How the WLCE Index measures

How rules account for women’s lives, and what women actually receive

Paired measurement:
one decision, measured twice

A rule provides a woman with leave, a pension, a doctor’s appointment or a loan. The Index compares it with a benchmark and with what she actually received.

The Women’s Life-Cycle Economy Index (WLCE Index), which Womafreesm is developing, will assess the content of rules and the results of applying them across women’s lives. For each decision, researchers will compare the rule with an external benchmark, then compare what it provides with women’s experience. A missing mechanism and data gaps are noted separately.

These two comparisons of one decision make up a paired measurement, the unit of the Index. An Unconsidered Point is a point in a woman’s life cycle where the system did not see her when it designed the rule. It arises in three cases, which can coincide: the rule falls short of the benchmark, the rule exists but the woman did not receive it, or there is no rule at all. The Index shows this as two distances: the Consideration Deficit, from the benchmark to the rule, and the Silent Denial, from the rule to what the woman received.

Take paid maternity leave in one country. The law is one side of the pair, and the benchmark here is at least 14 weeks under ILO Convention 183. What the women of that country who gave birth actually received is the other side. If the country has no such rule, its absence is checked against the sources and against other ways of securing the same entitlement, and the value is recorded under a coding rule that will be published before the pilot.

The institutional side records what the rule provides, by whom, for how long, on what conditions and in which document, with the source and the date. A missing rule is recorded in the same detail: what was searched for, in which sources and as of what date, so that the zero is a checkable fact.

The experience side records what happened to the woman who went through the decision: whether she knew about her right, whether she was able to use it, and what it cost her. Where no rule exists, this side records what happened to her without one and what she used to cover its absence.

A survey by itself shows what women report. A law by itself shows what is written in the text. Institutional consideration is the comparison of the two sides, and that is what the Index measures. The comparison can also show that the rule matches the benchmark and is delivered. A shortfall against the benchmark and a gap in application can occur in the same decision, and both are recorded separately.

Every paired measurement is described by three coordinates.

Life-cycle stage

When the decision came up in her life: pregnancy and birth, recovery after childbirth, care for children and older relatives, a career pause, return to work, menopause, later life and migration abroad.

Decision system

Where the rule operates: the labor market, healthcare, social protection and pensions, law, finance and insurance, AI and digital systems, the market. This coordinate says whom to address for a fix.

Decision Function

The action the system takes for a woman here: it grants access, allocates, compensates, protects or recognizes. A benefit can be approved (access) while the payment under it comes to zero (compensation).

A hypothetical example:
leave and a 14-week benchmark

Provided by the rule	Received in practice	What the comparison shows
14 weeks	14 weeks	On this measure the rule matches the benchmark and is delivered
14 weeks	8 weeks	A gap arises in applying the rule
8 weeks	8 weeks	The rule is delivered, but its content falls short of the benchmark
8 weeks	6 weeks	There is a shortfall against the benchmark and an additional gap in application

One measure, the length of maternity leave, against a 14-week benchmark. What the comparison of the two sides shows.

The first distance, between the benchmark and the rule, shows the Consideration Deficit. The second, between the rule and what was received, shows the Silent Denial. A single subtraction would hide the third row: the rule and the practice coincide, yet the woman receives less than the benchmark.

Why the Index reports two distances

One subtraction, rule minus practice, hides the worst case.

Where the law is weak and practice is just as weak, the difference is zero, and zero would read as “no gap.”

So the Index reports two distances. The first is measured on the rule itself. The second is measured on the path from the rule to the woman.

Between the entitlement and the outcome there is a chain of four links, and a loss is attributed to one of them.

Naming the link is what turns a finding into something a ministry or an employer can fix. The difference between what was provided and what was received does not by itself explain the cause: attributing a case to a link needs further information. Rules for cases where several links are affected, or where the data is insufficient to establish the cause, will be part of the methodology.

Recognition

the system sees that this woman qualifies.

Entitlement

the right is formally granted to her.

Operational capacity

the system is able to deliver it.

Take-up

she actually uses it.

Consideration Deficit

How far an institutional rule falls short of the benchmark before anyone tries to use it. The rule itself was designed for less than a whole life. This distance is measured on the rule alone, against the benchmark on the measurement card, before any woman is surveyed.

Example: the benchmark for maternity protection is at least 14 weeks of maternity leave under ILO Convention 183. A law that provides fewer weeks carries a Consideration Deficit before a single woman applies for leave.

Silent Denial

The entitlement exists, the woman did not receive it, and no refusal was ever issued. Measured as the signed distance between what the rule provides and what was received, attributed to the link where the loss occurred: recognition, entitlement, operational capacity, or take-up.

The sign matters. Positive: the woman received less than the rule provides. Negative: practice runs ahead of the law, as when an employer grants more leave than the law requires. That happens, and it has to stay visible.

Research on social benefits uses the concept of non-take-up: the situation in which people entitled to a benefit go without it. Its causes can lie in information, procedures, the work of organizations and a person’s circumstances. The WLCE Index methodology will describe separately how the Silent Denial measure relates to this concept and in which areas it applies.

