Measuring labour, and in global supply chains: A global Input-Output approach Technical Paper

Ali ALSAMAWI Tihana BULE www.Alliance87.org Claudia CAPPA Alliance8_7 Harry COOK Claire GALEZ-DAVIS #Achieve87 Gady SAIOVICI Copyright © 2019 International Labour Organization, Organisation for Economic Co-operation and Development, ­International Organization for Migration, and United Nations Children’s Fund

This is an open access distributed under the Creative Commons Attribution-Non Commercial 3.0 IGO License (https://creativecommons.org/licenses/by-nc/3.0/igo/). Users can reuse, share, adapt and build upon the original work non-commercially, as detailed in the License. The ILO, the OECD, IOM and UNICEF must be clearly credited as the owners of the original work. The use of the emblems of the ILO, the OECD, IOM and UNICEF are not permitted in connection with users’ work.

Attribution – The work must be cited as follows: Alsamawi et al (2019), Measuring , forced labour and human trafficking in global supply chains: A global Input-Output approach. International Labour Organization, Organisation for Economic Co-operation and Development, International Organization for Migration and United Nations Children’s Fund. Translations – In case of a translation of this work, the following disclaimer must be added, in the language of the translation, along with the attribution: This translation was not created by the ILO, the OECD, IOM or UNICEF and should not be considered an official translation of those Organizations. The ILO, the OECD, IOM and UNICEF are not responsible for the content or accuracy of this translation. Adaptations – In case of an adaptation of this work, the following disclaimer must be added along with the attribution: This is an adaptation of an original work by the ILO, the OECD, IOM and UNICEF. Responsibility for the views and opinions expressed in the adaptation rests solely with the author or authors of the adaptation and are not endorsed by the ILO, the OECD, IOM or UNICEF.

ILO ISBN: 978-92-2031-429-6 (Print); 978-92-2031-430-2 (web PDF) IOM ISBN: 978-92-9068-814-3 (Print); 978-92-9068-815-0 (web PDF) This document is published under the responsibility of the Secretary-General of the OECD. The opinions expressed and arguments employed herein do not necessarily reflect the official views of OECD member countries. This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.

The designations employed in this publication, which are in conformity with United Nations practice, and the presentation of material therein do not imply the expression of any opinion whatsoever on the part of the ILO, IOM or UNICEF concerning the legal status of any country, area or territory or of its authorities, or concerning the delimitation of its frontiers. The responsibility for opinions expressed and arguments employed herein do not necessarily reflect the official views of ILO, IOM or UNICEF member countries, and publication does not constitute an endorsement by their respective Board of Governors or the governments they represent. Reference to names of firms and commercial products and processes does not imply their endorsement by the ILO, the OECD, IOM or UNICEF, and any failure to mention a particular firm, commercial product or process is not a sign of disapproval. Neither the ILO, the OECD, IOM nor UNICEF guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use.

Funding for this report is partly provided to the ILO by the Department of Labor under cooperative agreement number IL‐30147‐16‐75‐K‐11 (MAP16 Project) (GLO/18/29/USA). Forty per cent of the total costs of this report is funded by the MAP16 project and financed with Federal funds, for a total of USD 61,000. One hundred per cent of the total costs of the MAP16 project is financed with Federal funds, for a total of USD 22,400,000. This material does not necessarily reflect the views or policies of the United States Department of Labor nor does mention of trade names, commercial products, or organizations imply endorsement by the United States Government. III

TABLE OF CONTENTS

FOREWORD...... V INTRODUCTION ...... 1 RATIONALE BEHIND USING THE GLOBAL INPUT-OUTPUT APPROACH ...... 2

PART 1: DEFINITIONS AND DATA SOURCES...... 4

1.1 OECD Inter-Country Input-Output Tables...... 4 1.2 Child labour...... 8 1.3 Trafficking for forced labour...... 10

PART 2: COMBINING THE DATASETS...... 12

2.1 From data sources to estimates...... 12 2.2 Assumptions and limitations...... 15 2.3 Additional robustness tests...... 17

PART 3: RESULTS...... 19

3.1 Child labour ...... 19 3.2 Trafficking for forced labour...... 22

CONCLUSIONS...... 24

BIBLIOGRAPHY...... 25

ANNEXES...... 27

Annex 1. OECD – ICIO 2018 geographical coverage and industry list...... 27

Annex 2. List of regions, country coverage, and target population coverage...... 29 A. Regional groupings, following the United Nations Statistical Division’s Standard Country or Area Codes for Statistical Use (M49)...... 29 B. List of countries used as underlying data for regional estimates...... 32 C. Sources of survey data for child labour ...... 33 D. Coverage in terms of target population...... 34 IV

TABLE OF FIGURES

Figure 1: Symmetric Input-Output table framework...... 4 Figure 2: Inter-Country Input-Output Framework...... 6 Figure 3: Schematic representation of the concepts of direct and indirect...... 7 Figure 4: From data sources to model estimates...... 13 Figure 5: Estimates of child labour for exported goods and services and domestic demand, by region, by different industry aggregates (2015)...... 16 Figure 6: Comparison of CL linked for exports between ICIO, more detailed national IO, and regional national IO datasets – (2011)...... 18 Figure 7: Estimates of child labour and value added for exported goods and services, and domestic demand, by region (2015)...... 20 Figure 8: Estimates of child labour for exported goods and services, direct and indirect, by region (2015)...... 21 Figure 9: Estimates of trafficking for forced labour and value added for exported goods and services, by region (2015)...... 22 Figure 10: Estimated trafficking for forced labour and value added for exported goods and services, direct and indirect, by region (2015)...... 23 V

FOREWORD

This technical paper accompanies the report on Ending child labour, forced labour and human trafficking in global supply chains jointly prepared by the ILO, OECD, IOM and UNICEF as a product of the Alliance 8.7 Action Group on Supply Chains (hereafter the Alliance 8.7 Report). The Alliance 8.7 Report responds to the Ministerial Declaration of the July 2017 meeting of the Group of Twenty (G20) Labour and Ministers, asking “the International Organisations in cooperation with the Alliance 8.7 for a joint report containing proposals on how to accelerate action to eliminate the worst forms of child labour, forced labour and modern in global supply chains including identifying high risk sectors, and how to support capacity building in the countries most affected”. It also responds to the Declaration on Child Labour, Forced Labour and Employment, November 2017, which called for “research on child labour and forced labour and their root causes (…) pay[ing] particular attention to supply chains”.

According to the 2016 ILO global estimates, there are a total of 152 million children in child labour and 25 million children and in forced labour in the world today. Governments, business, the financial sector and civil society must take strong action to address the root causes and determinants of these violations. The Alliance 8.7 Report and this technical working paper are a contribution to these efforts.

The authors would like to thank Colin Webb (OECD), Antoine Bonnet and Manpreet Singh (ILO), and Eileen Capilit (Independent Consultant) for their contributions and data support.

VI Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach 1 INTRODUCTION

This technical paper explains in detail the methodology and datasets used to produce the results published in Chapter 1 of the Alliance 8.7 Report on Ending Child Labour, Forced Labour and Human Trafficking in Global Supply Chains on how child labour, forced labour and human trafficking are linked with global supply chains. It is a result of a collaboration between OECD, ILO, IOM and UNICEF. This technical paper is also the first output of joint research and collaboration between the OECD Directorate for Financial and Enterprise Affairs (DAF) and the OECD Directorate for Science, Technology and Innovation (STI) on impacts of responsible business conduct (RBC) in global value chains, which falls under the DAF work stream on RBC and the STI work stream on the application the 2018 Inter-Country Input-Output (ICIO) infrastructure beyond Trade in Value Added database.

The social and environmental impact of firm participation in global supply chains is an area of increasing policy interest. On the one hand, global supply chains (GSCs) have the potential to generate growth, employment, skill development and technological transfer. On the other hand, decent work deficits (including child labour, forced labour and human trafficking) have been linked to economic activity supported by GSCs. The complexity and interconnectedness in the global markets presents a challenge for conventional statistics and accounting methods. For example, tracing back the origins of a final product or even its components requires capturing statistics not only in the market where the product is “consumed”, but also along its . Many times, such statistics – if they exist at all – are full of gaps.

While developing a consistent quantitative means of tracing social and environmental impacts in GSCs is very challenging, certain aspects can and have been measured. For example, analysis of CO2 emissions in the context of GSCs has already been integrated in the OECD Trade in Value-Added (TiVA) database.1 Similar methodology has also been applied to understand the role of skills in countries’ comparative advantage and industry performance in global value chains (see Grundke et al., 2017). When it comes to labour, ILO and the OECD have estimated labour content in trade (OECD, 2019a; Kizu et al, 2016) and share of jobs associated with global production. The basis for all this research are the Input-Output (IO) tables.

1 For example, CO2 embodied in foreign demand. See: http://oe.cd/io-co2.

2 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

RATIONALE BEHIND USING THE GLOBAL INPUT-OUTPUT APPROACH

The basis for the research in this technical paper are the OECD Inter-Country Input Output (ICIO) tables, the datasets from ILO, UNICEF and IOM on employment, child labour and human trafficking, and the results from the 2016 Global Estimates on Modern Slavery. IO tables are commonly used by national statistical offices to describe the relationship between producers and consumers within an at an industry level. They account for final and intermediate goods and services, allowing statisticians to identify and isolate the direct and indirect impact of, for instance, a specific industry into the whole economy. Several initiatives at the international level, including the OECD ICIO tables,2 have aimed to expand these tables to also analyse interdependencies between countries. These expanded datasets have provided researchers with tools to analyse several aspects of international trade and its impacts. Of particular relevance to this technical paper are attempts in literature to use global IO models to estimate the impact of international trade on social indicators, including country- specific analysis on child labour (see for example Alsamawi et al., 2017; Gomez-Paredes et al., 2016). This paper is a contribution to those and the aforementioned efforts on environmental and economic analyses.

Child labour standards, definitions and tools related to collection of child labour data are nowadays well-established statistical areas. In 1998, the Statistical Information and Monitoring Programme on Child labour (SIMPOC) was created at the ILO to improve gathering of statistical data on child labour. In 2008, the International Conference of Labour Statisticians (ICLS) adopted a resolution formalising the international criteria to measure child labour (ILO, 2008), in line with the ILO Conventions.3 Many countries include child labour data collection in their national statistical systems’ activities, for example through expanded age groups in labour force surveys (covering not only adults, but also children aged 5 years and older) or by including child labour modules in generic household or labour force surveys; others conduct ad-hoc national child labour surveys, many of which have been funded through technical co-operation projects. Every 4 years, the ILO publishes global estimates of child labour. The most recent estimates indicate that 152 million children were in this situation in 2016. Data on child labour are also collected at regular intervals as part of the Multiple Indicator Cluster Survey (MICS) programme, which was developed by UNICEF in mid-nineties to assist countries in collecting data on a numbers of wellbeing indicators for children and their . Data on child labour have been collected in MICS since 2000 in close to 150 surveys through a standard module questionnaire. The MICS module covers children 5 to 17 years old and includes questions on the type of work a child does and the number of hours he or she is engaged in it. Every year, UNICEF publishes global, regional and country-level estimates of child labour in its flagship publication, The State of the World’s Children.

When it comes to forced labour, measurement of forced labour is a relatively new statistical area. After about 15 years of pilot studies to measure the prevalence and characteristics of forced labour in different contexts or using different techniques, in 2018, the ICLS endorsed Guidelines to measure forced labour (ILO, 2018a). They reflect a framework to statistically measure the legal concept of ILO Convention 29. Forced labour is a rare statistical event, which often requires special (over)sampling techniques to reach a representative sample size to obtain robust statistical figures. Every 4 years the ILO publishes global estimates of forced labour. In 2017, jointly with the Walk Free Foundation and

2 More information about the OECD-ICIO, the full dataset, and methodology notes is available at http://oe.cd/icio. 3 The Resolution concerning measurement of child labour has been amended in 2018 to take into account the new statistical definitions on work and employment.

RATIONALE BEHIND USING THE GLOBAL IO APPROACH 3

IOM, estimates of modern slavery (including forced ) were published – this extended concept of modern slavery is out of the scope of this paper. According to the Global Estimates of Modern Slavery, on any given day in 2016, 25 million people were in forced labour in the world (ILO and Walk Free Foundation, 2017). Similarly, the measurement of human trafficking for forced labour is an area of ongoing efforts. In particular, the ILO, UNODC and IOM are working together on the development of joint survey tools to study and estimate the prevalence of trafficking for forced labour at both national and sectoral levels. This will lead to better statistical data allowing for deeper analysis of forced labour and trafficking in global supply chains.

There are currently no global or regional estimates of the prevalence of human trafficking. Relatively few examples of estimates related to human trafficking exist (IOM, 2018). Having said that, the 2017 Global Estimates of Modern Slavery estimated that out of the 40 million people that were victims of modern slavery in 2016, approximately 25 million people were in forced labour, including 16 million in forced labour in the private economy. Some national, but still experimental estimations, exist. For example, Multiple Systems Estimation (MSE) can be used to estimate the total number of (unidentified and identified) victims of trafficking at country level. MSE is a method based on the analysis of multiple lists of human trafficking cases provided by different actors in the counter- trafficking field, such as NGOs, law enforcement, international organizations and other authorities. Currently, this method cannot be applied globally. However, researchers are developing the method estimate that could potentially be used to cover approximately 50 countries around the world. Initial estimates are already available in several countries, including the UK and the Netherlands (UNODC, 2016). Finally, some sectoral estimates of human trafficking also exist using Respondent-Driven Sampling (RDS).4 While these are still limited, there is increasing evidence that using RDS in multiple waves could bring more insight and come closer to more representative-like study results.

