The Price of a Trafficked Person, 2000-2025

KJKey Judgments

Eight judgments carry this brief. Each is an analytic conclusion, distinct from the reported facts beneath it, and each carries an explicit confidence level per §1.2. A judgment can be wrong even where every underlying fact is correct.

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• 4: Doubtful (some internal contradiction, or later superseded).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. Research was fanned out across six parallel background subagents covering separate lanes: international agency economic estimates, the academic price literature and its critics, US case records and financial intelligence, the southeastern Arizona and Sonora corridor, non-US comparative price and debt data, and empirical transaction-price studies. Two of those lanes spawned further subagents of their own. The author supplied two primary sources directly, GAO-15-154 and the ICAT issue brief on trafficking against smuggling, both of which are used and cited.

Every load-bearing figure was then re-fetched and re-read by the compiler against its own primary source before it was allowed into the document. That re-verification is the reason for the dagger convention in Section 12: sources the compiler fetched and read directly are undaggered, sources reached only through a research lane are marked, and roughly half of Section 12 carries the mark. No source in Section 12 is graded on a lane's summary alone without that being disclosed.

A separate adversarial verification pass was then run on the completed draft by a subagent on a different model (Fable), with the downloaded source PDFs available to it locally so that quotations could be checked against the documents rather than against the web. It was instructed to find errors rather than to confirm the document, and specifically to test the document's many claims that something is absent from the record, since absence claims are the easiest kind to get wrong.

That pass returned thirteen corrections and nine unresolved doubts. Every one of them was then re-verified by the compiler against the primary source before being acted on, and none was taken on the verifying model's word. The corrections are recorded inline, marked Correction (multi-model verification pass):, rather than silently applied; additions are marked Addition (multi-model verification pass): likewise. They were not cosmetic. The material ones were: a figure pair attributed to Bales' fieldwork in Thailand that is in fact the Amazon gold-mining towns of Brazil, and relayed to him by a named local informant rather than observed (Section 5.6, Section 9); a claim that the UNODC 2024 report contained “exactly four” currency passages when it contains five, which the same draft had already quoted elsewhere (Section 3.4); a stated reason for the 2017-to-2022 break in the debt bondage series that the 2017 report itself refutes (Section 3.3); two misattributions of who told GAO what (Section 3.2); a swapped offender composition in the Dank study (Section 5.7); a rule described as rescinded that was only proposed for rescission (Section 12); and three arithmetic overstatements (Sections 5.9.7, 6, 9).

One of the pass's findings inverted a note this document had written about its own method. An earlier draft said that the ILO's Figure 5b text-layer ordering “turned out to be correct” on inspection; it does not, and the corrected note at Section 2.2 now says so and explains which two regions a naive extraction transposes.

The most consequential thing the pass surfaced was not an error but an omission: that Siddharth Kara was one of the two named external peer reviewers of the 2014 ILO report that used his data as the principal input for its sexual-exploitation estimate. That is recorded at Section 2.3, verified against the report's own acknowledgements.

The pass had one significant limitation, which is recorded rather than smoothed over. It could not fetch justice.gov, state.gov, dol.gov or the Internet Archive, so it was unable to check any Department of Justice or US Attorney's Office quotation, the restitution aggregates, or most Mexican and Arizona press figures. The two most load-bearing DOJ quotations, the US$5,000 sale in the Northern District of Texas and the unit-economics passage from the Southern District of New York, were instead verified by the compiler directly through a browser session and are quoted from the live pages. The remainder of the US case-record material carries the dagger convention in Section 12 and has not been double-checked by a second party.

Two further corrections were caught by the compiler's own arithmetic re-check before the verification pass ran, and are recorded in the same way, at Section 5.3 and Section 6.6. A research lane also self-corrected two attribution errors, on the journal in which Anne Gallagher published her critique of the Global Slavery Index and on the H-Net network that carried Laura Agustín's review of Kara; both are recorded at Section 6.6 rather than quietly fixed, because in the first case the mistaken journal turns out to be where the criticised modelling was itself published.

One constraint on this session should be recorded plainly: the web-search budget was exhausted at 200 of 200 calls partway through the research passes. Later discovery ran through direct URL fetches and bibliographic APIs rather than search. Several items in Section 13 are open because of that limit rather than because the material does not exist, and they are marked as such.

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

Built from this document's own text. A table rather than expansion on first use, because a reference document is read by jumping to a section from the contents, so “first use” is invisible to a reader who lands in the middle. Foreign bodies are given their own name in their own language where that is what the source uses.

Tradecraft and analysis

United States federal

International and intergovernmental

Mexico

Italy

Currency and units

Bibliographic apparatus in Section 12

2The output side: what a trafficked person is estimated to yield

Source line: International Labour Organization, Profits and Poverty: The economics of forced labour, first edition 2014 and second edition 19 March 2024. Both are published by the ILO in its own voice. Both are modelled estimates, not observed transactions, and Section 2.3 sets out what they actually rest on.

2.1 The headline series (PRIMARY/OFFICIAL: ILO 2014 and ILO 2024)

The ILO states the 2024 per-victim figure precisely in a footnote: “The exact value of profit per victim is US$9,995 annually.” It states the sectoral gap as “US$27,252 for the former against US$3,687 for the latter.”

Forced commercial sexual exploitation accounts for about one quarter of the people in privately imposed forced labour and 73 per cent of the profit. That ratio, rather than any single dollar figure, is the durable finding. The 2014 edition described the same sector as “six times more profitable than all other forms of forced labour”; the 2024 edition's own numbers give a ratio of 7.4 to 1.

2.2 Per-victim profit by region, 2024 (PRIMARY/OFFICIAL: ILO 2024, Figure 5)

Note on how this table was read, because it is a live trap: the per-victim panel is a bar chart whose category labels and values do not extract from the PDF text layer in the right order. The text layer lists the regions as Europe and Central Asia, Arab States, Asia and the Pacific, Americas, Africa against the values 21,248 / 20,318 / 15,119 / 5,879 / 5,199, which would pair Asia and the Pacific with US$15,119 and the Americas with US$5,879. That is wrong, and it inverts the report's own narrative, which says that in Asia and the Pacific “profits per victim are relatively low” while in the Americas “the high level of per-victim profit is a more important driver.”

The table above was therefore read off the rendered figure image rather than the text layer, and the pairings shown are the correct ones. Anyone rebuilding this table must render the figure; trusting the extraction will transpose two regions.

The 2014 edition's regional per-victim figures are recorded here only to show the shape of the earlier estimate, and are not comparable: Developed Economies US$34,800, Middle East US$15,000, Asia-Pacific US$5,000, Africa US$3,900. The ILO says why they cannot be compared: “it is not possible to have fully comparable results across different editions due to improvements in methodology and data availability.” The two editions also use different regional groupings entirely.

2.3 What these figures actually are, and what they are not

This subsection is a methodology statement rather than a finding, and it governs how every number above should be used.

Neither figure is a price, and neither is observed. The ILO per-victim number is an annual profit estimate produced by a model. For the four labour sectors it is a value-added calculation: the labour income share of value added per worker, minus the wages the victim actually receives, multiplied by the estimated number of victims in that sector and region. In the ILO's own words the profits “represent the difference between what the employers are actually paying the workers and what they would be paying them in the absence of forced labour under normal circumstances.” No transaction is observed anywhere in that chain, and nothing in it describes a sum anyone paid for a person.

The wage input is a proxy taken from a survey that does not identify trafficking victims. The wages of people in forced labour, in every region and every sector, are estimated from the KNOMAD-ILO Migration and Recruitment Cost Surveys of 2015 and 2016: 2,454 migrants across nine bilateral corridors in 2015, and 3,149 across ten corridors in 2016. The ILO states the limitation plainly: “the KNOMAD datasets do not inform whether a migrant is a victim of forced labour.” Respondents were classified as potentially in forced labour by proxy conditions, among them not having signed a contract before migrating, having the contract changed on arrival, having no rest day, or being an irregular migrant. Because complete data existed only for Europe and Central Asia, those wages were used as the base and scaled to the other regions by the wage ratios of free workers.

The ILO names the resulting bias in both directions and quantifies neither. It says the assumption that migrants and non-migrants in forced labour earn similar wages “may lead to an underestimation of the wages of non-migrants in forced labour and an overestimation of illegal profits from the use of forced labour.” It then states the opposite risk, that surveyed migrants “might not be the workers experiencing the worst working conditions,” which “would lead to an overestimation of the wages of people in forced labour and, thus, an underestimation of profits.”

The sexual-exploitation figure rests on one researcher's dataset, and the ILO says so twice. The revenue side of the forced commercial sexual exploitation estimate comes from the Global Sex Trafficking Metrics database, 2016, published as an appendix to Siddharth Kara, Modern Slavery: A Global Perspective (Columbia University Press, 2017). In the methodology section the ILO calls it the source “which remains virtually the only source of comparable data for multiple countries on the economics of forced commercial sexual exploitation and revenues from it,” and in Appendix 1 “the only source of comparable data for several countries that can be used to estimate the revenues from forced commercial sexual exploitation.”

The dependency is not new, and it is the same researcher across both editions. The 2014 edition's Table 2.8 carries a column headed “Monthly earnings per victim (Kara, 2009),” and the values in that column drive that edition's US$99 billion total and US$21,800 per-victim figure. The 2024 edition uses Kara's 2017 database for the same purpose. The 2014 table also records what those monthly earnings were taken to be, by region: Asia-Pacific US$1,485, Latin America and the Caribbean US$3,200, Africa US$1,300, Middle East US$6,510, Central and South-Eastern Europe and CIS US$5,040. Its resulting annual profit per victim ran from US$11,000 in Africa to US$80,000 in the Developed Economies.

And the researcher was one of the report's own peer reviewers. The ILO's 2014 acknowledgements record: “Sincere thanks and appreciation are due to the external peer reviewers, Mr Siddarth Kara, Director of the Program on Human Trafficking and Modern Slavery at the John F. Kennedy School of Government at Harvard University and Mr Stephen Bazen, Professor of Applied Econometrics, Aix-Marseille School of Economics. Their comments and feedback helped improve the report, and in particular the methodology.”

Assessment High confidence the most widely circulated figure on the profitability of sex trafficking, together with every institutional restatement of it, traces to one researcher's fieldwork rather than to independent institutional datasets. This does not make the figure wrong. It means that the ILO estimate and Kara's estimate are one data point rather than two, and that citing both as agreeing sources is double-counting. Section 9 treats this as the document's central correlation problem.

3The input side: what it costs to put a person into the system

There is no global estimate of acquisition cost comparable to the ILO's profit estimate. The reason is not that nobody has tried. It is that in the large majority of documented cases there is no acquisition purchase to price, because the person is not bought. What exists instead is a well-measured literature on what the victim pays, and that inversion is the structural finding at the centre of this brief.

3.1 Recruitment fees: the worker pays to enter (PRIMARY/OFFICIAL: ILO 2024, Section 4)

The ILO's 2024 edition treats recruitment fees as a second profit stream, and deliberately excludes it from the US$236 billion headline:

• Illegal profits from recruitment fees and related costs charged to international migrants in forced labour: US$5.6 billion annually, 15 per cent of all illegal profits generated from international migrants in forced labour.
• For forced labour exploitation specifically, recruitment fees are 26 per cent of total illegal profits from international migrants, against 11 per cent for forced commercial sexual exploitation.
• From the same KNOMAD data, “migrant workers would require an average of 3.93 months of work to pay off recruitment fees and related costs.”
• The spread is extreme and context-specific. ILO surveys measuring SDG indicator 10.7.1 “reveal countries with the SDG10.7.1 indicator as high as one year and a half and others as small as less than one month,” and “No significant trends have been identified by sectors across countries.”