Zero can mean two different things: a rule that works, or a rule and a practice that fail together.

A missing rule, a data gap and a missing benchmark

A rule that does not exist also decides her life.

The absence of a mechanism is recorded after the relevant sources and alternative ways of securing the same entitlement or opportunity have been checked. The value of the measure will be set by a coding rule published in advance. Gaps in the data and the absence of an approved benchmark are noted separately.

The methodology distinguishes four states:

The absence of a mechanism provided for by the methodology is confirmed

The result of the check under the established coding rule.

Information about the mechanism could be found or confirmed only in part

A data gap.

The measure has no approved benchmark in the WLCE Index methodology

An open methodological question; such a paired measurement is published as a gap with a note.

The measure falls outside the scope of the particular decision

Not applicable, with an explanation.

The coding rules for all four states will be published together with the preliminary formula before the pilot. The map of Unconsidered Points shows both confirmed absence of a mechanism and gaps.

Every indicator carries a measurement card with eight fields

Every indicator is entered in a registry: the list of everything the Index measures. Each entry carries a card with eight required fields. An indicator enters the calculation only when every field is filled in:

The institutional measure

The rule, its source and date: what it provides, by whom, for how long, on what conditions and in which document.

The experience measure

The survey question, with its exact wording and answer options, put to the women who went through the decision.

The Decision Function

What the system decides for a woman at that point: access, allocation, compensation, protection, or recognition.

The level

Whether the measure describes a woman, an organization, a product, or a country; each level is estimated apart.

Group for the calculation

Which women it concerns: whose situation it studies, whom the rule entitles, who applied, who received the result.

The benchmark

The value the rule is measured against, and the external document that justifies it, such as ILO Convention 183.

The missing-data rule

How a missing answer is counted in the calculation. The rule will be published along with the preliminary formula.

The tags

The life-cycle stage and the decision system, so results can be read by stage or by system from data coded once.

For each measure it is decided in advance who is included in the calculation and why.

The research looks separately at coverage by the rule, access to applying and receipt of the result: this shows at which stage women drop out of the process the rule provides for. Leave after childbirth is measured among those who gave birth. Pension credit for years of care among those who cared for a relative. Screening among women of the age it is addressed to. Credit among women who needed it, including those whom barriers kept from applying.

Take the group wider or narrower, and the figure describes other women while the error goes unnoticed.

Four levels,
each estimated on its own

A law exists at the level of a country. An HR policy exists at the level of an organization. An AI system exists at the level of a product. A woman’s answer exists at the level of a person. So each country-level paired measurement is estimated first, with its own margin of error, and only then enters the overall score that combines the separate measures.

Individual answers can be linked to the characteristics of a country, an organization or a product. The analysis takes into account which information belongs to each level and which observations are related: one country value repeated across thousands of questionnaires remains one country-level observation. An average of women’s answers describes women, not the text of a law, and a country average says nothing about an individual woman.

That is why the link between a rule and what happened to women is studied on pseudonymized woman-level data inside the secure research environment, with a model that keeps the levels apart.

Where the institutional side comes from

The institutional side is coded from published sources: each rule is transferred into the fields of the card from a named document.

Data from external databases is used after its match with the definition of the measure, the group studied and the observation period has been checked. For every record, a reference to the source and a description of the required conversions are kept. The program itself codes the rules that these databases do not cover.

World Bank Women, Business and the Law

Records what the law says in 190 economies and separately assesses supportive policies and enforcement: in the 2026 edition, global averages are 67 out of 100 for legal frameworks, 47 for supportive policies and 53 for expert assessments of enforcement.

ILO Global Care Policy Portal

Publishes more than 60 legal and statistical indicators for more than 180 countries. For each kind of leave it records how long it lasts, how much is paid, who pays for it and which workers it covers in law. The Index uses its data after checking it.

ILO Convention 183

Sets the benchmark for maternity protection: leave of at least 14 weeks (Article 4), cash benefits, the conditions for them and their financing (Article 6). It is the external document behind the 14-week benchmark in the maternity leave example above.

ILO 5R Framework for Decent Care Work

Sets the frame for unpaid and paid care: recognize, reduce and redistribute unpaid care work, and reward and represent paid care workers. Five words that decide which side of care a country’s policy actually touches: unpaid care, paid care or both.

EU AI Act, Annex III

The external basis for classifying AI systems: hiring and dismissal, access to public assistance and healthcare, credit scoring, risk assessment and pricing in life and health insurance, emergency triage. Articles 9, 10, 27 and 72 set the requirements.

NIST AI Risk Management Framework 1.0

A voluntary framework, not a legal requirement. It names the four core functions by which the Index describes an AI system: govern, map, measure and manage. Whether a system has these procedures is scored apart from what it actually does with women.

So the Index’s own audit of AI systems and the map of laws and rules are part of the calculation itself: they supply the institutional side of each pair.

The Index’s own methodology defines its coverage across countries.