Relating child labour, forced labour and human trafficking figures to global supply chains is a separate and significant challenge. A growing number of mixed methods (using both qualitative and quantitative approaches) and sectoral surveys are providing valuable insights into how these phenomena are involved in the global economy. Some businesses are also contributing to such insights, as they map the labour violation risks they are exposed to in the context of their human rights or social impact assessments and efforts. Nevertheless, the scope of these efforts has mostly been restricted to identifying child labour, forced labour or human trafficking in the production of goods and services of particular industries or in their main suppliers. These methods may miss collecting information on workers that are not in the immediate supply chain - for example, upstream suppliers of intermediate goods. Additionally, due to the complexity of global production networks, quantitative accounting of these relationships is not straightforward.

This technical paper describes the joint effort undertaken by the OECD, ILO, IOM and UNICEF to contribute to the body of work that can help fill this gap. It is the first application of the IO approach by international organisations as related to implementation of responsible business conduct principles and standards and decent work deficits. This is the first time that datasets from the OECD, ILO, IOM and UNICEF have been combined in this way, and the first time that this methodology has been applied in such a wide range of countries by international organisations. The paper is structured as follows: PART 1 describes the definitions and data sources used;PART 2 describes how the datasets were combined and the additional tests that were undertaken to probe the impacts assumptions and limitations; and PART 3 gives an overview of the results.

4 See for instance illegal gold mining in (Verité, 2016) and fishing in South-Eastern Asia (IOM, 2016).

4 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach PART 1: DEFINITIONS AND DATA SOURCES

1.1 OECD INTER-COUNTRY INPUT-OUTPUT TABLES

Wassily Leontief developed the first Input-Output (IO) table for the US economy in the 1930s and investigated the extension of the IO work to interregional analysis in 1953. He won the Nobel Prize for Economics in 1973 for his work on IOs. IO techniques have been used in literature in various applications since then, particularly in recent years (e.g. water, see Feng et al. 2011; materials, see Wiedmann et al. 2013; CO2, see Hertwich and Peters 2009; net primary production, see Haberl et al, 2007; employment, see Alsamawi, Murray and Lenzen, 2014).

The most significant feature of the IO model lies in the fact that it allows statisticians to identify and isolate the direct and indirect impact of a specific industry into the whole economy. IO tables in general consist of three main parts: 1. intermediate demand (Z) which describes, in monetary terms, the intermediate flows of goods and services within an economy;5 2. final demand (F) which shows the purchases of final goods and services by households and government as well as information about gross fixed capital formation, changes in inventories, and exports; and6 3. value added (V) which includes information about value-added by each industry and its components, e.g. compensation of employees, gross operating surplus, and other taxes on production (See FIGURE 1).

FIGURE 1: SYMMETRIC INPUT-OUTPUT TABLE FRAMEWORK

INTERMEDIATE DEMAND FINAL EXPENDITURE Final Direct consumption Exports purchases OUTPUT Industry 1 … Industry 36 and capital cross-border by non-­ formation residents Industry 1 … Intermediate matrix Final demand matrix Industry 36 Taxes less subsidies on intermediate and final products Total intermediate / final expenditure Value-added of which, compensation of employees of which, other net taxes on Value added matrix production of which, gross operation surplus Output (total production)

5 See: https://stats.oecd.org/glossary/detail.asp?ID=1431. 6 See: https://stats.oecd.org/glossary/detail.asp?ID=5526.

DEFINITIONS AND DATA SOURCES 5

However, national IO tables give an incomplete picture of the global economy. Questions related to who is boosting exports or who is the final consumer of domestic production are not possible to answer by looking only at the IO tables, which are country specific.

The OECD-ICIO tables attempt at resolving this issue. In simple terms, the OECD-ICIO tables can be considered “a global IO table”. They harmonize a number of national IO tables, reinforced and complemented with additional data sources7. They are built based on statistics compiled according to the 2008 System of National Accounts (SNA 2008) from national, regional and international sources and use an industry list based on the International Standard Industrial Classification (ISIC) Revision 4.

The latest edition of the tables (2018) is used for the purposes of this report. The published tables provide detailed data for 64 ,8 including all OECD, EU28 and G20 countries, most East and South-east Asian economies and a selection of South American countries. The edition also includes a category called Rest of the World. ANNEX 1 indicates the full list. Additionally, the authors have used additional unpublished data that cover a total of 198 economies in order to expand the analysis in this technical paper and the Alliance 8.7 Report. Thirty-six unique9 and harmonised industrial sectors are represented within the hierarchy, including aggregates for total manufactures and total services, as shown in ANNEX 1. The 2018 edition covers the period 2005 to 2015, with preliminary projections to 2016 for some indicators. For this paper, ICIO tables for the year 2015 have been used.

Other examples of currently available ICIO databases include: EORA, EXIOBASE, IDE-JETRO, and WIOD. Each of these tables has a different time series, industry details and country coverage. The choice of OECD-ICIO tables for this analysis is based on their acceptance and validation by OECD member countries, expectation to maintain the datasets and methodologies in the long-term, and strong institutional links with national statistics offices; and finally the availability of cross-border exports and direct purchases by non-residents shown separately.

The ICIO structure is similar to that of the national IO tables. The Z matrix or intermediate demand matrix of an ICIO (N x N dimensions, where N is the number of industries/products) hold the monetary flows of intermediate goods and services with an element Zi,j,c from supplying sector i, i =1… N, into a using sector j, j=1… N, and for country c. The F matrix (N x M dimensions, where M is the number of final demand categories) holds information on private (Households and Non-Profit Institutions Serving Households, NPISH) and public consumption (government consumption), gross fixed capital formation (GFCF), changes in inventory, and exports of final goods and services, for domestic and foreign final demands with an element Fi,c,l,k from supplying sector i, i=1… N in country c into final demand categories k, k =1… M and a foreign final demand countryl , where c ≠ l. Finally, the V matrix or primary input matrix (S x N dimensions, S being the number of primary input categories) holds information on value added and output (total production) with elements Vj,c and Xj,c respectively (see FIGURE 2).

7 Complementary data sources include international bilateral trade statistics, tourism satellite accounts and other national accounts constraints. 8 See: www.oecd.org/industry/ind/tiva-2018-countries-regions.pdf. 9 See: http://stats.oecd.org/wbos/fileview2.aspx?IDFile=c0787cf5-ec31-4130-8ddf-8667773e66ed.

6 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

FIGURE 2: INTER-COUNTRY INPUT-OUTPUT FRAMEWORK

FINAL CONSUMPTION DIRECT PURCHASES INTERMEDIATE DEMAND AND CAPITAL BY NON-RESIDENTS FORMATION OUTPUT Cou A Cou B Cou C Ind Ind Ind Ind Ind Ind Cou A Cou B Cou C Cou A Cou B Cou C 1 2 1 2 1 2 Ind 1 Cou A Ind 2 Ind 1 Cou B Ind 2 Ind 1 Cou C Ind 2 Taxes less subsidies .. ... on intermediate products ... on final products Value-added Output

Key: Cross-border flows ofintermediate goods and services

Domestic flows of intermediate goods and services

Cross-border flows of final goods and services

Domestic flows of final goods and services

Note: Taxes less subsidies on products are not part of the value added. Value added covers labour compensations, capital and other taxes less subsides on productions. White areas in the direct purchases block refer to zero elements.

In other words, the ICIO tables describe flows of intermediate and final goods and services among countries in monetary terms, hence allowing inter-industry and inter-country transactions to be recorded and analysed. This global interconnectedness captured by the ICIO tables means that the downstream use of an industry’s output by other industries, be they domestic or foreign, can be identified. Equally, ICIO tables can identify the inputs required for a particular industry from home or abroad. In other words, the ICIO tables allow for estimating how much input is required by each industry per unit of total output which is consumed either domestically or exported. For example, an increase in food processing supply may lead to an increase in demand for agricultural products, which in turn requires inputs from other upstream industries (e.g. electricity, fuel and chemical products). Through ICIO, all the total requirements needed to produce a product (both direct and indirect) can be determined. A schematic representation of the concepts of direct and indirect impacts is shown in FIGURE 3.

Direct impact captures the contribution of an industry in a specific country related to the production of goods and services for exports whereas indirect impact represents the contribution of other upstream industries that are incorporated in the production of goods and services for exports. ICIO also captures final and intermediate products. Final products are exported from country A to B to be finally consumed in country B (without additional transformation), whileintermediate products are exported from country A to country B, where they are either transformed for final consumption or exported to country C.

DEFINITIONS AND DATA SOURCES 7

FIGURE 3: SCHEMATIC REPRESENTATION OF THE CONCEPTS OF DIRECT AND INDIRECT

Apple Apples used Apple juice exported Children who picked DIRECT picking to produce juice to nal consumers the apples are "directly" FINAL [NO child labour] [child labour] – nal export linked in the nal export

Apple Apples used Apple juice exported Children who picked INDIRECT picking to produce juice to nal consumers the apples are "indirectly" FINAL [child labour] [NO child labour] – nal export linked in the nal export

Iron ore Iron ore used Steel exported Children working are DIRECT extracted for steel production to car manufacturers "directly" linked in INTERMEDIATE [NO child labour] [child labour] – intermediate export the intermediate export

Iron ore Iron ore used Steel exported Children working are INDIRECT extracted for steel production to car manufacturers "indirectly" linked in INTERMEDIATE [child labour] [NO child labour] – intermediate export the intermediate export

Within a country Outside a country

Note: Total exports equal intermediate plus final exports.

8 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

1.2 CHILD LABOUR

Child labour is any work that deprives children of their childhood, their potential and dignity, and that is harmful to physical and mental development. It is defined by the ILO Minimum Age Convention, 1973area). (No.14 This 138), measurement and the Worst framework Forms of isChild structured Labour inConvention, two main 1999areas (No.focusing 182), onand: (i)by thethe Unitedage of Nations the child; Convention and (ii) on the the activities Rights of of the the child. child This (their paper nature, uses theconditions statistical and definitions duration). for measurementActivities include of child economic labour as activities defined by(i.e. the paid International or unpaid Conference work for someoneof Labour whoStatisticians is not ain 2008member (the ofstandards the household, body in this work area). for10 Thisa measurement farm or framework business) is as structured well as inchildren’s two main areasengagements focusing on:in own(i) the-use age production of the child; of and services (ii) the (i.e.activities household of the childchores). (their It nature,is worth conditions noting andthat duration). the child Activities labour includedefinition economic used activitiesin this paper (i.e. paid only or unpaidrelates workto children for someone in economic who is not 15 aproductive member of activitiesthe household, covered work by for the a familySystem farm of orNational business) Accounts as well as(SNA). children’s In engagements particular, inthe own-use operational production definitions of services used (i.e. in thishousehold paper chores).use the It international is worth noting harmonization that the child of labour the definitionindicator ofused the in 2016this paper ILO onlyGlobal relates Estimates to children of inChild economic Labour. productive The following activities childrencovered by are the Systemconsidered of National to be asAccounts in child (SNA). labour:11 In particular, the operational definitions used in this paper use the international• Children harmonization aged 5-11 engaged of the indicator in economic of the 2017activity ILO for Global at least Estimates 1 hour of in Child the reference labour. week; The following children are considered to be as in child labour: • Children aged 12-14 engaged in economic activity for at least 14 hours in the • Childrenreference aged 5-11 week engaged; in economic activity for at least 1 hour in the reference week; • Children• Children aged 12-14 aged engaged 15-17 engagedin economic in activityeconomic for atactivity least 14 for hours at inleast the reference43 hours week;in the • Childrenreference aged 15-17 week engaged; in economic activity for at least 43 hours in the reference week; • Children aged 5-17 engaged in hazardous occupations and branches of economic activities. • Children aged 5-17 engaged in hazardous occupations and branches of economic activities. This20. paperThis uses paper datasets uses on datasets child labour on child from labour 65 nationally from 65 representative nationally representativesurveys (see ANNEX surveys 2). These(see Aincludennex 2).ILO-supported These include Labour ILO Force-supported Surveys Labouror Child Forcelabour SurveysSurveys, orUNICEF-supported Child Labour MultipleSurveys, Indicator UNICEF Cluster-supported Surveys Multiple (MICS)12 andIndicator USAID-supported Cluster Surveys Demographic (MICS) and16 Healthand USAID Surveys- (DHS).supported The datasetsDemographic from these and surveys Health contain Surveys information (DHS). onT heabout datasets 50% of from children these estimated surveys to becontain in child information labour globally. on about 50% of children estimated to be in child labour globally. 21. Within this set, 30 countries released their surveys with sectoral information related Withinto child this labour set, 30 ( countries) at ISIC released rev 3/3.1/4 their standardsurveys with codes sectoral. The informationchild labour related figures to childwere labour then harmoni at ISICsed rev to 3/3.1/4 match standardthe OECD codes.-ICIO’s The classificationchild labour figuress and rescaledwere then to harmonised the year 2015to match using the OECD-ICIO’sthe following classifications formula𝜔𝜔 : and rescaled to the year 2015 using the following formula: (ω) ω_(j,c,2015)= ω_(j,c (year) ) (〖UN〗_(pop,c,2015)/〖UN〗_(pop,c,(year))) where is the absolute𝜔𝜔!,!,!"#$ number= 𝜔𝜔!,! of, !"#$ children(𝑈𝑈𝑈𝑈!"! in ,!child,!"#$ /labour𝑈𝑈𝑈𝑈!"! in,!, (industry!"#$)) , country , and where is the absolute number of children in child labour in industry j, country c, and year of year of j,csurvey ; is the total population of children aged between 5 and 17 survey (year)!,! ; UN is the total population of children aged between 5 and 17 when expanding when expanding𝜔𝜔 thepop,c survey weights in country . 𝑗𝑗 𝑐𝑐 the surveyω weights 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦in country𝑈𝑈𝑈𝑈 c!"!. ,! 𝑐𝑐 14 More information can be found here: https://www.ilo.org/wcmsp5/groups/public/---dgreports/--- stat/documents/normativeinstrument/wcms_112458.pdf. Readers should note that the statistical definition was amended in 2018 to take into account new definitions of work and employment, which, in particular, break down the 2008 definition of children in employment into own-use production work by children and employment work by children (https://www.ilo.org/wcmsp5/groups/public/---dgreports/--- stat/documents/meetingdocument/wcms_667558.pdf); the 2008 definition is still used in this paper, however, due to data availability.