The ILO's own standard is that this figure should be zero. Its 2024 report states, in Box 3, that its General Principles and Operational Guidelines for Fair Recruitment and Definition of Recruitment Fees and Related Costs (2019) “state that 'no recruitment fees or related costs should be charged to, or otherwise borne by workers or job seekers' and call on governments to 'take measures to eliminate the charging of recruitment fees and related costs to workers and jobseekers.'”

Small precision, recorded because this document quotes both instruments elsewhere: the 2019 instrument's own text is “No recruitment fees or related costs should be charged to, or otherwise borne by, workers or jobseekers.” The 2024 report's rendering drops a comma and splits “jobseekers” into two words in the first clause while spelling it as one in the second. The quotation above is verbatim from the 2024 report, which is the document actually read for this section; anyone quoting the underlying 2019 principles should use their wording rather than the 2024 report's paraphrase of them.

Two features of this stream bear directly on the question this brief asks. First, the fee is paid by the person who will be exploited, not for them. Second, because the ILO excludes it, both the US$236 billion total and the US$9,995 per-victim figure understate the take by an amount the ILO declines to estimate globally, saying only: “The estimate therefore understates total illegal profits from forced labour.”

3.2 The same mechanism inside a lawful US visa programme (PRIMARY/OFFICIAL: GAO-15-154)

The Government Accountability Office examined the H-2A and H-2B temporary foreign worker visa programmes and found the recruitment-debt mechanism operating inside a legal channel, in the Mexico-to-United States corridor. Charging H-2 workers recruitment fees is already prohibited by regulation (20 C.F.R. §§ 655.22(g)(2)).

GAO's documented amounts, all from its own worker discussion groups:

GAO records the financing around those fees, and the attribution differs by clause in a way worth preserving. The loans and the collateral come from third parties, not from the workers themselves: “some NGOs and federal officials we interviewed told us that in order to pay these fees and other pre-employment expenses, such as transportation and passport fees, workers often had to take out loans, sometimes with high interest rates,” and “Federal officials and NGO representatives said, in some cases, workers may put up their own land or property as collateral.” Only the interest rate is the workers' own account: “some said the interest rate on their and others' loans was between 10 and 15 percent.” GAO then states the consequence in the statutory vocabulary: “Borrowing money at high interest rates to pay these fees can result in debt bondage, a possible indicator of human trafficking,” citing the definition at 22 U.S.C. § 7102(5).

The forestry group is the clearest single illustration of the input-output gap in any US federal document located during this pass. Those seven workers paid roughly US$1,250 each and all took out loans to do it. They were told they would earn US$10 per hour for eight hours a day and US$15 per hour above forty hours. On arrival they learned they would be paid by the amount of work completed rather than by the hour, with deductions for equipment and safety vests. GAO records the result: “After 2 weeks of working 10-11 hours per day, the workers said they earned $133 and owed $100 in rent.”

Grading note: these amounts come from GAO's own discussion groups, which GAO convened with the assistance of a migrant-rights organisation. GAO caveats them itself: “The results of these discussion groups cannot be generalized to the larger population of H-2A and H-2B workers.” A separate and similarly worded caveat in GAO's footnote 70, that such surveys “represent the views of the workers interviewed and cannot be generalized to the population of all H-2A and H-2B workers,” attaches to two NGO surveys GAO cites rather than to its own groups; an earlier draft of this section applied that second caveat to the discussion groups. Both say the same thing about generalisability. These amounts are cited as documented individual sums, not as population estimates, and are graded accordingly in Section 12.

3.3 Debt as the mechanism of control, and a definitional break in the series (PRIMARY/OFFICIAL: ILO/Walk Free/IOM 2017 and 2022)

Hold these apart. The apparent collapse from 51 per cent to 20.9 per cent in five years is a denominator and definitional change, not a measured fall in the use of debt. Two differences do the work:

• The denominator differs. The 2017 figure is a share of all victims of forced labour imposed by private actors, which includes forced commercial sexual exploitation. The 2022 figure is a share of adults in forced labour exploitation, a category that excludes forced commercial sexual exploitation.
• The definition differs. The 2017 edition uses a loose formulation, debt bondage as the condition “in which personal debt is used to forcibly obtain labour.” The 2022 edition attaches an unusually narrow two-part test, counting debt bondage only “(1) when a person is coerced work against their will to repay a debt with employer or recruiter… or (2) when debt is manipulated to compel a person to perform work tasks or accept work conditions that he or she would otherwise refuse.”

For completeness, the 2022 edition reports coercion by “systematic and deliberate withholding of wages” at “More than one-third (36 per cent) of adults in forced labour in the private economy,” and coercion through “the manipulation of debt” at about 5 per cent.

Assessment Moderate confidence the 2017 and 2022 debt bondage figures measure different things, and any narrative that treats them as a time series showing debt bondage in decline is comparing two definitions rather than two years. Neither edition states the size of any debt.

Sectoral detail, 2022 edition, Figure 13:

3.4 The absence at the centre of the record (PRIMARY/OFFICIAL: UNODC GLOTIP 2024)

The UNODC Global Report on Trafficking in Persons 2024, the flagship global report on the subject, covering 156 countries and cases detected between 2020 and 2023, contains no price for a trafficked person, no average value, and no estimate of acquisition cost anywhere in its 176 pages. A full-text sweep located five passages carrying a currency figure, four of them in the chapter on trafficking for forced criminality:

The USD 2,500 figure is one of the very few sums in any intergovernmental publication that is unambiguously a payment made for a person rather than by one. It is reported at one remove, as cases “reportedly paid,” and it belongs to a specific and recently emerged criminal economy rather than to trafficking generally.

The fifth passage sits in the Africa chapter rather than the forced-criminality one, and it is the smuggling-fee-to-debt conversion of Section 4.2 with a figure attached: victims of sexual exploitation “are reportedly forced to repay the equivalent of up to US$2,250 for the smuggling fee, suggesting that smuggling fees are inflated as a means of controlling trafficking victims through debt bondage.” The same passage records an unpriced sale immediately before it: a smuggler “introduces women, or 'sells' them, to a local community leader ('chairman'),” who “then hands them over to a pimp, for a fee.”

Two further unpriced sales appear in the same report, noted here because Section 11.1 turns on them: in Libya, detainees “are 'sold' by a prison guard to third parties who exploit men for forced labour,” and there are “private citizens who may pay a ransom for detainees and then force them to work in order to pay off what was paid.”

Separately, a footnote on the Horn of Africa route records that on reaching Yemen “migrants are met and captured by traffickers who pay a fee to the smugglers to 'buy' them and later exploit them for their own gain.” Provenance caution: that footnote is UNODC summarising a report by Free the Slaves, an advocacy organisation this document rejects for load-bearing use in Section 8. It is reproduced as an illustration of the practice, at the grade of the underlying advocacy source rather than at UNODC's, and no amount is attached to it in any case.

4Six transactions, one word

Almost every dispute about “what a trafficked person is worth” turns out on inspection to be two people using one word for six different transactions. This section defines them, because every figure elsewhere in this document is meaningless without the label.

4.1 The governing distinction: what the commodity is (PRIMARY/OFFICIAL: ICAT Issue Brief 01, 2016)

The Inter-Agency Coordination Group against Trafficking in Persons, which speaks for the UN system on this definitional question, draws the line at what is being sold. Its formulation is the cleanest available and it settles the vocabulary for the rest of this document:

Note what that first row does to the question this brief asks. Under the internationally agreed definitions, a price for a person exists only on the trafficking side. On the smuggling side the fee is a price for a service, and the person paying it is the customer, not the merchandise. Any figure that crosses this line without saying so has changed the subject.

4.2 The five distinctions that carry most of the confusion

Smuggling fee against trafficking debt. A smuggling fee is a price for a service between parties whose relationship is intended to end at the destination. A trafficking debt is a mechanism of control that begins there. The two are frequently the same underlying sum, converted, and ICAT describes the conversion as a recognised pathway rather than an anomaly: “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.”

UNODC records the same conversion with a number attached in its 2024 report, noting victims in North Africa “reportedly forced to repay the equivalent of up to US$2,250 for the smuggling fee, suggesting that smuggling fees are inflated as a means of controlling trafficking victims through debt bondage.”

Two features of this pathway matter for pricing. The debt is not set by the cost of anything: ICAT's phrasing is that the rate is set at a level “that makes the initial 'debt' impossible to pay off.” And the same person can be a customer at one hour and merchandise at the next without any transaction occurring in between, which means the moment a “price” attaches to them is not the moment money changes hands.

Purchase price against debt imposed. These move in opposite directions and are often confused because both attach a number to a person. The purchase price is paid for the victim by a criminal. The debt is charged to the victim. In the best-documented single case in this brief, both appear at once and differ by more than four times: Belgian authorities reported a group that “paid 2,800 USD for each woman recruited” and then “demanded the victims pay off a debt of 12,000 USD.”

Price against revenue. The single most common error in circulation is to quote an annual profit figure as though it were a sale price. The ILO's US$27,252 is an estimated annual yield from a victim in forced commercial sexual exploitation. UNODC's median observed purchase price is US$870. Both can be true at once, and they are answers to different questions.

Cost to society against market value. The UK Home Office's per-victim figure of £328,720 is regularly cited as though it valued a trafficked person. It measures the opposite. It is the public cost of one case of modern slavery in 2016/17 prices, dominated by a £271,190 estimate of physical and emotional harm to the victim, and the report contains no criminal-revenue content at all.

Recruitment fee against acquisition cost. The ILO's recruitment-fee data and UNODC's purchase-price data are both, loosely, “what it costs to get a person into the system.” They are borne by opposite parties. Section 9, Signal 1 treats the consequences.

Assessment High confidence the phrase “the average market value of a human being” does not denote a single measurable quantity. There is no clearing price, no exchange, and no market in the economic sense, and the six transactions above are not substitutable for one another. A figure offered without saying which of the six it is should be treated as uninterpretable rather than as high or low.

5What people are actually bought and sold for: the observed record

5.1 The only systematic transaction-price dataset located (PRIMARY/OFFICIAL: UNODC GLOTIP 2020)

The UNODC Global Report on Trafficking in Persons 2020 contains a section, “The traffickers' financial gains,” that has no counterpart in the 2022 or 2024 editions. It is the only place in the intergovernmental literature where prices paid for people are compiled from court records rather than modelled.

The data: “Between 2007 and 2017, 15 countries in six regions of the world reported information on the monetary value exchange between traffickers to recruit a victim.” Thirty cases, drawn from 489 case narratives that member states supplied to UNODC. Twenty-six involved women, two involved girls, one a man and one a baby.

Hold one distinction inside this dataset. The 23,600 USD maximum is the top of the 30-case recruitment series above. Elsewhere in the same report UNODC cites “examples where victims were sold for up to 25,000 USD each in countries where they were going to be exploited,” which is an exploitation-phase sale and a different quantity. The two are frequently quoted as though they were one figure.

The gap between the mean and the median, 3,662 against 870, is the shape of the distribution: most transactions are small, and a few large ones drag the average up by a factor of four. Anyone quoting a single “average price” from this dataset should quote the median.

The phase matters as much as the number. UNODC separates recruitment from exploitation, and the price rises along the chain:

• Recruitment, Central and South-Eastern Europe: “recruiting one victim (for 1,500 USD - 2000 USD) to be later sexually exploited.”
• Recruitment, South-East Asia for forced marriage: “reported to be 3,000 USD maximum,” which UNODC notes “is about the same as the annual minimum wage in the subregion.”
• Exploitation phase, East Asia: “a trafficked bride can be sold for about 10,000 USD.”
• Cross-border resale: “Victims recruited for a few thousand dollars in South-East Asia are sold for sexual exploitation in Australia or in Japan for 10,000-15,000 USD.”

