Global Slavery Index (Walk Free): Intelligence Source Guide

The Global Slavery Index is Walk Free's country-by-country estimate of modern slavery prevalence, government response and import risk. It is the most cited and most criticised dataset in the field, and you need to understand both facts before you use a single number from it.

At a glance

Source	Global Slavery Index (Walk Free)
Category	Conflict, Crime & Human Security › Human Trafficking & Child Protection
Homepage	https://www.walkfree.org/global-slavery-index/
Format	`HTML`
Access	Open — no account required
Disciplines	Human Intelligence, Supply Chain Intelligence
Mission domains	Human Trafficking, Forced Labour & Modern Slavery

Per-country modern-slavery prevalence. — as catalogued in the platform’s own source registry.

The Global Slavery Index is a periodic publication by Walk Free, an anti-slavery organisation founded in Australia and part of the Minderoo Foundation. Each edition provides, for essentially every country in the world, three distinct products that are often confused. The first is an estimated prevalence of modern slavery, expressed both as an absolute number of people and as a rate per thousand population, where modern slavery is an umbrella term covering forced labour, debt bondage, forced marriage, forced commercial sexual exploitation, human trafficking and slavery-like practices. The second is a vulnerability model: a composite score built from governance, inequality, conflict, discrimination and basic-needs indicators that attempts to explain why a country's population is exposed. The third is a government response assessment, scored across a set of milestones covering survivor support, criminal justice, coordination, risk-factor reduction and government supply chains, and reported on a graded letter scale. Recent editions add an import-risk analysis, estimating the value of at-risk products imported by G20 economies and naming the highest-risk product categories. The prevalence figures are derived from nationally representative household surveys conducted through the Gallup World Poll in a subset of countries, combined with data from the International Labour Organization and IOM, and extrapolated to countries where no survey was conducted using the vulnerability model.

The job this source does is give you a single, comparable, country-level number where none otherwise exists. That is genuinely useful and genuinely dangerous. It is useful because supply-chain due diligence, sanctions and trade policy, and NGO resource allocation all require a country-level risk ordering, and no government produces one. It is dangerous because the number's precision far exceeds its accuracy, and because the extrapolation means that for a large share of countries the estimate is a model output rather than a measurement. The most defensible use is comparative and ordinal: which regions and which countries sit at the high end, how a country's response score has moved across editions, and which product categories carry structural risk. The least defensible use is the one most commonly made: quoting a country's absolute victim count as if it were a statistic. For SUPPLYINT work the import-risk and government-response components are the more useful products, and they are more transparent than the prevalence estimate because they rest on documented policy assessment rather than on survey extrapolation. Treat the index as a structured argument about relative risk, sourced and auditable, rather than as a measurement.

Who publishes it, and why that matters

Walk Free is a privately funded advocacy organisation established by Andrew and Nicola Forrest through the Minderoo Foundation, an Australian philanthropic body whose wealth derives from mining. It is not a statistical agency, does not claim to be, and has an explicit campaigning mission to end modern slavery. Three implications follow. First, resourcing is unusually good for this field, which is why the surveys exist at all – commissioning nationally representative household modules across dozens of countries is expensive and no UN body was doing it at that scale before Walk Free paid for it. Second, the organisation has a stake in the number being large enough to sustain attention, and the methodological choices that widen the definition of modern slavery are made by an organisation with that interest, which is a reason for scrutiny rather than an accusation of bad faith. Third, Walk Free is now a formal partner in the joint global estimates produced with the ILO and IOM, which has moved parts of the methodology into an intergovernmental process with more external review. The organisation has responded to methodological criticism by publishing more of its method and by changing it between editions, which is creditable and simultaneously the reason editions are not comparable.

Provenance is the first question to ask of any dataset and the one most often skipped. Who collects it, what their incentive is, whether they publish a methodology, and whether they correct the record when they get something wrong all bear directly on how much weight a finding drawn from it can carry.

What a record actually contains

The fields you will be working with, what each one means, and whether it is something you can pivot on. Read the meanings carefully — more analysis is wrecked by misreading a field than by failing to find one, and a field that looks like an observation is often an inference.

