Economic and Debt Slavery in Mexico and Central America, 2015-2025

KJBottom Line Up Front

1How to read this document

1.1 Source grading: the NATO/Admiralty scale

Every source cited in Section 12 carries a two-character Admiralty grade: a letter for the reliability of the source itself, a number for the credibility of the specific information cited. The two axes are independent, a highly reliable source can carry a specific claim that is only possibly true, and a source of unproven reliability can still carry information later confirmed elsewhere.

Source reliability:

• A: Completely reliable. No history of inaccurate reporting (a federal agency's own press release or official statistics page, speaking in its own voice).
• B: Usually reliable. Minor history of inaccurate reporting, or reliable but not a primary party to the events described (an established wire service, a credentialed investigative outlet, a research institution with transparent methodology).
• C: Fairly reliable. Some history of inaccurate reporting (a regional news outlet relaying an official's statement rather than the agency's own release).
• D: Not usually reliable. Significant history of inaccurate reporting. (Not used as a load-bearing grade in this document; a source graded this low would not be cited for a factual claim.)
• E: Unreliable. Lacks a history of accuracy. (Not used in this document.)
• F: Reliability cannot be judged. (Used for a source with insufficient track record to grade, noted individually where it occurs.)

Information credibility:

• 1: Confirmed by other independent sources.
• 2: Probably true (source is reliable, information is plausible, but not independently confirmed by a separate source in this document).
• 3: Possibly true (some doubt; typically a single named source with no independent corroboration).
• 4: Doubtful (some internal contradiction, or later superseded).
• 5: Improbable (contradicted by other sources).
• 6: Credibility cannot be judged (insufficient information to assess).

The four-tier shorthand used in section subheadings (PRIMARY/OFFICIAL, NAMED INVESTIGATIVE, PEER-REVIEWED, SECONDARY/UNVERIFIED) is a quick reference only; the Admiralty grade in Section 12 is the authoritative, source-by-source grading. Rough correspondence: PRIMARY/OFFICIAL typically grades A1-A2; NAMED INVESTIGATIVE and PEER-REVIEWED typically grade B1-B2; SECONDARY/UNVERIFIED typically grades B3-C4 or lower.

1.2 Analytic confidence and the probability lexicon

This document distinguishes reported information (a fact directly stated by a source) from analytic judgment (this document's own conclusion about what the reported information means, which requires interpretation and could be wrong even if every underlying fact is correct). A judgment is labeled "Assessment:" and carries two things: a confidence level and, where the judgment is expressed as a likelihood rather than a certainty, a calibrated probability term.

Confidence level (a function of source quality, quantity, and independence, not of how strongly the judgment is worded):

• High confidence: judgment is based on high-quality information, ideally corroborated by multiple independent sources using different methods (alternative explanations were assessed and found significantly less likely).
• Moderate confidence: judgment is credibly sourced but not independently corroborated, or the sourcing is corroborated but shares a common origin, or the analysis is sound but the sourcing is not extensive enough for High confidence.
• Low confidence: judgment rests on a single source, fragmentary information, or reasoning this document could not independently verify. A Low-confidence judgment is not a weak claim dressed up, it is this document's own honest statement that the evidence does not yet support more than that.

Calibrated probability lexicon (the words themselves, not just contextual tone, carry the meaning):

• Almost certainly: roughly 95-99 percent probability.
• Highly likely / very likely: roughly 80-95 percent.
• Likely / probably: roughly 55-80 percent.
• Roughly even chance: roughly 45-55 percent.
• Unlikely: roughly 20-45 percent.
• Highly unlikely / remote: below 20 percent.

A RECTUMmendation (an unverified assertion presented as settled fact, this project's standing term for it) is not the same failure as a properly labeled Low-confidence Assessment. The former hides its own uncertainty; the latter states it plainly. This document aims to never do the former and to do the latter whenever a judgment, rather than a directly sourced fact, is being made.

1.3 What stays a plain fact, not an assessment

Not every claim in this document needs a confidence label. A directly quoted official statistic, a specific named prosecution's outcome, or a directly quoted on-the-record statement is reported information and is presented as such, with its Admiralty grade in Section 12 doing the epistemic work. Assessment labels and confidence levels are reserved for this document's own interpretive judgments, most concentrated in the Signals, Key Assumptions Check, and Analysis of Competing Hypotheses sections, plus the Bottom Line Up Front.

Sources rejected during research, and why, are logged in Section 8. Advocacy and mission-driven organizations, and partisan political-body factsheets, were excluded from load-bearing claims. Where an advocacy source was the only place a claim appeared, the claim was either dropped or re-sourced to a primary document; none ship here on advocacy sourcing alone.

1.4 Multi-model verification pass

This document was drafted by Claude Opus 5. Source location was parallelised across five background research agents, each given a separate area and the sourcing standard above, whose job was to find primary documents rather than to write. Nothing they returned was used as sourcing on its own account: every figure and quotation that carries a grade in Section 12's first group was independently retrieved and read by the drafting pass before being written, and everything they located but the drafting pass did not retrieve is segregated into Section 12's second group at a reduced grade and flagged at the point of use.

Three verification passes then ran against the completed draft, on models other than the drafting model.

The first was dispatched to Claude Fable 5.1 and terminated early on a session rate limit. Before it stopped it had rendered the Guatemalan Gini chart to an image and read it visually, which independently confirmed a correction the drafting pass had already made against itself: the 2014 value is the trough at 0.372 and the 2023 value rises to 0.421, so measured consumption inequality fell and then rose rather than declining throughout. That correction is printed at Section 3.1.

Two further passes then ran on Claude Sonnet 5, with divided remits. One took figures, tables, quotations and arithmetic; the other took legal reasoning, currency and supersession, the load-bearing negative findings, and the firewall.

The numeric pass retrieved every primary source in its remit, extracted the tables by word coordinate rather than by linear text, and confirmed the Mexican poverty series, the Guatemalan poverty and department tables, the Guatemalan household-access table, the CONEVAL jornalero figures, the child-labour figures, the IOM financing and collateral figures, the remittance-as-debt-service figures, the kidnapping figures and the enforcement table cell for cell; every block quotation word for word; and every piece of derived arithmetic. It found the T-visa table assembled from several separate annual reports without saying so, which is corrected at Section 7.8 and which turned out to expose something more useful than an error, namely that the agency systematically revises these counts between publications.

The legal pass produced the two corrections that changed what this document argues. The first is at Section 6.4. An earlier draft tested the Guatemalan land-secured migration debt against the 1956 Supplementary Convention alone, concluded it was not debt bondage, and let that conclusion stand for the wider question. That was an overreach: the same facts had not been tested against the ILO statistical standard, which imposes no requirement that labour be the thing pledged, nor against the Trafficking in Persons Protocol, which asks about means and purpose rather than about security. Section 6.4 now runs all three and reaches a narrower conclusion. The second is at Section 5.9, where the United States list of goods produced by child or forced labour had been read from a web listing that truncates alphabetically, giving Mexico four goods and one forced-labour designation. The published report gives fifteen goods and two, chile peppers and tomatoes, and the document's own Section 5.4 had been describing a customs order against Mexican tomato producers throughout.

The drafting pass also corrected itself in four places before verification: an automated summary that renamed three of the eleven ILO indicators, a first truncated reading of the same goods list, the Gini transposition above, and an Analysis of Competing Hypotheses that omitted the victims-per-trafficker effect UNODC advances in its own report, the omission of which had overstated the anomaly it was trying to explain.

Every correction and addition from the verification passes that appears in this document was independently re-verified against its primary source before being added, in keeping with the sourcing standard stated throughout, rather than taken on any model's word.

1.5 Abbreviations used in this brief

Every abbreviation the brief actually contains, with its expansion and domain. Foreign bodies are given their own name in their own language where that is the name the source uses, because a reader who goes looking for the institution needs the name it is filed under. This is a table rather than expansion on first use, because a reference document is read by jumping into the middle of it from the contents.

2What the poverty numbers measure, and Mexico's series

2.1 Terms of art, stated before the table because three of them change the numbers

This document uses three measures that are routinely reported as though they were one, and are not.

Multidimensional poverty is Mexico's official measure. A person is in multidimensional poverty if they have at least one of six social deprivations AND an income below the income poverty line. Both conditions are required. A person with four deprivations and an adequate income is not counted as poor by this measure; they are counted as "vulnerable by social deprivation," a separate category.

Income poverty (poblacion con ingreso inferior a la Linea de Pobreza por Ingresos, LPI) counts only the money. It ignores the deprivations entirely and is always the larger number. In 2024 the gap between the two was 5.8 percentage points, or 7.5 million people.

Extreme poverty requires three or more deprivations AND an income below the extreme income line, which is the monetary value of the food basket alone.

The distinction matters for this brief specifically because a household's exposure to coercive debt tracks the deprivation side, in particular the absence of social security, far more closely than it tracks the headline poverty rate. The headline rate can fall while the deprivation that produces the vulnerability does not.

2.2 The series (PRIMARY/OFFICIAL: INEGI, Comunicado de Prensa 118/25)

Mexico's official multidimensional poverty measurement, all persons, national, 2016 through 2024. Source: INEGI, "Pobreza Multidimensional," Comunicado de Prensa 118/25, 13 August 2025, Cuadro 1, computed from ENIGH 2016, 2018, 2020, 2022 and 2024.

The six deprivations, separately, same source and years:

And the income space:

2.3 What the shape of the series supports, and what it does not

The headline movement is real and large. Multidimensional poverty fell 14.3 percentage points between 2020 and 2024, from 43.9 to 29.6 percent, and the count fell from 55.7 to 38.5 million persons. Extreme poverty fell from 10.8 to 7.0 million over the same period.

Three things in the same table complicate reading that as a general improvement in the conditions this brief is about.

First, deprivation of access to health services moved in the opposite direction across most of the range. It stood at 15.6 percent in 2016, more than doubled to 39.1 percent by 2022, and was still 34.2 percent in 2024, well above where it started. In absolute terms the affected population went from 18.8 million in 2016 to 44.5 million in 2024. This document does not assert a cause. The timing is coincident with the replacement of the Seguro Popular scheme and the subsequent reorganisations of Mexican public health coverage, but establishing that as the mechanism would require the coverage administrative data rather than the household survey, and that is an open item (Section 13).

Second, the population classified as "vulnerable by social deprivation," meaning people with at least one unmet social right whose income is above the poverty line, rose across the whole period, from 25.3 percent in 2016 to 32.2 percent in 2024, and in absolute terms from 30.5 to 41.9 million. INEGI states the 2022 to 2024 movement directly: all six social deprivations fell, and the vulnerable-by-deprivation population still rose from 37.9 to 41.9 million. That is arithmetically consistent, because a person crossing the income line while retaining a deprivation moves into that category rather than out of it, but it means the fall in the headline rate is substantially an income event rather than a rights-access event.

Third, and most directly relevant here, roughly half the Mexican population has no social security access in every single year of the series. The indicator moves from 54.1 to 48.2 percent across eight years, a decline of 5.9 points. In 2024 that is 62.7 million people. Social security access in the Mexican measurement is the closest available proxy for formal employment with contributory benefits, and its absence is what turns an unexpected cost, a medical emergency or a crop failure into an event that must be financed privately. That is the entry point for every debt mechanism described in Section 5.

2.4 The institutional discontinuity in the series, and why it is not a break

The 2024 measurement was produced by a different institution than every prior year in the table. A constitutional reform published in the Diario Oficial de la Federacion on 20 December 2024, with secondary legislation entering into force on 17 July 2025, extinguished the Consejo Nacional de Evaluacion de la Politica de Desarrollo Social (Coneval) and transferred poverty measurement and social development policy evaluation to INEGI. Coneval had performed the measurement biennially from 2008 to 2022.

INEGI states that it carried out the 2024 measurement in adherence to the Lineamientos and the Metodologia para la Medicion de la Pobreza en Mexico issued by Coneval, applying the same criteria and processes, in order to preserve the continuity and consistency of the series that began in 2016. It publishes the calculation programs in Stata, R and Python, and states that standard errors are published for all indicators for the first time.

2.5 The states, and INEGI's own precision warning

In 2024 the five states with the highest multidimensional poverty were Chiapas at 66.0 percent, Guerrero at 58.1, Oaxaca at 51.6, Veracruz at 44.5 and Puebla at 43.4. The five lowest were Baja California at 9.9 percent, Baja California Sur at 10.2, Nuevo Leon at 10.6, Coahuila at 12.4 and Sonora at 14.1.