1015 MTheore concept information of child can labour be isfound broader in ILO,than the2008. SNA Readers and also shouldcan include note (depending that the statistical on the measurement definition framework was amended adopted in by 2018 countries) to take into account new definitions of work and employment, which, in particular, break down the 2008 definition of children in children engaged in own-use production of services (household chores). employment into own-use production work by children and employment work by children (see ILO, 2018b); the 2008 definition is still used in this paper, however, due to data availability. 16 Data on child labour have been collected in MICS on both economic activities (paid or unpaid work for someone who is not a member of the 11 The concept of child labour is broader than the SNA and also can include (depending on the measurement framework adoptedhousehold, by workcountries) for a family children farm engaged or business) in own-use and domestic production work (household of services chores (household such as cooking,chores). cleaning or caring for children). The 12MICS D atachild on labour child module labour also have collects been informationcollected in on MICS hazardous on both working economic conditions. activities (paid or unpaid work for someone who is not a member of the household, work for a family farm or business) and domestic work (household chores such as cooking, cleaning or caring for children). The MICS child labour module also collects information on hazardous working conditions.

• Children aged 15-17 engaged in economic activity for at least 43 hours in the reference week; • Children aged 5-17 engaged in hazardous occupations and branches of economic activities. 20. This paper uses datasets on child labour from 65 nationally representative surveys (see Annex 2). These include ILO-supported Labour Force Surveys or Child Labour Surveys, UNICEF-supported Multiple Indicator Cluster Surveys (MICS)16 and USAID- supported Demographic and Health Surveys (DHS). The datasets from these surveys contain information on about 50% of children estimated to be in child labour globally. 21. Within this set, 30 countries released their surveys with sectoral information related to child labour ( ) at ISIC rev 3/3.1/4 standard codes. The child labour figures were then harmonised to match the OECD-ICIO’s classifications and rescaled to the year 2015 using the following formula𝜔𝜔 : DEFINITIONS AND DATA SOURCES 9 where is the absolute𝜔𝜔!,!,!"#$ number= 𝜔𝜔!,! of !"#$ children(𝑈𝑈𝑈𝑈!"! in ,!child,!"#$ /labour𝑈𝑈𝑈𝑈!"! in,!, (industry!"#$)) , country , and year of survey ; is the total population of children aged between 5 and 17 !,! when expanding𝜔𝜔 the survey weights in country . 𝑗𝑗 𝑐𝑐 𝑦𝑦𝑦𝑦𝑦𝑦𝑦𝑦 𝑈𝑈𝑈𝑈!"!,! Additional22. Additional assumptions assumptions had to be madehad tofor bethe maderemaining𝑐𝑐 for the35 countries.remaining In 35 this countries. regard, countries In this whereregard, industry countries information where proportionsindustry information in a region Aproportions were available in werea region used to were estimate available child labour were byused industry to estimate for countries child that labour had released by industry only the for total countries number (normalizedthat had released to 100%). only The the rationale total fornumber using this (normalized approach isto that, 100%). based Theon available rational datae for with using sectoral this information,approach𝐴𝐴 is the that proportionality, based on ofavailable child labour data is withexpected sectoral to be information,similar across thecountries. proportionality Additional of information child labour on compensation is expected toof employeesbe similar per across industry countries. was incorporated Additional to avoid information bias to a specific on compensation country. The ofuse employeesof compensation per ofindustry employees was is incenvisionedorporated to tohelp avoid avoid bias bias to toward a specific large availablecountry. datasetsThe use inof a compensationregion. Moreover, of compensationemployees isof envisionedemployees’ datasetto help can avoid be used bias as to a wardproxy tolarge provide available an approximate datasets figuresin a region of total. employmentMoreover, percompensation industry. Thus, of foremployees’ countries where dataset only can total be child used labour as informationa proxy to was provide available, an approximate is as follow: figures of total employment per industry. Thus, for countries where only total child labour information was available, is as follow: ω 𝜔𝜔

! ! 1 ! 𝜔𝜔!,! 𝜕𝜕!,! !,! ! ! ! ! ! 𝜔𝜔 = ! + 𝑑𝑑𝑑𝑑 𝜔𝜔 ! 𝜔𝜔!,! 𝜕𝜕!,! ! ! 𝜔𝜔! ! 𝜕𝜕! ! ! + ! where represents the !amount! 𝜔𝜔! of the! 𝜕𝜕! compensation of employees for industry in

country !,.! It should 𝜕𝜕also be noted that , Northern America and Oceania are not included in the analysis𝑗𝑗 23. It should also be noted that Europe, Northern America and Oceania are not included due to lack𝑐𝑐 of available data. child labour in these regions is relatively marginal and therefore the in the analysis due to lack of available data. Child labour in these regions is relatively impacts on the overall results can be assumed to be minimal. The results for Eastern and South- marginal and therefore the impacts on the overall results can be assumed to be minimal. Eastern Asia should also be used with caution due to data limitations, notably the absence of child The results for Eastern and South-Eastern Asia should also be used with caution due to data labour data for its most populous country China. The full list of regions, country and population

coverage16 Data on childis included labour have in been ANNEX collected 2in. MICS on both economic activities (paid or unpaid work for someone who is not a member of the household, work for a family farm or business) and domestic work (household chores such as cooking, cleaning or caring for children). The MICS child labour module also collects information on hazardous working conditions.

10 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

1.3 TRAFFICKING FOR FORCED LABOUR

Human trafficking and forced labour are considered to be rare events, statistically speaking. Methodologies to capture reliable prevalence numbers are recent, and the availability of national datasets is lower than for child labour (see ANNEX 2). In addition, even for countries where there are national forced labour estimates, datasets rarely provide the sectoral distribution. Forced labour is often concentrated in “pocket” areas or sub-sectors (ILO and Walk Free Foundation, 2017), which would require specific statistical (over)sampling methods (ILO, 2018a) to provide reliable figures on sectoral distribution of forced labour that could then be linked with ICIO. Similarly, the measurement of human trafficking for forced labour is an area of ongoing efforts.

In the context of this research, an experimental effort was made to replicate the methodology adopted for the child labour analysis by: (i) modelling industry-level country estimates of victims with existing datasets and the results from the 2016 ILO global estimates on forced labour; and (ii) estimating the contribution of industries with trafficking for forced labour to global supply chains. The results presented should only be interpreted as indicative of the nature of this issue.

Several datasets were used in the context of this research: • The human trafficking data used in this exercise are 2006 to 2016 country aggregates from the Counter Trafficking Data Collaborative (CTDC) which include victim case data from IOM and partner organizations.13 Curated by IOM, CTDC is the first global data portal on human trafficking, with data contributed by multiple agencies. The data used in this report combine the three largest case-level “victim of human trafficking” datasets in the world, from IOM, Polaris and Liberty Shared. As for all administrative victim data collected by counter-trafficking orga- nizations, data on identified cases of human trafficking are best understood as a sample of the unidentified population of victims. This sample may be biased if some types of trafficking cases are more likely to be identified than others, but the extent of this bias is unknown. Nevertheless, there are few, if any, alternative sources of data on the distribution of human trafficking by industry across countries (IOM, 2018). • The regional results on forced labour in the private economy from the 2017 Global Estimates of Modern Slavery; • ILO harmonized datasets on employment by industry.14

These three datasets were merged as follows. First, it was assumed that all countries within a region had the same total prevalence of human trafficking for forced labour, using the results from the Global Estimate. The regional prevalence figures were distributed proportionally according to total employment figures to obtain an estimate of the total number of victims per country. Second, the country-level numbers were distributed by broad industry using human trafficking data from CTDC. Here, broad industry sectors refer to the 1-digit ISIC industry from Revision 4. Third, the 1-digit sectoral aggregates were distributed to the ISIC 2-digit level, using the distribution of within-industry adult employment. In total, 30 countries were used in the estimation. Note that Oceania, Central and Southern Asia, and Latin America and the Caribbean are not included because of low data availability. As for child labour, country coverage is indicated in ANNEX 2.

13 Accessed on 01/09/2019 and available at https://www.ctdatacollaborative.org/. 14 See the ILO Harmonized Microdata webpage: www.ilo.org/ilostat.

DEFINITIONS AND DATA SOURCES 11

While there are some national level estimates of prevalence, there are currently no global or regional prevalence estimates. Moreover, for the existing forced labour prevalence estimate, the sectoral data are not stable, particularly at lower levels of geographic disaggregation (i.e. country level) - hence the need to combine these datasets and provide estimates of trafficking for forced labour. In addition, because the forced labour estimate isn’t available at the country level, employment data had to be used to distribute the forced labour estimate.

It should be noted also that for both forced labour and human trafficking, there is not much sectoral data available at the ISIC 2-digit level, compared to child labour. This study exemplifies the need to bring the forced labour and human trafficking data up to the 2-digit level, in order to obtain the same level of quality detail as the child labour results over the coming years.