Field	Type	What it means	Pivot value
`Country`	string	The unit of analysis. Everything in the index is national, which means it cannot show you the sub-national concentration that actually characterises forced labour – a country's figure is an average over regions with radically different conditions.	ISO country code; join to trade, governance, conflict and labour-market data at national level only.
`Estimated prevalence per 1,000 population`	int	The headline rate. This is the comparable measure and the one to use if you use any, because the absolute count is simply this rate multiplied by population and inherits all its error plus population-estimate error.	Rank ordering across countries and regions; change across editions only if the methodology note says the editions are comparable, which it often does not.
`Estimated number of people in modern slavery`	int	The absolute figure that ends up in headlines. It is a point estimate produced without a published confidence interval at country level in most editions, which is the single strongest reason not to quote it as a statistic.	Nothing safely. If you must use it, present it as an estimate with the method named and the extrapolation status of the country stated.
`Survey status`	enum	Whether the country's estimate rests on a nationally representative survey conducted there, on a survey in a comparable country, or purely on model extrapolation. This is the most important field in the whole dataset and it is the one users most often fail to look up.	Confidence weighting for every downstream use; separate your analysis into surveyed and extrapolated groups before doing anything else.
`Vulnerability score and dimensions`	int	A composite index built from governance, inequality, conflict, discrimination and basic-needs indicators, reported overall and by dimension. Because the same vulnerability model feeds the extrapolation, vulnerability and prevalence are not independent for extrapolated countries.	Governance indicators, conflict data, migration policy, and the specific component indicators, which are more informative than the composite.
`Government response rating`	enum	A graded letter score summarising assessed action across milestones. Built from documented policy and legal analysis rather than survey work, which makes it the most auditable component of the index and arguably the most useful.	Named legislation, national action plans, national referral mechanisms, and the milestone-level detail behind the composite grade.
`Government response milestone scores`	array	The component assessments underlying the grade, covering survivor identification and support, criminal justice response, coordination and accountability, risk-factor reduction, and government procurement and supply chains. Far more actionable than the letter.	Specific policy gaps you can name in an advocacy or compliance product; comparison against the country's own national action plan commitments.
`Products at risk`	array	Named product categories assessed as carrying modern-slavery risk in G20 import flows, with source countries. Recent editions have highlighted electronics, garments, palm oil, solar panels and textiles among the highest-value at-risk imports.	Harmonized System commodity codes, supplier country lists, and the sectoral risk registers used in due-diligence regimes.
`Import value at risk`	int	The estimated annual value of at-risk products imported by a country or bloc. A modelled figure combining trade statistics with risk assessments, useful for prioritisation and unsuitable for any claim about actual tainted volume.	National trade statistics, customs data, and the specific supply chains behind the commodity codes.
`Region`	enum	The regional grouping used for aggregation. Regional aggregates are more robust than country figures because errors partially cancel, so regional statements are safer to make than national ones.	Regional comparison and trend framing, which is where this dataset is at its strongest.
`Edition year`	int	Which publication a figure comes from. Editions have used different methods and different definitional boundaries, so a figure without an edition is meaningless and a cross-edition comparison without a compatibility check is an error.	The edition's own methodology annex, which is the only place the comparability question is answered.
`Forced marriage component`	int	Where reported separately, the portion of the estimate attributable to forced marriage as opposed to forced labour. Forced marriage constitutes a very large share of the global figure, and analysts who read the headline as forced labour misunderstand the composition entirely.	Family law regimes, minimum-age-of-marriage legislation, and the very different intervention set that forced marriage requires.

Coverage — and what is not in it

Coverage is nominally universal – essentially every country receives a prevalence estimate, a vulnerability score and a government response rating – and this universality is itself the thing to interrogate, because it is achieved by modelling where measurement was impossible. Survey-based estimates exist for a substantial but minority set of countries, concentrated where the Gallup World Poll operates and where a household survey on this subject is feasible and safe. Countries that are closed, at war, or where asking these questions endangers respondents receive extrapolated figures, and those are frequently the countries with the highest estimates, which is an uncomfortable structural feature rather than an accident. The government response component covers nearly all countries and rests on documentary policy analysis, so its coverage is more evenly grounded. Import-risk analysis covers G20 economies and a defined set of product categories rather than all trade. Temporally, editions have appeared irregularly since 2013 with substantial methodological revision between several of them, and the joint global estimates produced with the ILO and IOM operate on their own cycle. Update rhythm is therefore multi-year, not annual, and any monitoring workflow built on this source needs to accept that resolution.

Known blind spots

Absence of evidence here is not evidence of absence. These are the conditions under which Global Slavery Index (Walk Free) will not show you something that is nevertheless real:

Countries where no survey could be conducted receive modelled figures, and those are disproportionately the closed, conflict-affected and authoritarian states where the problem is plausibly worst. The estimate for such a country is a prediction from correlates, not an observation, and it will move when the model changes rather than when the country does.
Sub-national concentration is entirely invisible. Forced labour clusters in specific sectors, regions and supply chains, and a national rate averages a fishing fleet, a mining district and a capital city into one number that describes none of them.
State-imposed forced labour is definitionally and practically hard for household surveys to capture, because the people subject to it are frequently in detention, in institutions, or in regions where enumerators do not go and where respondents cannot answer honestly.
Household surveys miss people who are not in households. Institutionalised populations, people in closed worksites and labour camps, homeless populations and mobile migrant workers are exactly the groups at highest risk and the hardest for a residential sampling frame to reach.
Forced marriage dominates a large part of the global figure and is measured very differently from forced labour, so a single national number blends two phenomena with different causes, different demographics and different remedies. The blend is not visible in the headline.
Cross-edition comparison is unsafe more often than it is safe. Methodology, definitional scope and data inputs have changed between editions, and apparent movement in a country's figure frequently reflects a change in method rather than a change in conditions.
The vulnerability model and the prevalence extrapolation are not independent, so for extrapolated countries the finding that vulnerable countries have high prevalence is partly an artefact of construction rather than an empirical result.
Government response scores reward documented policy, which advantages states with good drafting and bureaucratic capacity over states with poor drafting and better practice. A high score is evidence of a legal and institutional framework, not of outcomes for survivors.
The import-risk product is built on trade statistics and sectoral risk judgements, not on supplier-level evidence, so it cannot identify a tainted supply chain and should never be treated as having done so.