UNODC's own comparators are deliberately chosen and worth reproducing, because they are the agency's way of saying the number is small: recruiting one victim in Central and South-Eastern Europe for 1,500 to 2,000 USD “is worth the same to criminals as selling 80-100 grams of methamphetamines, or as little as one pistol in the illegal firearms market or the value of two kilos of illegally traded live glass (juvenile) eels.” At the exploitation end, a victim sold for 10,000 USD in East Asia is “cheaper than the value of 15 kilograms of ivory or than 40 grams of crystal methamphetamines sold to the final consumer.”

UNODC states plainly that this contradicts the received figures. “While these findings are not representative of the dynamics of the trafficking in persons market worldwide, they do challenge many of the market value estimations commonly provided.” It names the outlier it is challenging: an analysis of US legal proceedings finding “that intermediaries may charge up to 100,000 USD per person recruited,” of which UNODC says “these figures may be sensational outliers, and that the actual monetary transaction to recruit a trafficking victim appears to be far less than expected.”

And it says recruitment may not even be good business. “Comparing the monetary returns described in the analyzed court cases to national average incomes, suggests that recruiting trafficking victims may not be more profitable than an average job in the legal sphere. A single trafficker recruiting victims in Central and South-Eastern Europe and selling them to groups in Western Europe would have to recruit at least 20 women per year to reach the average annual salary in their own countries.”

Its closing formulation is the finding stated as plainly as an agency will state it: “The relatively low income made by single traffickers show how little the victims are valued in the illicit market of trafficking. Traffickers trade their victims as commodities. Cases reveal that victims are 'priced' at as little as a few hundred dollars, equivalent to a few grams of methamphetamine.”

The denominator is the caveat. Thirty of 489 case narratives carried a purchase price, and 49 of 489 carried an estimate of proceeds. Roughly six per cent of the court cases UNODC assembled recorded what anyone paid for a person. UNODC does not claim the sample is representative and neither does this document.

5.2 Earnings per victim in the same dataset

From the same section: “In the large majority of reported cases, however, traffickers earned less than 5,000 USD per victim (in 12 cases out of 16 cases reporting this information). Only 3 cases reported monthly earnings per victim above 10,000 USD per victim.” Total proceeds per case were “generally in the range of a few thousand US dollars, except for some cases in which they exceeded 100,000 USD,” those exceptions being large organisations exploiting many victims over years. UNODC describes “One of the highest proceeds in the court cases collected” as 1.1 million USD, from a group trafficking children for forced pickpocketing across Western and Southern Europe over two and a half years. It is not the largest sum in the section: the same passage records that “US law enforcement traced tens of millions of dollars to an organization involving 35 offenders who have trafficked hundreds of women from South-East Asia to several US cities.”

UNODC closes with a statement that bears directly on every headline number in Section 2: “Estimating the global size of the trafficking in persons market in terms of illegal profits remains challenging given the lack of a reliable estimate of the global number of victims.”

5.3 The worked example: one case with both sides on the record (PRIMARY/OFFICIAL: UNODC GLOTIP 2020, Belgian authorities)

This is the cleanest input-output pair in any official source located.

The women were “originally 'bought' in South-East Asia from another group that specialized in recruiting women with the promise of a job in Europe.” The debt imposed is 4.3 times the purchase price.

The arithmetic of the case is internally consistent and worth setting out, because it shows what UNODC's profit figure is actually made of. Fifty victims at a US$12,000 debt each is US$600,000 receivable; fifty acquisitions at US$2,800 each is US$140,000 paid out; the difference is US$460,000, against UNODC's stated estimate of “about 450,000 USD.” UNODC's own qualifier confirms the construction: it says the estimate considers “only the recruitment costs.” So the per-victim figures are an acquisition cost of US$2,800 against roughly US$9,000 of profit over two years, and acquisition is about 31 per cent of gross rather than a negligible fraction of it.

That 31 per cent matters, because it is far higher than the ratios in every other case in this brief. This is the one documented instance where buying people is a substantial share of the cost base, and it is also the only case where the buyers paid a genuine market price to an arm's-length supplier rather than acquiring someone through deception, family or debt. Assessment (Confidence: Low): acquisition cost may be a meaningful expense specifically where a criminal group buys from an unrelated group at arm's length, and negligible everywhere else. One case cannot establish that, and it is recorded as a hypothesis.

UNODC also states what the debt is for, and it is not repayment: “In some cases, however, the debt serves more as a coercive instrument than as a source of income, as it bonds the victim to the trafficker and discourages them from escaping or seeking alternatives.”

5.4 European casework figures (NAMED OFFICIAL: Europol, Eurojust)

Europol opens its assessment of this subject by disclaiming the ability to do it: “Based on the information available to Europol, it is challenging to estimate the overall profits derived from THB activities in the EU. Some Member State (MS) contributions do provide assessments of profit, but in most cases the data is incomplete and lacks a consistent methodology for calculation. This makes any estimate of THB profits extremely unreliable.” It adds: “No evaluation of profits derived from THB activities in the EU specifically is available.” The profit figures Europol does reproduce it attributes to ILO and open sources, not to its own casework.

The case-level amounts it does publish:

Note the sham-marriage line specifically. It is one of the few places where the amount paid to the person and the amount paid for the person appear side by side in an official document, and they differ by a factor of five to thirty-three.

5.5 The Nigeria to Italy corridor: where the debt is the asset (RESEARCH INSTITUTION: UNICRI 2010)

UNICRI's study of Nigerian women trafficked to Italy records the debt range and its provenance: “the amount of the debt varies from 10,000 to 30,000 euro, with picks as high as 50/60 thousand all the way to 80/100,000,” citing work by the Direzione Distrettuale Antimafia of Naples (2008). Some victims “stated to have paid only around 3,000 euro for the trip to the Algerian, Tunisian or Libyan coats. Once in Italy, they stipulated another debt… which, added to the first loan, reached up to 30/40,000 euro.”

The mechanism by which the debt inflates is currency fraud, and it is documented in the victims' own words: the debt “is usually very high since very often the maman swindle the women by lending money in the Nigerian currency, the Naria, but pretending to be repaid in euro.” One victim: “We agreed for a loan of 45,000 Naira that, once in Italy, became, according to her calculations 35,000 euro. Not knowing the exchange rate I thought it was a fair deal.” UNICRI's own footnote puts 45,000 naira at roughly 270 euro as of June 2009. The stated debt is therefore about 130 times what was actually lent.

Against that, the earnings: “During the week I made about 70-80 euro per day and, sometimes up to 100. On Sundays, C. came to take all my money, approximately 600-700 euro.”

The structurally important finding is what the study does not contain. Victims in it are repeatedly sold between exploiters, in their own testimony: “C. had sold me to F., her sister. F. told me that she was my new maman and that from that time on I was supposed to pay her. She added that the money I had already given C. did not reduce my debt.” And: “T. laughed and said that S. had sold me to her and that she was the one that I had to repay.” No sale price is recorded for any of these transfers.

Assessment Moderate confidence in this corridor the tradable asset is the debt rather than the person, which is why a purchase price does not appear even in a study built on victim testimony about being sold. What changes hands between exploiters is the right to collect. This is a plausible partial explanation for why purchase-price data is scarce in every jurisdiction, and it is offered as a hypothesis rather than a demonstrated general rule; it is tested in Section 11.2.

5.6 Field-observed prices in the academic literature (PEER-REVIEWED/ACADEMIC: Bales 1999-2012, Kara 2016)

Kevin Bales' Disposable People reports prices he encountered in fieldwork rather than derived from a model:

Bales puts one input beside its output directly: “The girl who 'cost' $150 can be sold for sex up to ten times a night and bring in $10,000 per month.” That ratio, roughly 1 to 67 in the first month, is the disposability argument in a single sentence.

Siddharth Kara's published summary figures, from The Solutions Journal, 22 February 2016:

• “the global weighted average acquisition cost of a trafficked sex slave is approximately $1,900 and the global weighted average monthly profits generated through the commercial sexual exploitation of that slave are approximately $2,400.”
• “old-world slaves could be purchased for a global weighted average of $12,000 or more in today's dollars and generated roughly 15 percent to 20 percent in annual return on investment,” against “today's slaves sell for a global weighted average of $400” generating “anywhere from 200 percent to 500 percent or more in annual return on investment.”
• Sample as stated in that article: “ten years of research across twenty countries,” having “directly interviewed more than one thousand current and former slaves of all kinds.”

Kara's US$1,900 acquisition figure is the closest thing in the literature to a global average purchase price for a sex trafficking victim, and it sits between UNODC's observed median of US$870 and its observed mean of US$3,662. That is a genuinely interesting convergence and it is examined, with its limits, in Section 9, Signal 2.

5.7 The United States: observed prices from 142 convicted offenders (FEDERALLY FUNDED RESEARCH: Dank et al., NIJ 2014)

The largest body of observed US pricing comes from a National Institute of Justice funded study across eight cities: San Diego, Seattle, Dallas, Denver, Washington DC, Kansas City, Atlanta and Miami. Its qualitative base is “interviews with 119 stakeholders and 142 convicted offenders,” the offenders comprising 73 pimps and traffickers, 36 sex workers and 33 people convicted of child pornography offences.

Prices charged to buyers are recorded per respondent rather than averaged, and the spread is very wide: from US$15 to US$350 per fifteen minutes, US$30 to US$590 per half hour, and US$40 to US$1,000 per hour across the eight cities. By act, the study's cross-site summary is that “Prices for these women and girls depended on the sex act and ranged from $60 for oral sex to $150 to $200 for intercourse.” One respondent describes the pricing strategy explicitly: “In the ad, I would charge $80 for a half-hour. My partners would charge $350 [for in-calls], $250 for out-calls. But who makes more money? Neiman Marcus or Walmart?”

Quotas imposed on victims, which are the closest US equivalent of the Mexican cuota in Section 7.8:

Only “Eighteen percent of respondents reported that they imposed a quota,” which is a useful corrective to the assumption that quotas are universal. Where they exist they are enforced by violence: “on the weekends I have to make $1,000 or $1,500 … you didn't come home until you had it … if they didn't come back with it they were going to get beat.”

How the money divides depends almost entirely on the venue, and this is where the “what is she paid” side of the question gets its clearest answer:

• Pimp-controlled street and internet work: nothing. “Pimp-controlled women and girls are forced to hand all earnings over to their pimp. Some might get a daily allowance of $80 to $100 to cover their hotel room and food for the day.” The study records an officer's phrasing of the same point: “she's carrying around a couple hundred dollars in her pocket … it's not her money.”
• Escort service on contract: a real split. “if the fee was $300, then they would pay the escort service $125, and keep $175, plus tips.” In San Diego, “There is usually a 60-40, or in some cases an 80-20, split between the girl and the agency.”
• Asian massage parlours in Washington DC: the house takes a US$60 to US$80 house fee, the worker nominally keeps a US$120 tip, and the operator then reclaims it through charges for the right to live there, “anywhere from $50 a day to $650 a week,” plus fines: “you did not put your towels in the laundry basket so that is a $50 fine.”
• Latino brothels: “These women keep half of the money charged, but may also be charged fees for condoms or other items … She was charged $10 for 50 condoms.”

An independent CUNY dissertation study of 85 third parties in Harlem, conducted separately, puts the mean take at 60.3 per cent with a modal split of 50/50, and reports pimp income with a mean of US$71,599 against a median of US$8,550. The eight-fold gap between that mean and median is the same distributional shape as UNODC's price data, and the author's own caveat is that respondents “may embellish their earnings.”

And the two per-person prices. In 350 pages built on 142 offender interviews and eight cities of case knowledge, exactly two transactions transfer a person for a stated sum:

• Denver: “In one case, a girl ran away from one pimp to another and the first pimp paid the other pimp $400 to keep her.”
• Kansas City, a Chinese smuggling and exploitation route: “By the time they reached the United States, several months passed in which they would have accrued a debt of approximately $50,000 per person. Since an individual's family often times would have only raised about $5,000 to fund the trip, the individual would be in debt to the snakehead for $45,000. At this point, the snakehead would sell the individual to a 'business owner' for $45,000.”