Figure 4: From data sources to model estimates

12 PART 2: COMBINING THE DATASETS DRAFT – FINAL DRAFT – FINAL DRAFT – FINAL DRAFTDRAFT – –FINAL FINAL DRAFT – FINAL Part 2 Part 2DRAFT – FINAL Part 2 DRAFT – FINAL Part2.1.Part Combining 2 2 the Datasets Part 2 2.1. Combining the Datasets 2.1.30. Combining 2.1 PartThe F 2ROMICIO the com DATA Datasetsbined SOURCES with a social TO indicator ESTIMATES (in this case child labour and trafficking Part 2 2.1.2.1. Combining Combining the the Datasets Datasets 2.1. Combining the Datasets for forced labour) allows to estimate how30. the indicatorThe ICIO is com linkedbined with with the a socialproduction indicator of (in this case child labour and trafficking goods30. 2.1.and CombiningTheservicesThe ICIO ICIO combined for the comthe Datasetswithbineddomestic a social with andindicator a social foreign (in indicator this markets. case child (i nIn labourthis the case exampleand traffickingchild of labour child for forced andlabour, labour)trafficking to 2.1. Combining the Datasets 30.30. TheThe ICIO ICIO com combinedbined with with a asocial social forindicator indicator forced (i labour)n(i nthis this case caseallows child child to labour labourestimate and and howtrafficking trafficking the indicator is linked with the production of DRAFT – FINAL 30. The ICIO combined with a socialcapture indicatorfor allforced allows direct(in this labour)to and estimatecase indirect allowschild how labourthe imto pacts indicatorestimate and that traffickingis rhowlinkedippl e the withthroughout indicatorthe production the is complexlinkedof goods with and supply servicesthe chainsproduction for the of of forfor forced forced labour)30. labour) Theallows allows ICIO tocom to estimate binedestimate with how ahowgoods social the the indicatorandindicator indicator services (in is this is linkedfor caselinked the child withdomestic with labour the the productionandand production trafficking foreign of markets. of In the example of child labour, to for forced labour) allows to estimate how goodsthe indicator domesticand services isandDRAFT linked foreign for– FINAL the withmarkets. domestic the In production the and example foreign ofof child markets. labour, In to thecapture example all direct of andchild indirect labour, to 30. The ICIO combined with goodsthea goodssocial entire and indicator and economy servicesfor services forced (in ,thisfor a for labour)global casethe the domestic child allowsdomesticLeontief labour to and estimate inverseand and foreigncapture foreigntrafficking how need markets. allthemarkets. s direct indicatorto be In andIncomputed, the isthe indirectlinkedexample example with im of pacts theof child childproduction that labour, labour, rippl of toe to throughout the complex supply chains of goods and services for the domestic and foreigncapture markets.impacts all direct thatIn the rippleand example indirectthroughout of im thechildpacts complex labour, that supply rippl to chainse throughout of the entire the economy, complex a global supply Leontief chains of for forced labour) allows to estimatecapturecapture how all allthe inversedirectgoods directindicator B andand needsand services indirectis indirect tolinked be forcomputed, im with theimpacts pactsdomestic the with thatproduction thatthe andrippl rentireippl foreigne ethroughoutof economythroughout markets., Inthea theglobalthe complex examplecomplex Leontief supplyof supply child inverse labour,chains chains to ofneed of s to be computed, with Part 2 capture allgoods direct and and services indirect for imthepacts domestic that and rtheippl foreign entiree throughoutcapture markets. economy allPart directtheIn , 2 thea complex globaland example indirect Leontief supply of imchildpacts inversechains labour, that𝐁𝐁 r of ipplto neede throughouts to be computed, the complex with supply chains of thethe entire entire economy economy, a, aglobal global Leontief Leontief inverse inverse need need!𝟏𝟏s tos to be be computed, computed, with with the entire captureeconomy all, directa global and Leontief indirect im inversepactswhere that need risippl theanse to entirethroughoutidentity be computed,2.1.economy Combining matrix the, complexa globalwith the Datasets Leontief supplyand chainsinverse is of a matrixneeds to beof computed,technical with coefficients with DRAFT – FINAL 𝐁𝐁 = (𝐈𝐈 − 𝐀𝐀)𝐁𝐁 𝐁𝐁 2.1. Combining the Datasets the entire economy, a global Leontief inverse needs to be computed, with !𝟏𝟏 elements 30. , whereThe ICIO combinedrepresentwhere with a𝐁𝐁 social𝐁𝐁 theindicatoris𝐁𝐁 anamount! (i𝟏𝟏identityn this case in child thematrix labour intermediate and traffickingand matrix is a matrix of technical coefficients with where𝐁𝐁𝐈𝐈 is an identityfor forced matrix labour)𝑁𝑁𝑁𝑁𝑁𝑁 allows to estimateand𝐀𝐀 how! theis𝟏𝟏! 𝟏𝟏indicatora matrix is linked of with technical the production coefficients 𝐁𝐁of = (𝐈𝐈 − 𝐀𝐀 with) wherewhere is where isan an identityI identityis an identity matrix matrix matrix (NxN)andand and𝐁𝐁 = Ais( 𝐈𝐈is is −a a a𝐀𝐀matrix )matrix!𝟏𝟏 of oftechnical of technical technical coefficients coefficients coefficients with elements with with 𝐁𝐁!where𝟏𝟏 !"is,! angoods identity and services matrix for 𝐁𝐁 the 𝐁𝐁 =domestic=(𝐈𝐈(−𝐈𝐈and and−𝐀𝐀 foreign𝐀𝐀) ) is markets.a matrix In the of example technical of child coefficientslabour, to with 30. The ICIO comwherebined with is an a socialidentity indicator matrix (i n this caseofand supplying child is labour !"a ,! matrixindustry and𝑍𝑍 traffickingof technical coefficients, into!",! elementa using 𝐁𝐁 =with (sindustry𝐈𝐈 − 𝐀𝐀) , where, and for countryrepresent the amount in the intermediate matrix 𝐁𝐁 = (element𝐈𝐈 − 𝐀𝐀 𝑎𝑎) s =! 𝟏𝟏 capture!,! all direct, where and𝑍𝑍 indirect impactsrepresent that𝐈𝐈 rippl ethe throughout amount the complex in the supply intermediate𝑁𝑁𝑁𝑁𝑁𝑁 chains of 𝐀𝐀matrix𝐙𝐙 Part 2 where is an identity matrix elementelementands s a 𝐈𝐈 elementij,cis =a Z_(ij,c)⁄X_(j,c)matrixs 𝑋𝑋 theof entire technical, ,where, where economywhere , Z_(ij,c) , wherecoefficientsa 𝑁𝑁𝑁𝑁𝑁𝑁global represent representLeontiefrepresent represent inversewith𝐀𝐀 the the theamount need amountthe amounts toamount inbe computed,the in intermediatein inthe thethe with intermediate intermediate matrix Z ofmatrix matrix for forced labour) allows to estimate how the indicator is (slinkedee 𝐁𝐁Millar= 𝐈𝐈with𝐈𝐈(𝐈𝐈 − andthe𝐀𝐀) Blair,production𝐈𝐈 !"2010,! offor 𝑁𝑁𝑁𝑁𝑁𝑁more𝑁𝑁𝑁𝑁𝑁𝑁 information𝑁𝑁𝑁𝑁𝑁𝑁 𝐀𝐀 𝐀𝐀 𝐀𝐀). The matrix𝑍𝑍!",! holds the direct links elements , where represent the amount!",! !,!!,!,𝑍𝑍! in,! the!" ,intermediate! of!" matrix,! supplying !",! industry , into!",! a using industry , and for country 𝐈𝐈 elements 𝑁𝑁𝑁𝑁𝑁𝑁 , where of𝐀𝐀represent supplyingof supplying the industry amount!",!𝑖𝑖 industry, 𝑖𝑖 in𝑍𝑍= the1 … intermediate𝑁𝑁 , into, into!"matrix, !a ausing using 𝑎𝑎 𝐁𝐁 industryindustry= 𝑗𝑗, 𝑗𝑗 = !1,!… 𝑁𝑁 , and, 𝑍𝑍and for countryfor country 𝑐𝑐 𝐙𝐙 goods and services for the domestic and𝐈𝐈 foreign markets. In of𝑁𝑁𝑁𝑁𝑁𝑁betweentheof supplying examplesupplying industries𝐀𝐀!,!! , !,of !industry𝑎𝑎, ! industrychild𝑍𝑍 = whilst𝑍𝑍𝑎𝑎 labour, = the !to ,!matrix !,! , into,B! ,into! !,contain!,𝑍𝑍! ,!a 𝑍𝑍 ausing using s all industry industrydirect and indirect𝑋𝑋 links., and, and Infor forthis country country regard,𝐙𝐙 𝐙𝐙 2.1. Combining the Datasets!" ,! 𝑎𝑎 supplying𝑎𝑎 == industry! i,,!! 𝑋𝑋i,! = 1…𝑋𝑋 N,𝑍𝑍 into𝑍𝑍 a (susing ee Millarindustry j,and j !=𝟏𝟏 1…Blair, N, and2010 for countryfor more c (see information Millar and𝐙𝐙 𝐙𝐙). The matrix holds the direct links of supplying!",! industry𝑍𝑍 !",! , into!",! a using(see Millarindustry(see Millarand Blair, and𝑋𝑋where𝑋𝑋 Blair, 2010 is an,2010 andidentityfor forformore matrix more country information information 𝐁𝐁and = ( 𝐈𝐈 − is𝐀𝐀 )).a matrixTheThe ofmatrixmatrix technical𝐀𝐀 coefficientshold holds thes with thedi rect di linksrect links capture all direct and indirect 𝑎𝑎ofim supplyingpacts= that!",! industry rippl!,!𝑍𝑍 e throughout𝑍𝑍 (sthe the(s,ee into!"ee multiplier ,!Millarcomplex Millara using and and supplyindustrymatrix Blair, Blair, chains can2010 2010 be offor forwritten more , moreand informationforas information country 𝐙𝐙 ). ).The The matrix matrix𝑖𝑖, 𝑖𝑖 =wherehold hold1 …s 𝑁𝑁 sthe the isdi directarect vector links links 𝑗𝑗, 𝑗𝑗 = 1 … 𝑁𝑁 𝑐𝑐 𝑎𝑎 = !,! 𝑍𝑍 Blair, 2010 for more 𝑖𝑖information)., 𝑖𝑖 𝑖𝑖=, 𝑖𝑖1=…1 𝑁𝑁The… 𝑁𝑁 betweenmatrix A holds industries𝐙𝐙 the direct whilst 𝑗𝑗links,𝑗𝑗,𝑗𝑗 𝑗𝑗= between the=1 …1 matrix𝑁𝑁… industries𝑁𝑁 B contain whilst thes𝑐𝑐 all direct𝑐𝑐 and indirect links. In this regard, the entire economy, a(s globalee Millar Leontief and Blair,inverse 𝑋𝑋2010 need for𝑋𝑋 s moreto be computed,informationbetween with).between industriesThe matrixindustries whilstelement whilst𝐈𝐈shold the sthematrix the matrix di , Brectwhere Bcontain𝑁𝑁𝑁𝑁𝑁𝑁 contain links represents s𝐀𝐀 allall thedirect amount and and in theindirect indirect intermediate links. links. matrix In this In regard,this regard, (see Millar and Blair, 2010 for betweenmorebetween information industries industriesmatrix B). containsThe whilst whilst 𝑖𝑖matrix,𝑖𝑖 ,𝑖𝑖 allthe=𝑖𝑖 thedirect= 1matrix 1… matrixhold …and𝑁𝑁𝑁𝑁!" s indirect, !Bthe Bcontain containdi links.rect s links Insall thisall direct directregard, and𝑗𝑗 the,and𝑗𝑗!,𝑗𝑗 indirect𝟏𝟏multiplier=𝑗𝑗 indirect=11……𝑁𝑁 matrixlinks.𝑁𝑁 links. canIn In bethis this written regard, regard, as𝑐𝑐 𝑐𝑐 𝐀𝐀 30. The ICIO combined with a social indicator (in this case child labourof supplying and!",! industry trafficking𝑍𝑍 the multiplier, into!",! a using industry matrix can 𝐀𝐀,be and𝐀𝐀 forwritten country as where is a vector 𝑖𝑖, 𝑖𝑖 = 1 … 𝑁𝑁 (diagonalizedthe multiplierthe) withmultiplier 𝑗𝑗 anmatrix, 𝑗𝑗 element= matrix1 …can 𝑎𝑎𝑁𝑁 canbe= bewritten written!,! as as𝑍𝑍. 𝑐𝑐 where where is𝐙𝐙 a isvector a vector between industriesbetween industries whilst the whilst matrix𝑖𝑖, 𝑖𝑖 the= 1 matrixB… contain𝑁𝑁 B contains all directs all direct and andindirect𝑗𝑗,(see𝑗𝑗 indirect =Millar1 … links.and𝑁𝑁 links. Blair, In 𝑋𝑋 2010In this this for regard,𝛒𝛒 moreregard,𝐁𝐁 =information𝑐𝑐 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 ). The𝑽𝑽𝑽𝑽 matrix 𝐁𝐁𝐀𝐀 hold𝐀𝐀 s the direct𝛒𝛒 links for forced labour) allows to estimate how thethe theindicator multiplier multiplier is linkedmatrix matrix with can can thebe be productionwritten written !as ,!as of !𝟏𝟏 where where is isa avector vector !𝟏𝟏 𝐁𝐁 between𝐀𝐀 industries𝐀𝐀 whilst𝑖𝑖, 𝑖𝑖 the= 1matrix… 𝑁𝑁 B contains all direct and𝑗𝑗, 𝑗𝑗 indirect= 1!…𝟏𝟏𝑁𝑁 links. In this regard,𝑐𝑐 the multiplierthe multiplier matrix canmatrix be canwritten be written as as (diagonalized) with anwhere elementwherewhere!,! 𝑉𝑉isis a( isdiagonalized avector a vectorvector (diagonalized)𝛒𝛒 𝐁𝐁. =) with𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 with an𝑽𝑽𝑽𝑽 an element element 𝐁𝐁 𝛒𝛒 . goods and services for the domestic and!𝟏𝟏 foreign31. markets. (diagonalizedSimilarly, In the )exampleto with estimate thean multiplierofelement childtheρ matrixamount labour,= can be ofto written! ,!child as𝛒𝛒 𝐁𝐁.labour = 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 linked!𝑽𝑽𝑽𝑽𝟏𝟏!𝟏𝟏 with𝐀𝐀 where 𝐁𝐁exported is a vector goods𝛒𝛒 𝛒𝛒and𝐁𝐁 = 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑽𝑽𝑽𝑽 𝐁𝐁 𝛒𝛒 (diagonalized(diagonalized) with) with an an element element!𝟏𝟏 𝑋𝑋 !.,! . !,! where is an identity matrix and is a matrix of technical coefficients!𝟏𝟏 with !,! !𝛒𝛒𝑉𝑉,!𝐁𝐁𝛒𝛒𝐁𝐁== 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑽𝑽𝑽𝑽𝑽𝑽𝑽𝑽!𝟏𝟏 𝐁𝐁 𝐁𝐁 𝛒𝛒𝛒𝛒 capture all direct(diagonalized and indirect 𝐁𝐁 =) (with 𝐈𝐈im−pacts an𝐀𝐀) element that r ippl e throughout. the complex(diagonalized supply) with chains an elementρ =of !,! . !,! 𝑉𝑉 (diagonalized) with an element services𝛒𝛒𝐁𝐁. =𝛒𝛒ρ 31.𝐁𝐁 ,𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑̂B= t=he diag(〖VX〗^(-1𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑followingSimilarly,𝑽𝑽𝑽𝑽 𝑽𝑽𝑽𝑽 𝐁𝐁 equationto 𝐁𝐁 estimate was 𝛒𝛒ρ! the, !̂ 𝛒𝛒! ,used!amount!,!𝑉𝑉 : of𝑋𝑋 child𝛒𝛒𝐁𝐁 = labour 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑽𝑽𝑽𝑽 linked 𝐁𝐁 with 𝛒𝛒ρρ_(j,c)=exported= goods V_ ! ,!and ρ 𝑉𝑉=𝑉𝑉31. !Similarly,!,! to estimate the amount of child labour linked with exported goods and the entire economy, a global Leontief inverse !, !need31.!s,! to beSimilarly, computed, to with estimate !,! !the,! amount!,! of𝑉𝑉 child labour linked with exported 𝑋𝑋goods and elements , where represent the amount!,! in 𝑉𝑉the intermediateSimilarly,services to estimate ,matrix the31. following the Similarly, ρamountρ equationto= = estimate of child wastheρ !labour ,!amount! =used,!𝑋𝑋 linked :of ! ,!child with labour exported linked with goods exported and goods services and (S), the 𝐈𝐈 𝑁𝑁𝑁𝑁𝑁𝑁 𝐀𝐀 !,! ρ𝑉𝑉31.31.