Write the blind spot into the product. A statement that something “was not observed in Global Slavery Index (Walk Free)” is defensible; a statement that it “did not happen” is not, and the difference is what survives cross-examination.

Access, licensing and what you may do with it

Access model: Open — no account required

The index is published as a website with country pages, downloadable data files and a full report with a methodology annex, and there is no API. Practically, the route is to download the data files for the edition you need and to read the methodology annex before touching them. That annex is not optional reading here in the way it sometimes is elsewhere – it is where the survey coverage, the extrapolation approach, the definitional boundaries and the comparability statement live, and every serious limitation of the data is disclosed there rather than concealed. The joint global estimates published with the ILO and IOM are a separate document with their own methodology and should be collected alongside. Country pages carry the response-milestone detail that the downloadable summaries sometimes compress, so for advocacy or compliance work aimed at a specific country, read the page as well as the file. Archive what you download: earlier editions have been reorganised on the site over time, and reproducing a figure from a superseded edition later can be surprisingly difficult.

Licence

Walk Free publishes the index for public use and the reports and data are freely downloadable, with reuse conditions stated on the site. Recent practice has been to release under a Creative Commons arrangement requiring attribution, but the terms have not been identical across editions and you should read the notice on the specific edition you are using rather than assuming. Attribution should name Walk Free, the index and the edition year, because the edition is load-bearing information rather than a citation nicety. Note separately that the underlying survey microdata is not released – what you get is the modelled and aggregated output – so independent replication of the prevalence estimates is not possible from the public materials, which is one of the substantive criticisms made of the index and a fact you should state if you are relying on it for anything consequential. Where the index incorporates ILO and IOM material, that material carries its own terms.

Rate limits and fair use

Not applicable; this is a set of documents and files rather than a service. Fetch the edition you need once, cache it with the edition year and download date, and check for a new edition on a yearly cadence. Editions appear at multi-year intervals, so anything more frequent is wasted effort. If you are building a country risk product that refreshes continuously, this source is a slowly-varying input and should be modelled as such rather than polled.

Licensing changes, and it changes without warning. A dataset that was free for research this year may not be free for commercial or evidential use next year. Confirm the current terms before you build a dependency on it, and record the terms you relied on alongside the data — the licence in force at the time of collection is part of the provenance.

Collecting it

How Global Slavery Index (Walk Free) is actually pulled, in the order you would set it up. Prefer the bulk or export interface over per-item lookups wherever one exists: it is kinder to the publisher, faster for you, and gives a reproducible snapshot rather than a series of point-in-time answers you cannot reconstruct later.

Method	Format	Cadence	Notes
Edition data download	CSV	Per edition	The prevalence, vulnerability and government-response tables. Store with the edition year in the filename and treat editions as separate datasets rather than as versions of one series.
Full report and methodology annex	HTML	Per edition	Mandatory companion to the data. The annex states survey coverage, the extrapolation method and the comparability position, and without it the numbers cannot be responsibly interpreted.
Country page harvest	HTML	Per edition	Country pages carry the milestone-level government response detail and the narrative assessment, which is where the actionable policy content lives for advocacy and compliance work.
Joint global estimates	HTML	Per estimate cycle	The ILO, Walk Free and IOM global estimates are a separate publication with their own method. Collect them alongside and keep them clearly distinguished from the index in your data model.
Import-risk tables	CSV	Per edition where published	Product categories, source countries and at-risk import values for G20 economies. The most directly usable component for supply-chain risk registers, and the one needing the clearest caveat about what it does not show.

Ingesting it into the platform

Every step below is idempotent and cursor-based: interrupt one and it resumes from where it stopped rather than duplicating rows or losing progress. Collection is recorded per source, so a feed that quietly stops publishing shows up as a stale timestamp instead of silently thinning your coverage.