For extreme poverty the same year: Chiapas 27.1 percent, Guerrero 21.3, Oaxaca 16.3, Veracruz 8.8, Puebla 7.3. Lowest: Baja California 0.4 percent, Nuevo Leon 0.5, Aguascalientes 0.6, Coahuila 0.8, Colima 1.0.

INEGI attaches a precision warning to that second list which is worth carrying, because it is the kind of caveat normally dropped in reproduction. The agency states that the lowest extreme-poverty state percentages have only a moderate level of precision, with a coefficient of variation in the range of 15 to 30 percent, and that Baja California specifically has a low level of precision, with a coefficient of variation of 30 percent or more. The 0.4 percent figure for Baja California should not be treated as a point estimate.

Two state-level divergences run against the national trend. Educational lag fell nationally from 19.4 to 18.6 percent between 2022 and 2024 but rose in six states, among them Chiapas from 31.1 to 34.0 percent, Oaxaca from 29.1 to 30.5, and Veracruz from 25.5 to 26.3. And the food access deprivation fell in every state except Chihuahua.

Sonora, the Mexican state on the other side of the corridor this project's geography uses, is among the five least-poor states in the country at 14.1 percent multidimensional poverty in 2024. That is a useful corrective: the corridor's Mexican terminus is not itself a poor place by national standards. The poverty that feeds it is imported from the south.

3The Central American series, and the measurement gap

3.1 Guatemala: twenty-three years, no change (PRIMARY/OFFICIAL: INE, ENCOVI 2023)

Guatemala's Instituto Nacional de Estadistica presented ENCOVI 2023 on 21 August 2024, the first national living-conditions survey since ENCOVI 2014, fielded 6 August to 27 December 2023. Poverty is measured by consumption against a national poverty line, not by the Mexican multidimensional method, and the two are not comparable.

In 2023 that is 9.7 million people in poverty, of whom 2.8 million are in extreme poverty and 6.9 million non-extreme. ENCOVI 2023 is the first round to estimate separate urban and rural poverty lines: general urban Q15,911.2 and general rural Q12,700.1 per person per year; extreme urban Q6,381.2 and extreme rural Q5,781.5, all in quetzales of October 2023.

Two cautions on comparing this with Mexico. The instruments differ, consumption against a national line in Guatemala versus a joint income-and-deprivation construct in Mexico, so the levels are not comparable even where the directions are. And the nine-year gap between ENCOVI 2014 and ENCOVI 2023 spans the pandemic entirely, so the series cannot say what happened in between.

A third caution belongs to the source itself. The ENCOVI 2023 results deck states on one page that comparison with earlier rounds required simulating the 2023 updates onto prior years' consumption, converting to October 2023 quetzales and applying the new 2023 poverty lines. The note under its own trend chart then says the opposite: that the data for each year are the estimates reported in each survey as conducted. The deck describes a retro-projection and charts the original series. The table above reproduces the charted figures, and any 2014 to 2023 comparison should be made in the knowledge that the producing agency has not stated unambiguously which basis it is on.

3.2 Guatemala by department, and the highland concentration

Poverty incidence by department, 2023, all twenty-two:

The spread is a factor of 4.2 between Alta Verapaz at 90.3 percent and the department of Guatemala at 21.6. By characteristic: rural 66.3 percent against urban 46.6; ages 0 to 17, 64.7 percent; no schooling 68.0 percent, primary 60.8, secondary 38.1, tertiary 9.9.

Huehuetenango, at 81.2 percent the third-poorest department, is also by a wide margin the principal exit point for Guatemalan migration: 33.6 percent of returned migrants left through La Mesilla and a further 18.3 percent through Gracias a Dios, both in Huehuetenango, with San Marcos supplying another 30.0 percent through Tecun Uman and El Carmen (Section 6).

3.3 The ethnicity gap, and where the figure actually lives

ENCOVI 2023's official monetary-poverty release contains no ethnicity breakdown at all. In a country where 37.5 percent of the population self-identifies as Maya, the flagship poverty publication does not report poverty by ethnic group.

The figure exists, but in a different instrument published by a different set of institutions: the multidimensional poverty index (IPM-Gt), updated on ENCOVI 2023 data by MIDES, INE, SEGEPLAN and MINFIN with UNDP technical support, published June 2025. It gives national multidimensional poverty incidence of 57.7 percent, intensity 45.1 percent, index value 0.260, against 61.6 / 48.6 / 0.299 on ENCOVI 2014. On ethnicity it reports that among the Maya population three of every four people are multidimensionally poor, 75.6 percent, against 46.1 percent among the mestizo population, a ratio of 1.6. Self-identification in ENCOVI 2023: ladino or mestizo 60.3 percent, Maya 37.5, Xinca 1.3, Afro-descendant 0.9, Garifuna 0.1.

The same publication states that comparing 2023 with the 2018 results computed on ENCOVI 2014 shows no significant reduction in the incidence of multidimensional poverty, and that rural incidence, 76.3 percent against urban 40.8, fell 6.2 points since 2014 while the urban figure showed no statistically significant change.

3.4 What poverty consists of, materially (PRIMARY/OFFICIAL: INE, ENCOVI 2023)

The most useful table in ENCOVI 2023 for this brief is not the headline rate but the household-access table, because it states what the condition actually consists of. Percentage of households with access to or possession of each, by poverty status, 2023:

Two features matter beyond the obvious. Mobile telephony is the one service that does not stratify sharply: 67.2 percent of extreme-poor households have a mobile phone against 77.5 percent of non-poor ones, a gap of ten points, where drainage differs by forty-nine points and residential internet by twenty-eight. A household can be without drainage, without piped water, cooking on firewood, and still be reachable by phone. That is the material precondition for a remittance economy and for a debt serviced from abroad.

And 97.2 percent of extreme-poor households cook with firewood, against 57.8 percent of non-poor households. Firewood dependence is near-universal at the bottom and remains the majority condition even at the top.

3.5 Honduras, El Salvador and Nicaragua: three different instruments and one blackout

Honduras measures poverty as a percentage of households, not of people, and this single fact accounts for most of the apparent gap between its national figures and every international series. INE Honduras reports households in poverty at 73.61 percent (2021), 64.1 (2023), 62.86 (2024) and 60.14 (2025), with extreme poverty at 53.7, 41.5, 40.1 and 38.26 across the same years. Rural extreme poverty in 2025 was 48.89 percent of households, and the Gini stood at 0.495. Household per-capita income rose from L2,671 (2021) to L4,864 (2025).

The unit difference is quantifiable. For 2024 the Honduran household poverty rate of 62.86 percent corresponds to a person rate of 67.09 percent, a gap of 4.2 points. The direction is worth noting: converting Honduras to a person basis makes it look poorer, not less poor, because poor households are larger.

Two cautions. The 2020 and 2022 rounds are marked "dato no disponible" and no survey was fielded in those years, and INE has retired its pre-2021 poverty publications, so no official Honduran poverty figure for 2015 to 2020 was obtainable in this pass.

El Salvador's survey moved institutions. DIGESTYC was dissolved by a law approved in August 2022 and the Oficina Nacional de Estadistica y Censos (ONEC) was launched inside the Banco Central de Reserva on 15 November 2022; the EHPM is now published by ONEC. The old DIGESTYC domain no longer resolves. El Salvador's extreme poverty line is the food basket and nothing else: at 2022 values, USD 61.95 per person per month urban and USD 40.13 rural, which is USD 2.06 and USD 1.34 a day.

Households in poverty, from ONEC's own rounds: 26.2 percent (2020), 24.6 (2021), 26.6 (2022), 27.18 (2023), 25.80 (2024), 22.47 (2025). The 2025 round, published 27 March 2026, reports 22.5 percent of households nationally, 24.8 rural and 21.1 urban, with multidimensional poverty at 20.1 percent of households and mean monthly household income of USD 763.81.

Two cautions specific to El Salvador. Its household rate of 27.18 percent for 2023 corresponds to a person rate of 30.3 percent, a gap of 3.1 points. And the World Bank's current published figure for El Salvador is the 2023 round, which is now two rounds behind ONEC. That is an ingestion lag at the World Bank, not a gap in Salvadoran measurement, and it is a materially different thing from the Nicaraguan case below. A brief that describes both as "stale data" would be wrong about one of them.

Nicaragua is statistically dark. The last consumption-based national poverty measurement is the EMNV of October 2016, fielded on 2,800 dwellings over three weeks, national-level only, with no urban/rural breakdown, published July 2017. It reported general poverty of 24.9 percent and extreme poverty of 6.9 percent. It was never harmonised into the World Bank's comparable series, whose last Nicaragua datapoint remains EMNV 2014.

The consequence, stated precisely: there has been no monetary poverty measurement of post-2018 Nicaragua by anyone. Not INIDE, not the World Bank, not CEPAL. Nicaragua appears in none of CEPAL's country poverty or inequality tables in Panorama Social 2025, while still being folded into the regional headline by projection from a 2014 survey. What INIDE publishes instead is a set of five unmet-basic-needs component indices, marked preliminary, drawn on a sampling frame built from the 2005 census, in a country that has since experienced mass emigration.

3.6 Four countries, four incompatible definitions

The single most important methodological point in this section is that the four national measures cannot be laid side by side. Guatemala uses a consumption aggregate against a cost-of-basic-needs line, person-based. Honduras uses household income against a canasta, household-based, on a 1978 expenditure basket. El Salvador uses household income against a food basket for extreme poverty and twice the food basket for relative poverty, household-based. Nicaragua used consumption against a national line, person-based, a decade ago.

CEPAL publishes its own harmonised estimates alongside the official national figures, and the divergence is the clearest available demonstration:

Honduras's official extreme-poverty rate is more than double CEPAL's, 40.1 against 19.9, and that is very largely because one counts households and the other counts people. Mexico's official column here is not the multidimensional headline of 29.6 percent but the income-line figure of 35.4 percent, which is the income-only analogue.

3.7 What is measured everywhere: remittances (PRIMARY/OFFICIAL: World Bank)

Where national poverty measurement is patchy, remittance data is not, because central banks capture it as a balance-of-payments item. Personal remittances received as a percentage of GDP, indicator BX.TRF.PWKR.DT.GD.ZS, retrieved from the World Bank Indicators API:

Two readings, and the distinction matters. Mexico's 3.7 percent is not a small remittance economy; in absolute terms Mexico receives more than the other four combined. The percentage is low because the economy is large. A ratio and a volume are different measures and this is a ratio.

Nicaragua's near-doubling from 14.58 percent in 2020 to 26.63 in 2024 is the largest movement in the table, and Nicaragua is also the country whose domestic poverty measurement stopped. The one economic indicator that is well measured for Nicaragua describes people who have already left.

3.8 One series that moves against the headline

Severe food insecurity, percentage of population, FAO's Food Insecurity Experience Scale, retrieved through the World Bank:

Guatemalan severe food insecurity rose in every reported year from 2016 to 2023, from 17.0 to 21.3 percent, across the same period in which INE's official poverty rate fell from 59.3 to 56.0 percent.

And Nicaragua has no FIES series at all, in any year. The country absent from the poverty tables is also absent from the food-security tables.

4The legal architecture: five different things, all of them called slavery in ordinary speech

This is the section the rest of the document depends on. "Debt slavery" is not a legal category anywhere in the jurisdictions covered here. It is a plain-English compression of at least five distinct statuses, each with its own definition, its own evidentiary requirement and, in Mexico, its own sentencing range. A scene, a statistic or a prosecution that satisfies one of them frequently does not satisfy the others, and the gap between them is where most of the analytic difficulty in this subject lives.

4.1 Forced labour, as defined by treaty (PRIMARY/OFFICIAL: ILO Convention No. 29)

The Forced Labour Convention, 1930 (No. 29), Article 2(1), in the operative text:

"For the purposes of this Convention the term forced or compulsory labour shall mean all work or service which is exacted from any person under the menace of any penalty and for which the said person has not offered himself voluntarily."

Article 2(2) then excludes five categories: compulsory military service of a purely military character; normal civic obligations; work exacted as a consequence of a judicial conviction, provided it is supervised by a public authority and the person is "not hired to or placed at the disposal of private individuals, companies or associations"; emergencies including war, fire, flood, famine, earthquake and epidemic; and minor communal services performed in the direct interest of the community.

Two elements are therefore required and neither is poverty: a menace of a penalty, and the absence of voluntary offer. A person who takes brutal work at a starvation wage because there is no other work has not, on this definition, been subjected to forced labour. This is the single most important boundary in the brief and it is the one most often erased in popular treatment of the subject.

Note on wording. The ILO's own 2025 revised indicator booklet renders the definition as "all work or service which is exacted from any person under the threat of a penalty and for which the person has not offered themself voluntarily." That is a modernised paraphrase, not the treaty text, and the booklet's own second element is headed "Menace of any penalty," which is the binding formulation. This brief quotes the Convention from the Convention.