= 32.Similarly, Similarly,!,! In order to to toestimate estimateestimate the the amount directamount servicesand of𝑋𝑋 of𝑋𝑋 indirectchild child 𝑋𝑋labour labourcases, the followinglinkedof linked child with labourwith equation exported exportedthat arewas goodslinked usedgoods : andwith and !",31.! Similarly,31. Similarly, to estimate to estimate theρ amount= the amount ofservices! ,!child of(𝑋𝑋𝐒𝐒 following )childlabour labour, t equationhelinked following linkedservices waswith used:with exported,equation the exported following was goodsequation goods used was and and used: : of supplying!",! industry𝑍𝑍 , into!",! a using industry servicesservices𝑋𝑋 gross exports,, andt,he the followingfor ,following equation country equation (1) equation can be was rewrittenwas used used: as: follows: 𝑎𝑎 = !,! services𝑍𝑍 , the following equation𝐁𝐁 was used: ( 𝐒𝐒) 𝐙𝐙 services𝑋𝑋 , the following equation was used: (𝐒𝐒) (see Millar and Blair, 2010 for more information). The matrix! 𝟏𝟏 hold(s the) direct links (𝐒𝐒) where is an identity matrix and is a matrix( () ) 𝐒𝐒of technical coefficients with between industries whilst𝑖𝑖, 𝑖𝑖 the= 1matrix… 𝑁𝑁 B(𝐒𝐒 contain) s all direct𝐁𝐁 = ( and𝐈𝐈𝑗𝑗−, 𝑗𝑗 𝐀𝐀indirect=) 1 … 𝐒𝐒𝑁𝑁 links.𝐒𝐒 In this regard,𝑐𝑐 (𝐒𝐒) ………..……………... the multiplierelement matrixs can be written as , where representwhere thewhere 𝐀𝐀 amount whereiswhere a isdiagonalized ain diagonlized theis a a intermediatediagonalized wvectorhere indicator matrixis a diagonalized 𝐒𝐒indicator ofmatrix 𝐒𝐒 of !"#$%& = ,vector 𝐒𝐒withindicatorof𝛒𝛒 𝐁𝐁 vector== 𝐒𝐒 𝐅𝐅zer 𝛒𝛒of𝛒𝛒=, o 𝐁𝐁vector ρwith 𝛒𝛒 elements ,𝐅𝐅𝐅𝐅𝐁𝐁 with 𝐅𝐅(……… 𝟏𝟏 , anwith) an ..( (𝟏𝟏 𝟐𝟐element an ))inelement) element the off diagonal. , where ca ptures,, wherewhere all 𝐈𝐈 𝑁𝑁𝑁𝑁𝑁𝑁 𝐀𝐀 where is a diagonalized indicator of vector , with an element !,! , where !",! the direct impacts. child labour has account the same is a vector dimensions that holds the as amount in equation of child labour! ,(1)! ,per𝜔𝜔 but industry only!,! and includes the…… ….. 𝑍𝑍 !𝟏𝟏 child labour account 𝛒𝛒 is a vector that holds the amount… 𝛒𝛒……. .of child labourρ = per industry𝑋𝑋!𝐒𝐒,! !,! =and 𝛒𝛒 country𝐁𝐁 𝐅𝐅 (𝟏𝟏) (diagonalized)of with supplying an element!",! industry . , into!",! a usingchild industry labourchild 𝐁𝐁 account……… ρlabour. . account countryis, aand vector for is, aand countrythat vector 𝐒𝐒𝐁𝐁 holddescribesw that =heres holdthe𝛒𝛒 the 𝐁𝐁… …… amountsamount𝐅𝐅is… …the . …. a. . amountofdiagonalized child (of𝟏𝟏) labourchild of child intensities labour ! indicator,labour!!,! that per𝜔𝜔 perhold𝜔𝜔 industry 𝐁𝐁of industrythe vector andand , with an element , where 𝑎𝑎 = !,! 𝛒𝛒𝐁𝐁 =𝑍𝑍 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑽𝑽𝑽𝑽𝐒𝐒 w= 𝐁𝐁directhere 𝛒𝛒𝐁𝐁 𝐅𝐅impact is 𝛒𝛒a ( diagonalized(i.e.𝟏𝟏) children fromindicator industry𝜔𝜔 of𝐙𝐙 vector working𝛒𝛒 𝛒𝛒, within the an production elementρ ρ =of= products !!,!,! exported, where where 𝑋𝑋 is a diagonalized indicator of vector𝛒𝛒……….. , Nwith, 𝛒𝛒and an elementj,camount!"#$%& describes of child 𝐒𝐒 thelabour 𝐒𝐒 amount =recruited= 𝛒𝛒 𝐁𝐁𝛒𝛒 of per,𝐁𝐁 𝐅𝐅 wherechild one𝐅𝐅 unit labour (of𝟏𝟏 (total)𝟏𝟏 intensities) output in a gthativen supplyinghold the industry amount𝑋𝑋𝑋𝑋 of child (see Millar and Blair, 2010!,! for more informationwcountrywherehere). The is is countryamatrix adiagonalized ,diagonalized and hold𝐒𝐒 , anddescribes sindicator indicatorthe 1𝑥𝑥𝑥𝑥 di describesrect theofρ !of ,! linksvectoramount vector the amount , ofwith, withchild of an anchild elementlabour element labour intensities intensities that that holdhold, where, wherethethe !,! 𝑉𝑉 𝐒𝐒 = 𝛒𝛒𝐁𝐁by 𝐅𝐅 industrylabour(𝟏𝟏 )recruited). Therefore, andper countryoneω indirect unit. hereof totalimpactis !only,! childoutput represented can labourin be a bycalculatedgiven export account supplying category as of followindustry finalis a demand. s:vector and Whilecountry. that the ! hold,!F heres the is amount of child labour!,! per𝜔𝜔 industry and where is a!, !diagonalized𝑖𝑖, 𝑖𝑖 = 1 … 𝑁𝑁indicator of vectorchild, labourwithamount𝑗𝑗, 𝑗𝑗an =account ofelement1 child… 𝑁𝑁 labour is𝜔𝜔 a recruitedvector𝜔𝜔 thatper 𝑐𝑐, one holdwhere units the of totalamount output of in child a given labour !supplying,! !,! !per,𝜔𝜔! industry𝛒𝛒 and ρ = !,! 31. Similarly,between to estimate industries childtheρ whilst amount labour= the accountof matrix! ,!child Blabouris containa vector linkeds thatall directwithhold sonlyexported theand representedamount indirect goods𝜔𝜔 of vector links.bychild and export capture labourIn !!,!, !thiscategorys the per directregard, industry impacts,of final 𝑖𝑖 the 𝛒𝛒and multiplierdemand. Whiledescribes the the vector amount of childcaptures labour the direct!,! 𝛒𝛒 𝑋𝑋 childamountchild labour labour of (1 x child𝛒𝛒 𝛒𝛒account account )labour ρ recruitedis isa avector vectorρ 𝐅𝐅per= that !onethat,! hold unithold!,!s ofsthe the total amount 𝛒𝛒amount output of of inchild child a g ivenlabour !labour,!!, ρ!ρsupplying per𝜔𝜔 per=𝜔𝜔 industry industry industry and and 𝑋𝑋 country , and describes thecountry amountand of country child,1𝐀𝐀 𝑥𝑥𝑥𝑥and . !labourthat,! here are directlyintensities isρdescribes only and𝜔𝜔 representedindirectly that countrythe𝑋𝑋 hold requiredamount 𝛒𝛒by 𝛒𝛒 the toexport sati ofsfy ,one childcategoryand unit oflabour finalof ρ describesfinaldemandρ intensities= demand..= Hence, the the !While ,that!amount!,!𝑋𝑋 holdthe of the child labour intensities that hold the services , tthehe following multiplierchild labour equation matrix account wascan used be is:written a vector as that andhold countrys the𝛒𝛒𝛒𝛒1 impacts,amount𝑥𝑥𝑥𝑥. here𝑖𝑖 theof multiplieris ρchildwhere only labour𝛒𝛒 !represented ,! Bis describesper a vectorindustry bythe exportamount and ofcategory child labour of thatfinal are demand. directly and 𝑋𝑋While𝑋𝑋 indirectly the 𝜔𝜔 countrycountry 𝛒𝛒 , ,and and outcome describes ρdescribes of this= calculation the the amount ! , !amountis a matrix of (of child childthat𝛒𝛒𝐁𝐁 presents labour labour𝜔𝜔 the domesticintensities intensities child labour that that in hold hold the the 𝛒𝛒 amount of child labour recruited per oneamount unitrequired vectorof of total child to output capturesatisfy laboursourcing ins𝜔𝜔one the arecruited unitg industrydirectiven of supplyingfinalimpacts, (directly per 𝑋𝑋demand. amount onefrom the industry industry unitmultiplier Hence, of of andchild totalthe indirectly outcome labour outputdescribes from other ofrecruitedin thisthe upstreama amountgcalculationiven perindustries) supplying of one S child is unita labourmatrix industryof total output in a given supplying industry country , and describes!,! thevector amount captureof child!s𝟏𝟏 the labour direct𝜔𝜔𝜔𝜔𝐅𝐅 intensities !impacts,,! the !"#!$%&'that multiplier hold the ……… describes..(𝟑𝟑) !the,! amount of child labour (diagonalized) with an element1𝑥𝑥𝑥𝑥 ρ amount. amount Where of of (Nxl)childthat child1 is𝑥𝑥𝑥𝑥are labour that thelabour directly presentsoff recruitedin recruitedcountrydiagonal ρand the indirectly into domesticper finalmatrixper𝐒𝐒 one destination onerequired childandunit𝐒𝐒of unit country= labourcountryof, to𝛒𝛒ofwith total𝐁𝐁 sati .total1𝑁𝑁𝑁𝑁𝑁𝑁 𝐅𝐅in 𝑥𝑥𝑥𝑥 s zero. )fy sourcingoutput output one hereelements unit in industry isρin a of onlyag ivenfinalgin iveni therepresented( directlydemandsupplying supplyingmain from. diagonal.Hence, industryindustryby industry exportthe category of final demand. While the ( ) and country. 𝜔𝜔 here is only represented𝛒𝛒𝐁𝐁 =and 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 bycountry export𝑽𝑽𝑽𝑽𝐅𝐅 . category 𝐁𝐁 here !,!! ,!isof onlyfinal𝛒𝛒 representeddemand.𝑖𝑖 While by theexport𝑖𝑖 category of final demand. While the 𝐒𝐒 amount of child labour recruited per onethat unit arecapture of directlytotal1i𝑥𝑥𝑥𝑥1 outcomeands𝑥𝑥𝑥𝑥 alloutput indirectlytheand𝛒𝛒 ofindirect ρinindirectlythisρ afrom calculationg ivenimpacts. other requiredsupplying upstream is a tomatrix industries)industry sati (sfy inone thatcountry unit presents cof into finalthe final domestic demand destination child. Hence, labourcountry theinl . vector capture!,s! the direct! ,!impacts,andand the country countrymultiplier. . here here describes is isonly only the represented 𝑐𝑐 representedamount of childvectorby by exportlabour export capture𝑙𝑙 category category𝛒𝛒𝐁𝐁s the ofdirect of final final impacts, demand. demand. the While multiplierWhile the the describes the amount of child labour 1𝑥𝑥𝑥𝑥 ρ !,! 𝑉𝑉 vector𝛒𝛒 𝐁𝐁captures the direct impacts, the multiplier𝐁𝐁 describes the amount of child labour𝐁𝐁 31. andSimilarly, country to. estimate here is𝐅𝐅 theonlyρ amount =represented of ! ,!childoutcome by labourexport of sourcingthis linkedcategory calculation withindustry of exported final (directly is demand. a goodsmatrix from Whileindustry and( thethat and𝛒𝛒 𝐁𝐁 presentsindirectly𝐅𝐅 thefrom domestic other upstream child industries)labour in that are directly and indirectly𝑋𝑋 requiredvectorvector33. to capture saticaptureAsfydditionals one thes𝐅𝐅 the unitdirect directanalysis of impacts,final impacts, was demand carried 𝐒𝐒the the multiplier. that multiplierHence, out areon 𝑁𝑁𝑁𝑁𝑁𝑁 thethe directly ) structural describes describes and layerindirectly the the d amountecomposition amount required of of child child (i.e. to labour satitiers).laboursfy one unit of final demand. Hence, the vector capture𝛒𝛒s the direct impacts, thesourcing multiplierthat areindustry in directly 𝐅𝐅country describes𝐅𝐅 ( directly and into the finalindirectly amountfrom destination industry ofrequired child country andlabour to . indirectlysati sfy one from unit other of final upstream demand industries). Hence, the services , theoutcome following of equationthis calculation…… was….. used is :a that matrixthat are Estimatingare (directly directly𝛒𝛒 that 𝛒𝛒 the𝐁𝐁and presentsand contributions indirectly indirectly the𝑖𝑖 domestic requiredof required other child upstreamoutcometo to laboursati sati𝛒𝛒s𝑖𝑖fy sinindustriesoffy one thisone unit calculationunit intoof of final exportingfinal demand isdemand a industry matrix. .Hence, Hence, ( is verythe thethat 𝛒𝛒 𝐁𝐁 presents the domestic child labour in that are directly𝐒𝐒𝐅𝐅 and= 𝛒𝛒indirectly𝐁𝐁 𝐅𝐅 required(𝟏𝟏) in tocountry outcomesatisfy intoone of thisfinalunit calculation destinationof final𝐒𝐒 demand country is a . matrixHence, .𝑁𝑁𝑁𝑁𝑁𝑁 () the that𝛒𝛒𝐁𝐁 presents the domestic child labour in where is a diagonalized indicatorsourcing of industry vector (directly, with anfrom elementoutcome industryoutcomevaluable 𝛒𝛒 𝛒𝛒of andof this thisindirectlyto calculationgive calculation𝑐𝑐 more ,from where insight other is isa upstream aintomatrix matrix thesourcing (industries)total ( 𝑙𝑙impact that industrythat𝛒𝛒 𝐁𝐁𝛒𝛒 presents𝐁𝐁. presentsHowever, ( directlythe the thedomestic domestic impact from industrychildin child each labour labour layer and inof indirectlyin from other upstream industries) 𝛒𝛒 𝐒𝐒 sourcing𝑁𝑁𝑁𝑁𝑁𝑁) industry𝑖𝑖 (directly from industry𝑖𝑖 and indirectly from other upstream industries) outcome(𝐒𝐒) ofin countrythis calculation into final destination is a matrix sourcingcountrysourcing ( . industrythat industry𝛒𝛒𝐁𝐁 presents! ,! (directly ( directlythe domestic from from industry industrychild labour and and indirectlyin indirectly from𝑗𝑗 from other other𝐒𝐒 upstream upstream industries) industries)𝑖𝑖𝑁𝑁𝑁𝑁𝑁𝑁) child labour account is a vector that holds 𝑖𝑖the amount of childin countrylabour!𝑖𝑖,!𝑐𝑐 per 𝜔𝜔 into industry final destinationand 𝐒𝐒 countryin𝑙𝑙 country .𝑁𝑁𝑁𝑁𝑁𝑁 ) into final destination country . 𝛒𝛒 sourcing industry (directly 𝛒𝛒 from industryinin country country andρ indirectly = into into final final from! ,destination! destination other𝐒𝐒 𝐒𝐒 upstream country country industries) . 𝑁𝑁𝑁𝑁𝑁𝑁 .𝑁𝑁𝑁𝑁𝑁𝑁 ) ) 𝑖𝑖 𝑖𝑖 country , and describes the𝑐𝑐 amount 𝐒𝐒of child labour𝑁𝑁𝑁𝑁𝑁𝑁 intensities)𝑙𝑙 that𝑋𝑋 hold𝑖𝑖 the 𝑖𝑖 in country into final destination country . 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑐𝑐 𝑙𝑙 amount of child labour𝜔𝜔 recruited per one𝑖𝑖 unit of total output in 𝑖𝑖a given supplying𝑐𝑐 industry 𝑙𝑙 !,! 𝐒𝐒 = 𝛒𝛒𝐁𝐁 𝐅𝐅 ………..(𝟏𝟏)𝑐𝑐 𝑐𝑐 𝑙𝑙 𝑙𝑙 and country1𝑥𝑥𝑥𝑥. w herehere isρ isonly a diagonalized represented𝑐𝑐 indicatorby export of category vector 𝑙𝑙,of with final an demand. element While the , where !,! vector capturechilds the labour direct account impacts, the is multipliera vector that holddescribess the amountthe amount of child of child labour!, !labour per𝜔𝜔 industry and 15 𝐅𝐅 𝛒𝛒 𝛒𝛒 ρ = !,! that are directlycountry and indirectly, and required describes to satisfy the one amount unit of of final child demand labour. Hence,intensities the that𝑋𝑋 hold the 𝛒𝛒 outcome of thisamount calculation of child islabour a matrix𝜔𝜔 recruited ( per that 𝛒𝛒one𝐁𝐁 presents unit of thetotal domestic output inchild a g ivenlabour supplying in industry !,! sourcing industryand country (directly1𝑥𝑥𝑥𝑥. from here industry isρ only represented and indirectly by fromexport other category upstream of final industries) demand. While the 𝐒𝐒 𝑁𝑁𝑁𝑁𝑁𝑁) in country intovector final destination captures the country direct impacts,. the multiplier describes the amount of child labour 𝑖𝑖 𝐅𝐅 𝑖𝑖 𝑐𝑐 that are directly and indirectly𝑙𝑙 required to satisfy one unit of final demand. Hence, the outcome𝛒𝛒 of this calculation is a matrix ( that𝛒𝛒𝐁𝐁 presents the domestic child labour in 15 sourcing industry (directly from industry and indirectly from other upstream industries) in country into final destination𝐒𝐒 country .𝑁𝑁𝑁𝑁𝑁𝑁 ) 𝑖𝑖 𝑖𝑖 15 15 𝑐𝑐 𝑙𝑙 15 15 1515 15