Register each edition as its own dataset — Add the index in sources.php with the edition year as part of the dataset identity in datasets.php, so that no query can silently mix editions and every displayed figure carries the edition it came from.
Load prevalence with its survey status — Import the prevalence table through import.php with the survey-versus-extrapolated status as a mandatory column. If the status is missing for a country, the record loads flagged rather than clean, because a modelled figure and a measured one must never be indistinguishable downstream.
Separate the three products — Prevalence, vulnerability and government response are different measurements with different reliability and should be separate tables joined on country and edition, not merged into a single country score at load time.
Normalise country identity — Map to ISO 3166 codes with explicit handling of disputed and changed entities, so the index joins cleanly to trade, conflict and governance data in country.php without silently dropping the jurisdictions that matter most.
Feed the country risk model as one weighted input — Push the vulnerability and government-response components into country-risk.php as inputs among several rather than as a dominant score, and record their weight explicitly so that a change of edition does not move the composite without anyone noticing.
Build the supply-chain view — Load import-risk product categories against Harmonized System codes so that a commodity in a client's supply chain resolves to a risk category, and expose that through the platform's supply-chain and human-rights surfaces rather than as a standalone table.
Tag and cross-link — Use resolve-tags.php to attach forced-labour, trafficking and supply-chain tags so that index-derived country context appears alongside case-level and enforcement data on the same jurisdiction rather than in isolation.
Enforce the caveat at render time — Configure reports.php so any output using a prevalence figure emits the edition, the survey status and the estimate language automatically. This is the one control that prevents the most common misuse, and it has to live in the platform rather than in analyst discipline.

Registered sources and their last-collected state are listed in sources.php, and the scheduled chain that keeps them current is in automation.php.

How it is wrong, and how to tell

Every dataset is wrong in characteristic ways. Knowing which ways is the difference between using a source and being used by one, and it is the part of source evaluation most often skipped because it is the part that takes work.

This is a serious piece of work with a serious methodological problem, and both halves of that sentence are true. The serious work: Walk Free funded nationally representative household surveys on a subject nobody was surveying, publishes its methodology in detail, discloses which countries are extrapolated, has revised its approach in response to criticism, and has moved part of the estimation into a joint process with the ILO and IOM. The government response component in particular is transparent, documented and checkable. The serious problem: the headline prevalence figures are extrapolated for many countries, are published without country-level uncertainty intervals in a form the public can see, rest on a definitional umbrella that mixes forced labour with forced marriage, and cannot be independently replicated because the microdata is not released. The index has been criticised on exactly these grounds by academic specialists and by people sympathetic to its aims, and the criticism is substantially correct. The honest assessment is that the index is the best available country-level ordering and a poor country-level measurement, that its response and import-risk components are stronger than its prevalence component, and that anyone using the headline numbers as statistics rather than as estimates is misusing them.

Characteristic false positives

Extrapolated estimates read as measurements. A country with no survey receives a figure derived from its vulnerability profile, which means the number tells you about governance and inequality indicators as much as about slavery, and it will change when the model is revised even if nothing in the country has.
Cross-edition comparison across a methodology change. Apparent rises and falls between editions frequently reflect definitional or method revisions, and the annex will usually say so, but the number in the spreadsheet will not.
Forced marriage and forced labour conflated. A large part of the global estimate is forced marriage, which has different demographics, different causation and entirely different remedies. Products that quote the total while discussing labour exploitation are describing a different phenomenon from the one they are counting.
Government response grade mistaken for effectiveness. The grade measures documented policy, legislation and institutional architecture. A state can score well on paper while identifying almost no victims and prosecuting almost no one, and several do.
Import-risk figures presented as tainted trade. The at-risk import value is trade volume in categories assessed as risky, not an estimate of goods actually produced with forced labour, and the distinction disappears within one repetition in most reporting.
National averages applied to sub-national decisions. A sourcing decision about a specific region or a specific sector cannot be made from a country rate, and doing so both over-restricts low-risk suppliers and under-flags high-risk ones in the same country.
Population-driven artefacts in absolute counts. Because the count is a rate multiplied by population, the largest countries dominate every absolute ranking, which produces headlines about which countries have the most victims that are mostly statements about which countries have the most people.
Circularity between vulnerability and prevalence. For extrapolated countries the two are mechanically related, so using vulnerability to explain prevalence in those countries is not an empirical finding and any model that does so will look better than it is.

None of these make the source unusable. They make it a source that requires corroboration before an assertion built on it goes into a product, which is true of every source and admitted by few.

Ageing

Editions age in blocks rather than continuously, and the components age at different rates. Government response ratings age fastest and most meaningfully: legislation passes, national action plans expire, referral mechanisms are established or defunded, and a rating from a superseded edition can be wrong within a year of a legal change. Vulnerability scores age at the speed of their input indicators, which is slow for structural measures like inequality and fast for conflict, so a vulnerability score for a country that has entered war is stale immediately. Prevalence estimates age most ambiguously – the underlying phenomenon changes slowly, but the estimate is tied to a survey round and a model vintage, so its currency is the currency of the edition rather than of the world. Import-risk analysis ages with trade patterns, which reconfigure within a few years. In practice a stale use looks like a compliance document citing a prevalence figure from an edition two cycles back, or a country risk score still reflecting a government response rating that predates a major legislative reform. Re-baseline your whole country layer when a new edition appears, and never let two editions coexist in one product.

What this source feeds

A source is only worth what it lets you conclude. These are the disciplines that collect through it, the mission domains it serves and the data points it yields — every one is a tag, so you can follow any thread from here into the rest of the library.