Assessment Moderate confidence the Kansas City figure is the most analytically revealing number in this brief. The sale price is exactly the debt balance. What the buyer purchases is not a person in the abstract but a receivable of a stated size, and the person is the security for it. That is the same structure UNICRI documents on the Nigeria to Italy corridor in Section 5.5, arrived at on a different continent through a different criminal network and a different research method, and it offers a mechanism for why purchase prices are so rarely recorded: where the debt is the asset, the debt figure is the price, and it is booked as a debt rather than as a sale.

Grading note: this study is an Urban Institute product, and this document excludes named policy think tanks. It is admitted because it is a federally funded NIJ grant report (award 2010-IJ-CX-1674, NIJ document 245295) with a stated methodology, and it is cited in that form. Its market-size estimates, which range from US$39.9 million in Denver to US$290 million in Atlanta for 2007, are model outputs and are not used here as observations.

5.8 Court awards are not prices (PEER-REVIEWED: Cusveller and Kleemans 2018)

A study of all 190 injured-party claims in Dutch first-instance human trafficking judgments from 2013 and 2014 found an average claim submitted of EUR 68,657, with a minimum of EUR 740 and a maximum of EUR 1,065,000, and an average amount actually awarded of EUR 36,927.

The reason this belongs in a document about prices is a caution rather than a figure. Awards in these cases are set by judicial convention rather than measured from evidence. In prostitution-related claims the amounts awarded “varied between €100 and €500 per day worked,” and the authors record one case where the gap is explicit: “Although one trafficked woman stated that she had to earn €1,000 each day and €2,000 per day at weekends, she was awarded compensation of only €500 per day, which was stated to be the usual amount awarded per day in case law.”

Assessment High confidence court-derived money figures in trafficking cases are floors set by legal convention, not measurements of what was taken. Any dataset built from awarded compensation will understate actual earnings by an unknown margin, and should never be read as a market price.

5.9 The United States federal record: what is prosecuted and what is priced (PRIMARY/OFFICIAL: DOJ, FinCEN, GAO, USSC)

5.9.1 The two completed sales located in the United States record

A woman sold for US$5,000, with a written contract. In October 2020 the US Attorney's Office for the Northern District of Texas charged two men over the sale of a 19-year-old woman. According to the criminal complaint, the first defendant “advertised her as a 'slave' on a fetishism website, where he offered to sell her to the highest bidder. Mr. Hubert, screen name 'The Darkest Lord,' offered $5,000.” The transaction is then confirmed twice over from the buyer's side: “Mr. Hubert informed him that he had purchased the victim for $5,000 and demanded $5,000 back in exchange for her safe return. He even sent the victim's father a 'contract,' signed by both Mr. Juarez and Mr. Hubert, as proof of the 'sale.'”

Note the resale price. Offered back to her own family, the price was identical to the purchase price, US$5,000, which is the behaviour of a party recovering an outlay rather than pricing an asset.

Grading note: this is a criminal complaint, which the Department states is “merely an allegation of wrongdoing, not evidence,” with both defendants presumed innocent. It is cited as a documented allegation carrying a stated figure, not as a proven transaction.

A person sold for US$45,000, being the exact balance of their own debt. The Kansas City snakehead case set out in Section 5.7. Note what this one is and is not: it comes from an NIJ-funded grantee report relaying a Kansas City law enforcement account of how a smuggling route operated, not from a charging document or a court record. It is a documented law-enforcement description, not a federal record of a completed prosecution, and it is the only one of the two that is not a case filing.

5.9.2 Price formation, in the one case where the negotiation is on the record

The clearest evidence anywhere in this brief of how a price is actually arrived at comes from an inchoate case, where the buyers were undercover FBI agents. A man planning to sell his 14-year-old daughter “boasted to his co-worker that he would sell the child to a pimp he knew in Memphis, Tenn., and hoped to get as much as $40,000 because she was 'a young, clean virgin.' Ultimately, his contact in Memphis offered him only $8,000 for the girl.”

Three things are visible here that no aggregate figure shows: an asking price, a counter-offer at one fifth of it, and an explicit premium claimed for virginity and youth. The gap between US$40,000 asked and US$8,000 offered is the difference between a seller's belief about value and a buyer's, and it is a caution against treating any single reported price as a market rate.

5.9.3 The structural silence in the Department of Justice record

The Department of Justice repeatedly describes traffickers buying and selling people, and then does not say for how much. Documented instances of that exact gap:

• Northern District of Texas, 2016: “Some of the member of the group bought and sold the girls and women they were trafficking amongst themselves.” No figure.
• Southern District of Georgia, Operation Blooming Onion indictment: one defendant “unlawfully traded or sold foreign workers” with another. Across 54 pages, the only dollar figure is the US$200,000,000 the enterprise is alleged to have made.
• Eastern District of North Carolina: a release headlined “Jacksonville Man Sentenced… for Selling Women as Sex Slaves All Across America” contains no dollar figure at all.

On a close reading of roughly 27 Department trafficking releases, two carried a purchase price, about seven per cent. Restitution and forfeiture, by contrast, are quantified in nearly every sentencing release.

Assessment High confidence the Department of Justice reliably publishes what a court orders and not what a trafficker charged. The absence of purchase prices in its releases is a property of what a press release is for, not evidence that the transactions are rare. Section 11.2 weighs this against the competing explanations.

A search caution worth recording: a US Attorney's release titled “Five Who Purchased and Sold Children for Sex” describes sex buyers purchasing sex acts, not anyone purchasing a child. Any keyword search on this question will surface it, and it does not mean what the headline appears to mean.

5.9.4 Unit economics, and the answer to “what is she paid” (PRIMARY/OFFICIAL: USAO SDNY, 2013)

The single most complete statement of the revenue split in any located official document describes the Tenancingo to New York corridor, which is the same network treated from the Mexican side in Section 7.8:

“In a typical day, a Mexican sex trafficking victim in New York has sexual intercourse with 20 to 30 customers. Each customer usually pays $30-$35 for 15 minutes of sex. Of that $30-$35, $15 typically goes to either the driver who transported the woman to the client, or to the residential brothel where the woman worked. The other $15 goes to the victim, who is then typically forced to give all of it to the trafficker.”

Worked through: at 20 to 30 customers a day and US$30 to US$35 each, the gross is roughly US$600 to US$1,050 a day, which is US$219,000 to US$383,000 a year at seven days a week. The victim's nominal share is US$15 per act, roughly half the gross, and her actual share is zero. That last clause, “who is then typically forced to give all of it to the trafficker,” is the answer to the input-side half of the question this brief was asked: on the record of a US Attorney's Office, in this network, the person is paid nothing.

This matches the Dank study's finding for pimp-controlled work in Section 5.7 exactly, arrived at through offender interviews in eight other cities rather than through a prosecution. Two independent routes, same answer.

5.9.5 Debt in the US federal record (PRIMARY/OFFICIAL: FinCEN, DOJ, NIJ)

Correction and discrepancy, recorded rather than resolved: FinCEN's 2020 advisory states a single figure of US$55,000 per victim for the Thai trafficking case it describes. The Department of Justice release on the same underlying prosecution states a range, “an exorbitant 'bondage debt' of between $40,000 and $60,000.” FinCEN's figure sits inside the Department's range but is presented as a point estimate. Anyone citing the US$55,000 should know it is a single-case figure reported two different ways by two US agencies.

The Florida case states the coercive purpose of the debt in the government's own voice, which is unusually direct: the ringleader “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.”

A third independent sighting of the debt-as-asset structure. The NIJ labour trafficking study records that for some victims, “their smuggling debts were bought by employers (commonly agriculture contractors, some of whom had arranged for their smuggling), who then trafficked them.” That is the same mechanism as the Kansas City sale in Section 5.7 and the Nigeria to Italy transfers in Section 5.5: what changes hands between criminals is the debt, and the person follows it. Three sightings, on three continents, in three unrelated bodies of research.

5.9.6 What FinCEN publishes, and what it does not

FinCEN's suspicious-activity data gives the only US figures on the scale of the money, as distinct from prices:

• “approximately 12,100 BSA reports identifying at least $7.1 billion in suspicious activity” over January 2020 to December 2021.
• Of that, “2,311 BSA reports referencing CVC in connection with OCSE and human trafficking, totaling over $412 million,” with a single June 2021 report accounting for US$163,755,867 of it.

Two negatives are worth recording. FinCEN's ten human-trafficking red-flag indicators contain no dollar thresholds at all; the only dollar-denominated red flags in its entire trafficking corpus are the structuring thresholds of US$3,000 and US$10,000, which describe evasion of reporting rules rather than the price of anything. And FinCEN's qualitative statement on recruitment fees is deliberately unquantified: “Recruitment fees can range from hundreds of dollars to tens of thousands of dollars, and take years to repay.”

5.9.7 What US law says a person's exploitation is worth, and what courts actually order

Congress wrote a per-person valuation formula into statute. Under 18 U.S.C. § 1593(b)(3), the “full amount of the victim's losses” includes “the greater of the gross income or value to the defendant of the victim's services or labor or the value of the victim's labor as guaranteed under the minimum wage and overtime guarantees of the Fair Labor Standards Act.”

That is the only place in this entire brief where a government defines what a trafficked person's exploitation is worth. It is a floor set by labour law, and no federal body publishes an aggregate of amounts ordered under it.

What is published, case by case, in the Attorney General's annual report to Congress, is the restitution actually ordered. Aggregated from those case tables:

These aggregates were computed from the reports' case-by-case appendices, which publish no totals of their own. They are derived figures and are labelled as such.

Three observations. The median is stable at roughly US$25,000 to US$36,000 per convicted defendant across the four sampled years. The mean runs between three and thirteen times the median, depending on the year (3.4 in FY2018, 9.9 in FY2021, 12.7 in FY2022, 7.8 in FY2023), the same heavily skewed shape as every other distribution in this brief. And in FY2023 there were 145 restitution awards against 289 convictions, meaning roughly half of convicted federal traffickers have no restitution order recorded, despite section 1593 making restitution mandatory.

Per-victim variance is enormous where it is itemised. In the one located release that breaks restitution down by survivor, US$1,510,300 was ordered across seven people: US$508,000, US$40,000, US$91,300, US$252,000, US$264,000, US$10,000 and US$345,000. A fifty-fold spread within a single case is a caution against any per-victim average.

Assessment High confidence restitution figures are not prices and are not measurements of what was taken. They are court-ordered sums, roughly half of eligible defendants never receive an order at all, and collection is negligible: GAO found that the Department of Justice “collected $1.5 billion (4 percent), of the $34 billion ordered from fiscal years 2014 through 2016,” with 91 per cent of outstanding criminal restitution debt classed as uncollectible.

Note on the sentencing data: there is no US Sentencing Commission product on human trafficking offences. The Commission has no trafficking offence category, folding section 1591 cases into “Sexual Abuse,” so no federal sentencing statistic isolates trafficking.

5.10 China: the one peer-reviewed study with observed purchase prices (PEER-REVIEWED: Shen 2016)

Anqi Shen's study of female perpetrators in internal child trafficking in China rests on semi-structured interviews with ten women randomly selected from the forty-four incarcerated for child trafficking in one Chinese prison as of July 2013, all ten of whom agreed to take part. It is small, single-site and offender-sourced, and the author says so. It is also the only peer-reviewed study located that records prices actually paid for people.

The paper supplies its own conversion, “10 yuan = approximately US$1.57,” which is what the figures above use.

Two things in it bear on this brief's wider argument. The first is the wage context that makes the numbers legible: one respondent's alternative was mixing sand on a building site at 30 to 40 yuan a day, and she said of the courier fee that “300-400 yuan for undertaking a train journey to transport a baby was so tempting that I had no hesitation to take the opportunity.” The second is the author's own caution against the profitability assumption, which converges with UNODC's: the respondents “could potentially receive payments in the range of 300/400 to 25,000 yuan for their services. However, the reality is that the amount they could eventually receive depended on a number of factors and nothing was guaranteed.”