4.2 Debt bondage, as defined by treaty and domesticated in Mexican criminal law

The 1956 Supplementary Convention on the Abolition of Slavery defines debt bondage at Article 1(a) as the status arising from a debtor pledging personal services, or those of a person under their control, as security for a debt, where either the value of those services as reasonably assessed is not applied to liquidating the debt, or the length and nature of the services are not limited and defined.

Mexico's Ley General para Prevenir, Sancionar y Erradicar los Delitos en Materia de Trata de Personas, Article 12, fraccion I, tracks that definition almost word for word:

"Por deudas: La condicion que resulta para una persona del hecho de que un deudor se haya comprometido a prestar sus servicios personales, o los de alguien sobre quien ejerce autoridad, como garantia de una deuda, si los servicios prestados, equitativamente valorados, no se aplican al pago de la deuda, o si no se limita su duracion ni se define la naturaleza de dichos servicios."

("By debt: the condition that results for a person from the fact that a debtor has undertaken to render their personal services, or those of someone over whom they exercise authority, as security for a debt, if the services rendered, equitably valued, are not applied to payment of the debt, or if their duration is not limited and the nature of those services is not defined.")

The analytic weight in both texts sits on the two disjunctive conditions at the end. The debt itself is not the offence. The offence is a debt whose repayment terms are either not honestly credited or not bounded. That is a test about accounting and about definiteness, not about severity, and it can be satisfied by an arrangement that involves no violence whatsoever.

Article 12 also carries a second limb, servitude "por gleba" (by the soil), covering a person prevented from changing their condition of living or working on land belonging to another, obliged to render services on it whether paid or unpaid without being able to leave, or subject to property rights over land that carry rights over the persons on it.

4.3 Labour exploitation, which requires no coercion at all

Mexican law separates a third thing from both of the above. Article 21 of the same statute:

"Existe explotacion laboral cuando una persona obtiene, directa o indirectamente, beneficio injustificable, economico o de otra indole, de manera ilicita, mediante el trabajo ajeno, sometiendo a la persona a practicas que atenten contra su dignidad, tales como: I. Condiciones peligrosas o insalubres...; II. Existencia de una manifiesta desproporcion entre la cantidad de trabajo realizado y el pago efectuado por ello, o III. Salario por debajo de lo legalmente establecido. IV. Jornadas de trabajo por encima de lo estipulado por la Ley."

("Labour exploitation exists when a person obtains, directly or indirectly, an unjustifiable benefit, economic or otherwise, unlawfully, through the labour of another, subjecting the person to practices that offend their dignity, such as: I. Dangerous or unhealthy conditions...; II. The existence of a manifest disproportion between the quantity of work performed and the payment made for it, or III. A wage below what is legally established. IV. Working days beyond what is stipulated by law.")

Fraccion IV, and an aggravated penalty where the victims belong to indigenous or Afro-Mexican peoples and communities, were added by reform published 7 June 2024.

This definition is purely objective. It is satisfied by below-minimum wages, by excessive hours, or by a manifest disproportion between work and pay, with no requirement of a menace, a penalty, or an absence of consent. It is, in other words, the offence that the conditions described in Section 5 actually fit, and it is a materially lesser offence than the ones they are popularly described as.

Article 22 defines trabajos forzados separately and does require coercion: force or threat of force, including the force of a criminal organisation; grave harm or its threat placing the person in a condition of vulnerability; or abuse or threatened denunciation of the person's irregular migration status to the authorities, or any other abuse of law or legal process, causing the person to submit to unjust conditions. Article 23 then reproduces the C029 Article 2(2) exceptions almost exactly, including the judicial-conviction and normal-civic-obligation carve-outs.

4.4 The sentencing asymmetry, stated plainly

Setting the Mexican penalties side by side is the clearest single statement of how the legal system ranks these conducts. All from the Ley General, text in force as last reformed 14 November 2025.

4.5 One word doing a great deal of work: enganchar

Article 10 lists the conduct verbs constituting trafficking in persons: "captar, enganchar, transportar, transferir, retener, entregar, recibir o alojar" ("to recruit, to hook/engage, to transport, to transfer, to retain, to hand over, to receive or to house").

The second verb, enganchar, is the same word as el enganche, the historical and still-current Mexican and Central American system of advance-payment labour recruitment described in Section 5, and enganchador is the standing term for the recruiter who operates it. The statute does not define the verb, and this document found no Mexican appellate authority construing it during this pass. The word is doing double duty in ordinary Mexican Spanish as both the neutral act of signing someone on and the specific practice of binding them with an advance, and the statute borrows it without resolving which it means.

4.6 The statistical definition, which is what actually gets counted

Legal definitions decide prosecutions. A separate and narrower instrument decides what appears in a statistic. The 20th International Conference of Labour Statisticians adopted Guidelines concerning the measurement of forced labour in Geneva, 10 to 19 October 2018. For statistical purposes:

"a person is classified as being in forced labour if engaged during a specified reference period in any work that is both under the threat of menace of a penalty and involuntary. Both conditions must exist for this to be statistically regarded as forced labour."

The Guidelines list, among the elements of coercion, "debt bondage or manipulation of debt" alongside threats of violence, restrictions on movement, withholding of wages, withholding of documents, and abuse of vulnerability through denial of rights, threats of dismissal or deportation.

Bonded labour then gets its own operational three-limb test at paragraph 11, and this is the most useful single definition in the brief because it is the only one written to be measured rather than argued:

"Bonded labour is a form of forced labour in which the job or activity is associated with (i) advance payments or loans or excessive fees from recruiters and/or employers to the worker or to a person's family members; (ii) a financial penalty, meaning that the terms of repayment are unspecified at the outset and/or in contravention of laws and regulations regarding the amount of interest or other repayment conditions, or the job or activity is under-remunerated (in relation to legal regulations or the labour market); and (iii) some form of coercion until a worker or family member has repaid the loan or payment advance."

Limb (i) is the advance. Limb (ii) is the indefiniteness or the illegality of the terms, or under-remuneration. Limb (iii) is the coercion. All three are required. The practical consequence is that the advance-payment recruitment systems described in Section 5 satisfy (i) routinely and (ii) frequently, and whether they constitute bonded labour turns entirely on (iii), which is the limb that a household survey is least able to observe.

4.7 The eleven indicators, and why they are not a definition

The ILO's operational instrument for frontline identification is a set of eleven indicators, unchanged in name across the 2012 first edition and the 2025 revised edition: abuse of vulnerability, deception, restriction of movement, isolation, physical and sexual violence, intimidation and threats, retention of identity documents, withholding of wages, debt bondage, abusive working and living conditions, and excessive overtime.

The 2025 edition's debt bondage entry, at 3.9, is worth quoting at length because it names the specific mechanisms this brief documents in Section 5:

"Debt bondage arises when people are coerced into working against their will to repay a debt to an employer or recruiter, or when the debt is deliberately manipulated to ensure it cannot be repaid. This can occur when repayment terms are not clearly defined, when workers are forced to pay inflated prices for food and accommodation, when illegal wage deductions are imposed, when accounts are falsified, or when extremely high interest rates are illegally applied. The result is that workers become tied to an employer or creditor for an indefinite period, ranging from a single season to many years, or even across generations. In some cases, debts are inherited, with children recruited to work in exchange for repayment of their parents' loans."

The ILO is explicit that the indicators are not themselves a definition and that no single indicator establishes forced labour; the booklet describes them as red flags prompting further investigation, adapted to context, sector and legal framework.

"Forced to pay inflated prices for food and accommodation" is the company store. "Illegal wage deductions" and "accounts are falsified" is the ledger. Inherited debt is the generational limb. The ILO's own operational text, in other words, describes the tienda de raya without naming it.

4.8 Treaty status of the states covered

Parties to the 1956 Supplementary Convention on the Abolition of Slavery, adopted 7 September 1956, entered into force 30 April 1957, 124 parties:

4.9 What can be measured, and the number that does not exist

The definitions above are worth having only if something can be counted against them. This subsection states what the global measurement apparatus does and does not produce, because the gap between the two is where most public confusion about this subject originates.

The authoritative global figure is the joint ILO, Walk Free and IOM Global Estimates of Modern Slavery, published 12 September 2022 for a reference period of 2017 to 2021. It gives 27,577,000 people in forced labour worldwide, of whom 23,657,000 are in privately-imposed forced labour and 3,920,000 in state-imposed forced labour, alongside 21,993,000 in forced marriage, for a combined modern slavery figure of 49,570,000.

For this brief's purposes the single most useful figure it publishes is the debt bondage share. Among adults in forced labour exploitation, the share in situations of debt bondage is 20.9 percent worldwide, 19.7 percent of men and 22.0 percent of women. By sector: mining and quarrying 43.1 percent, agriculture 31.0, construction 27.4, services excluding domestic work 19.4, domestic work 19.3, other 14.5, manufacturing 14.0.

Two cautions on that figure. The denominator is adults in forced labour exploitation, which is the privately-imposed non-sexual category, and not the 27.6 million total; applying 20.9 percent to the total would be wrong. And the estimate is global. The Global Estimates publish the private-and-state split, the sectoral split and the debt bondage share at global level only. Table 2, the sole regional table, carries only the three totals, so there is no published figure for debt bondage in the Americas, and none for forced labour in the private economy in the Americas.

The number that does not exist. No United Nations body publishes a country-level forced labour or modern slavery estimate for Mexico, Guatemala, Honduras, El Salvador or Nicaragua. The Global Estimates contain no country estimate for any country at all, and the ILO explains why in its own methodology report. A leave-one-out cross-validation, at Table 29, gives an average country estimate of 105,400 adults in forced labour exploitation against a root mean cross-validation deviation of 99,300, or 94.3 percent, where the equivalent figure for the global estimate is 1.2 percent. The report's conclusion:

"These results suggest that the performance of the imputation models is reasonably adequate for global estimation, but highly inaccurate for country estimation. In other words, one may say that the imputation models produce highly volatile estimates at the country level, but relatively stable estimates at the global level."

The same report states that the global estimates "should not be considered as hard figures."

The consequence for everything that follows is worth stating once, plainly. There is no measured prevalence of forced labour or debt bondage for any country in this brief. Every figure in Sections 5 and 7 that appears to bear on scale is a count of institutional activity, detections, prosecutions, inspections, certifications, and each measures the capacity and priorities of the counting institution at least as much as it measures the underlying conduct.

5Debt on the origin side: the enganche, the ledger and the store

Section 4 established that debt bondage requires an advance, indefinite or dishonest repayment terms, and coercion. This section sets out what the public record shows about each of those three limbs inside Mexico and Guatemala, and it reaches sharply different conclusions for the two countries, for reasons that are about evidence rather than about the countries.

5.1 Who the Mexican agricultural workforce is (PRIMARY/OFFICIAL: CONEVAL, August 2024)

CONEVAL's study of the jornalero agricola population, published August 2024 on ENIGH 2022 data, is the only recent official quantification.

For context, the national poverty rate in 2022 was 36.3 percent against 60.5 percent for this group.

An important negative finding attaches to this population. The Encuesta Nacional de Jornaleros Agricolas has not been fielded since 2009. The widely repeated figures of 2,040,414 jornaleros of whom 40 percent are indigenous come from that 2009 survey and are now seventeen years old. CONEVAL's 23.7 percent and the 2009 survey's 40 percent rest on different definitions and are not comparable. Both CONEVAL and the CNDH have formally recommended that the survey be repeated, and this pass found no evidence that it has been.

5.2 The enganche, described by the Mexican state (PRIMARY/OFFICIAL: CNDH Recomendacion General 36/2019)

Mexico's national human rights commission issued Recomendacion General 36/2019, "Sobre la situacion de la poblacion jornalera agricola en Mexico," on 20 May 2019. Paragraph 21.2 distinguishes two forms of the recruitment system:

"Existen dos tipos de 'enganche': 'abastecimiento' y 'coyotaje'. El primero solo garantiza la promocion y traslado de mano de obra para el agricultor, de la zona de origen a la zona de trabajo sin garantizar la contratacion. ... En el 'coyotaje', el 'enganchador' suministra la mano de obra al productor y se traslada con el grupo de jornaleros, quedando a cargo de estos y con el control sobre el contrato. Al termino de la contratacion, el enganchador recibe la paga de todo el grupo a su cargo, a quienes paga sus respectivos salarios despues de descontar los gastos de transporte, alimentos, prestamos y una comision por sus servicios. Este tipo de 'enganche' es mas comun en las zonas indigenas, donde la poblacion presenta grandes carencias y limitada experiencia en la migracion."