Collected by these intelligence disciplines

Serves these mission domains

Yields these data points

Location / Coordinates

How each sector uses Global Slavery Index (Walk Free)

The same dataset is worked very differently depending on who you are, what authority you hold, and what you are ultimately producing. A military analyst is supporting a commander’s decision; a journalist is meeting a publication standard; an NGO caseworker is protecting a person. The records are shared — the constraints, thresholds and outputs are not.

🎖 Military and defence

Relevance is mainly through the contractor and host-nation labour supply chain, and through stabilisation contexts where forced labour is part of the conflict economy. Vulnerability dimensions covering conflict, displacement and governance failure identify where deployed operations are most likely to encounter exploitation in their own logistics tail, and the government response rating tells you whether the host state has any functioning referral pathway to route a case into. Use the country layer to set the baseline expectation for labour standards in contracted services, and pair it with the ILO indicator framework for actual screening, because the index gives you the country risk level and no operational detail whatsoever.

🕵 National intelligence

As a SUPPLYINT and ECONINT input this is a country-risk layer, best used to prioritise where deeper collection is warranted rather than as reporting in itself. Its strongest analytical use is comparative and structural: which states have deteriorating response ratings, which vulnerability dimensions are moving, and where import risk concentrates for an economy of interest. It also supports assessment of state-imposed forced labour indirectly, since those cases are precisely where the index's measurement fails and where the gap between a low reported figure and a high vulnerability score is itself a signal. Never put an extrapolated prevalence figure in an assessment without labelling it as modelled.

👮 Law enforcement

Limited direct operational value and real strategic value. The government response milestones tell you where a partner jurisdiction's institutional weaknesses lie – whether it has a national referral mechanism, whether victim identification is functioning, whether the criminal justice response exists on paper only – which shapes what cooperation is realistic and where mutual legal assistance is likely to stall. For domestic prioritisation, the vulnerability dimensions point to the population groups and sectors most exposed. It contains nothing case-level, nothing about individuals and nothing you can act on tactically.

🔍 Private investigation and corporate security

This is the most common professional use of the index and it is a legitimate one, provided it is used at the right resolution. For supply-chain due diligence and enhanced client screening, the country prevalence ordering, the government response grade and the import-risk product categories together give you a defensible basis for prioritising which suppliers, jurisdictions and commodities require deeper investigation. Regulators increasingly expect a documented, risk-based prioritisation, and this is one of the few sources that supplies one. What it cannot do is clear or condemn a specific supplier, and a due-diligence file that stops at the country score has not conducted due diligence.

📰 Journalism and OSINT media

The index is a reliable generator of headlines and an unreliable source of facts, and the gap between those is where most bad reporting happens. Absolute victim counts by country are almost always the wrong story – they are population-driven and, for many countries, modelled. The better stories are in the components: which governments' response scores have fallen, which product categories carry the highest at-risk import value into your country, and which countries' figures are extrapolated rather than measured, which is a legitimate story in itself. Always name the edition and always say estimate. Specialists will and do challenge reporting that does not.

🌍 NGO, humanitarian and human rights

For advocacy organisations the government response milestones are the operational product: they give you a documented, comparable account of what a government has and has not done, which is directly usable in submissions, shadow reports and campaign material. The vulnerability dimensions help target programming towards the groups the model identifies as exposed. Use the prevalence figures sparingly and honestly – the field has been damaged by advocacy that quoted contested estimates as settled facts, and that damage lands on the credibility of organisations doing careful work. Where you operate in a country whose estimate is extrapolated, your own field evidence is better than the index and you should say so.

🎓 University and research

The index is best treated as an object of study as much as a source of data. The prevalence estimates should not be used as a dependent variable in cross-country regressions without confronting the extrapolation problem head-on, because for extrapolated countries the estimate is a function of the same governance and inequality covariates that typically appear on the right-hand side. The government response scores are more defensible as a measured construct and are underused. Read the methodology annex of each edition and the substantial critical literature on the index before citing, record the edition explicitly, and note that the survey microdata is not publicly available, which places a real limit on replication.

Playbook: working Global Slavery Index (Walk Free) end to end

A repeatable sequence from first pull to finished product. Each phase states what you are trying to establish, not merely what to click — the objective is a defensible chain of reasoning, not a completed checklist.

Phase 1 — Decide which of the three products you actually need

Prevalence, vulnerability and government response are different things with different reliability, and most analytical requirements are met by the second or third rather than the first. Naming which one you need at the outset prevents the default drift towards the headline number, which is the weakest component.

Phase 2 — Read the edition's methodology annex before the data

The annex tells you the survey coverage, the extrapolation approach, the definitional boundary and whether the edition is comparable to its predecessor. Every serious limitation is disclosed there. An analyst who works from the spreadsheet alone will make errors the publisher explicitly warned against.

Phase 3 — Split countries into surveyed and extrapolated

Build the two lists before any analysis and keep them separate throughout. Findings that hold in both groups are findings; findings that appear only in the extrapolated group are properties of the model. This single step eliminates a large share of the mistakes made with this source.