Assessment Low confidence a purchase price of roughly US$314 for an infant, in one Chinese province in or before 2013, is consistent with the low end of UNODC's observed distribution and with Bales' field observations, but it is one figure from one interview in a study of ten people and it supports no average of any kind. It is recorded because observed purchase prices are extremely rare, not because this one is representative.

Note on what this does and does not change: this study is interview-based rather than built on police, prosecution or court files. The document's finding that file-based research documents flows rather than lump-sum sale prices is unaffected by it, and if anything is reinforced, since the price here surfaced only when a researcher sat down with the offenders themselves.

6The price-collapse claim, tested

The best-known statement about the economics of modern slavery is that the price of a human being has collapsed: roughly $40,000 in the nineteenth century against roughly $90 today, and that the collapse is why victims are now treated as disposable. The claim appears constantly and is usually attributed to Kevin Bales. This section tests it, because it is exactly the kind of well-travelled figure this document exists to grade.

6.1 What Bales actually wrote (PEER-REVIEWED/ACADEMIC: Bales, Disposable People)

Chapter 1 of Disposable People (University of California Press, 3rd edition 2012) states: “By 1850 an average field laborer sold for $1,000 to $1,800. This was three to six times the average yearly wage of an American worker at the time, a cost equivalent to around $40,000 to $80,000 today. Despite their high price, slaves generated, on average, profits of only about 5 percent each year.”

Two things follow. The historical figure is a range, $40,000 to $80,000, not the single $40,000 that circulates. And the figure “$90” does not appear anywhere in that chapter. A full-text search of it returns no match.

Bales' old-slavery-against-new-slavery contrast in the same chapter is entirely qualitative: high purchase cost against very low purchase cost, low profits against very high profits, slaves maintained against slaves disposable. It carries no dollar figures at all.

6.2 The methodology behind the modern figure

Bales' own working paper on the question, “Testing a Theory of Modern Slavery” (2006), does not compute a monetary global average. He substitutes ratios to oxen, land and annual wages because monetary equivalence could not be established, and he disclaims the result repeatedly in his own words, calling it “a rough and preliminary review,” “a tentative and exploratory test,” “an, admittedly rough, comparison of equivalences,” and “a target, a straw man.” His individual field observations run from $10 to a few hundred dollars, rarely over $1,000. No published table of his yields $90.

6.3 The historical half of the comparison does not have one answer either (RESEARCH INSTITUTION: Williamson and Cain, MeasuringWorth)

Samuel H. Williamson and Louis P. Cain, working from Fogel and Engerman, Kotlikoff's New Orleans slave market data, Soltow, and the Historical Statistics of the United States, put the 1850 average price of an enslaved person at $400 and the 1860 average market value at “around $800.” The present-day equivalent of that $400 depends entirely on which measure is chosen, and the three defensible answers differ by a factor of about seventeen:

The authors give the range themselves, “the value in 2020 of $400 in 1850 (the average price of a slave that year) ranges from $14,000 to $240,000,” and they state a recommendation: “the labor or income value, using the unskilled wage is the best measure of the value of a slave's services in today's prices. That $400 would be $114,000 in today's price.” Across the antebellum period they put the range at “$60,000 (in 1809) to $184,000 (in 1859),” and the 1860 real price at “$25,000 in current dollars.”

Note that the commonly quoted $40,000 matches none of the three measures.

6.4 The modern half has no government or intergovernmental source at all

Williamson and Cain, needing a present-day comparator, reach for an advocacy organisation: “There are several organizations such as Anti-Slavery International that will point out that in many places today, slaves sell for as little as (or even less than) $100!” That sentence is the whole problem in miniature. An economic-history project with a rigorously sourced historical price series has no equivalent modern series to compare it against, because none exists.

The check across bodies that might publish one:

6.5 Why the largest victim dataset in the world cannot answer this question (PRIMARY/OFFICIAL: IOM CTDC codebook)

The Counter-Trafficking Data Collaborative holds records on more than 222,000 identified victims across 197 or more countries. It has 27 variables, and not one of them is monetary. Debt appears only as meansDebtBondageEarnings, a binary flag taking the value 1 or null, and even that merges two distinct things: the codebook says the variable “merges two possible means of control: debt bondage and take earnings.”

Two further constraints on anyone tempted to mine it: the published dataset is synthetic, generated under differential privacy, so its records “do not correspond to actual individuals” and no individual case can be cited from it; and because debt bondage is merged with wage theft in the current version, even the binary cannot isolate debt bondage prevalence.

Assessment High confidence the price-collapse claim as popularly stated is not supported in the form it is usually given. Its historical half is real but has three legitimate present-day values spanning a factor of seventeen, none of them $40,000. Its modern half has no government, intergovernmental or peer-reviewed population source; the best available modern figures are Kara's US$400 global weighted average and UNODC's US$870 observed median, both of which are an order of magnitude above $90. The underlying directional claim, that acquisition cost has fallen enormously relative to the yield a victim generates, survives all of this comfortably and is separately supported by the ratio evidence in Sections 2, 5.1 and 5.3. What does not survive is the specific pair of numbers.

Assessment Moderate confidence the reason the modern half cannot be sourced is not neglect. It is that governments collect recruitment costs, because SDG indicator 10.7.1 obliges them to, and profits, because the ILO models them, and public costs, because treasuries budget against them. Nobody collects sale prices, because sale prices appear in evidence bundles rather than in statistical returns. UNODC's 2020 exercise is the exception that shows the rule, and it was not repeated.

6.6 What the peer-reviewed literature says about both authors (PEER-REVIEWED: Weitzer, Gallagher, Guth et al., Tyldum and Brunovskis)

This document uses Bales and Kara throughout, because between them they are the origin of most quantitative claims in this field. It would be a RECTUMmendation to use them without recording that both have been criticised in peer-reviewed venues, specifically on the point of where their numbers come from.

On Bales. Ronald Weitzer, in The ANNALS of the American Academy of Political and Social Science, writes that “Kevin Bales claims that 'the number of slaves in the world today is 27 million'… Bales says the figure is 'a good guess' but offers no evidence or even a rough idea of how he arrived at it,” and that Bales' Thailand estimate of thirty-five thousand sex slaves is offered “without any documentation.” Weitzer notes the consequence: the 27 million figure “appears in the most recent U.S. State Department's Trafficking in Persons (TIP) Report.”

On Kara. Weitzer addresses the profit split directly: “Kara goes further, claiming $36 billion in profits from sex trafficking alone. He also proclaims: 'Only 4.2 percent of the world's slaves are trafficked sex slaves, but they generate 39.1 percent of slaveholders' profits.' To sustain such precise assertions about the proportion who are sex slaves and the profits derived from them in comparison to other slaves would require hard evidence on both dimensions, data that do not exist.”

Laura Agustín, reviewing Kara's 2009 book for H-Net, is more specific about the mechanism, and her description matters because these are the tables underlying the ILO figure: “each table posits general assumptions that must be accepted to believe what is inferred from them. For example, Massage Parlor Economics, Kathmandu, assumes four slaves per parlor, averages ten sex acts per day, one of ten customers buys a condom, one slave is re-trafficked every six months, and 50 percent 'tip' per thirty sex acts, going on to give an average price per sale of sex… We have no idea where these figures came from.” She also identifies a conflation directly relevant to this brief's Section 4: Kara “claims that 'sex slaves' are the best earners for masters because they are sold 'literally thousands of times before they are replaced', conflating an owner's sale of a slave to another owner with a slave's sale of sexual services to customers.”

On the wider practice of recycling. Tyldum and Brunovskis, in International Migration, state the structural problem this document's correlation standard is built to catch: “numbers are given weight not based on the methods used to arrive at them… but based on the authority of the person or organization that provided the estimate.” And critically: “Even if several independent actors present similar numbers to estimate the number of trafficking victims, this should not be taken to indicate that the number is correct, as key respondents in the same field may be influenced by each other, or the same sources of information.”

On the Global Slavery Index, which this document excludes as advocacy output but which is widely treated as data: Anne Gallagher records that its prevalence structure rests on 470 affirmative responses out of 29,000 people surveyed across 25 countries, with the remaining 139 countries assigned figures by extrapolation, subject to unspecified “geopolitical adjustments” where results did not “better reflect known realities.” A separate peer-reviewed methodological review in Social Inclusion concludes that “the Index cannot be validated or replicated.”

Assessment High confidence the criticism is directed at the provenance and precision of these authors' global aggregates, not at their fieldwork observations. Their individually observed figures, which is what this document mostly uses them for in Section 5.6, are not what the critiques attack. Their global weighted averages and profit splits, which is what the ILO relies on, are. That distinction should be carried wherever either author is cited.

7Regional focus: southeastern Arizona and the Sonora corridor

Jurisdictional note. The corridor treated here runs from the Sonoran towns of Altar, Sásabe, Naco and Sonoyta into Santa Cruz, Cochise and Pima counties, with Tucson and Phoenix as the transfer points. Federal prosecutions fall to the US Attorney's Office for the District of Arizona. Everything in this section is a smuggling, plaza-fee or coerced-earnings figure. Section 7.5 records what is not here, which is the more important half.

7.1 The smuggling fee curve, 2005 to 2025 (mixed grades, see the caution below)

Caution on reading this as a series. These are not the same kind of number. The 2005 DHS figure is an unweighted administrative average across apprehension records; everything from 2013 onward is a reported price from interviews, charging documents or officials. The 2022 EMIF figure is a national survey average, not a corridor figure. The rise is real and large, but plotting these ten rows as one curve would be a methodological error, and the 2024 Nogales quotes of US$1,000 to US$1,200 sit an order of magnitude below the 2025 InSight Crime ceiling in the same corridor, which is itself evidence that “the price” is not one number even within one year and one crossing point.

The 2005 baseline is unusually well corroborated, and it is the only point on the curve that is. The DHS paper reports four independently produced estimates of the same quantity for the same year, from four different institutions using four different methods:

The paper also reports real growth in smuggling costs of 5.3 per cent a year from 2000 to 2006 in the DHS data and 8.0 per cent a year from 1993 to 2000 in the EMIF data, and it notes the divergence between administrative and survey figures directly, observing that “apprehended migrants may tend to more frequently report the price paid to smugglers to get to the border as opposed to the full smuggling price for the entire trip.”

7.2 The plaza fee: what the controlling organisation charges

InSight Crime supplies the only volume arithmetic located: on an estimate of as many as 30,000 migrants a month passing through Altar at a US$100 quota, “that is $3 million just from that quota, just in Altar, just in one month.”

Negative finding, stated as a finding. Federal charging documents in the District of Arizona describe the piso mechanism repeatedly, including that contraband and crossings must be approved or taxed and that non-payers “would be held hostage, beaten, tortured, killed,” but no located indictment or Department of Justice release states a per-migrant plaza rate in dollars. Every quantified plaza figure in the table above comes from journalism.

7.3 What the organisation pays its own people (PRIMARY/OFFICIAL: USAO District of Arizona)

This is a distinct category from the fee the migrant pays, and conflating the two is the single largest distortion available in this corridor. These are wages per head paid to operatives.

Assessment Moderate confidence the foot-guide rate of US$1,000 per head appears unchanged across District of Arizona releases from 2024 to 2026 while the retail crossing fee roughly doubled over a comparable window. If both observations hold, the margin between what the migrant pays and what the labour costs has widened substantially, and the additional revenue is accruing above the guide layer rather than to it. This is offered at Moderate confidence because the retail figures and the wage figures come from different source types and neither is a systematic series; it is a pattern worth testing, not a demonstrated finding.