("Two types of 'enganche' exist: 'supply' and 'coyotaje'. The first only guarantees the promotion and transport of labour for the farmer, from the zone of origin to the zone of work, without guaranteeing hiring. ... In 'coyotaje', the 'enganchador' supplies the labour to the producer and travels with the group of day labourers, remaining in charge of them and with control over the contract. At the end of the engagement, the enganchador receives the pay of the entire group in his charge, and pays them their respective wages after deducting the costs of transport, food, loans and a commission for his services. This type of 'enganche' is more common in indigenous zones, where the population has great deprivations and limited experience of migration.")

The analytically decisive clause is the third sentence. The worker is not paid by the employer. The employer pays the enganchador, who pays the worker after deducting transport, food, loans and his own commission. That is a structure in which the worker's earnings are netted against a running account they do not control, which is limb (ii) of the ICLS test before any question of coercion arises. The same paragraph adds that the employer usually does not sign contracts, or signs a collective contract with a union operating in the zone without the jornalero's involvement.

The Commission stated the consequence more directly in Recomendacion 209/2022, concerning thirty-eight indigenous jornaleros in Villa de Arista, San Luis Potosi, at footnote 5:

"El prestamo sobre el salario opera en forma recurrente en el trabajo agricola, pues los centros de trabajo no pagan las jornadas semanales trabajadas a cada trabajador, sino que hacen los pagos a los contratistas o enganchadores y estos no entregan esas cantidades a los trabajadores, por lo que estos se ven obligados a solicitar a su enganchador prestamos en pequenas cantidades de dinero para sufragar sus gastos elementales, como comida o incluso medicamentos, cantidades que el enganchador les deducen al momento de pagarles sus salarios."

("The loan against wages operates recurrently in agricultural work, because the workplaces do not pay each worker for the weekly days worked, but make the payments to the contractors or enganchadores, and these do not hand those amounts to the workers, so the workers find themselves obliged to ask their enganchador for loans in small amounts of money to cover their elementary expenses, such as food or even medicines, amounts that the enganchador deducts from them at the moment of paying their wages.")

5.3 The store, and what the state has and has not measured

The historical term tienda de raya appears in none of the CNDH instruments located in this pass. The mechanism is described without the label. Recomendacion 70/2016, concerning Villa Juarez, San Luis Potosi, paragraph 8:

"hay una tienda en el mismo rancho que es propiedad del contratista que lleva a los trabajadores desde sus lugares de origen, donde expende productos excesivamente caros, por lo que los 100 pesos que les dan a los jornaleros cada semana como prestamo sobre su sueldo, no les alcanza para comprar casi nada, que al final de la temporada les pagan su salario, pero les descuentan mucho de lo que consumen en la tienda"

("there is a store on the same ranch owned by the contractor who brings the workers from their places of origin, where he sells excessively expensive products, so the 100 pesos they give the day labourers each week as a loan against their wages is not enough to buy almost anything; at the end of the season they are paid their wage, but much of what they consume in the store is deducted.")

Set that against the ILO's own operational text at Section 4.7, which lists among the mechanisms of debt bondage that "workers are forced to pay inflated prices for food and accommodation" and that "illegal wage deductions are imposed." The Mexican case narrative and the ILO indicator describe the same arrangement.

5.4 The one case where a government named it debt bondage (PRIMARY/OFFICIAL: U.S. Customs and Border Protection)

On 21 October 2021 CBP issued a Withhold Release Order against Agropecuarios Tom S.A. de C.V. and Horticola Tom S.A. de C.V. and their subsidiaries, barring their fresh tomatoes from entry into the United States. The finding:

"CBP identified at least five of the International Labor Organization's indicators of forced labor during its investigation, including abuse of vulnerability, deception, withholding of wages, debt bondage, and abusive working and living conditions."

5.5 The wage floor, and its very recent history (PRIMARY/OFFICIAL: CONASAMI)

Mexico has a profession-specific statutory minimum wage for agricultural day labourers, and the fact worth carrying is how new it is.

Pesos per day. The category "Jornalero(a) agricola" did not exist in the professional minimum wage schedule before 2022; the 2015, 2020 and 2021 tables contain no such entry. From 1 January 2026 the category is renamed "Jornalero(a) agricola y/o trabajador(a) del campo."

5.6 The inspection gap, stated by three governments

The enforcement machinery is the weakest link in the Mexican picture, and three separate official sources say so.

Mexico has ratified the Protocol of 2014 to the Forced Labour Convention, deposited 11 June 2023, in force for Mexico 11 June 2024, promulgated by decree in the Diario Oficial on 6 May 2024. It has not ratified ILO Convention 81 on labour inspection, nor Convention 129 on labour inspection in agriculture. It has bound itself to the prohibition without binding itself to the inspection standard, and the unratified convention is the one that covers the sector where the conduct is documented.

The CNDH's 2018 study of labour authorities found that eleven labour authorities did not report or did not carry out inspections in agricultural workplaces at all, that only nine reported detecting non-compliance in agricultural workplaces, that only seventeen keep any register of the jornalero population and none of those disaggregate it by age, and that only five reported having become aware of criminal conduct and referred it to the public prosecutor. Recomendacion General 36/2019 found at paragraphs 93 to 99 that inspections are conducted in offices, food-safety areas or packing sheds, and that visits to the fields themselves happen only very exceptionally.

The United States Department of State's 2025 Trafficking in Persons report states that Mexico's 510 labour inspectors had a limited mandate for oversight of informal businesses and farms, which employed more than half of Mexican workers, that a 24-hour advance notice requirement for routine inspections hampered their effectiveness, and that labour inspectors identified four victims in 2024.

5.7 What Mexico cannot tell you, and what it can

Two data findings shape everything above.

Mexico's headline crime statistic cannot answer the labour-versus-sexual question. The SESNSP incidence series records trafficking under a single undifferentiated modalidad, with no split by exploitation type or by statutory article, and covers state jurisdiction only. Federal-jurisdiction data is published separately and less completely.

The CNDH's own diagnostics do disaggregate, and are the best Mexican-source figures located. For 15 June 2012 to 31 July 2017 the Commission identified 5,245 victims, of which 629 were labour exploitation and five were forced labour or forced services, against 1,376 for exploitation of prostitution and other sexual exploitation. It reports that all labour-related categories together represent about 17 percent of victims. The sex composition inverts between the two streams: in the labour sphere men and boys are 53 percent of victims and women and girls 47, where in the sexual sphere the split is roughly 95 percent female.

For 1 August 2017 to 31 July 2021 the Commission reports that the federal prosecutor's office opened 363 investigation files, of which five were for labour exploitation under Article 21, ten for trafficking under Article 10 combined with labour exploitation, and two for trafficking combined with forced labour under Article 22. Seventeen labour-related federal files in four years.

5.8 Child labour in Mexican agriculture (PRIMARY/OFFICIAL: INEGI, ENTI 2022)

INEGI's Encuesta Nacional de Trabajo Infantil 2022, released 5 October 2023, found 3.7 million children aged 5 to 17 in child labour, a rate of 13.1 percent, 1.7 points above 2019. Of these, 2.1 million, or 7.5 percent of the age group, were in prohibited occupations.

Within prohibited occupations the agricultural sector accounted for 33.0 percent, rising from 29.0 percent in 2019, and is the largest single sector, ahead of services at 23.2 percent and commerce at 21.5. The share is 39.0 percent for boys and 17.7 for girls. Thirty-four percent of children in prohibited occupations worked more than 28 hours a week, and 37.4 percent were unpaid.

Applying the sector share to the prohibited-occupation total gives roughly 693,000 children in prohibited agricultural occupations in 2022. That multiplication is this document's own arithmetic and is not published by INEGI in that form.

5.9 Guatemala: the mechanism is historical, the evidence is not contemporary

The Guatemalan side of this section reaches a different conclusion, and the difference is a finding rather than an asymmetry in the phenomenon.

The vocabulary of Guatemalan agricultural debt recruitment, habilitacion, enganche, cuadrilla contracting, mozos colonos, is well attested in peer-reviewed historical scholarship covering roughly 1824 to 1947. That literature establishes the mechanism and its terminology. It cannot support a claim about 2015 to 2025.

For the brief's actual date range, no source meeting this document's sourcing standard was located describing habilitacion or enganche as a live debt mechanism on Guatemalan coffee or sugar fincas. Contemporary description exists, but in the publications of advocacy and mission-driven organisations, which are excluded from load-bearing use here and are named in Section 8 so that a later pass knows where to route around them.

Three official findings bear on why the record is thin.

First, the US Department of Labor's 2024 List of Goods Produced by Child Labor or Forced Labor, eleventh edition, published 5 September 2024, lists the following for the five countries in scope, taken from the list's own country table:

The forced-labour column is populated for other countries on the same page of the table, including Ethiopia, Mali, Nepal, India and Niger, so the blanks for Guatemala, Honduras, El Salvador and Nicaragua are genuine absences rather than an extraction failure.

Second, the ILO Committee of Experts observed in its 2025 report, on Guatemala's compliance with Conventions 81 and 129, that section 281 of the Guatemalan Labour Code still requires labour inspectors to notify their presence by accrediting their identity and appointment, without exception, and asked the government to bring the provision into conformity so that inspectors with proper credentials may enter workplaces at any hour of the day or night. Guatemalan inspectors must announce themselves before an inspection.

Third, the Committee set out the resulting arithmetic in its 2023 report: the 2018 employment survey found 297,408 boys and 99,071 girls under fourteen engaged in child labour, and between January 2018 and May 2022 the labour inspectorate detected 136 cases of child labour in private enterprises. The Committee asked the government to investigate the cause of the difference. Roughly 396,000 children by survey against 136 cases in four and a half years of inspection.

6Financing the journey: the best-quantified debt in the brief

The migration debt is the one mechanism in this subject that has been measured properly, once, by a UN body, at national scale, with the collateral and the interest rate both reported. That single instrument carries most of this section.

6.1 What the crossing costs (PRIMARY/OFFICIAL: UNODC; SEGOB/UPM via CNDH)

UNODC's Global Study on Smuggling of Migrants 2018, at page 99, remains the current global reference. No later edition of the global study has been published, though UNODC's Observatory on Smuggling of Migrants has since issued narrower regional and topical products which do not supersede it and are not used here.

"A 2016 report cited a price range of US$6,000-8,000 for being smuggled from northern Mexico into the United States. According to field studies conducted among migrants returned to their origin countries, in recent years, Mexican migrants may have paid around US$5,000 to be smuggled to the United States, whereas Central American migrants may have paid on average US$7,000 to be smuggled across Mexico and finally to the United States."

The same study puts the land route to North America at 735,000 to 820,000 migrants smuggled per year and 3.7 to 4.2 billion US dollars in annual smuggler revenues, and notes that fees vary with the safety of the method, the distance, the difficulty of the crossing and the perceived wealth of the migrant, that some research indicates women are charged more than men, and that some smugglers sell more expensive packages including a set number of crossing attempts.

A Mexican federal figure, from the Unidad de Politica Migratoria's 2020 study of flows associated with migrant smuggling from the northern triangle, reported by the CNDH: an average of USD 5,862 for Hondurans, Guatemalans and Salvadorans for transit through Mexico plus the crossing, USD 2,330 for the crossing alone, and USD 4,559 for Mexicans crossing the border.

Note the direction of the difference between the two sources. UNODC puts Central Americans above Mexicans, roughly 7,000 against 5,000; the Mexican government study puts them at 5,862 against 4,559. Both agree Central Americans pay more, and the ratio is similar, 1.4 and 1.29. The levels differ by around fifteen to twenty percent. That is a reasonable degree of convergence between two independently produced estimates using different methods, and Section 9 treats it as a moderate correlation rather than a strong one.

6.2 How the journey is financed (PRIMARY/OFFICIAL: IOM Guatemala, 2022 survey)

The International Organization for Migration's Encuesta sobre migracion internacional de poblacion guatemalteca y remesas 2022, published June 2023, asked returned migrants and remittance recipients how they paid for the trip. From page 41, Grafica 29:

A formal bank loan finances 3.2 percent of journeys. A moneylender finances 17.6 percent, five and a half times as many.

6.3 What is pledged, and at what rate

The load-bearing paragraph of the brief, from the same page:

"El 19.8% de las personas migrantes retornadas que obtuvieron creditos con prestamistas tuvo que dar en garantia bienes inmuebles (el 10.6 %, las escrituras de su vivienda, y el 9.3 %, las escrituras de su terreno). El 67.5 % de estas personas pago el 5.0 % de interes mensual sobre el monto prestado; el 26.0 % pago el 10.0 % de interes mensual; el 4.4 % pago el 20 % de interes mensual, y el 1.9 % pago un 15.0 % mensual por concepto de intereses. El monto promedio del credito que se adquiere con prestamistas asciende a Q125,000 (USD 16,700)."