Phase 4 — Decompose the estimate where the edition allows

Separate forced labour from forced marriage wherever the edition reports the split. They have different demographics, different sectors and different remedies, and an analysis of labour exploitation built on a total that is heavily forced marriage is measuring the wrong thing.

Phase 5 — Use the rate, not the count

Work in prevalence per thousand for any comparison. Absolute counts are rates multiplied by population and produce rankings dominated by the largest countries, which tells you about demography rather than about exploitation. Reserve counts for the one legitimate purpose of conveying scale, with the estimate caveat attached.

Phase 6 — Pull the government response milestones, not the grade

The letter grade compresses away everything useful. Go to the milestone-level assessment for each country in scope and extract the specific findings – whether a national referral mechanism exists, whether victim identification is funded, whether public procurement is covered. That detail is what makes an advocacy or compliance product concrete.

Phase 7 — Cross-check the response assessment against primary sources

Verify the index's account of a country's legal framework against the actual legislation, the national action plan, and independent country assessments such as the Trafficking in Persons Report and regional monitoring bodies. The index's assessment is usually right and occasionally lags a legal change, and the lag is where your product would be embarrassed.

Phase 8 — Map import risk to your own commodity exposure

Take the at-risk product categories, resolve them to Harmonized System codes, and intersect with the commodities your organisation or client actually imports. The output is a prioritised list of sectors needing supplier-level work, not a conclusion about any supplier. Be explicit about that boundary in the deliverable.

Phase 9 — Triangulate every country of interest against case-level and legal sources

Put the index's country picture beside the CTDC case data, ILO country-level supervisory comments, national referral statistics and the TIP report narrative. Where they agree you have a robust picture; where they disagree the disagreement usually reveals whether the country has a measurement problem, an identification problem or a political problem.

Phase 10 — Go sub-national before making an operational decision

The index cannot support a decision about a region, a sector or a facility. Once the country layer has told you where to look, move to sector-specific sources – labour inspection reports, sectoral studies, worker-voice data, trade-union reporting – and treat the country score as having done its job by pointing you there.

Phase 11 — Version-lock your product to one edition

Record the edition in the product, refuse to mix editions in a single table, and plan a re-baselining exercise when a new edition appears rather than partially updating. Mixed-edition country layers produce differences between countries that are artefacts of publication timing and are almost impossible to detect later.

Phase 12 — State the contested status of the source explicitly

If your product will be read by anyone with subject expertise, say that the index is widely used and methodologically contested, name the extrapolation issue, and explain why you are still using it. This costs a paragraph and buys credibility. Presenting the figures without that acknowledgement is the fastest way to have a good analysis dismissed.

The platform ships this as a step-checked workflow in playbooks.php, so progress is recorded against a case rather than held in someone’s head.

What to pair it with

No single source carries a finding. These are the datasets that corroborate, extend or contradict this one — and a source that contradicts is worth more than one that agrees, because it is the only thing that will tell you when you are wrong.

Source	Relationship	What it adds
ILO Forced Labour	corroborates	The joint global estimates that Walk Free co-produces with the ILO and IOM, plus the standards and supervisory machinery that give the response assessment its legal reference points.
Counter-Trafficking Data Collaborative	contradicts	Case-level data on identified victims. Built by an incompatible method, so where the two disagree about a country the disagreement is diagnostic of identification capacity rather than of error.
Trafficking in Persons Report	corroborates	The US State Department's annual country assessment and tier ranking, the main independent check on the index's government response scoring and generally more current.
US DOL List of Goods Produced by Child Labor or Forced Labor	extends	Country-and-commodity pairs assessed as produced with forced or child labour, which is more granular than the index's product categories and directly usable in sourcing risk registers.
UNODC Global Report on Trafficking in Persons	corroborates	Detection and conviction statistics reported by states, providing the criminal-justice counterpart to the index's prevalence and response measures.
Walk Free reports	prerequisite	The organisation's wider publication set, including the methodology documents and thematic reports that explain what a given edition's figures were built from.
US Customs and Border Protection forced labour enforcement	extends	Actual trade enforcement actions against named entities and commodities, which converts country-level risk into concrete supply-chain consequences.

Legal, ethical and operational constraints

The index contains no personal data and using it presents no privacy issue. The legal exposure is entirely on the output side, in two forms. First, defamation and commercial harm: the index rates governments and identifies at-risk product categories, and repeating those assessments about a named company, supplier or country in a commercial product can attract legal challenge, particularly where you have converted a country-level risk score into an assertion about a specific entity. Keep the resolution honest – the index supports a statement that a sector in a country carries elevated risk, and does not support a statement that a supplier uses forced labour. Second, regulatory reliance: modern slavery reporting regimes in several jurisdictions, and forced-labour import bans in others, require documented risk assessment. Using the index as one input into that assessment is normal and appropriate; using it as the entirety of the assessment is very likely inadequate under those regimes, because they expect supplier-level enquiry proportionate to identified risk. Attribution obligations under the publication licence apply to any reuse. Finally, be careful about how country-level findings feed migration and border policy arguments, since prevalence figures for origin countries have been used to justify measures that harm the people the figures describe.