7.4 Where the fee becomes a debt, and then a ransom (PRIMARY/OFFICIAL: USAO District of Arizona)

• Smuggling fees charged and prosecuted: “Torres-Centeno informed the Colombian nationals they would need to travel to Sonoyta, Sonora, Mexico and pay a fee between $1,200.00 USD to $2,000.00 USD per person to cross into the United States.” Elsewhere, migrants “agreed to pay between $2,200 and $2,500 per person.”
• The stash house as the enforcement point: “the people being smuggled were detained at the stash house until their smuggling fees were paid,” and co-conspirators “sometimes threatened the families or sponsors of the people being smuggled if the smuggling fees were late.” In that case agents recovered over US$1.4 million in cash, and ledgers documenting “the smuggling of more than 7,000 people.”
• Hostage-taking with a stated figure: a relative “was contacted by a male who said they would be held until $8,000 per person was paid in cash.” A separate 2025 case records a US$135,000 ransom demand for a Guatemalan citizen.
• Aggregate proceeds: one ring “collected over $110,000 in smuggling fees” and forfeited them; three separate Tucson and Phoenix organisations each had over US$1 million seized or traced.

This is the ICAT conversion of Section 4.2 happening in one federal district: a contractual fee, unpaid, becomes a debt, and the debt becomes a hostage demand made to a third party.

7.5 What southeastern Arizona does not have: two verified silences

No Arizona sex trafficking prosecution states a price. A sweep of Arizona Attorney General trafficking releases located none carrying a dollar figure for a victim, a debt, or per-victim proceeds. The only Arizona state figure located is an aggregate political claim from a 2022 Attorney General release that human smuggling “brought in more than $14 million a day for cartels and gangs,” which carries no stated methodology and is not used as a fact in this document. The largest Arizona commercial-sex figure in the federal record, over US$500 million earned by the Backpage principals, is advertising revenue from an enterprise conviction and is not a trafficking price; it should never be presented as one.

No sale price for a person exists in the border record located. InSight Crime documents the practice of migrants being taken and transferred between organisations in this corridor without attaching a figure, describing “rip crews targeting migrants that might assault them or kidnap and sell them to other polleros, or other polleros who may kidnap the migrants and force them to pay to travel with them.” A documented Ecuadorian case runs a migrant from a US$6,000 smuggling fee, through a handover to a criminal organisation, to a US$4,000 ransom demanded of the family, with no price paid between criminals recorded at the handover.

Assessment High confidence in this corridor the transfer of people between criminal organisations is documented but unpriced in the open record. The quantified evidence is entirely of fees paid by migrants, taxes paid to controlling organisations, wages paid to operatives, and ransoms demanded of families. This mirrors the global pattern in Sections 5 and 6 rather than departing from it.

7.6 The one quantified purchase-price evidence, and it is not at the border

The only located figures for a person actually bought in Mexico are from the interior, and they are reported by Mexican outlets rather than by any government body:

Assessment Low confidence these figures are the only quantified purchase prices located anywhere in Mexico, they are geolocated to the interior rather than the border, they rest on individual reported cases, and one of the four carries no attribution at all. They are recorded here because their absence elsewhere is itself significant, not because they support a regional average. No Mexican federal body publishes a sale price. Section 7.7 evidences that.

7.7 The Mexican official record: what is and is not published

Two Mexican primary documents were obtained and parsed in full:

• The Comisión Nacional de los Derechos Humanos, Diagnóstico sobre la Situación de la Trata de Personas en México 2021, 119 pages: contains no sale prices, no quotas and no revenue estimates. Its only currency figures are prosecutorial budgets.
• The Senate's Instituto Belisario Domínguez, “Trata de personas en México: Algunas cifras,” Mirada Legislativa 220 (August 2022), 20 pages: contains no monetary figures at all.

What the Unidad de Inteligencia Financiera publishes is laundering flow rather than prices. Its typology on trafficking and pimping records one subject who “de 2002 a 2019 realizó 206 depósitos en sus cuentas por $50,859,222,42 M.N y retiros por $111,712,707.82 M.N desconociendo el origen y destino de los recursos” (“from 2002 to 2019 made 206 deposits into his accounts totalling 50,859,222.42 pesos and withdrawals of 111,712,707.82 pesos, with the origin and destination of the funds unknown”), and sent abroad “$28, 731,140.00 USD” through a company to a jurisdiction “considerado un paraíso fiscal” (“considered a tax haven”). None of it is attributable to a per-person transaction.

The best-quantified Mexican input-side mechanism located is a recruitment loan, from an IOM Mexico forum record: “los reclutadores o los contratistas les ofrecen prestamos, 'te doy $2000 pesos', y así primero te forzó a irte y segundo, pues ya las personas llegan endeudadas. Hay un porcentaje importante del 66.6% de las personas que contestaron esta encuesta que dijeron que tuvieron que pagar el enganche a su reclutador contratista” (“recruiters or contractors offer them loans, 'I'll give you 2,000 pesos', and so first they forced you to leave and second, people arrive already in debt. A significant 66.6 per cent of those who answered this survey said they had to pay the down payment to their recruiter or contractor”). Note the word doing the analytic work: enganche, a down payment, is the vocabulary of hire purchase, and the survey records two thirds of respondents paying one.

7.8 Coerced earnings: the quota, and what it reveals about valuation

The clearest quantified picture of what an exploited person is made to produce in Mexico comes from the Tlaxcala and Puebla pimping networks. The strongest source is journalism built directly on federal judicial sentences from 2015 to 2025:

That last row is the closest thing in the Mexican record to a documented handover payment, and it should be read against the row above it. On the figures in this table, a person handed over for 1,000 pesos was made to generate up to 90,000 pesos a month. The ratio is roughly ninety to one in the first month, and it is the same shape as Bales' Thai observation in Section 5.6 and the Belgian case in Section 5.3, arrived at by an entirely separate route.

Note also the charge-back: one victim's quota of 6,500 pesos included “el cobro de mil 500 pesos por la promoción” (“the 1,500-peso charge for the promotion”), meaning she was billed for her own advertising out of the quota she was forced to meet.

8Sources rejected this session, and why

8.1 Sources whose status changed during research, in both directions

Walk Free, promoted in one specific context. Walk Free is a mission-driven organisation and its Global Slavery Index is excluded here. However, the Global Estimates of Modern Slavery (2017 and 2022) are joint publications of the ILO, Walk Free and IOM, published by the ILO in the ILO's own series, using the ILO's survey infrastructure. Those two documents are used and graded as ILO publications, with the co-authorship stated. The distinction drawn is between a document an advocacy organisation produces and one it co-signs inside a UN agency's methodology. Readers who reject that distinction should discount the debt bondage prevalence figures in Section 3.3 accordingly; nothing else in this brief depends on them.

The Urban Institute, admitted on a narrow basis. Research institutions are citable under this document's standard and named policy think tanks are not, and the Urban Institute sits close to that line. Its study of the underground commercial sex economy was federally funded by the National Institute of Justice and published as a grant report with a stated methodology. Where it is used it is cited as the NIJ-funded grant report and its funding is stated.

Anti-Slavery International's claim survived, its sourcing did not. The claim that people are sold today for very small sums, which appears in advocacy material as “$100 or less,” is broadly corroborated by primary sources this document did fetch: UNODC's observed median of US$870 and minimum of US$36, and Bales' field observations of US$12 to US$150. The advocacy figure is directionally right. It is not cited, because primary confirmation exists and the primary confirmation gives different numbers. This is exactly the case the standing standard anticipates: cite the primary, note the agreement here rather than in the citations.

A journalistic source promoted above its usual grade. The Mexican quota figures in Section 7.8 come from a Milenio investigation, which is regional reporting and would ordinarily be graded C. It is graded higher because the investigation is built directly on federal judicial sentences from the Consejo de la Judicatura Federal covering 2015 to 2025, which makes it a named investigative treatment of primary court records rather than an outlet relaying an official's statement. The grade reflects what the piece is made of, and the underlying sentences are named as the open item to fetch (Section 13).

A source rejected for one purpose and used for another. The 2018 UK Annual Report on Modern Slavery contains a Home Secretary's foreword saying men's lives in Libya “were bought for a few hundred pounds.” That is a ministerial statement about third-party footage, not a government estimate, and it is not used as a data point. The document is not otherwise rejected.

9Signals: genuine cross-source correlation

The standard applied here: 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 study is one data point, not several. On this subject that standard bites unusually hard, because a small number of original datasets underlie a very large volume of confident restatement.

Signal 1: the acquisition cost is paid by the victim, not by the trafficker (STRONG)

Claim: across the documented record, the money that moves a person into exploitation is overwhelmingly paid by that person or their family, in advance, and not by anyone acquiring them.

Assessment High confidence Assessment (Confidence: High; almost certainly true).

Four institutions, four methods, no shared data:

• The ILO, from the KNOMAD migration cost surveys, puts recruitment fees and related costs charged to international migrants in forced labour at US$5.6 billion a year and an average of 3.93 months of work to repay.
• The US Government Accountability Office, from its own worker discussion groups in a lawful US visa programme, documents fees of more than US$350 and about US$1,250 paid by Mexican H-2B workers, at interest those workers put at 10 to 15 per cent. That such fees are financed by borrowing, and sometimes secured on the worker's own land, is attributed by GAO to NGOs and federal officials rather than to the workers it spoke to; the distinction is preserved at Section 3.2.
• IOM Mexico, from a forum survey, records 66.6 per cent of respondents saying they had to pay an enganche, a down payment, to their recruiter or contractor, with recruiters offering loans expressly so that “people arrive already in debt.”
• UNICRI, from victim testimony on the Nigeria to Italy corridor, documents debts of EUR 10,000 to 30,000 constructed out of a journey the victim was already paying for.

These are a UN labour agency's statistical model, a US federal audit body's qualitative fieldwork, an intergovernmental migration agency's survey, and a UN research institute's testimony collection. They do not cite each other and they were built for different purposes. The convergence is on the direction of payment, which is the strongest available finding in this document.

What makes this strong rather than merely consistent: the four sources disagree substantially on amounts, which are context-specific and range from US$350 to EUR 30,000, while agreeing completely on who pays. Agreement on structure despite disagreement on magnitude is what independent corroboration looks like.

Signal 2: where a purchase price exists at all, it is small (MODERATE)

Claim: the price one criminal pays another for a person, where any price is recorded, is typically in the hundreds to low thousands of dollars.

Assessment Moderate confidence Assessment (Confidence: Moderate; likely).

• UNODC, from 30 court cases across 15 countries, 2007 to 2017: median US$870, mean US$3,662, 26 of 30 under US$5,000, domestic cases averaging US$250.
• Kara, from his own fieldwork: a global weighted average acquisition cost of approximately US$1,900 for a trafficked sex slave, and US$400 as a global weighted average across all forms.
• Bales, from separate fieldwork a decade earlier: about US$150 paid per person to a recruitment agent in Thailand, US$800 to US$2,000 for a girl sold to a brothel, about US$50 as the initial loan for a bonded agricultural labourer.
• Mexican reported cases: 6,000 and 20,000 pesos for children, roughly US$300 to US$1,000 at the relevant exchange rates.
• Shen, from prison interviews in China: 2,000 yuan, about US$314, paid for an infant, within a range of participant payments running from about US$47 to about US$3,925.
• Dank and colleagues, from US case knowledge: US$400 paid by one pimp to another to keep a runaway, and US$500 quoted to return a woman to a previous pimp.

Rated Moderate rather than Strong for three reasons. The sources are measuring somewhat different transactions across different sectors, continents and decades. Two of them, Kara and Bales, are individual researchers rather than institutions, and their global figures are extrapolations rather than observed transactions. And UNODC's dataset, the only institutional one, covers 30 cases and UNODC itself says the findings “are not representative of the dynamics of the trafficking in persons market worldwide.”