("19.8% of returned migrants who obtained credit from moneylenders had to give real property as security (10.6%, the deeds to their dwelling, and 9.3%, the deeds to their land). 67.5% of these people paid 5.0% monthly interest on the amount lent; 26.0% paid 10.0% monthly; 4.4% paid 20% monthly; and 1.9% paid 15.0% monthly in interest. The average amount of the credit acquired from moneylenders comes to Q125,000, USD 16,700.")

The monthly rates compound to roughly the following annual equivalents: 5 percent monthly is about 80 percent a year, 10 percent monthly about 214 percent, 15 percent about 435 percent, and 20 percent about 791 percent. Those annualisations are this document's own arithmetic and are not in the source, which reports monthly rates only.

Two observations the source does not make.

The average credit of Q125,000, about USD 16,700, is substantially larger than every smuggling fee in Section 6.1. It is a loan size and not a price, and it must not be presented as the cost of a coyote. What the gap between the two suggests is that the borrowing covers more than the fee, or that it has grown through accrued interest by the time it is measured, or both. This document cannot distinguish those and does not assert either.

And the interest rates are a finding by contrast rather than in isolation. Guatemalan formal-sector personal lending runs at annual rates; the informal rates above are monthly. The distance between a regulated annual rate and an 80-to-791-percent effective annual rate is the finding, and it is available from this one instrument only.

6.4 What the collateral does and does not satisfy, across three different tests

This is the sharpest analytic point available in the brief. It turns on the treaty text quoted at Section 4.2, and it must be run against three instruments rather than one, because they do not agree with each other.

Test one, the 1956 Supplementary Convention. The Convention reaches "a pledge by a debtor of his personal services or of those of a person under his control as security for a debt." What the Guatemalan highland arrangement pledges is not personal services. It is a house deed or a land deed. On the plain text of Article 1(a), and on the Mexican domestication at Article 12 fraccion I which tracks it almost word for word, a migration loan secured on land is not debt bondage in the sense that instrument defines, however coercive its economic effect and however ruinous its rate.

Test two, the ICLS statistical standard, which the first test does not control. Correction (multi-model verification pass): an earlier draft of this subsection stopped at test one and let the reader conclude that the arrangement escaped the analytic apparatus generally. It does not. The ICLS operational definition of bonded labour quoted at Section 4.6 contains no requirement that the security be labour. It requires (i) an advance, loan or excessive fee, (ii) a financial penalty, meaning repayment terms unspecified at the outset or contrary to law on interest, or under-remuneration, and (iii) coercion until the loan is repaid.

On the IOM survey's own figures the highland arrangement satisfies limb (i) plainly and limb (ii) on its face, given monthly interest rates of 5 to 20 percent and, in the documented cases, no defined repayment schedule. What is open is limb (iii). So the same facts that fall outside the treaty definition may already satisfy the measurement definition, and whether they do turns entirely on coercion, which Section 11.2 records as genuinely unresolved. The document holds both results at once rather than reporting the first and suppressing the second.

Test three, the Trafficking in Persons Protocol, which the earlier draft omitted entirely. Addition (multi-model verification pass): Article 3(a) of the Palermo Protocol defines trafficking as recruitment, transportation, transfer, harbouring or receipt of persons

"by means of the threat or use of force or other forms of coercion, of abduction, of fraud, of deception, of the abuse of power or of a position of vulnerability or of the giving or receiving of payments or benefits to achieve the consent of a person having control over another person, for the purpose of exploitation. Exploitation shall include, at a minimum, the exploitation of the prostitution of others or other forms of sexual exploitation, forced labour or services, slavery or practices similar to slavery, servitude or the removal of organs."

Neither the word debt nor the phrase debt bondage appears anywhere in that definition. Debt enters only twice and both times indirectly: as a route to "abuse of a position of vulnerability," which is a listed means, and through "practices similar to slavery" and "servitude" as listed purposes, which import the 1956 Convention by reference.

That structure matters here. The Protocol does not ask what was pledged. It asks whether a means was used and whether the purpose was exploitation. A household that has already signed over a house deed and must accept whatever an employer or recruiter subsequently imposes in order not to lose it is a candidate for the vulnerability limb, and if the creditor or a downstream employer then converts the obligation into coerced work the purpose limb is met as well. That is precisely the transition the ICAT doctrine at Section 6.7 describes, and the IOM survey's own category "paid with their work in the United States," at 3.4 percent, is where it would show up in the data.

What the three tests together support. Assessment (Confidence: Moderate; likely): the arrangement is not debt bondage on the 1956 definition, may well be bonded labour on the ICLS measurement definition depending on a coercion question this brief cannot resolve, and is not excluded from the Palermo Protocol by the fact that property rather than labour was pledged. A reader should take from this that the instruments diverge, not that the conduct escapes all of them.

The practical consequence for language is worth stating plainly, and it is narrower than the earlier draft implied. Describing the highland lending system as "debt slavery" borrows the authority of a legal term for something the principal slavery instrument does not cover. That is a real objection to the phrase. It is not a finding that the arrangement is lawful, nor that no international framework reaches it, and this document does not make either claim.

6.5 Who uses a smuggler, and where they leave from

From the same survey: 66.2 percent of returned migrants and remittance beneficiaries used a coyote for their transit. Among Guatemalans currently abroad the figure is 87.6 percent, against 5.0 percent by their own means and 3.4 percent on a tourist visa. Of an estimated 2,141,401 Guatemalans abroad, 1,884,918 travelled with a smuggler.

Exit points, returned migrants: La Mesilla, Huehuetenango 33.6 percent; Tecun Uman, San Marcos 19.5; Gracias a Dios, Huehuetenango 18.3; El Carmen, San Marcos 10.5; La Aurora international airport 12.3; El Naranjo, Peten 4.4; Ingenieros, Quiche 1.4. Huehuetenango and San Marcos together account for 81.9 percent of land exits.

6.6 Remittances as debt service

The survey is explicit that the early remittance stream is repayment. On page 71 it reports the 2022 monthly average received per recipient at USD 831, against USD 379 in 2016, and states that amounts are highest in the first three years after the sender travelled "debido a que tienen que pagar la deuda del viaje" ("because they have to pay the debt of the trip"). Page 70 states that the principal commitment is to pay the debt of the journey acquired before leaving. Page 78 puts remittances applied to "pago de la deuda del viaje" at 1.0 percent of the total, USD 153,812,934.

Set that 1.0 percent beside Section 3.7, where remittances are 19.1 percent of Guatemalan GDP. The share of remittances explicitly designated for travel-debt repayment is small; the share of the national economy that remittances represent is very large; and the survey's own explanation of why early-year remittances run high is the debt. These are consistent if the debt is repaid quickly and the transfers continue for other purposes afterwards.

6.7 Where a smuggler becomes a trafficker (PRIMARY/OFFICIAL: ICAT, 2016)

The UN inter-agency coordination group against trafficking in persons published an issue brief in October 2016 addressing exactly the boundary this section sits on. The passage that matters:

"after an individual has been smuggled into his or her destination country, the smuggler may become a trafficker by imposing a condition of debt bondage on the individual, a practice similar to slavery. The smuggler may tell the individual that he or she owes a large amount of money for the 'smuggling fee,' and that in order for the individual to pay off the debt he or she must work, live, and eat at a specific location designated by the smuggler. The individual may be charged for rent and food at a rate that makes the initial 'debt' impossible to pay off. The individual is left in a state of debt bondage and thus becomes a victim of human trafficking."

The brief also sets out the structural distinctions: the relationship between smuggler and migrant is a commercial transaction that usually ends after the border crossing; smuggling need not involve force, coercion, deception or abuse of power; smugglers commit a crime against the state and traffickers commit a crime against individuals; in smuggling the commodity is a service and in trafficking it is a person.

6.8 Debt created by force: kidnapping for ransom (PRIMARY/OFFICIAL: CNDH)

A separate mechanism produces a debt without any lending at all. Mexico's CNDH investigated kidnapping of migrants and reported on 15 June 2009, covering September 2008 to February 2009:

"el monto de rescate que se pide a las victimas va, en general, de 1,500 a 5,000 dolares. El promedio de los montos exigidos a las victimas identificadas en esta investigacion es de 2,500 dolares por persona."

("the ransom amount demanded of victims runs, in general, from 1,500 to 5,000 dollars. The average of the amounts demanded of the victims identified in this investigation is 2,500 dollars per person.")

The Commission documented 198 kidnapping events involving 9,758 victims in those six months, drawn from 238 sources, mostly victim testimony, an average of 33 events a month, more than one a day, and more than 1,600 people kidnapped per month. It projected roughly 400 events and 18,000 victims a year, and estimated the illicit proceeds from the identified cases at about USD 25 million.

The Commission is explicit that this is a floor and not an estimate of the whole: it states that the 198 cases cannot be considered an exhaustive account, that the nature of the offence and the limits of the investigation make full knowledge practically impossible, and that the figure is a minimum with a larger dark figure above it.

The ransom is paid by the family, which converts it into household debt in the origin country. A testimony recorded in the Commission's 2011 report makes the conversion explicit: a demand of ten thousand dollars, met by the whole family pooling money, described by the victim as "casi doscientas mil lempiras" of debt.

7Regional focus: the debt that crosses

This section follows the obligation north. It is organised around a single question: when the debt described in Sections 5 and 6 arrives in the United States with the person who owes it, what does the American legal system do about it?

7.1 Jurisdictional overview

Two routes carry the debt across, and they are legally opposite.

The unlawful route is smuggling, prosecuted under the immigration title, 8 U.S.C. 1324, and where a person is held for payment, under the hostage-taking statute, 18 U.S.C. 1203. The lawful route is the H-2A and H-2B temporary agricultural and non-agricultural work visas, regulated by the Department of Labor under 20 CFR part 655, where the charging of recruitment fees to workers is prohibited outright.

Both routes generate the same household obligation and the same leverage over the person who owes it. Only one of them is a crime at the point the debt is created.

7.2 The Arizona finding, stated negatively

No prosecution in the District of Arizona charged under the forced labour chapter, 18 U.S.C. 1589, 1590, 1592 or 1594, involving debt, was located in this pass. The migrant-labour and migrant-holding prosecutions located in that district are charged as smuggling, harbouring, transporting and unlawful employment.

The clearest example is United States v. Roa-Joachin and others, CR-21-01176-TUC-JGZ-BGM, arising from a 22 April 2021 search of two Phoenix stash houses at which more than thirty undocumented people were found. Angelico Roa-Joachin, 43, and Rafael Hernandez-Garcia, 35, were each sentenced on 21 April 2023 to 78 months by Judge Jennifer G. Zipps, with Luis Alberto Calixto-Pegueros receiving 37 months and Alexis Mejia-Zamora 27. The convictions were for conspiracy to transport and harbour illegal aliens for profit, and for Roa-Joachin also reentry of a removed alien. The DOJ release describes the mechanism: the individuals "were often held in overcrowded conditions and were not permitted to leave until their families or sponsors paid the smuggling organization for their release." The court found that three of the defendants sexually coerced or assaulted at least one woman held at the houses, which produced a sentencing enhancement rather than a separate charge.

7.3 The scale of the lawful route (PRIMARY/OFFICIAL: DOL Office of Foreign Labor Certification)

H-2A positions certified by the Department of Labor, by fiscal year:

H-2A certifications grew 2.85 times across the decade.

On nationality, the two federal series cannot be joined. DOL counts positions and does not record worker nationality: the H-2A disclosure record layout contains no worker citizenship or country-of-origin field, and recruiter information is a yes-or-no flag with no recruiter name or country. The State Department counts visas issued by nationality but does not break H-2A from H-2B in its by-nationality annual tables, only in a separate detail workbook.

From that workbook, H-2A and H-2B visas issued, selected years:

Mexico holds roughly nine in ten H-2A visas worldwide throughout the decade. The Central American participation is almost entirely H-2B and it is recent and steep: Guatemalan H-2B visas rose from 2,754 to 10,154 and Honduran from 650 to 7,058 between 2015 and 2024, with Honduras and El Salvador negligible before FY2021.

7.4 The rule that prohibits the debt, and the gap in it (PRIMARY/OFFICIAL: 20 CFR 655.135)

The regulation governing H-2A employers prohibits worker-paid recruitment fees in terms that reach the exact mechanism this brief describes. Paragraph (j), in force text:

"Comply with the prohibition against employees paying fees. The employer and its agents have not sought or received payment of any kind from any employee subject to 8 U.S.C. 1188 for any activity related to obtaining H-2A labor certification, including payment of the employer's attorney fees, application fees, or recruitment costs. For purposes of this paragraph (j), payment includes, but is not limited to, monetary payments, wage concessions (including deductions from wages, salary, or benefits), kickbacks, bribes, tributes, in kind payments, and free labor."

