptions

Intersectionality just means people can belong to more than one disadvantaged group.

Multiple membership is the premise, not the claim. The framework's assertion is that the positions interact, so the combined effect is not the sum of the separate effects, which is a statement about interaction rather than mere co-occurrence (Bowleg, 2012).

Intersectionality is only a critical theory and cannot be studied quantitatively.

It began as legal and feminist theory, but a substantial quantitative program now estimates intersectional effects with interaction terms and multilevel models of intersectional strata (Bauer, 2014; Evans et al., 2018).

The framework is about individual identity.

The units that intersect are systems of power, racism, sexism, classism, not personal attributes; the framework analyzes structures whose effects register in individual lives, not identities in isolation (Cho et al., 2013).

Adding interaction terms to a regression is automatically an intersectional analysis.

An interaction term is necessary but not sufficient. Without adequately populated strata and an explicit stance toward the categories, a model can wear intersectional language while testing nothing the framework requires (Bauer et al., 2021).

Glossary

Additive model.

A model that predicts a group's outcome as a baseline plus a separate main effect for each disadvantaged position, constraining any interaction to zero.

Anticategorical approach.

McCall's approach that treats social categories as unstable fictions to be deconstructed rather than taken as units of analysis.

Interaction.

The departure of an observed joint outcome from the sum of the separate main effects; the statistical form of the intersectional claim.

Intercategorical approach.

McCall's approach that provisionally adopts existing categories in order to compare relationships of inequality across the full grid of intersections.

Intersectionality.

The framework holding that axes of social differentiation are mutually constituting, so a position at their intersection is distinct from the combination of its parts.

Intracategorical approach.

McCall's approach, closest to Crenshaw's original, that focuses in depth on a single neglected group at a particular intersection.

MAIHDA.

Multilevel analysis of individual heterogeneity and discriminatory accuracy; a model nesting individuals within intersectional strata and partitioning outcome variance.

Main effect.

The average effect of one disadvantaged position estimated across levels of the others; the building block an additive model sums.

Matrix of domination.

Collins's image of oppression as an interlocking structure of race, class, and gender rather than a single-axis hierarchy.

Single-axis framework.

The default approach intersectionality critiques, in which one social category is studied at a time and others are controlled away as noise.

Social cognition.

The processing of information about other people and the self as a social object; the psychology the framework asks to treat targets as multiply situated.

Sociology.

The systematic study of social life under which the Medical Subject Headings file the intersectional framework as an indexing classification.

Stratum.

A group defined by a specific combination of categories across several axes, such as low-income disabled women; the unit crossed in intersectional models.

Variance partition coefficient.

In a multilevel model, the share of total outcome variance lying between strata rather than within them; a summary of how intersectional an outcome is.

Key Researchers

Greta R. Bauer (Western University). Epidemiologist and biostatistician who moved intersectionality from theory into population-health methodology and led the field's systematic review of its quantitative applications. ORCID - Faculty Page - Google Scholar

Lisa Bowleg (George Washington University). Applied social psychologist who pressed psychology and public health to treat intersectionality as more than an additive list of identities. ORCID - Faculty Page - Google Scholar - Wikipedia

Elizabeth R. Cole (University of Michigan). Professor of psychology and women's and gender studies whose 2009 article set the agenda for how psychology should actually conduct intersectional research. ORCID - Faculty Page - Google Scholar

Patricia Hill Collins (University of Maryland, College Park). Distinguished University Professor emerita of sociology whose Black feminist thought supplied the matrix-of-domination framing and who later mapped intersectionality's definitional dilemmas. Faculty Page - Wikipedia

Kimberlé Williams Crenshaw (UCLA School of Law; Columbia Law School). Legal scholar and critical race theorist who coined the term intersectionality and gave it its canonical statement in the 1991 essay Mapping the Margins. Faculty Page - Wikipedia

Leslie McCall (The Graduate Center, CUNY). Presidential Professor of sociology and political science who classified intersectional research into its anticategorical, intracategorical, and intercategorical approaches. Faculty Page - Wikipedia

Frequently Asked Questions

What is the intersectional framework?
It is an analytic approach holding that social positions such as race, gender, and class are mutually constituting rather than independent, so that a person's experience at their intersection cannot be reduced to the sum of the separate categories (Crenshaw, 1991).

Who coined the term intersectionality?
The legal scholar Kimberlé Crenshaw introduced it to describe how Black women's discrimination claims fell between single-axis race doctrine, framed around Black men, and sex doctrine, framed around white women (Crenshaw, 1991).

How is intersectionality different from just having several identities?
Holding multiple memberships is only the premise. The framework's actual claim is that the positions interact, so their combined effect is more or less than the sum of their separate effects, a statement about statistical interaction rather than mere co-occurrence (Bowleg, 2012).

What are McCall's three approaches?
The anticategorical approach deconstructs social categories, the intracategorical approach focuses on a single neglected group at one intersection, and the intercategorical approach compares inequality across the full grid of category combinations (McCall, 2005).

How did intersectionality enter psychology?
Elizabeth Cole's 2009 article translated the framework into concrete design questions for psychological research, supported by a best-practices literature on sampling, measurement, and analysis (Cole, 2009; Warner, 2008).

Can intersectionality be studied with statistics?
Yes. A quantitative program estimates intersectional effects using interaction terms and, increasingly, multilevel models that nest individuals within intersectional strata and partition the outcome variance (Bauer, 2014; Evans et al., 2018).

What does it mean for combined disadvantage to be more than additive?
It means the doubly-disadvantaged group's outcome departs from the sum of the separate main effects; a positive interaction indicates compounding beyond additivity, a negative one a combination less severe than the parts would predict (Bauer, 2014).

Why do critics worry about the term's breadth?
As intersectionality spread across disciplines it came to name an identity, a population, a relationship, and a method at once, and scholars have asked whether so widely borrowed a term retains the precision of its original formulation (Collins, 2015; Nash, 2008).

References