What keeps it from being rated lower is the sheer spread of independent origins now pointing at the same order of magnitude: a UN agency reading court files across 15 countries, two individual researchers doing field interviews two decades apart, a criminologist interviewing incarcerated offenders in a Chinese prison, US law enforcement describing local practice in Denver and Washington DC, and Mexican reporting on interior cases. Six routes, no shared data, all landing in the hundreds to low thousands of dollars. The two outliers in the whole brief that sit far above that band, the US$23,600 top of UNODC's range and the US$45,000 Kansas City sale, are both explicable: the first is the extreme tail of a skewed distribution, and the second is a debt assignment rather than a sale of a person, as Section 5.7 sets out.

What earns it Moderate rather than Weak: UNODC's observed median of US$870 and Kara's independently derived global weighted average of US$400 to US$1,900 bracket each other closely, having been produced by a UN agency reading court files and an individual researcher doing field interviews. Neither cites the other. On the contrary, UNODC's section exists partly to challenge the higher estimates then in circulation.

Signal 3: the ratio between acquisition and yield is very large, and four separate routes produce the same shape (STRONG)

Claim: what a person generates for an exploiter exceeds what it cost to acquire them by one to two orders of magnitude, and does so within months.

Assessment High confidence Assessment (Confidence: High; almost certainly true).

Five routes: a European prosecution summarised by a UN agency, two separate accounts in Bales on two continents, Mexican court sentences reported by an investigative outlet in the 2020s, and a cross-institutional comparison of two independent datasets. The correlated finding is not a value but a shape: acquisition is a small fraction of yield, and the yield recurs while the acquisition cost does not.

Be precise about the size of the effect rather than reaching for a round phrase. The ratios span from roughly 3 to 1 up to roughly 90 to 1, which is a spread of about one and a half orders of magnitude among themselves, and the honest summary is “a single-digit to two-figure multiple within the first year,” not “one to two orders of magnitude” in every case.

Note the honest weakness in the last row: it compares a modelled annual profit to an observed purchase price, which are not the same kind of number, and it is the weakest of the five. The two Bales rows and the Tlaxcala row are the strongest, because in each a single observer reports both sides of the same transaction.

Signal 4: the Arizona-Sonora smuggling fee rose sharply, but only one year of it is corroborated (NARROW)

Claim: the price of being moved across this corridor rose substantially between 2005 and 2025.

Assessment Moderate confidence Assessment (Confidence: Moderate; likely, for the direction; Low for any specific multiple).

The 2005 baseline is genuinely and unusually well corroborated. Four independently produced datasets, from DHS apprehension records, El Colegio de la Frontera Norte's EMIF survey, Princeton's Mexican Migration Project and the Mexican Migration Field Research and Training Program, put the average smuggling cost for Mexican nationals that year between US$1,182 and US$1,970. Four institutions, four sampling frames, one narrow range. That is the strongest single-year correlation anywhere in this document.

Nothing after 2005 has that. The 2022 EMIF national average of US$6,937 is a survey figure but is national rather than corridor-specific. Everything else on the curve is a reported price from an interview, a charging document or an official's statement. Rated Narrow because the endpoints are real and well-sourced while the path between them is not a series, and because prices quoted within the same corridor in the same year differ by more than an order of magnitude, from US$1,000 quoted at Nogales in 2024 to US$20,000 quoted for the Sonoran Desert in 2025.

Signal 5: tested and found absent, no corroborated border sale price (ABSENT)

Claim tested: that a price for a person changing hands between criminal organisations at the US-Mexico border can be documented.

Assessment High confidence Assessment (Confidence: High) that no such figure exists in the located open record.

This was searched for directly and specifically. The transfer of people between organisations in the corridor is well documented, including by InSight Crime describing migrants kidnapped and sold between polleros. No source located, in either country, in any register, attaches a dollar figure to any such transfer. The quantified evidence is entirely fees, taxes, wages and ransoms.

This absence is reported as a finding. It is consistent with the global picture in Sections 5 and 6, and Section 11.2 examines the competing explanations for it.

Signal 6: tested and found weak, the price-collapse pairing (WEAK)

Claim tested: that the widely repeated “$40,000 then, $90 now” comparison is corroborated.

Assessment High confidence Assessment (Confidence: High) that it is not corroborated, and Section 6 sets out why in detail. The historical half is well sourced but has three defensible modern equivalents spanning a factor of seventeen, none of which is $40,000. The modern half has no institutional source at all. The two named originators of modern-slavery price economics, Bales and Kara, disagree with each other on both halves. On the historical half Bales' book gives $40,000 to $80,000 against Kara's $12,000, a factor of three to seven. On the modern half Bales' talk gives $90 against Kara's $400, a factor of four and a half. The often-quoted “roughly thirty times” is not the gap between the two authors; it is Kara's own old-against-new ratio, $12,000 against $400.

The directional claim survives and is separately supported by Signal 3. The specific numbers do not.

Named specifically as NOT correlation

These are findings this document relies on that rest on a single source each. They are not weaker facts for it, but they must not be read as corroborated.

• The ILO's US$27,252 per-victim figure for forced commercial sexual exploitation, and the US$172.6 billion built on it. The revenue side comes entirely from Kara's Global Sex Trafficking Metrics database, and the ILO says so. The ILO figure and Kara's figure are not two sources agreeing. They are one dataset appearing twice, and the same dependency exists in the 2014 edition through Kara 2009. Any document citing “the ILO says” and “Kara says” as mutual corroboration is double-counting. This is the single most important epistemic caveat in this brief, because that dependency underlies 73 per cent of the headline US$236 billion.
• UNODC's entire price dataset. Thirty court cases, one report, 2020, never repeated in 2022 or 2024. Every price statistic in Section 5.1 is that one exercise.
• The Belgian case in Section 5.3. One case, reported by one national authority to one UN report.
• The UK Home Office's £328,720. One study, one country, 2016/17 prices, and a cost-to-society figure rather than a price.
• The Nigeria to Italy debt range of EUR 10,000 to 30,000. Reaches this document through a UNICRI footnote citing a single 2008 Direzione Distrettuale Antimafia study. The underlying Italian government document was not fetched.
• The Mexican quota figures. One journalistic compilation, albeit one built on federal court sentences.
• The GAO H-2 fee amounts. GAO's own discussion groups, which GAO states “cannot be generalized to the population of all H-2A and H-2B workers.”
• Every plaza fee figure for the Arizona corridor. All from journalism; no federal charging document located states a rate.

10Key Assumptions Check

A Key Assumptions Check lists the things this document takes as true without having verified them. They are flagged not because they are doubted but because they are unverified, which is a different thing. If any of them is wrong, specific findings above change, and each entry says which.

10.1 That dollar figures spanning 2000 to 2025 are being compared like with like. Almost no source in this brief adjusts for inflation, and the few that do adjust to different base years: the ILO to 2021 constant prices, MeasuringWorth to 2020 dollars, UNODC's price dataset explicitly “based on current prices” across a span from 2007 to 2017. A US dollar in 2007 is worth roughly 40 per cent more than one in 2025. Every comparison in this document between an older and a newer figure is therefore understating the older one. If wrong: the Arizona smuggling fee curve in Section 7.1 overstates the real rise, perhaps substantially. UNODC's mean of US$3,662 pools nominal prices across eleven years and is not a price in any single year. Signal 4 is affected; Signals 1, 2 and 3 are not, because they rest on ratios and structures rather than levels.

10.2 That the court-case samples reflect what happens outside court. UNODC's 30 cases, the 190 Dutch judgments, the 37 Taiwanese proceedings and the 122 US case records are all samples of detected and prosecuted trafficking. UNODC says explicitly that its findings “are not representative.” Prosecuted cases may systematically differ from undetected ones in exactly the dimension being measured: a case with a documentable transaction is a case that leaves an evidentiary trail. If wrong in the likely direction: observed prices are biased toward the low end, because small, sloppy, local operations are easier to detect than large disciplined ones. Signal 2 would then understate typical prices. This is the most consequential assumption in the brief.

10.3 That “trafficking” denotes the same thing across sources. The ILO measures “forced labour,” a labour-law category. UNODC measures conduct meeting the Palermo Protocol definition. US prosecutions apply the TVPA. “Modern slavery” is an umbrella term with no legal definition, which the Global Estimates say directly. Peer-reviewed criticism holds that Bales' definition “conflates slavery with other forms of exploitation.” If wrong: the per-victim figures in Section 2 and the prices in Section 5 may have different denominators, and dividing one by the other, as Signal 3's final row does, would be comparing populations. This is why that row is flagged as the weakest of the five.

10.4 That the ILO's victim counts are approximately right. Every per-victim figure in Section 2 is a total divided by a victim count. The count comes from the 2021 Global Estimates. If the denominator is wrong, every per-victim figure moves inversely and proportionally, without any error appearing in the arithmetic. If wrong: US$9,995 and US$27,252 move by the same proportion. Note that the direction is counter-intuitive: if there are more victims than estimated, the per-victim profit falls. Peer-reviewed criticism of global victim counts is extensive and is recorded in Section 6.6.

10.5 That offenders and victims report money truthfully. The Dank study's weekly intake figures, the Harlem dissertation's income figures, and the corridor fee reports all rest on self-report by people with reasons to exaggerate or minimise. The Harlem study says so: respondents “may embellish their earnings.” The Dank study built a validation rule precisely because it did not trust direct estimates, taking the lower of a direct estimate and a bottom-up reconstruction whenever they diverged by more than 25 per cent. If wrong: Section 5.7's earnings figures move, in an unknown direction. The per-act prices are less exposed, because they are corroborated across many independent respondents in different cities.

10.6 That an absent figure was never collected, rather than collected and withheld. This document repeatedly treats the absence of a price as evidence about the phenomenon. An alternative is that agencies hold such figures and do not publish them, for investigative or victim-protection reasons. Nothing located confirms or refutes this. If wrong: the “structural silence” findings in Sections 5.9.3, 6.4 and 7.5 would be findings about publication policy rather than about the criminal economy, and Section 11.2's hypothesis H3 would be favoured over H2.

10.7 That the corridor reporting accurately relays what was said. Most Arizona and Sonora figures in Section 7 come from journalists reporting what migrants, officials or smugglers told them. This document could not independently verify any individual quoted price. If wrong: Section 7.1 and 7.2 lose their quantitative value. Section 7.3 does not, because those figures come from federal charging documents.

11Analysis of Competing Hypotheses

ACH exists to guard against anchoring on the first plausible explanation and then collecting only the evidence that confirms it. Each question below is genuinely contested by the sourcing, and each is stated as a question rather than as a conclusion looking for support.

11.1 Why does the record almost never contain a price paid for a person?

This is the central puzzle of the brief. People are demonstrably bought and sold: UNODC, Europol, UNICRI, the Department of Justice, the NIJ studies and victim testimony all describe it. Almost nobody records the amount.

H1: Such transactions are genuinely rare, and the record is accurate. Evidence for: UNODC found a price in only 30 of 489 case narratives. Prosecutions describe sale in qualitative terms far more often than in quantitative ones. Most trafficking involves recruitment by deception or coercion of someone already known to the trafficker, where no purchase occurs at all: the IOM CTDC variable set records recruiter relationships as intimate partner, friend and family, and Dank's respondents said “the exchange of money is not essential to establish a business relationship.” Evidence against: the same records repeatedly describe sale, resale and trading as routine. The Nigeria to Italy testimony has victims sold repeatedly between maman. Denver officials described pimps who “are sharing girls they're selling girls back and forth to each other.”

H2: Transactions are common, but the priced instrument is the debt, not the person. Evidence for: this is the strongest-supported hypothesis and it has four independent sightings. The Kansas City case where the sale price is exactly the debt balance, US$45,000. UNICRI's Nigeria to Italy corridor, where victims are demonstrably sold between exploiters and what transfers is the obligation to repay, with no price recorded for any transfer. The NIJ labour study, where “smuggling debts were bought by employers… who then trafficked them.” And GLOTIP 2024's Libya material, where “private citizens who may pay a ransom for detainees and then force them to work in order to pay off what was paid,” which is the same structure with the ransom standing in for the purchase. On this account there is no missing price, because the transaction is booked as an assignment of a receivable and the person is the security. Evidence against: it does not explain the Texas US$5,000 sale or the Denver US$400 payment, in neither of which is a debt mentioned. It may describe one common structure rather than the general case.