Note the last two words. The regulation's own definition of a prohibited payment includes free labor, which is to say it already contemplates the worker discharging the fee by working it off. Paragraph (o) separately prohibits the employer from holding or confiscating a worker's passport, visa or other immigration or government identification document, absent a written, unsolicited and revocable request from the worker.

The structural gap is in who is bound. Paragraph (j) binds the employer and its agents. The foreign recruiter in the origin country is reached only derivatively, through paragraph (k), which requires the employer to prohibit the recruiter contractually from seeking payments. The worker's actual counterparty in Guatemala or Michoacan is not directly regulated by the Department of Labor. The prohibition operates on the American end of a transaction that happens at the other end.

7.5 The fullest documented case (PRIMARY/OFFICIAL: DOJ, S.D. Ga.)

United States v. Patricio and others, Southern District of Georgia, Waycross Division. A 54-count, 53-page indictment against 24 defendants was unsealed on 22 November 2021, the product of a multi-year investigation styled Operation Blooming Onion. The statutes charged include 18 U.S.C. 1349 and 1341 (mail fraud conspiracy and mail fraud), 1594 (conspiracy to engage in forced labor), 1589 (forced labor), 1956(h) (money laundering conspiracy) and 1512 (witness tampering).

The workers were nationals of Mexico, Guatemala and Honduras, brought in on fraudulently obtained H-2A visas. The government's description of the mechanism:

"The conspirators required the workers to pay unlawful fees for transportation, food, and housing while illegally withholding their travel and identification documents, and subjected the workers 'to perform physically demanding work for little or no pay, housing them in crowded, unsanitary, and degrading living conditions, and by threatening them with deportation and violence.'"

Workers were required to dig onions with their bare hands and paid 20 cents for each bucket harvested. They were held in fenced work camps with little or no food, limited plumbing and no safe water. The indictment alleges rape, kidnapping, and threats to kill workers or their families, and that in many cases workers were sold or traded between conspirators. At least two workers died as a result of workplace conditions. The operation freed more than 100 individuals. The organisation is alleged to have profited by more than 200 million dollars, laundered through cash purchases of land, homes, vehicles and businesses and through a casino, and to have petitioned for more than 71,000 foreign workers.

Sentences included Javier Sanchez Mendoza Jr., 360 months for conspiracy to engage in forced labor, whom the government described as having recruited and unlawfully charged more than 500 Central American citizens to obtain H-2A visas before withholding their identification papers and threatening them and their families at home. The case closed on 12 June 2026 with restitution across all defendants exceeding 1.3 million dollars.

7.6 The clearest statement of the mechanism in any US government document

A separate prosecution states the coercive logic more plainly than any other source located. In the Middle District of Florida, in the Los Villatoros Harvesting matter, the Department of Justice described the conduct of a recruiter sentenced on 27 October 2022:

"Cabrera successfully recruited approximately 40 workers to work for the criminal enterprise. He charged them fees of between $1,000 to $2,000 prior to coming to the United States to work for LVH, lied to them by telling them that LVH would reimburse them after their arrival... Cabrera understood that the workers had gone into heavy debt to pay the fees he had charged them, and that he and his co-conspirators could use those debts to coerce the workers into continuing to work for LVH."

The organiser, Bladimir Moreno, was sentenced on 29 December 2022 to 118 months with more than 175,000 dollars in restitution, on RICO conspiracy and conspiracy to commit forced labor, for conduct in 2015 to 2017 involving Mexican H-2A workers. The Department's description of his coercion included "imposing debts on the workers," confiscating passports, degrading living conditions and threats of arrest and deportation.

7.7 Enforcement capacity against the lawful route (PRIMARY/OFFICIAL: DOL Wage and Hour Division)

Wage and Hour Division enforcement in agriculture, H-2A block, by fiscal year, from the Division's published dataset:

Total agricultural compliance actions across all statutes, a different and broader measure, fell from 1,361 in FY2015 to 649 in FY2025.

7.8 What the relief system counts, and the inversion it reveals

The United States operates a relief mechanism for trafficking victims, and its numbers are the most surprising in the brief.

T nonimmigrant status, principal applicants (T-1), from USCIS's own published workbook, data as of fiscal year 2026 quarter 3:

In fiscal year 2025 USCIS received its highest ever number of T visa applications, approved 1,398 and denied 2,362. Denials exceeded approvals, which had not previously happened in the programme. Principal applications were adjudicated in a mean of 21.4 months and a median of 21.2. The pending queue went from 20,351 at the end of FY2024 to 45,300 at the end of FY2025 and to 81,695 by the third quarter of FY2026.

Note on vintage, because USCIS's own publications disagree with each other, and the disagreement is systematic rather than incidental. The table above is taken entirely from the agency's applications workbook, which carries every fiscal year from 2008 in one place. The agency's individual annual reports to Congress, each covering a single year, give different figures for the same rows:

USCIS states a caveat this document carries: most applications approved or denied in a fiscal year were received in previous years, so the approval and denial counts in a row are actions taken in that period and do not represent an adjudicative trend for that year's receipts. An approval rate must not be computed by dividing approvals by receipts in the same row.

The statutory cap on T-1 approvals is 5,000 per fiscal year under 8 U.S.C. 1184(o)(2), and USCIS states in its own applicant characteristics fact sheet that the cap has never been reached since the programme was created. The highest annual approval total in the programme's history is 3,786, in FY2024, which is 75.7 percent of the cap. The binding constraint is adjudication capacity, not the cap.

Continued Presence, the law-enforcement-initiated route, is smaller still: in fiscal year 2025, 392 initial grants and 133 extensions, 525 approved applications in total across the United States.

The inversion is in what the identification system finds. The Department of Health and Human Services certifies foreign national trafficking victims for federal benefits, and the labour share is the large majority in both years located.

For FY2023, from the Attorney General's annual report: of 582 adults who received certification letters, 444 or 76 percent were labour trafficking, 63 or 11 percent sex trafficking, 58 or 10 percent both, and 17 or 3 percent unknown. For FY2024: of 495 adults certified, 75 percent labour trafficking, 11 percent sex trafficking, 10 percent both, 4 percent unspecified; and of 2,642 children issued eligibility letters, 74 percent labour trafficking.

Caveat on the FY2023 figures: they are published as a pie chart rather than a table, and the four values extract cleanly while their assignment to the four legend categories does not. The assignment above rests on magnitude and on the fact that it reproduces the FY2024 distribution almost exactly. A reader needing the FY2023 split for its own sake should read the figure visually rather than rely on this transcription. On country of origin, the Attorney General's report gives the top five countries of foreign national victims served by the HHS Trafficking Victim Assistance Program in FY2023 as Honduras with 349 clients, Guatemala with 317, Mexico with 315, El Salvador with 168 and the Philippines with 106. The four countries in this brief's scope therefore supplied 1,149 clients, against a programme total that the same report's by-type figures sum to 1,573, which is roughly 73 percent. The programme's own by-type split for FY2023 gives 1,010 clients, 64 percent, in the largest category, which on the same magnitude reasoning as above is labour trafficking.

Set that beside the prosecution figures. From the Attorney General's annual report for fiscal year 2023, the two years it states directly:

Labour trafficking is 4.3 percent of prosecutions brought in FY2022 and 6.6 percent in FY2023. The wider series located in this session's research, running FY2015 to FY2024, keeps labour trafficking between roughly 3 and 8 percent of prosecutions in every year with a published split, but only the two years above were verified against the report by the drafting pass.

One feature of that table cuts against reading the prosecution count as a measure of scale, and it is visible within the verified years. In FY2022 seven labour trafficking prosecutions produced forty-six charged defendants, about 6.6 defendants per case, where 155 sex trafficking prosecutions produced 264 defendants, about 1.7 per case. In FY2023 the effect is absent: 12 labour prosecutions produced 19 defendants and 169 sex prosecutions produced 239. The multi-defendant character of labour trafficking conspiracies is therefore real but not stable year to year, and a single year's prosecution count understates the enforcement activity it represents by a factor that varies.

7.9 A right that exists on paper and not in operation

One regulatory development inside the focal period bears directly on whether an indebted worker can act on their own behalf, and its status is genuinely unusual.

Because agricultural workers are excluded from the National Labor Relations Act's definition of employee, the Department of Labor in 2024 imported analogous protections by regulation. The Farmworker Protection Rule, "Improving Protections for Workers in Temporary Agricultural Employment in the United States," 89 FR 33898, published 29 April 2024 and effective 28 June 2024, added at 20 CFR 655.135(h)(2) a protection for activities related to self-organization and for concerted activity, together with a right to refuse to attend an employer meeting whose primary purpose is to communicate the employer's opinion on protected activity, and at 655.135(m) a right to a designated representative at an investigatory interview.

Those provisions were then litigated. A preliminary injunction issued in the Southern District of Georgia on 26 August 2024 against the whole rule in seventeen states; a preliminary injunction issued in the Eastern District of Kentucky on 25 November 2024 covering four further states and a set of grower associations and their members; and on the same day the Southern District of Mississippi stayed 655.135(h)(2) and (m) nationwide under 5 U.S.C. 705, declining to limit the relief to the parties. The Eastern District of North Carolina upheld the rule on 5 May 2025 and that decision is on appeal. A rescission was proposed at 90 FR 28919 on 2 July 2025 and, as of the compilation date of this brief, has not been finalised.

8Sources rejected this session, and why

Two entries changed status during research and both are worth recording.

The ILO / Walk Free / IOM joint Global Estimates were admitted; Walk Free's own Global Slavery Index was not. These are routinely treated as the same product and they are not. The joint Global Estimates are published by the ILO, a UN body, and are used here for their global and regional figures and, more importantly, for their methodology report. Walk Free's separately published Global Slavery Index takes the joint regional estimates as a starting point and produces national estimates independently, by Walk Free's own account of its method. Every country-level "modern slavery in Mexico" figure in circulation is the second product, not the first. This document uses the first and declines to give the second. The distinction is exactly the sort this brief exists to hold apart, and it is recorded here rather than buried in a citation.

A newspaper investigation was excluded on availability rather than on standard. The most widely cited account of company-store debt in Mexican agro-export camps is a bylined 2014 newspaper series which would meet this document's sourcing standard as named investigative journalism. It could not be fetched: the publisher's domains were unreachable to the tools used across repeated attempts, and it falls marginally outside the date range in any case. It is therefore absent rather than rejected, and no claim here rests on it. This distinction matters because the prevalence claims that trace to it are widely repeated, and their absence from this brief should not be read as a judgment that they are wrong.

One further note. A journalism collaboration carrying detailed Guatemalan court records on land lost to migration lenders was located but carries no individual byline and is published by a platform rather than a wire service or credentialed outlet. Its underlying court records are independently checkable and its account is consistent with the IOM survey's collateral findings, but nothing in Section 6 rests on it and it is not cited.

9Signals: genuine cross-source correlation

The standard applied here is the one this document states in its own voice: a correlated signal requires two or more independently produced sources, from different institutions, using different methods, arriving at the same finding without one citing or reprinting the other. Several outlets covering one press release is one data point, not several. Two bodies analysing the same underlying dataset is a real but weaker correlation than two independent bodies of evidence, and is rated as such.

Genuine correlations found this session

Signal 1 (STRONG). Four institutions, on three different legal instruments, independently describe the same debt arrangement.

The ILO's 2025 indicator text describes debt bondage arising where "workers are forced to pay inflated prices for food and accommodation, when illegal wage deductions are imposed, when accounts are falsified" (Section 4.7). Mexico's CNDH, investigating a labour complaint in San Luis Potosi, describes a contractor-owned store on the ranch selling at excessive prices, a weekly hundred-peso loan against wages, and end-of-season deduction of store consumption (Section 5.3). US Customs and Border Protection, applying the ILO framework to a named Mexican tomato producer in an import-admissibility investigation, found abuse of vulnerability, deception, withholding of wages, debt bondage and abusive working and living conditions (Section 5.4). And the US Department of Justice, prosecuting an H-2A visa fraud and forced labour conspiracy in Georgia, describes conspirators requiring workers "to pay unlawful fees for transportation, food, and housing while illegally withholding their travel and identification documents" (Section 7.5).

What makes this strong rather than moderate is the independence of the instruments: a UN standard-setting booklet, a national ombudsman's case recommendation, a customs enforcement order and a federal indictment. None is derived from the others; three of the four are adversarial or investigative proceedings with their own evidentiary requirements; and they converge on a mechanism specific enough to be recognised item by item.