15

15

COMBINING THE DATASETS

13

FigureFigure 44:: FromFrom datadata sourcessources toto modelmodel estimatesestimates FIGURE 4: FROM DATA SOURCES TO MODEL ESTIMATES

• National child labour datasets

• Results from Alliance 8.7 Global Estimates on Modern Slavery National input-output tables • Counter-Traf cking Data Collaborative (CTDC) • ILO Harmonized Microdata (by industry) OECD ICIO tables

Estimate / harmonize CL by industry to match OECD ICIO classi cation

• Determination of required inputs (from home & abroad) to generate a unit of total output Estimate the amount of CL • Determination of total requirements per one unit of total output, (direct & indirect) to produce a product per industry, and per country [ICIO tables built from national IO tables, import tables, bilateral trade statistics, and others]

Estimates of units of inputs required per unit of total output (consumed domestically or exported) [from ICIO] will be coupled with the estimates of CL per unit of total output to measure the number of children supporting domestic & foreign demand for goods & services

INDICATORS e.g. estimate of CL linked for exports & domestic demand

In order to estimate the direct and indirect cases of child labour that are linked with gross exports, 32.32. InIn orderorder toto estimateestimate thethe directdirect andand indirectindirect casescases ofof childchild labourlabour thatthat areare linkedlinked withwith equation (1) can be rewritten as follows: grossgross exportsexports,, equationequation (1)(1) cancan bebe rewrittenrewritten asas followfollows:s:

where is~ a diagonlized matrix of !"#$%&, with zero elements………..(𝟐𝟐 )in the off diagonal.~ captures all wherewhere Bis is a a diagonlized diagonalized matrix matrix of Bof𝐒𝐒𝐒𝐒, !"#$%&with, with zero == elementszer 𝛒𝛒𝛒𝛒 𝐁𝐁o𝐁𝐁 elements 𝐅𝐅𝐅𝐅 ………in the..( 𝟐𝟐off )in diagonal. the off Bdiagonal. captures all the ca pturesdirect all the impacts.direct impacts. has the same has dimensions the same as dimensions in equation (1),as inbut equation only includes (1) ,the but direct only impact includes (i.e. the the direct impacts.S^ has the same dimensions as in equation (1), but only includes the directchildren impact𝐁𝐁𝐁𝐁 from (i.e. industry children i working from in industry the𝐁𝐁𝐁𝐁 production working of products in the exported production by industry of products i).𝐁𝐁𝐁𝐁 Therefore, exported direct impactdirect (i.e. children!"#$%& from industry working in the production of products exported 𝐒𝐒 !"#$%& byby industryindirectindustry impact ).). Therefore,Therefore, can 𝐒𝐒 be calculated indirectindirect as impact impactfollows: cancan bebe calculatedcalculated asas followfollows:s: 𝑖𝑖 𝑖𝑖 𝑖𝑖𝑖𝑖 Where is the off diagonal matrix!"#!$%&' of , with zero……….. (elements𝟑𝟑) in the main diagonal. WhereWhere isis thethe off off diagonal diagonal matrix matrix𝐒𝐒 𝐒𝐒of!"#!$%&' , with of zero==, 𝛒𝛒𝛒𝛒 withelements 𝐁𝐁𝐁𝐁 𝐅𝐅𝐅𝐅 zero……… in.. (theelements𝟑𝟑) main diagonal. in the main captures diagonal. all the captures allB the indirect impacts. B B captureindirects all impacts. the indirect impacts. 𝐁𝐁𝐁𝐁⃛ 𝐁𝐁𝐁𝐁 ⃛ 𝐁𝐁𝐁𝐁 33.33. AAdditionaldditional analysisanalysis waswas carriedcarried outout onon thethe structuralstructural layerlayer ddecompositionecomposition (i.e.(i.e. tiers).tiers). EstimatingEstimating thethe contributionscontributions ofof otherother upstreamupstream industriesindustries intointo exportingexporting industryindustry isis veryvery valuablevaluable toto givegive moremore insightinsight intointo thethe totaltotal impactimpact.. However,However, thethe impactimpact inin eacheach layerlayer ofof 𝑗𝑗𝑗𝑗 𝑖𝑖𝑖𝑖

14 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

DRAFT – FINAL Additional analysis was carried out on the structural layer decomposition (i.e. tiers). Estimating the contributions of other upstream industries j into exporting industry i is very valuable to give more productioninsight into is the vague total impact.. Using However, structural the impact layer in eachdecomposition layer of production technique, is vague. Usingwe ares­ tructural able to estimatelayer decomposition the amount, technique,for instance, we are of able child to estimatelabour in the each amount, layer for of instance, the production, of child labour and in with directeach and layer indirect of the production, information. and withThe direct Leontief and indirect inverse information. matrix The can Leontief be decomposed inverse matrix into B an infinitecan be number decomposed of layers into anwhere infinite number of layers where 𝐁𝐁

𝟐𝟐 𝟑𝟑 𝟒𝟒 In this paper, four layers have been estimated,𝟐𝟐 and𝟑𝟑 for completeness,𝟒𝟒 the rest of layers can In thisIn this paper, paper, four four layers layers havehave𝐁𝐁 beenbeen= 𝐈𝐈 + estimated,estimated,𝐀𝐀 + 𝐀𝐀 +and and 𝐀𝐀 for +f orcompleteness, 𝐀𝐀 completeness,+ ⋯ the rest the ofrest layers of layerscan be can be estimatedestimated as as

𝟐𝟐 𝟑𝟑 𝟒𝟒 𝟐𝟐 𝟑𝟑 𝟒𝟒 34. Thus, the direct and !"#$indirect impact by each tier of the production can be estimated 34. Thus,Thus, the direct the anddirect indirect and𝐁𝐁 !"#$impactindirect= by𝐁𝐁 impacteach− (𝐈𝐈 tier+ by 𝐀𝐀of + theeach𝐀𝐀 production +tier 𝐀𝐀 of+ thecan 𝐀𝐀 productionbe) estimated canas follows: be estimated as follows:

𝟐𝟐 𝟑𝟑 𝟒𝟒 𝟐𝟐 𝟑𝟑 𝟒𝟒 wherewhere Qrepresents represents aa setset of matricesmatrices that that hold ho theld the amount amount of child of !"#$childlabour labourassociated associated in each stage in each where represents a set𝐐𝐐 of= matrices𝛒𝛒(𝐈𝐈 + 𝐀𝐀 that+ 𝐀𝐀 ho+ld the𝐀𝐀 +amount 𝐀𝐀 + of 𝐁𝐁 !"#$child)𝐅𝐅 labour associated in each stageof theof theproduction production process. process. While in While some cases in some the first cases two-three the first tiers two of -threproductione tiers occupy of production almost 𝐐𝐐 occupythe whole 𝐐𝐐almost impact, the wholethis may impact, not necessarily this may be notthe case.necessarily be the case.

2.2. Assumptions and Limitations

35. It is important to recognise that the data and methodology both present limitations and that assumptions had to be made, and do impact the results. Data limitations for child labour include lack of data coverage for certain regions; lack of specific sectoral information in child labour surveys; limited coverage of children for some regions. Data limitations for trafficking for forced labour include the facts that forced labour is considered to be a rare event, statistically speaking and that methodologies to capture reliable prevalence numbers are recent. The availability of national datasets is lower than for child labour, and therefore less statistical confidence can be granted to the results in this technical paper for trafficking for forced labour. The main limitation for the human trafficking data was outlined earlier: data on identified cases of human trafficking are best understood as a sample of the unidentified population of victims (as for all administrative victim data collected by counter-trafficking organizations). This sample may be biased if some types of trafficking cases are more likely to be identified than others, but the extent of this bias is unknown. Nevertheless, there are few, if any, alternative sources of data on the distribution 22 of human trafficking by industry across countries.22 It is also necessary to reiterate that the trafficking for forced labour data that are “plugged” into the ICIO model are experimental estimates, based on the procedure described above. 36. There are also limitations related to the methodology. A number of assumptions had to be made due to data limitations. Notably, data was not available on the share of child labour between domestic and export markets by industry. Therefore, each unit of production in a given industry (whether it is part of global supply chains or not) is assumed to use the same amount of child labour. The implication of this assumption results in an underestimation of child labour in global supply chains in industries and countries where child labour is disproportionately concentrated in export production, and an overestimation in industries and countries where child labour is disproportionately concentrated in domestic production.

22 IOM’s GMDAC (2018) Data Bulletin Series: Informing the Implementation of the Global Compact for Migration. International Organization 22 IOM’s GMDAC (2018) Data Bulletin Series: Informing the Implementation of the Global Compact for Migration. International Organization for Migration (IOM): Geneva. Accessed on 12/07/2019 and available from https://publications.iom.int/system/files/pdf/gmdacbulletins.pdf. for Migration (IOM): Geneva. Accessed on 12/07/2019 and available from https://publications.iom.int/system/files/pdf/gmdacbulletins.pdf.

15

COMBINING THE DATASETS 15

2.2 ASSUMPTIONS AND LIMITATIONS

It is important to recognise that the data and methodology both present limitations and that assump- tions had to be made, and do impact the results. Data limitations for child labour include lack of data coverage for certain regions; lack of specific sectoral information in child labour surveys; limited coverage of children for some regions. Data limitations for trafficking for forced labour include the facts that forced labour is considered to be a rare event, statistically speaking and that methodologies to capture reliable prevalence numbers are recent. The availability of national datasets is lower than for child labour, and therefore less statistical confidence can be granted to the results in this tech- nical paper for trafficking for forced labour. The main limitation for the human trafficking data was outlined earlier: data on identified cases of human trafficking are best understood as a sample of the unidentified population of victims (as for all administrative victim data collected by counter-trafficking organizations). This sample may be biased if some types of trafficking cases are more likely to be identified than others, but the extent of this bias is unknown. Nevertheless, there are few, if any, alternative sources of data on the distribution of human trafficking by industry across countries (IOM, 2018). It is also necessary to reiterate that the trafficking for forced labour data that are “plugged” into the ICIO model are experimental estimates, based on the procedure described above.

There are also limitations related to the methodology. A number of assumptions had to be made due to data limitations. Notably, data were not available on the share of child labour between domestic and export markets by industry. Therefore, each unit of production in a given industry (whether it is part of global supply chains or not) is assumed to use the same amount of child labour. The implication of this assumption results in an underestimation of child labour in global supply chains in industries and countries where child labour is disproportionately concentrated in export production, and an overestimation in industries and countries where child labour is disproportionately concentrated in domestic production.

In the absence of detailed information, it is not possible to understand which countries and industries might suffer from underestimation or overestimation out of the whole sample. However, because it is well-known that has the most own-use production of goods (which by definition are not exported), we would expect this effect to be most pronounced in agriculture. FIGURE 5 looks at the difference between export and domestic markets through the lens of different industry aggregates. Agriculture and other services which are domestic- oriented (e.g. financial services) are presented next to total numbers as well as other industries like manufacturing which are more broadly linked with exports. There is variation between regions as to the impact this assumption may have on the broader analysis. In order to be more precise, separate data would be needed that differentiates firms into exporters and non-exporters, which is currently not available. Therefore, in the absence of detailed information, it is not possible to be precise on the effect of this assumption on the results either in terms of overestimation or underestimation. Future datasets in this regard could make the data more precise.