H3: Transactions are common and priced, but prices do not survive into published documents. Evidence for: the Department of Justice quantifies restitution and forfeiture in nearly every sentencing release and purchase prices in about seven per cent of them, which is a pattern in what press releases are for. Prices live in indictments, sentencing memoranda and evidence bundles rather than in statistical returns. UNODC found prices only when it went and read 489 case narratives, which no agency does routinely. Evidence against: the Dutch, Taiwanese and US studies that did read court files found debts, fees and daily earnings in abundance and purchase prices almost never. If the figures were in the files, file-based research should have surfaced them.

Assessment Moderate confidence H2 and H3 are both substantially supported and are not mutually exclusive; H1 is the weakest. The evidence most likely reflects a criminal economy in which acquisition often involves no purchase at all (deception, family, intimate partner), and in which where a transfer does occur it is frequently structured as a debt assignment rather than a sale, compounded by a publication regime that surfaces court-ordered sums rather than criminal ones. The absence of a claim in any source about which of these dominates is not evidence for any of them, and this document does not resolve it.

11.2 Is trafficking as profitable as the headline figures suggest?

H1: Yes. It is a very high-margin criminal enterprise. Evidence for: the ILO's US$236 billion and US$27,252 per victim. The ratio evidence in Signal 3, where yield exceeds acquisition by roughly three to ninety times within the first year. The SDNY unit economics implying US$219,000 to US$383,000 gross per victim per year with zero paid to the victim. Backpage's US$500 million. Evidence against: the ILO figure is a model whose largest component rests on one researcher's data, as Section 2.3 establishes.

H2: No. Margins are high per victim but absolute incomes for most participants are modest. Evidence for: UNODC states it outright, that “recruiting trafficking victims may not be more profitable than an average job in the legal sphere,” and that a recruiter in Central and South-Eastern Europe “would have to recruit at least 20 women per year to reach the average annual salary in their own countries.” In its data, traffickers earned under US$5,000 per victim in 12 of 16 cases. The Harlem dissertation puts median pimp income at US$8,550, against a mean of US$71,599, and states that “the pedestrian pimp is not generating over $1 million a year with underage workers.” The Dank study's per-city cells rest on as few as one respondent. Evidence against: median-versus-mean effects mean most participants can earn little while the aggregate is still enormous, so this is not strictly inconsistent with H1.

H3: The question is unanswerable at the aggregate level. Evidence for: UNODC says “Estimating the global size of the trafficking in persons market in terms of illegal profits remains challenging given the lack of a reliable estimate of the global number of victims.” Europol says any estimate of EU trafficking profits is “extremely unreliable” and that no EU-specific evaluation exists. Peer-reviewed criticism holds that it is “impossible to satisfactorily count (or even estimate) the number of persons involved in or the magnitude of profits within an illicit, clandestine, underground economy at the macro level.”

Assessment Moderate confidence H1 and H2 are both true of different things, and H3 is correct about the aggregate. Per-victim margins are genuinely extreme, which is well corroborated by Signal 3 through four independent routes. Aggregate global profit figures are model outputs of contested reliability, which two of the agencies best placed to produce them decline to produce. The reconciliation is that this is a low-barrier, high-margin, low-absolute-income criminal economy with a small number of large operators, which is the same shape as the mean-median gaps that appear in every distribution in this brief.

11.3 Did the price of a person actually collapse, and is that why victims are treated as disposable?

H1: Yes, as commonly stated. Prices fell from tens of thousands to under a hundred dollars, and disposability followed. Evidence for: the directional claim is well supported. Acquisition cost is a trivial fraction of yield in every documented case in Signal 3. Evidence against: the specific numbers do not survive, as Section 6 demonstrates in detail. There is no modern price series. The best modern figures, Kara's US$400 global weighted average and UNODC's US$870 observed median, are five to ten times the quoted “$90.”

H2: The comparison is category error. Historical prices were purchase prices in a lawful property market with clear title; modern figures are mostly debts, fees and one-off payments in a market with no title at all. Evidence for: the antebellum series exists precisely because slaves were legally alienable property, recorded in sale documents. The modern equivalents are, per Section 4, five other kinds of transaction. Buying a person conferred a defensible legal asset; paying US$400 for a runaway confers nothing enforceable, so the two sums are not measuring the same right. Kara's own comparison, US$12,000 old-world against US$400 today, uses “global weighted average” for both halves without establishing that the underlying transactions are comparable. Evidence against: the disposability consequence follows either way. Whether or not the two prices are commensurable, the practical result, that a person can be replaced more cheaply than they can be maintained, is documented directly.

H3: The change is in the ratio, not the price, and the ratio is what drives disposability. Evidence for: Bales' own book supplies both halves of this: 1850s slaves cost the equivalent of “three to six times the average yearly wage of an American worker” and “generated, on average, profits of only about 5 percent each year,” against modern ratios of 3 to 1, 31 to 1, 67 to 1 and 90 to 1 in Signal 3. That is a change in return on acquisition of two to three orders of magnitude, and it is expressible without either half of the disputed dollar pair.

Assessment Moderate confidence H3 is the defensible form of the claim, and H2 correctly identifies why H1 as usually stated cannot be sustained. The finding that survives every objection in this document is that the return on acquiring a person has risen by orders of magnitude, because acquisition cost collapsed toward zero while yield did not. Stating it as a ratio requires no inflation adjustment, no commensurability between legal and illegal property markets, and no reliance on a modern price series that does not exist. Stating it as “$40,000 then, $90 now” requires all three.

12Sources Cited

Numbering runs continuously. Each entry carries an Admiralty grade. A dagger (†) marks a source reached through this session's research passes and quoted from their extraction, but not independently re-fetched by the compiler of this document. Undaggered entries were fetched, extracted and read directly during compilation. The distinction is recorded because this document's own standard forbids grading a source you did not fetch, and roughly half of what follows would fail that test if it were not marked.

12.1 Primary and official: international bodies

12.2 Primary and official: United States federal

12.3 Primary and official: Mexico

12.4 Federally funded research reports

12.5 Peer-reviewed

12.6 Individual researchers, cited as such

12.7 Research institutions

12.8 Named investigative journalism

12.9 Regional and local reporting

12.10 Out-of-scope material, used only where noted

12.11 Located and excluded

Sources located during research and excluded from load-bearing use are listed with reasons in Section 8 and are not renumbered here.

13Open items for a follow-up research pass

RESOLVED during this pass

• RESOLVED: whether any intergovernmental body has ever published prices paid for people. Yes, once. UNODC GLOTIP 2020, 30 court cases, 15 countries, 2007 to 2017. Landed at Section 5.1. The initial sweep of the 2024 edition suggested the answer was no, which would have been a significant error; the 2020 edition was found only on a second pass.
• RESOLVED: what the ILO's per-victim sexual exploitation figure rests on. Kara's Global Sex Trafficking Metrics database, stated by the ILO in both the 2014 and 2024 editions, with the 2014 methodology text obtained verbatim. Landed at Section 2.3.
• RESOLVED: whether the “$40,000 against $90” pairing is Bales' published position. It is not. Verified directly against chapter 1 of Disposable People. Landed at Section 6.1.
• RESOLVED: the apparent collapse in debt bondage prevalence between the 2017 and 2022 Global Estimates. Definitional, not real. Landed at Section 3.3.
• RESOLVED: Shen (2016) on Chinese internal child trafficking, flagged by a research pass as the most likely peer-reviewed source of a per-child sale price and reported unreachable after HTTP 403 on two tools. Recovered from an Internet Archive capture, fetched and read directly. Landed at Section 5.10 and as a sixth route in Section 9, Signal 2.
• RESOLVED, partially: the ILO 2024 regional per-victim figures. The text-layer extraction ordering was verified against the rendered figure image. What remains unresolved is that the ILO publishes no table for this chart, so the values are only readable from the bar labels.

Still open

1. Obtain Kara's Global Sex Trafficking Metrics database (2017), appendix to Modern Slavery. This is the highest-value open item in the brief: it underlies 73 per cent of the ILO's headline figure and nobody outside the ILO appears to have examined it. Next step: a library copy of the Columbia University Press edition, since it exists only in print. A second route is Kara's Northwestern Journal of International Human Rights 9, no. 2 (2011) article, attempted twice this pass (Cloudflare interstitial, then HTTP 403); try an institutional proxy or the SSRN copy.
2. Fetch the Direzione Distrettuale Antimafia of Naples (2008) study underlying the EUR 10,000 to 30,000 Nigeria to Italy debt range. It currently reaches this document through a UNICRI footnote. A candidate URL is appinter.csm.it
3. Pull the Consejo de la Judicatura Federal sentences (2015 to 2025) behind the Mexican quota figures in Section 7.8, so those figures rest on the judgments rather than on an investigative summary of them.
4. Obtain District of Arizona indictments and sentencing memoranda from PACER. Press releases are demonstrably the wrong instrument for this question: they quantify what a court orders, not what a trafficker charged. This is the single most likely route to a documented sale price in the corridor, and it was not attempted this pass.
5. Search Pima County and Maricopa County Attorney records. Only shallowly searched before the session's web-search budget was exhausted at 200 of 200 calls.
6. Marcus et al. (2014), The ANNALS 653, and Marcus et al. (2016), Sociological Perspectives 59. Attempted three times across two passes, HTTP 403 each time, no open-access copy and no repository copy per Unpaywall. The 2016 paper carries the Atlantic City earnings and split data, which would be a genuinely independent check on Section 5.7's US splits.
7. The FTC has not been searched at all for job-placement and recruitment-fee fraud enforcement with stated amounts. The budget ran out first.
8. Whether the ILO will publish a third edition of Profits and Poverty using something other than Kara's data. No indication either way was located.
9. Whether GLOTIP will repeat the 2020 price exercise. It was not repeated in 2022 or 2024. It may simply not exist again in published form, and if so, the 2020 dataset is the only one the world has.
10. The Home Office's £328,720 per-victim cost figure was not independently re-fetched by this document's compiler and is marked accordingly at source. It is used only as a contrast case in Section 4, so nothing load-bearing depends on it, but it should be verified before reuse.

A note on the search budget

This session exhausted its web-search allocation at 200 of 200 calls partway through the research passes. Later discovery ran through direct URL fetches and bibliographic APIs (Crossref, OpenAlex, Unpaywall, Semantic Scholar, Europe PMC) rather than search. Items 5, 6 and 7 above are all consequences of that limit rather than of the material not existing.

14Note on the eventual World-Building Document

This document is reference material and nothing more. It names no character, faction, organisation or location from Vampires of Tucson or any other project, and it draws no line from any real event, price, person or organisation described here to any fictional one.

It does not represent that any fictional organisation operates the way any real criminal network described in these sources operates, and it should not be read as establishing that it could. The real networks documented here are real. The people whose prices, debts and quotas appear in these pages were real people, and several of the figures in this brief are the sums for which actual children were actually sold. That is the reason the sourcing standard was applied as strictly as it was, and it is the reason the firewall exists.

Where the eventual World-Building Document draws on this material, that will be a separate, later and clearly fictional-license step. It should be treated as such rather than presented as continuous with the evidentiary standard used here. Nothing in a work of fiction inherits the grading in Section 12, and no reader of that fiction should be given cause to think it does.

The document argues no policy position, in any direction, on immigration, border enforcement, sex work, labour migration, criminal sentencing, or any other contested question its sources touch. Where its sources argue such positions, that has been noted and excluded from load-bearing use rather than adopted.