Signal 2 (MODERATE). Two governments independently price the crossing, and agree on the ratio more than the level.

Rated Moderate rather than Strong for three reasons: both ultimately rest on surveys of returned migrants, which is a shared method rather than two independent bodies of evidence; the UPM figures were read through the CNDH's reproduction rather than from the UPM study itself; and the periods differ.

Signal 3 (MODERATE). Two national criminal-justice systems, measured separately, both prosecute labour exploitation as a small minority of trafficking.

Two different countries, two different statutory schemes, two different reporting bodies, the same order of magnitude on comparable measures.

Rated Moderate rather than Strong for two reasons. Both are prosecution counts, which share a common structural origin: both measure what a criminal justice system charges, and both would be depressed by the same evidentiary difficulty, so this is one mechanism appearing twice rather than two independent lines of evidence. And the measures are not quite the same object, the Mexican figure being investigation files opened and the American one prosecutions brought.

A caution that applies to both and is visible in the American data: prosecution counts understate labour trafficking enforcement by a variable factor, because labour cases are multi-defendant. In FY2022 seven labour prosecutions carried forty-six defendants. The correlation above is between two counts that share this distortion, which is a further reason not to rate it Strong.

Signal 4 (NARROW). Two countries' labour inspectorates are required to announce themselves, documented by two different bodies.

Rated Narrow because it is two observations rather than a body of evidence, and because the two requirements are not identical: Guatemala's is a statutory identification requirement and Mexico's an advance-notice period. They are the same in effect and not in form.

Tested and found weak or absent, reported because the test is part of the finding

Signal 5 (WEAK). Poverty and emigration do not line up cleanly at the departmental level in Guatemala.

The expectation is that the poorest departments supply the most migration. The data does not support that in any simple form. Alta Verapaz at 90.3 percent poverty and Quiche at 86.4 are poorer than Huehuetenango at 81.2, yet Huehuetenango dominates the exit statistics with 51.9 percent of land exits through its two crossings. San Marcos, at 53.8 percent poverty and below the national rate, supplies a further 30.0 percent.

Part of this is an artefact: exit-point data records where people cross, not where they are from, and border departments over-register. But that artefact cannot be assumed to explain the whole, and this document could not obtain the departmental origin breakdown that would settle it. The honest statement is that the relationship between departmental poverty and emigration is not demonstrated by the data located, and that the two poorest departments in the country are not the leading documented exit points.

Signal 6 (DIVERGENT, reported as a finding). Three instruments give three directions for Guatemalan poverty over the same decade.

INE's national consumption measure has poverty falling from 59.3 percent in 2014 to 56.0 in 2023. CEPAL's harmonised income measure has it rising from 50.5 to 54.5 across the same two years. FAO's experiential food-insecurity scale has severe food insecurity rising in every reported year from 17.0 percent in 2016 to 21.3 in 2023. And the same INE release has measured consumption inequality rising, Gini 0.372 to 0.421.

This is not a correlation and it is not a contradiction. Three instruments measuring different things can move in different directions consistently, and the widening inequality reported by the first source is a plausible reconciliation of the first with the third. It is recorded here because a brief that reported only the falling national headline would give a materially misleading picture of Guatemala over the focal period, and because the divergence is itself a strong argument for the terms-of-art discipline this document applies throughout.

Named specifically as NOT correlation (single-sourced, included for what they are)

These findings are not weaker facts for resting on one source each. They must simply not be read as corroborated.

1. The entire quantitative account of migration lending in Section 6.3 rests on one instrument: the IOM Guatemala 2022 survey. The collateral share, the split between house deeds and land deeds, the distribution of monthly interest rates and the average credit size all come from a single paragraph of a single national survey in a single country in a single year. No second measurement of informal migration-lending terms in Mexico or Central America was located, from any source, and this document's search for one is recorded as a negative finding in its own right. This is simultaneously the most valuable and the most exposed material in the brief.

1. The CBP Withhold Release Order naming debt bondage at a Mexican producer is one enforcement action by one agency against two named companies. It establishes that a government made that finding once. It establishes nothing about prevalence.

1. The CNDH kidnapping and ransom figures come from one national ombudsman's investigations, and the Commission itself states its counts are a floor with a larger unmeasured figure above them.

1. CONEVAL's jornalero population figures, including the count of 2.3 million and the indigenous share of 23.7 percent, come from a single 2024 study. The dedicated national survey of this population has not been fielded since 2009.

1. The finding that no Guatemalan, Honduran, Salvadoran or Nicaraguan good is listed for forced labour rests on one instrument, the US Department of Labor's 2024 list. It is a well-constructed instrument with a public methodology, and it is one instrument.

10Key Assumptions Check

A Key Assumptions Check states the propositions this document depends on that it has not verified. They are flagged not because they are doubted but because they are unverified, which is a different thing. Each entry names the assumption, why it could not be verified, and what would change above if it were wrong.

Assumption 1: that INEGI's 2024 poverty measurement is comparable to CONEVAL's 2016-2022 series. INEGI states that it applied CONEVAL's own methodology, criteria and processes unchanged, and publishes replication code. No independent replication by a body outside INEGI was located. The agency asserting continuity is the agency that inherited the function from a body that was legally extinguished four weeks before publication. If this is wrong, the entire Section 2 series breaks at the 2022-2024 boundary and the headline fall of 6.8 points becomes uninterpretable.

Assumption 2: that ENCOVI 2023 can be compared with ENCOVI 2014. The source contradicts itself on this point, describing a retro-projection methodology on one page and stating on another that each year's figures are as reported in that year's survey. This document reproduces the charted figures. If the series is in fact a back-cast, the 59.3 to 56.0 movement is not two independent measurements and Guatemala's twenty-three-year flatness, which is one of this brief's headline findings, would need restating in weaker terms.

Assumption 3: that the IOM survey's returned-migrant sample describes the borrowing population generally. The survey asks returned migrants and remittance beneficiaries about financing. Returned migrants are, by definition, people whose migration ended, and a large share of those returned involuntarily. If the borrowing terms faced by people who successfully remained abroad differ systematically, the interest-rate distribution and collateral share in Section 6.3 describe the unsuccessful rather than the typical case. Nothing located allows this to be tested, and it is the most consequential unverified assumption in the brief because so much rests on that single instrument.

Assumption 4: that detected, charged and certified counts bear some stable relationship to underlying conduct. Every enforcement and identification figure in Sections 5 and 7 measures institutional activity. UNODC states plainly that its data are detected cases and attributes movement partly to improved detection capacity. If detection efficiency changed materially over the focal period in any of the jurisdictions covered, apparent trends in these series are artefacts. This document generally treats them as capacity measures rather than prevalence measures, but the Signal 3 correlation would weaken considerably if detection efficiency moved differently in Mexico and the United States.

Assumption 5: that the absence of contemporary documentation of Guatemalan finca debt reflects detection rather than absence. Section 5.9 states both readings and declines to choose. This document's structure nonetheless leans on the detection reading in places, because it groups Guatemala with Mexico as an origin-side debt jurisdiction. If advance-payment recruitment on Guatemalan fincas has genuinely declined to insignificance since the mid-twentieth century, then Section 5's Guatemalan material describes a historical system and the brief's origin-side debt findings are effectively Mexican findings only.

Assumption 6: that the corridor geography established in the companion origin-regions document still holds. Section 7 is framed around southern Arizona and the Sonora corridor because the companion document established that this corridor carries Mexicans and Guatemalans and essentially no one else. That finding rests on FY2024 encounter data. If the corridor's composition has shifted materially since, the regional framing of Section 7 is misdirected, though its substantive findings about charging practice and H-2A debt would be unaffected.

11Analysis of Competing Hypotheses: contested points

Analysis of Competing Hypotheses guards against anchoring on the first plausible explanation and then collecting only confirming evidence. Each contested question below is stated as a question, the in-scope evidence is set out, competing hypotheses are listed, and the evidence is weighed against each rather than for the favoured one.

11.1 Why is labour trafficking a large majority of identified victims and a small minority of prosecutions?

What the in-scope sourcing supports: in the United States, roughly 75 percent of adults certified by HHS as foreign national trafficking victims in FY2024 were labour trafficking victims, and roughly 73 percent of victims served were Mexican or northern-triangle nationals; labour trafficking was 4.3 percent of federal trafficking prosecutions brought in FY2022 and 6.6 percent in FY2023. In Mexico, labour-related files are 17 of 363 federal investigations across four years. UNODC reports that globally 42 percent of detected victims in 2022 were trafficked for forced labour, while only 17 percent of trafficker convictions were for forced labour against 70 percent for sexual exploitation.

H1: Labour trafficking is harder to prove. Forced labour under 18 U.S.C. 1589 requires proof of a coercive means, and the coercion in a debt case is typically financial and psychological rather than physical. Article 22 of the Mexican statute similarly requires force, grave harm or abuse of legal process. A prosecutor with the same facts can charge visa fraud, harbouring, or wage violations at a lower evidentiary burden, as happened in Patricio, where forced labour was charged alongside mail fraud and money laundering, and in Roa-Joachin, where holding people until payment was charged as harbouring for profit.

H2: The two systems reach different populations. HHS certification flows from victim-services and immigration-relief channels; prosecution flows from law enforcement referrals. DOJ's own grantee data give a very different profile from HHS's, 67 percent sex trafficking against 17 percent labour, on the same national population in the same year. If the intake channels differ that sharply, the two figures may not be in tension at all.

H3: Sexual exploitation is genuinely more prevalent and the identification figures are skewed. Detection of sexual exploitation is more likely to arise from policing activity that generates cases regardless of victim cooperation, while labour cases surface through service providers.

H4: Part of the gap is arithmetic, because forced labour cases carry more victims per offender. This is UNODC's own explanation and it is the one an outside reader is least likely to reach unaided. Addition (multi-model verification pass): the Global Report on Trafficking in Persons 2024 states that the inconsistency between the share of detected victims by type and the share of traffickers convicted "can be explained in part by the fact that, according to the available data, forced labour involves, on average, more victims per trafficker than sexual exploitation," and reports that on an analysis of 920 court cases "one trafficker, on average, can exploit one to two victims for sexual exploitation, while in forced labour, the number of victims can rise to two to three victims, depending on the sector."

Weighing. H4 must be applied before the others, because it removes part of the gap without requiring any behavioural explanation at all: if a forced labour case yields roughly twice the victims of a sexual exploitation case, then equal prosecutorial effort produces half the defendants per victim, and a victim-share of 42 percent against a conviction-share of 17 percent is less anomalous than it first appears. It does not remove the whole gap. A factor of two on victims per case does not account for a factor of two and a half between detection and conviction shares, and it says nothing at all about the American HHS-versus-prosecution divergence, where the comparison is between certified victims and prosecutions rather than between victims and convictions.

H3 is the weakest: it does not explain why the same inversion appears in UNODC's global detection data, nor why the CNDH finds a male-majority victim profile specifically in the labour stream. H2 has direct documentary support in the DOJ-versus-HHS divergence and cannot be dismissed. H1 has the strongest mechanical support, in the statutory structure itself and in the observed charging choices in both countries.

11.2 Has advance-payment debt recruitment on Guatemalan plantations ended, or merely stopped being documented?

What the in-scope sourcing supports: the mechanism and its vocabulary are richly documented in peer-reviewed history through 1947. For 2015-2025, no source meeting this document's standard describes it as live. The US Department of Labor lists six Guatemalan goods for child labour and none for forced labour. Guatemalan labour inspectors must announce themselves under section 281 of the Labour Code. The inspectorate detected 136 child labour cases in four and a half years against roughly 396,000 children found by survey.

H1: The practice persists and the inspection system cannot see it. A regime requiring inspectors to identify themselves before entry, with a detection rate three orders of magnitude below the surveyed rate for a related and more visible violation, is not an instrument capable of documenting a debt arrangement.

H2: The practice has genuinely declined. Guatemalan agriculture has changed substantially since the mid-twentieth century; the finca labour system that produced habilitacion was tied to a landholding structure and a legal regime, including vagrancy laws, that no longer exist. The absence of contemporary evidence may simply be evidence of absence.

H3: The practice persists and is documented, but only in sources this document excludes. Contemporary description exists in the publications of advocacy organisations, which are named in Section 8. If those accounts are accurate, the gap is in this brief's sourcing standard rather than in the record.