Operational security

Downloading published reports and data files reveals essentially nothing beyond an interest in the topic, and there is no query interface to leak a research question. The meaningful exposure is reputational and political rather than technical. If your organisation publishes a country risk product built on this index, you are adopting a contested methodology in public and inheriting the criticism attached to it, and several governments respond to the index's ratings with direct pushback against organisations that cite them. In jurisdictions where anti-slavery advocacy is treated as political interference, holding or circulating index-derived country assessments can itself create exposure for local staff and partners. Think about who in your organisation is named on a product that rates a government's response before you publish it. Internally, keep the edition and the survey status attached to every figure, because the most damaging leak here is not of data but of a modelled number escaping into a public document as a fact.

Two rules that hold regardless of jurisdiction. Collection that is lawful is not automatically proportionate, and a dataset assembled for one purpose does not carry consent for another. Where the records concern identifiable people, the question is not only whether you may hold the data but whether holding it serves the purpose you are accountable for.

Is it earning its place?

Sources accumulate. Feeds get added during an incident and are never reviewed again, and a decade later the pipeline is carrying dead weight that nobody dares remove. These are the measures that show whether Global Slavery Index (Walk Free) is contributing anything, and they are worth baselining now so the answer is available later.

Share of countries in your product whose figures are extrapolated rather than survey-based, tracked and disclosed. If you cannot state this number, you do not know how much of your country layer is model output.
Number of products that mix editions, which should be zero. Any non-zero value indicates a versioning failure that will produce artefactual differences between countries.
Frequency with which your government-response layer disagrees with the current Trafficking in Persons Report assessment, which measures how stale the index component has become.
Proportion of supply-chain risk decisions where the country score triggered supplier-level enquiry rather than substituting for it, which is the test of whether the source is being used at the right resolution.
Time from a new edition's publication to a completed re-baselining of your country layer, which should be measured in weeks rather than being allowed to drift across a cycle.
Count of externally published figures that carry the edition year and the estimate language, as a check that the render-time caveat is actually firing.
Whether your analysts can state, unprompted, the forced marriage share of the headline global figure – a fast diagnostic of whether the team understands what the number contains.

Beware of volume. Indicator counts rise easily and say almost nothing. Unique contribution — findings this source produced that no other source in your stack would have — is the measure that matters, and it is usually far lower than anyone expects.

Tradecraft notes

The distinctions that separate a competent analyst from a fast one:

Always check the survey status of a country before quoting its figure. The difference between a measured and a modelled estimate is the difference between evidence and inference, and the spreadsheet presents them identically.
The government response component is the best part of the index and the least used. It is documented, auditable, tied to identifiable policy facts and directly actionable in advocacy and compliance work, whereas the prevalence figure is contested and rarely changes a decision.
Never compare editions without reading both methodology annexes. Method revisions have produced apparent movements larger than any plausible real change, and a country whose figure halved between editions almost certainly experienced a definitional change rather than a policy triumph.
Remember what modern slavery includes here. Forced marriage is a large component, and analysts working on labour exploitation who use the total are counting a population that includes many people whose situation their intervention does not address.
Absolute counts rank populations. Use the rate for every comparison, and treat any list of the countries with the most victims as a list of large countries unless it has been population-adjusted.
Watch for circularity when modelling. If you regress the index's prevalence on governance and inequality indicators, remember that for extrapolated countries those indicators generated the estimate, so a strong fit is partly mechanical.
Use the index to prioritise, never to conclude. Its correct place in a workflow is the first screen that decides where to spend investigative effort, and its incorrect place is the last page of the assessment.
Say out loud that the source is contested. Acknowledging the criticism and explaining your handling of it makes the product more credible with expert readers, not less, and it protects you when a specialist raises the extrapolation issue after publication.
Where you have field evidence for a country that conflicts with an extrapolated estimate, your field evidence wins. The model was never intended to override observation, and treating it as authoritative in a country where you have direct knowledge is a failure of judgement rather than of the source.

Questions analysts actually ask

Are the prevalence numbers real measurements?

For some countries, yes – they rest on nationally representative household surveys conducted there. For many others they are extrapolated from a vulnerability model calibrated on surveyed countries. The published tables do not visually distinguish the two, so you have to look up the survey status yourself, and doing so should be the first step of any use.

Why is the index so heavily criticised?

Mainly on four grounds: extrapolation to unsurveyed countries, the absence of published country-level uncertainty, a broad definitional umbrella that combines forced labour with forced marriage, and non-release of the survey microdata, which prevents independent replication. The criticism is substantially correct and does not mean the index is useless – it means the headline numbers are estimates rather than statistics.

Can I compare a country across editions?