16 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

FIGURE 5: ESTIMATES OF CHILD LABOUR FOR EXPORTED GOODS AND SERVICES AND DOMESTIC DEMAND, BY REGION, BY DIFFERENT INDUSTRY AGGREGATES (2015)

100% 7% 9% 2% 12% 9% 14% 9% 12% 22% 26% 26% 31% 33% 39 47%

98% 93% 88% 91% 86% 91% 91% 88% 78% 74% 74% 69% 67% 61% 53%

0% TOTAL AGRI + REST TOTAL AGRI + REST TOTAL AGRI + REST TOTAL AGRI + REST TOTAL AGRI + REST SERVICES* SERVICES* SERVICES* SERVICES* SERVICES*

Sub-Saharan Latin America and Northern Eastern and South- Central and Africa the Caribbean and Western Asia Eastern Asia Southern Asia

Note: “Total” refers to Linkedtotal toindustries; domestic production “Agri and + consumptionservices” refers agricultureLinked toand exports some domestic services (utilities, construction, telecoms, IT services, finance & insurance, real estate, other business services, public administration, education, health, private households); “Rest” refers to mining, manufacturing and the main and remaining services which have tourism-like characteris- tics. For more information, see Annex 1. Source: Authors’ calculation based on (a) child labour data from the 65 country datasets used in the 2016 ILO Global Estimates of child Labour (including ILO-supported national surveys on child labour or child labour modules in national Labour Force Surveys; UNICEF-supported Multiple Indicator Cluster Surveys (MICS); and USAID-supported Demographic and Health Surveys); and (b) OECD Inter-Country Input-Output (OECD-ICIO) tables (2018 edition).

Another inherent assumption made is that outputs across the whole industry are produced in the same way with the same inputs. Difference in capital and labour intensity of production and the size of the firms involved across the industry are not fully accounted for, with the exemption of the fact that additional information on compensation of employees per industry was incorporated to avoid bias to a specific country (see section 1.2). Compensation of employees was envisioned as a proxy to provide an approximate figures of total employment per industry.

These are the reasons why the results in Chapter 1 of the Alliance 8.7 were not presented in absolute values. They should rather be considered as snapshots that allow more insight into the main charac- teristics of the phenomena linked to global supply chains in each region. They represent a starting point for further investigation and a foundation for cooperation and concerted action on the part of stakeholders of global supply chains. The reader should take into account the fact that the prevalence and extent of child labour (and forced labour and human trafficking) vary greatly across regions. Once more data are available by country and by industry, it will be possible to refine and update the results.

COMBINING THE DATASETS 17

2.3 ADDITIONAL ROBUSTNESS TESTS

Additional exercises were completed to analyse and probe the assumptions and limitations of the data and the methodology. Their purpose was to examine the extent to which detailed datasets may lead to a better estimates and results.

Aggregation vs Disaggregation – A perspective from a detailed national IO table

The extent to which detailed IO datasets lead to a better estimate of demand-based analyses has been investigated in literature (see Lenzen, 2011, for an example related to environmental indicators). Using a dataset available for Brazil in 2011, an additional test was conducted to look at how child labour was linked with domestic and foreign final demand using 149 industries. The difference between using this detailed set versus the 36 industries in the ICIO database was about 3%, with respective results for child labour linked with exports standing at 17.3% for the ICIO data (see PART 3) and 14.1% for the detailed national data. This difference is because of how domestic and foreign demand are considered and may be due to a misallocation of international trade (traded vs non-traded industries) in the 36 industry analyses. For instance, hotel and restaurant industries have the highest difference compared with other industries due to mostly serving domestic demands.

Regional national IO

Another perspective can be gained by using regional national IO tables. These tables give additional information on exports by each region within a country. Consider two regions (e.g. north and south) and consider that both produce the same products. However, production in the north involves chil- dren, while in the south it does not. In that case, if the production in the north purely serves the domestic market and if all the exports come from south, then there will be no prevalence of child labour linked in global supply chains. The opposite pattern holds for the south.

The regional national IO table for Brazil includes 27 regions and 149 industries for the year 2011 in purchasing prices. The ICIO and the detailed national IO is in basic prices. FIGURE 6 shows the differences between ICIO, detailed national IO, and the regional IO in Brazil. The differences stand at 17.3% for the ICIO data, 14.1% for the detailed national data, and 14.2% for the regional national IO. The impact of the fact that regional national IO is in purchasing prices is only expected to be felt in margin-based industries like wholesale and retail.

18 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

FIGURE 6: COMPARISON OF CL LINKED FOR EXPORTS BETWEEN ICIO, MORE DETAILED NATIONAL IO, AND REGIONAL NATIONAL IO DATASETS – BRAZIL (2011)

250000

200000

150000

100000

50000

0 FIN ITS TEL ELQ FOD TEX PET CEQ TRQ TSP REA OBZ OTS PAP HTR EDU AGR RBP PVB GOV HTH CON PVH FBM MET MEQ MTR OTM MNE EGW WRT MNS WOD CHM MNN MMM

PER71% CHILD From ICIO From national IO From regional IO

Note: Results driven from the detailed and regional tables were aggregated into 36 ICIO industries for comparison. More information about industry names is available in Annex 1. Source: a) OECD-ICIO model, b) The Brazilian Institute of Geography and Statistics (IBGE) c) Guilhoto et. al., (2019).

PART 3 19 PART 3: RESULTS

Chapter 1 of the Alliance 8.7 report presents the results of the analysis for both child labour and trafficking for forced labour.

3.1 CHILD LABOUR

The extent to which child labour within a region is estimated to contribute to exports to other regions varies across regions (see FIGURE 7). While the results demonstrate that a child in child labour is far more likely to be involved in production for the domestic economy, there is a non-negligible risk that this child will be contributing to global supply chains, with 9 to 26 per cent of child labour estimated to be linked to exports across regions. Furthermore, a narrow focus on eliminating child labour, forced labour and human trafficking within the production settings that form part of global supply chains – without addressing the common set of legal gaps and socio-economic pressures at their root – risks simply displacing the abuses into sectors of the local economy that are not linked to global supply chains, meaning in turn that our ultimate goal, ending all forms of child labour, forced labour and human trafficking, regardless of where they occur, would be no closer. While the unique complexities of global supply chains present special challenges, efforts to end child labour, forced labour and human trafficking in global supply chains cannot be divorced from broader efforts towards ending these abuses generally.

Regional variation also exists in terms of whether child labour is disproportionately concentrated in industries that contribute to global supply chains. As shown below, part of the child labour estimated to contribute to exports is contributed indirectly through upstream industries, making due diligence efforts and visibility/traceability challenging. While in most regions, the proportion of children esti- mated to be linked to exports is close to the proportion of value added that is exported, this is not always the case. For instance, in Latin America and the Caribbean, 16% of value added goes to the export sector, against 22% of child labour. This could suggest that the sectors in which children work tend to be export sectors more often than in other regions, or that the value added in the export sector is relatively low compared to the domestic sector. These results should be considered with the caveats mentioned above regarding the distribution of child labour between domestic and export markets.

20 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

FIGURE 7: ESTIMATES OF CHILD LABOUR AND VALUE ADDED FOR EXPORTED GOODS AND SERVICES, AND DOMESTIC DEMAND, BY REGION (2015)

100% 12% 9% 12% 15% 16% 18% 15% 22% 26% 23%

88% 91% 88% 85% 84% 82% 85% 78% 74% 77% CHILD VALUE LABOUR ADDED

Linked to exports

Linked to domestic production 0% and consumption CHILD VALUE CHILD VALUE CHILD VALUE CHILD VALUE CHILD VALUE LABOUR ADDED LABOUR ADDED LABOUR ADDED LABOUR ADDED LABOUR ADDED

Sub-Saharan Latin America and Northern Africa Eastern and South- Central and Africa the Caribbean and Western Asia Eastern Asia Southern Asia

Source: Authors’ calculation based on (a) Child labour data from the 65 country datasets used in the 2016 ILO Global Estimates of Child Labour (including ILO-supported national surveys on Child labour or Child labour modules in national Labour Force Surveys; UNICEF-supported Multiple Indicator Cluster Surveys (MICS); and USAID-supported Demographic and Health Surveys); and (b) OECD Inter-Country Input-Output (OECD-ICIO) tables (2018 edition); (c) Value Added data from OECD (Annual National Accounts & Structural Analysis Databases), UN main aggregates and UN National Accounts Official Country Data.

The empirical analysis also provides insights into where child labour is concentrated along supply chains. The results in FIGURE 8 indicate that, across regions, between 28 and 43 per cent of the child labour estimated to contribute to exports does so indirectly, through preceding tiers of the supply chain (e.g. extraction of raw materials or agriculture). The values for each region represent the aggregation of countries with available trafficking for forced labour data. In other words, a child in child labour who is contributing to exports in Eastern and South-Eastern Asia is more likely to be contributing to exports indirectly, in preceding tiers of the supply chain, than a child in child labour in other regions. Nevertheless, across all regions, there is a significant risk that a child in child labour who is contributing to exports will be contributing indirectly, in upstream industries of the supply chain where risk may be more difficult to identify and mitigate. These results make clear that efforts against child labour in global supply chains will be inadequate if they do not extend beyond immediate suppliers, i.e. downstream suppliers closer to final production and also cover actors in preceding tiers of supply chains, including those involved in upstream production activities such as raw material extraction and agriculture serving as inputs to other industries.

RESULTS 21

FIGURE 8: ESTIMATES OF CHILD LABOUR FOR EXPORTED GOODS AND SERVICES, DIRECT AND INDIRECT, BY REGION (2015)

100%

20% 31% 29% 28% 37% 38% 40% 39% 43% 40%

80% CHILD VALUE 69% 71% 72% LABOUR ADDED 60% 60% 61% 57% 63% 62%

Indirect

Direct 0% CHILD VALUE CHILD VALUE CHILD VALUE CHILD VALUE CHILD VALUE LABOUR ADDED LABOUR ADDED LABOUR ADDED LABOUR ADDED LABOUR ADDED

Sub-Saharan Latin America and Northern Africa Eastern and South- Central and Africa the Caribbean and Western Asia Eastern Asia Southern Asia

Source: Authors’ calculation based on (a) Child labour data from the 65 country data sets used in the 2017 ILO Global Estimates of Child Labour (including ILO-supported national surveys on Child labour or Child labour modules in national Labour Force Surveys, UNICEF-supported Multiple Indicator Cluster Surveys (MICS), and USAID-supported Demographic and Health Surveys); (b) OECD ICIO tables (2018 edition); and (c) value added data from the OECD (Annual National Accounts and Structural Analysis Databases), United Nations main aggregates and United Nations national accounts official country data.

22 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

3.2 TRAFFICKING FOR FORCED LABOUR

FIGURE 9 indicates that the share of trafficking for forced labour contributing to exports varies across regions. Across all regions, the estimated share of trafficking for forced labour present in exports is lower than the share of value added these industries contributed to exports. This means that indus- tries with higher trafficking for forced labour prevalence are less likely to contribute to global supply chains. Nevertheless, a non-negligible part of trafficking for forced labour does contribute to global supply chains and further industry-level analysis and comparison is needed to better understand and address the risks.

Pending further industry-level analysis, these results are partly driven by the role played specifically by trafficking into construction and support services such as domestic work and cleaning. Domestic work is often not well captured by ICIO tables and does not contribute to value added production processes but rather supports the activities of private households. While construction clearly plays a role in value added production processes, it is a structural zero in terms of its direct contribution of exports due to the way it is defined and recorded by national accounts for national IO tables.

FIGURE 9: ESTIMATES OF TRAFFICKING FOR FORCED LABOUR AND VALUE ADDED FOR EXPORTED GOODS AND SERVICES, BY REGION (2015)

100% 4% 9% 10% 16% 24% 9% 17% 17% 28% 26%

96% 91% 90% 91% 83% 83% 84% 76% TRAFFICKING VALUE 74% FOR FORCED ADDED 72% LABOUR

Linked to exports

Linked to domestic production 0% and consumption TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED LABOUR LABOUR LABOUR LABOUR LABOUR Sub-Saharan Northern Northern Africa Eastern and South Europe Africa America and Western Asia -Eastern Asia

Sources: Author’s calculation using (a) CTDC non-k-anonymized data between 2006 and 2016; (b) results of the Alliance 8.7 Global Estimates of Modern Slavery; (c) ILO Harmonized Microdata (by Industry); (d) OECD Inter-Country Input-Output (OECD-ICIO) tables (2018 edition); and (e) Value Added data from OECD (Annual National Accounts & Structural Analysis Databases), UN main aggregates and UN National Accounts Official Country Data.

RESULTS 23

Preliminary results from FIGURE 10 show that, across all regions, there is significant risk that a person trafficked for forced labour who is contributing to exports will be contributing indirectly, in upstream industries, where risk may be more difficult to identify and mitigate.

Assessing trafficking for forced labour contextualised with the value added contributing indirectly to exports indicate different regional patterns. Across all regions, while the levels of indirect value added in exports are similar, there is great variation in the estimate of trafficking for forced labour indirectly exported. These differences could be explained by the fact that trafficking for forced labour is concentrated in specific industries and regions.

FIGURE 10: ESTIMATED TRAFFICKING FOR FORCED LABOUR AND VALUE ADDED FOR EXPORTED GOODS AND SERVICES, DIRECT AND INDIRECT, BY REGION (2015)

100%

29% 29% 35% 21% 37% 41% 38% 38% 52% 65%

TRAFFICKING VALUE FOR FORCED ADDED 71% 71% 69% LABOUR 65% 63% 59% 62% 62% 48% Indirect 35%

Direct 0% TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE TRAFFICKING VALUE FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED FOR FORCED ADDED LABOUR LABOUR LABOUR LABOUR LABOUR Sub-Saharan Northern Northern Africa Eastern and South Europe Africa America and Western Asia -Eastern Asia

Based on (a) Counter-Trafficking Data Collaborative non-k-anonymized data between 2006 and 2016; (b) the 2017 ILO-Walk Free Foundation Global Estimates of Modern Slavery; (c) ILO harmonized microdata (by industry); (d) OECD ICIO tables (2018 edition); and (e) value added data from OECD (Annual National Accounts and Structural Analysis databases), United Nations main aggre- gates and United Nations national accounts official country data.

24 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

CONCLUSIONS

The results show that child labour and trafficking for forced labour is a problem affecting the whole of the global supply chain, and that a significant share of child labour and trafficking for forced labour occurs upstream, in the production of raw materials and other inputs to final exports products, making due diligence efforts, including visibility and traceability, challenging. Across regions, between 28 and 43 per cent of the child labour estimated to contribute to exports does so indirectly, through preceding tiers of the supply chain (such as extraction of raw materials or agriculture). Company due diligence beyond immediate suppliers could thus present one of the most significant opportunities to eradicate these abuses.