Weighing. H1 and H3 are compatible and mutually reinforcing; H2 stands against both. Nothing located resolves them. The DOL forced-labour listing is the strongest evidence for H2, since it is a US government instrument that does examine Guatemalan agriculture and does list Guatemalan goods, just not for forced labour, and its methodology would be expected to pick up a widespread debt-peonage system. Against that, the same instrument lists no forced-labour good for any of the four Central American countries, which raises the possibility that it is calibrated in a way that under-detects the category regionally rather than that the category is absent in four countries at once.

11.3 Did the material conditions that produce debt vulnerability in Mexico improve between 2020 and 2024?

What the in-scope sourcing supports: multidimensional poverty fell from 43.9 to 29.6 percent and extreme poverty from 10.8 to 7.0 million. Over the same period, health-access deprivation stood at 34.2 percent against 15.6 percent in 2016; social security deprivation ended at 48.2 percent, 62.7 million people; the population vulnerable by social deprivation rose from 30.0 to 41.9 million; and the average number of deprivations among the extreme poor did not improve, standing at 3.8 in both 2022 and 2024.

H1: Yes, and the deprivation indicators lag. Income improvements arrive first and rights-access follows; the vulnerable-by-deprivation increase is an arithmetic consequence of people crossing the income line while retaining a deprivation, which is movement in the right direction.

H2: Partly. The income improvement is real; the exposure to catastrophic private cost is not improved and in the health dimension is much worse than at the start of the series. The mechanism by which a household enters coercive debt runs through an uninsured shock, and the two indicators governing that, health access and social security, are the two that did not improve on the decade.

H3: The measure moved rather than the conditions. The 2024 figure was produced by a different institution using an inherited methodology, and the health indicator's construction changed with the replacement of Seguro Popular by INSABI and then IMSS-Bienestar, with ENIGH 2024 dropping the old affiliation categories and adding new ones.

Weighing. H3 has real force on the health indicator specifically, where the survey instrument demonstrably changed, and this document accordingly does not treat the health series as a clean measurement across the whole range. It has much less force on the headline: the methodology, source survey and replication code are stated to be unchanged, and the deprivation indicators move smoothly across the institutional boundary rather than discontinuously. H1 is a reasonable reading of the vulnerable-by-deprivation rise taken alone but does not address social security, which moved only 5.9 points across eight years and left nearly half the population uncovered.

12Sources Cited

Numbered continuously. Each entry carries an Admiralty grade: a letter for source reliability, a number for the credibility of the specific information cited. The grade reflects whether this document's drafting pass actually retrieved and read the source. Sources located during research but not retrieved by the drafting pass are segregated into their own group at grade B or C and are flagged in the body at the point of use, in keeping with the rule that a source not fetched is not graded as though it were.

Primary/official: Mexico

Primary/official: Guatemala

Primary/official: United States federal

And, from the same agency, the underlying series: "Number of Form I-914, Application for T Nonimmigrant Status, By Fiscal Year, Quarter, and Case Status, Fiscal Years 2008-2026," data as of FY2026 Q3. uscis.gov . Both were retrieved and read. They differ on FY2025 applications received, 34,650 in the report against 37,177 in the workbook; the discrepancy is recorded at Section 7.8.

The country table in the published report is the authority and is what Section 5.9 reproduces. The Department's web listing at /agencies/ilab/reports/child-labor/list-of-goods-print truncates alphabetically on retrieval and produced a wrong Mexican entry in two earlier attempts; do not use it. The report file host refuses plain command-line requests and requires browser-style headers.

Primary/official: international organisations

Entries 19 and 20 are used for global and regional figures and for methodology only. Walk Free's separately published Global Slavery Index, and all country-level modern slavery estimates derived from it, are excluded; see Section 8.

Used for the 2022 detection and conviction shares by form of exploitation, and for the victims-per-trafficker analysis at Section 11.1. The country annexes for Mexico and the Central American states were not retrieved by the drafting pass and remain at entry 44.

Sources located during research but not retrieved by the drafting pass

These carry a reduced grade and are flagged at the point of use in the body. They are listed so that a follow-up pass knows exactly what to fetch, and so that no reader mistakes them for material this document verified.

Companion documents

1. Vampires of Tucson project reference. "Origin Regions and Northbound Trafficking Routes: Factual Reference, 2015-2025." Referenced for the corridor composition finding underlying Section 7's regional framing and for the Roa-Joachin case record.
2. Vampires of Tucson project reference. "Border and Cartel Trafficking: Factual Reference, 2014-2024." Referenced for the smuggling-versus-trafficking charging distinction established there.

Sources located, evaluated and excluded from load-bearing use

Enumerated with reasons in Section 8, and not repeated here.

13Open items for a follow-up research pass

Resolved during this session

RESOLVED. The verbatim text of Article 1(a) of the 1956 Supplementary Convention. Two hosts returned HTTP 403 (ohchr.org and refworld.org) and an early draft carried the definition as a paraphrase under a verification flag. Resolved by retrieving the authentic UNTS text from the United Nations Treaty Collection and reading Article 1 directly. The paraphrase was replaced with the treaty text, and the finding that the Convention reaches a pledge of personal services rather than of property, which is the basis of the analysis at Section 6.4, rests on that verbatim text.

RESOLVED. The correct ILO indicator list. An automated summary of the 2025 revised edition returned three renamed indicators and a different ordering. Resolved by reading the document's own numbered section headings, 3.1 through 3.11, which confirm the classic names survive the revision. Recorded as a correction at Section 4.7.

RESOLVED. The Guatemalan Gini series. A linear text extraction produced a monotonic decline; a coordinate-based re-extraction of the same chart placed 0.372 under 2014 and 0.421 under 2023. Resolved in favour of the coordinate reading, which is corroborated by the direction of the underlying distributional finding. Recorded as a correction at Section 3.1.

RESOLVED. The full Mexican entry on the DOL TVPRA list. Two retrievals against the Department's web listing returned only four alphabetically consecutive Mexican goods and recorded chile peppers as the sole forced-labour designation. The verification pass identified this as a truncation and the published report was then retrieved directly, which required browser-style request headers because the Department's file host refuses plain command-line requests. The report's country table gives Mexico fifteen goods with two forced-labour designations, chile peppers and tomatoes. Recorded as a correction at Section 5.9, where the full table for all five countries now appears, and the negative finding for the four Central American states is confirmed against the same table.

RESOLVED, partially. Whether Mexico's 2024 poverty measurement is institutionally continuous with the earlier series. The legal chain is now fully documented: constitutional reform DOF 20 December 2024, secondary legislation in force 17 July 2025 extinguishing Coneval and transferring the function to INEGI, first INEGI measurement published 13 August 2025. What remains unresolved is whether the resulting figures are statistically comparable, which no body outside INEGI has independently tested. Carried forward as Key Assumption 1.

Still open

1. A second measurement of informal migration-lending terms. The entire quantitative account at Section 6.3 rests on one paragraph of one IOM survey. Next step: retrieve the Honduran Encuesta Nacional de Migracion y Remesas 2023, produced by INE Honduras with IOM, and establish whether it carries financing, coyote-use and collateral items comparable to the Guatemalan instrument. If it does, the brief gains its first corroboration for the single most exposed finding it contains. If it does not, that absence should be stated in Section 9 as a completed test.

1. Judicial construction of "enganchar" in Article 10 of the Mexican trafficking statute. No Mexican appellate authority construing the term was located, and the Low-confidence assessment at Section 4.5 turns on that gap. Next step: a search of Suprema Corte and Tribunales Colegiados criteria in the Semanario Judicial de la Federacion for the term. It is possible the question has simply never been litigated, in which case that is the finding.

1. Whether advance-payment debt recruitment persists on Guatemalan fincas. The largest substantive gap in the brief, set out at Section 11.2 and unresolved between three hypotheses. Two routes were attempted and failed: NORMLEX returned HTTP 403 on every attempt, so the ILO Committee of Experts' direct requests to Guatemala under Convention 29, which are not printed in the published report body, could not be read; and Guatemala's Procurador de los Derechos Humanos site returned HTTP 403 with a bot challenge. Next steps, in order of promise: the CEACR direct requests via a browser session; the Guatemalan Ministerio Publico's own releases for labour-trafficking convictions in agriculture; the PDH's Informes Anuales Circunstanciados for 2023 and 2024. It is possible this material does not exist in published form from any non-advocacy source, in which case the honest resolution is that the question cannot be answered to this document's standard.

1. National poverty figures for Honduras before 2021. INE Honduras has retired its pre-2021 poverty publications and the World Bank's national-line series for Honduras also begins at 2021. Next step: an archival retrieval of EPHPM reports for 2015 to 2019. Note that any figure recovered will be on a household basis and on a basket whose vintage is itself an open question, per the verification flag at Section 3.5.

1. Prevalence of company-store arrangements in Mexican agro-export camps. The existence of the mechanism is documented by the Mexican state in individual cases; no government source measures how common it is. Next step: request or locate STPS inspection findings disaggregated by finding type, if such disaggregation exists. This was attempted and no national series was found, and it is possible the measurement simply does not exist.

1. The UPM 2020 smuggling-price study itself, rather than the CNDH's reproduction of its figures. Section 6.1 and Signal 2 both rest partly on figures read at one remove.

1. The ICAT 2016 issue brief, whose doctrinal statement at Section 6.7 is load-bearing for the smuggling-to-trafficking boundary and was not retrieved by the drafting pass. The UNODC Mexico mirror returned HTTP 404 on the attempt made; the icat.un.org original returned HTTP 403 to earlier attempts in this session.

1. DOL and State Department H-2A series, at Section 7.3, to be re-pulled directly from the agency files before any individual figure is relied on.

1. HHS certification splits and DOJ prosecution-by-type counts, at Section 7.8, likewise. Note a related reporting gap worth pursuing on its own: 22 U.S.C. 7103(d)(7)(I) requires the Attorney General to report the number of persons charged or convicted under each of the forced labour and peonage statutes and the sentences imposed, and recent annual reports do not appear to contain a per-statute table. Establishing whether that table has genuinely lapsed would materially sharpen Section 11.1, because it is the data that would most directly test the evidentiary-difficulty hypothesis.

1. An independent replication of INEGI's 2024 poverty figures. Carried from Key Assumption 1. Next step: check whether any Mexican academic body has re-run the published Stata, R or Python replication code against the ENIGH 2024 microdata.

1. Nicaragua. Not an open item so much as a standing negative finding, recorded here so a later pass does not repeat the search. There has been no monetary poverty measurement of post-2018 Nicaragua by INIDE, the World Bank, CEPAL or FAO. The country's last consumption-based measurement was fielded in October 2016. This was established by enumerating the producing institutions' own publication pages rather than inferred from an absence of search results, and it should be treated as settled unless INIDE publishes.

1. UNODC's guidance note on abuse of a position of vulnerability, and the Special Rapporteur's 2016 report on debt bondage. Both were located in the verification pass and neither was retrieved by the drafting pass; the UNODC note returned HTTP 404 on the attempt made. They matter because Section 6.4's third test, whether the Palermo Protocol reaches a property-secured migration debt, rests on this document's own reading of the treaty text with no interpretive authority behind it. Retrieving the guidance note would either raise that assessment above Low confidence or refute it. This is the highest-value single retrieval on the list.

1. Treaty status of El Salvador and Honduras under the 1956 Supplementary Convention. Section 4.8 records El Salvador as a signatory that never ratified and Honduras as not a party, read from the UN Treaty Collection status page in a single pass. Both are consequential claims, since they would mean two of the four Central American states covered are not bound by the principal debt bondage instrument, and both are the kind of claim easily produced by misreading a status table. Next step: re-check against the Treaty Collection's per-state depositary record and against each state's own ratification record.

14Note on the eventual VoT World-Building Document

This document is reference material and nothing else. It names no Vampires of Tucson character, faction, house, bloodline or location, and it draws no line whatsoever from any real person, organisation, region or event described here to any fictional one. No entity in this brief is a model for, an origin of, or a key to anything in the series.

It also does not represent that any real system described here operates the way a fictional organisation does, or that the mechanisms documented in Sections 5 through 7 map onto a fictional economy in any respect. The advance-payment recruitment systems, the smuggling debts, the prosecutions and the statistical series in this brief are described because they are what the public record contains, and for no other reason.

Where the eventual World-Building Document draws on this material, that is a separate, later step taken under clearly fictional licence. It should be treated as such: as invention informed by research, not as a continuation of the sourcing standard used here. Nothing in a fictional document inherits the Admiralty grades, the confidence labels or the correlation standard applied in this one, and no claim in a fictional document should be read as resting on the sources cited in Section 12.

The reverse also holds and is the more important direction. Nothing invented for the fiction may be written back into this document. If a later pass adds material here, it arrives by the same route everything else did: a public source, fetched, graded and cited, or it does not arrive at all.