Only after checking the methodology annexes of both editions for a comparability statement. Methods and definitions have changed between several editions, and apparent movements frequently reflect those changes. Where the annex says editions are not comparable, they are not, whatever the spreadsheet allows you to compute.

Is this the same as the ILO global estimates?

Related but distinct. Walk Free is a partner in the joint global estimates published with the ILO and IOM, and those estimates feed into the index, but the index is Walk Free's own publication with its own country-level modelling, vulnerability index and government response assessment. Cite whichever you actually used and do not treat the two as interchangeable.

Can I use it for supplier due diligence?

As a first screen, yes, and regulators generally accept a documented risk-based prioritisation. As the whole of your due diligence, no. The index operates at country and product-category resolution and cannot say anything about a specific supplier, facility or shipment, and a compliance file that stops there is unlikely to satisfy the regimes that require proportionate enquiry.

What does the government response grade actually measure?

Documented policy and institutional architecture, assessed against defined milestones covering survivor support, criminal justice, coordination, risk-factor reduction and government supply chains. It measures what a state has put in place, not what happens to survivors, and states with strong legal drafting and weak implementation score better than the reverse.

How often is it published?

Irregularly, at multi-year intervals rather than annually. Build your workflow around edition-based re-baselining rather than continuous updating, and treat the index as a slowly-varying structural input into any country risk model.

Which component should I use if I only use one?

The government response assessment, and specifically its milestone-level detail. It is the most transparent, the most auditable, the most directly actionable and the least contested part of the publication, and for most professional purposes it changes decisions in a way the prevalence estimate does not.

How should I present a figure to a non-specialist audience?

As an estimate, with the edition year, the rate rather than the count where a comparison is being made, and a sentence stating whether the country was surveyed or modelled. If that framing makes the finding sound weaker than you wanted, the finding was weaker than you wanted.

Standards, formats and interoperability

What this source speaks natively, and what it has to be translated into before a partner can consume it. Work that arrives in a recognised format is easier to defend, easier to hand over and easier to automate against:

The Palermo Protocol definition of trafficking and the ILO forced labour conventions, which supply the legal boundaries the index's umbrella term sits across.
ILO indicators of forced labour, which the survey instruments draw on for question design and which are the operational counterpart to the index's country scores.
Sustainable Development Goal target 8.7 on eradicating forced labour and modern slavery, the policy framework in which the index positions itself.
ISO 3166 country codes for joining the country layer to trade, governance and conflict data.
Harmonized System commodity codes, the necessary bridge between the import-risk product categories and any real supply chain.
Modern slavery reporting regimes in the United Kingdom, Australia, Canada and elsewhere, and the emerging EU forced-labour import framework, which are the regulatory contexts where this data most often gets used.
The OECD Due Diligence Guidance for Responsible Business Conduct, which sets the expectation of proportionate, risk-based enquiry that country-level scores are meant to trigger rather than satisfy.

References

Primary documentation and authoritative references for this source. Publishers revise and retire material, so treat the retrieval date as part of the citation and re-check before relying on any of it in a formal product.

Global Slavery Index — Walk Free. The index itself: country pages, data downloads and the report. Go to the methodology annex before the numbers.
Walk Free — Walk Free, Minderoo Foundation. The publisher's wider work, including thematic reports and the documentation explaining how editions relate to the joint global estimates.
ILO forced labour and modern slavery — International Labour Organization. The joint global estimates and the legal standards. The authoritative reference for what forced labour means as a matter of international law.
Trafficking in Persons Report — US Department of State. Annual country assessments and tier rankings, the main independent check on the index's government response component and usually more current.
List of Goods Produced by Child Labor or Forced Labor — US Department of Labor, ILAB. Country-commodity pairs with documented evidence, the granular complement to the index's product-category risk analysis.
Forced labour trade enforcement — US Customs and Border Protection. Where country and commodity risk becomes an enforceable import restriction, and therefore where the compliance consequences of this analysis actually land.
UNODC — United Nations Office on Drugs and Crime. The Global Report on Trafficking in Persons, giving state-reported detection and conviction figures as an independent measurement of a related construct.
Counter-Trafficking Data Collaborative — IOM and partners. Case-level victim data. The most useful contrast set, because its disagreements with the index's country picture are informative rather than merely inconvenient.
IOM — International Organization for Migration. Co-producer of the joint global estimates and the largest provider of victim assistance, whose operational reporting grounds the modelled figures in casework.

Link integrity: every reference above was verified with a live request when this page was generated. Where a publisher had moved or withdrawn a document, the link was repointed at a preserved copy in the Internet Archive and marked as archived. Anything with no reachable copy anywhere had its link removed rather than left to rot — the source is still credited, it simply cannot be linked.

Put it into practice

The Quantus Intel threat intelligence platform operationalises this source: it loads each edition as a separate versioned dataset, keeps survey status attached to every country figure, feeds the response and vulnerability components into country risk as weighted inputs rather than a headline, and stamps the edition and estimate language onto anything that leaves the platform.. Browse the full source catalogue, or follow any tag above into the rest of the library.