How the WLCE Index compares with other indices

Nine subindices come from one matrix

The base layer is a matrix: a table with life-cycle stages down the side, decision systems across the top, and a paired measurement at each intersection.

The nine subindices are thematic measures calculated from this common system of measurements. Each one gathers the paired measurements of one theme, such as care load, across every stage and system, so each subindex traces back to the same measurements. Their working names are listed in the glossary; final names and definitions will be published with the pilot methodology note.

Both views will be published, the matrix and the nine subindices, so that a reader can ask either “Where in her life does she fall out of the calculation?” or “How does this country score on care load?” Every indicator carries a stage tag and a system tag, so both views come from the same data, coded once.

Five decisions fixed in the formula before the global wave

A preliminary formula and its verification plan will be published before the pilot; version 1, prepared after the pilot, is fixed before the first global survey wave, so that it can be criticized before the first result appears. After that, any change becomes a new numbered version with a public explanation of the reasons.

Still being developed, and due in the preliminary formula with a worked numerical example: how the Consideration Deficit and the Silent Denial are converted to a 0 to 100 scale and combined within one paired measurement, how a negative value of the Silent Denial measure is handled, how a missing benchmark and data gaps are handled, and how double counting of one measure across several subindices is prevented.

Fixed scale end points

100 is full consideration, 0 is a systemic blind spot. The values behind both are fixed in advance for all waves, so changing the scale cannot move a score. Comparability also rests on the measures, definitions, sample and procedures: any change is a new version.

Equal weights

Every subindex carries the same weight: a failure in maternity protection weighs as much as one in the labor market. A second set of weights from an expert panel will be published next to it, showing the effect. Weights are a value choice, announced in advance.

Averaging inside, geometric mean between

Within a subindex, measures are averaged. Between subindices a geometric mean is used, so a high score in one area only partly offsets a low score in another. A subindex of zero brings the total to zero; how zeros are handled is set by the preliminary formula.

Floors for critical life stages

A country that falls below the minimum set in the methodology note for a critical life stage gets a separate status alongside its score, so that the shortfall does not disappear into an average. A high score in any other subindex cannot offset that status.

Factor analysis as a diagnostic only

It looks for questions that duplicate each other and folds related ones into one scale. The weights between subindices come from the equal-weights rule above; factor analysis leaves them as they are. Weights from data would reward correlation, not importance.

A score you can take apart is a score you can defend.

Every score can be taken apart. A country score can be traced back to the individual indicators it came from, and the Index is checked against existing external indicators through correlation and regression: where it agrees with them and where it diverges.

What will be published for each country

The score, with two uncertainty intervals
An interval that accounts for the way participants were selected, the sample size and the weights applied, and a range of results under the tested calculation variants.
A range of possible positions
Instead of one exact position.
Probability of landing in a group of countries
The share of tested calculation scenarios in which the country falls into each group (tier) of the comparison.
The map of Unconsidered Points
By life-cycle stage: where women’s circumstances fall out of the calculation.
The link where the loss occurred
For each finding, the link in the chain: recognition, entitlement, operational capacity, or take-up.

Where too few women have gone through a decision for the estimate to be reliable, the result will be published with the note “insufficient data for a reliable estimate.”

A country enters the comparison under four conditions: a probability sample (a separate sample of that country’s women, selected at random and apart from the open survey), sufficiency of data, its quality and comparability.

The open survey gives breadth, subgroups and rare situations; the probability sample gives the country estimate. A country that has not passed the check will get a profile and a map of Unconsidered Points, and its place in the comparison waits until the conditions are met.

What each number actually says

The Index says exactly what it measured, and every number comes with a note on the material it rests on.

One survey wave shows the state at the time of the survey: what the rule provides and what the women it concerns received. Repeat surveys will make it possible to study change over time. Whether the causal effect of a particular decision can be estimated will depend on the study design and the assumptions of the analysis.

An AI system’s paperwork and its behavior are scored separately: whether the system has the required procedures is one score, and what it actually does with women is another. Meeting a regulator’s requirements counts toward the first score, and the second shows whether the two match.

The Year 1 pilot tests the instrument itself, including whether the questions measure what they claim, whether they work the same across countries and languages, and how stable countries’ positions are.

For each country the Index will publish a score with its uncertainty intervals, a range of possible positions and the probability of landing in a group of countries, rather than one exact place in a table: that is what keeps the number honest. The weights are a value choice, announced in advance in the methodology note, and open to challenge there and through the report of an external statistical auditor, which will be published alongside the Index.

What results the pilot will be judged by How to review the methodology before data exists

FAQ

1. What is the unit of measurement in the WLCE Index?

A paired measurement: one institutional decision measured twice, as the rule is written and as the woman who went through it experienced it. Where the rule is below the benchmark, fails to reach her or is missing, the paired measurement records an Unconsidered Point, the Index’s unit of observation: a point in a woman’s life cycle where the system did not see her when it designed the rule.

2. What is a Decision Function in the WLCE Index?