The results presented in the report break new ground by offering an initial quantitative picture of the presence of child labour and trafficking for forced labour in global supply chains. They also provide a critical foundation for further data collection efforts aimed at generating a more granular picture of the extent, nature and location of these violations in global supply chains. Collaboration with the private industry, and among governments, social partners and other stakeholders, can further enhance data availability and transparency as well as promote the harmonization of statistical standards and tools, and thereby also contribute to devising better-targeted approaches. 25

BIBLIOGRAPHY

Alsamawi, A.; et al. 2017. The social footprints of global trade (Springer, New , USA). Available at: http://doi. org/10.1007/978-981-10-4137-2.

Alsamawi, A.; Murray, J., and Lenzen, M. 2014. “The Employment Footprints of Nations”, in Journal of Industrial Ecology, Vol. 18, Issue 1, pp. 59-70. Available at: http://doi.org/10.1111/jiec.12104. Feng, K.; et al. 2011. “Comparison of bottom-up and top-down approaches to calculating the water footprints of nations”, in Economic Systems Research, Vol. 23, Issue 4, pp. 371-385. Available at: https://doi.org/10.1080/0 9535314.2011.638276. Gomez‐Paredes, J.; et al. 2016. “Consuming childhoods: An assessment of child labor’s role in Indian production and global consumption”, in Journal of Industrial Ecology, Vol. 20, Issue 3, pp. 611–622. Available at: https://doi. org/10.1111/jiec.12464. Guilhoto, J.J.M. ; et al. (2019) “Sistema Interestadual de Insumo-Produto do Brasil, Estimação em Condições de Informação Limitada para o ano de 2011: Uma Aplicação do Método SUIT”, in Economia Aplicada, in printing. Haberl, H.; et al. 2007. “Quantifying and mapping the human appropriation of net primary production in earth’s terrestrial ecosystems”, in Proceedings of the National Academy of Sciences of the United States of America, Vol. 104, Issue 31, pp. 12942–12947. Available at: https://doi.org/10.1073/pnas.0704243104.

Hertwich, E. G. and Peters, G.P. 2009. “Carbon footprint of nations: A global, trade-linked analysis”, in Environmental Science & Technology, Vol. 43, Issue 16, pp. 6414–6420. Available at: https://doi.org/10.1021/es803496a. ILO. 2008. Resolution concerning statistics of child labour, 18th International Conference of Labour Statisticians (ICLS), Geneva, November-December 2008 (Geneva). Available at: www.ilo.org/wcmsp5/groups/public/---dgre- ports/---stat/documents/normativeinstrument/wcms_112458.pdf.

ILO. 2017. Global estimates of child labour: Results and trends, 2012-2016 (Geneva). Available at: www.ilo.org/ wcmsp5/groups/public/---dgreports/---dcomm/documents/publication/wcms_575499.pdf.

ILO. 2018a. Measurement of forced labour, 20th International Conference of Labour Statisticians (ICLS), Geneva, 10-19 October 2018, ICLS/20/2018/Room document 14 (Geneva). Available at: www.ilo.org/wcmsp5/groups/ public/---dgreports/---stat/documents/meetingdocument/wcms_636050.pdf.

ILO. 2018b. Resolution to amend the 18th ICLS Resolution concerning statistics of child labour, 20th International Conference of Labour Statisticians (ICLS), Geneva, 10-19 October 2018, ICLS/20/2018/Resolution IV (Geneva). Available at: www.ilo.org/wcmsp5/groups/public/---dgreports/---stat/documents/meetingdocument/ wcms_667558.pdf.

ILO and Walk Free Foundation. 2017. Global Estimates of Modern Slavery: Forced Labour and , in collaboration with International Organisation for Migration (Geneva, ILO). Available at: www.ilo.org/wcmsp5/ groups/ public/---dgreports/---dcomm/documents/publication/wcms_575479.pdf.

IOM. 2016. Report on Human Trafficking, Forced Labour and Fisheries Crime in the Indonesian Fishing Industry (Geneva). Available at: www.verite.org/wp-content/uploads/2016/11/Verite-Report-Illegal_Gold_Mining-2.pdf.

IOM. 2018. Data Bulletin Series: Informing the Implementation of the Global Compact for Migration (Geneva). Available at: https://publications.iom.int/system/files/pdf/gmdacbulletins.pdf.

Kizu, T.; Kuhn, S. and Viegelahn, C. 2016. Linking jobs in global supply chains to demand, ILO Research Paper No. 16 (Geneva, International Labour Office). Available at: www.ilo.org/wcmsp5/groups/public/---dgreports/--- inst/ documents/publication/wcms_512514.pdf.

Lenzen, M. 2011. “Aggregation versus disaggregation in input–output analysis of the environment”, in Economic Systems Research, Vol. 23, Issue 1, pp. 73-89. Available at: https://doi.org/10.1080/09535314.2010.548793. Miller, R.E. and Blair, P.D. 2010. Input-output analysis: Foundations and extensions (Englewood Cliff, Prentice-Hall, New Jersey).

26 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

UNODC. 2016. Multiple Systems Estimation for estimating the number of victims of human trafficking across the world, Research Brief (Paris.) Available at: www.unodc.org/documents/data-and-analysis/tip/TiPMSE.pdf.

Verité. 2016. The Nexus of Illegal Gold Mining Supply Chains: Lessons from Latin America (Massachusetts). Available at: www.verite.org/wp-content/uploads/2016/11/Verite-Report-Illegal_Gold_Mining-2.pdf.

Wiebe, K. and Yamano N. 2016. Estimating CO2 Emissions Embodied in Final Demand and Trade Using the OECD ICIO 2015: Methodology and Results, OECD Science, Technology and Industry Working Papers, No. 2016/05 (OECD Publishing, Paris). Available at: www.oecd-ilibrary.org/docserver/5jlrcm216xkl-en.pdf?expires=1574677697&id=i d&accname=guest&checksum=0645E593EF31E7950ED669DF72D67308.

Wiedmann, T.; et al. 2013. “The material footprint of nations. Reassessing resource productivity”, in Proceedings of the National Academy of Sciences of the United States of America, Vol. 110, Issue 36, pp. 1–6. Available at: https:// doi.org/10.1073/pnas.1220362110. 27

ANNEXES

ANNEX 1. OECD – ICIO 2018 GEOGRAPHICAL COVERAGE AND INDUSTRY LIST

1 AUS Australia 23 NLD Netherlands 45 COL Colombia

2 AUT Austria 24 NZL New Zealand 46 CRI Costa Rica

3 BEL Belgium 25 NOR Norway 47 HRV Croatia

4 CAN Canada 26 POL Poland 48 CYP Cyprus

5 CHL Chile 27 PRT Portugal 49 IND

6 CZE Czechia 28 KOR Republic of Korea 50 IDN

7 DNK Denmark 29 SVK Slovakia 51 KAZ Kazakhstan

8 EST Estonia 30 SVN Slovenia 52 MYS Malaysia

9 FIN Finland 31 ESP Spain 53 MLT Malta

10 FRA France 32 SWE Sweden 54 MAR

11 DEU Germany 33 CHE Switzerland 55 PER Peru

12 GRC Greece 34 TUR Turkey 56 PHL Philippines

13 HUN Hungary 35 GBR United Kingdom 57 ROU Romania

14 ISL Iceland 36 USA United States of America 58 RUS Russian Federation

15 IRL Ireland 37 ARG 59 SAU Saudi Arabia

16 ISR Israel 38 BRA Brazil 60 SGP Singapore

17 ITA Italy 39 BRN Brunei Darussalam 61 ZAF South Africa

18 JPN Japan 40 BGR Bulgaria 62 THA Thailand

19 LVA Latvia 41 KHM Cambodia 63 TUN Tunisia

20 LTU Lithuania 42 CHN China 64 VNM Viet Nam

21 LUX Luxembourg 43 HKG China, Hong Kong SAR 65 ROW Rest of the World

Taiwan Province of the 22 MEX Mexico 44 TWN People’s Republic of China

28 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

3-char Code ISICA Labels Short labels code 0 DTOTAL TOTAL Total TOT 1 D01T03 01,02,03 Agriculture, hunting, forestry and fishing Agriculture AGR Mining and extraction of energy producing 2 D05T06 05,06 Mining, energy MNE products Mining and quarrying of non-energy 3 D07T08 07,08 Mining, non-energy MNN producing products 4 D09 09 Services to mining and quarrying Mining, services MNS 5 D10T12 10,11,12 Food products, beverages and Food products FOD Textiles, textile products, leather and 6 D13T15 13,14,15 Textiles & apparel TEX footwear 7 D16 16 Wood and products of wood and cork Wood WOD 8 D17T18 17,18 Paper products and printing Paper & printing PAP 9 D19 19 Coke and refined petroleum products Coke & petroleum PET 10 D20T21 20,21 Chemicals and pharmaceutical products Chemicals CHM 11 D22 22 Rubber and plastics products Rubber & plastics RBP 12 D23 23 Other non-metallic mineral products Non-metal minerals NMM 13 D24 24 Basic metals Basic metals MET 14 D25 25 Fabricated metal products Fabricated metals FBM 15 D26 26 Computers, electronic and optical products ICT & electronics CEQ 16 D27 27 Electrical equipment Electrical equipment ELQ 17 D28 28 Machinery and equipment, nec Machinery MEQ 18 D29 29 Motor vehicles, trailers and semi-trailers Motor vehicles MTR 19 D30 30 Other transport equipment Other transport TRQ Manufacturing nec; repair of machinery 20 D31T33 31,32,33 Other manufacturing OTM and equipment Electricity, gas, water supply, sewerage, 21 D35T39 35to39 Utilities EGW waste and remediation services 22 D41T43 41,42,43 Construction Construction CON Wholesale and retail trade; repair of motor 23 D45T47 45,46,47 Wholesale & retail WRT vehicles 24 D49T53 49to53 Transport, storage and postal services Transport & storage TSP 25 D55T56 55,56 Accommodation and food services Accommodation & food HTR Publishing, audiovisual and Publishing & broad- 26 D58T60 58,59,60 PVB broadcasting activities casting 27 D61 61 Telecommunications Telecoms TEL 28 D62T63 62,63 IT and other information services IT services ITS 29 D64T66 64,65,66 Financial and insurance activities Finance & insurance FIN 30 D68 68 Real estate activities Real estate REA 31 D69T82 69to82 Other business sector services Other business services OBZ Public admin. and defence; compulsory 32 D84 84 Public admin GOV social security 33 D85 85 Education Education EDU 34 D86T88 86,87,88 Health and social work Health HTH Arts, entertainment, recreation and other 35 D90T96 90to96 Other services OTS personal service activities 36 D97T98 97,98 Private households with employed persons Private households PVH

ANNEXES 29

ANNEX 2. LIST OF REGIONS, COUNTRY COVERAGE, AND TARGET POPULATION COVERAGE

A. REGIONAL GROUPINGS, FOLLOWING THE UNITED NATIONS STATISTICAL DIVISION’S STANDARD COUNTRY OR AREA CODES FOR STATISTICAL USE (M49)*15

Sub-Saharan Africa

Angola Madagascar Botswana Burundi Mauritania Cabo Verde Mauritius Mayotte Mozambique Namibia Comoros Niger Congo Nigeria Côte d’Ivoire Réunion Democratic Republic of the Congo Rwanda Djibouti Sao Tome and Principe Equatorial Guinea Senegal Eritrea Seychelles Eswatini Gabon South Africa Gambia, the South Sudan Ghana Guinea Guinea-Bissau United Republic of Tanzania Kenya Zambia Zimbabwe

Northern Africa and Western Asia

Algeria Libya Armenia Morocco Azerbaijan Oman Bahrain Qatar Cyprus Saudi Arabia Egypt State of Palestine Georgia Sudan Iraq Syrian Arab Republic Israel Tunisia Jordan Turkey

* Taken from https://unstats.un.org/sdgs/indicators/regional-groups/, last accessed on August 8, 2019.

30 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

Kuwait United Arab Emirates Lebanon Yemen

Central and Southern Asia

Afghanistan Maldives Bhutan Pakistan India Sri Lanka Iran (Islamic Republic of) Tajikistan Kazakhstan Turkmenistan Kyrgyzstan

Eastern and South-Eastern Asia

Brunei Darussalam Malaysia Cambodia Mongolia China China, Hong Kong SAR Philippines China, Macao SAR Republic of Korea Democratic People’s Republic of Korea Singapore Indonesia Thailand Japan Timor-Leste Lao People’s Democratic Republic Viet Nam

Latin America and the Caribbean

Anguilla Guadeloupe Antigua and Barbuda Guatemala Argentina Guyana Aruba Haiti Bahamas Honduras Barbados Jamaica Belize Martinique Bolivia (Plurinational State of) Mexico Bonaire, Sint Eustatius and Saba Montserrat Brazil Nicaragua British Virgin Islands Panama Cayman Islands Paraguay Chile Peru Colombia Puerto Rico Costa Rica Saint Kitts and Nevis Cuba Saint Lucia Curaçao Saint Vincent and the Grenadines Dominica Sint Maarten (Dutch part) South Georgia & the South Sandwich Islands Suriname El Salvador Trinidad and Tobago Falkland Islands (Malvinas) Turks and Caicos Islands French Guiana United States Virgin Islands Grenada Uruguay Venezuela (Bolivarian Republic of)

ANNEXES 31

Northern America

Bermuda Greenland Canada United States of America

Europe

Åland Islands Latvia Albania Liechtenstein Andorra Lithuania Austria Luxembourg Belarus Malta Belgium Monaco Bosnia and Herzegovina Montenegro Bulgaria Netherlands, the Channel Islands Norway Croatia North Macedonia Czechia Poland Denmark Portugal Estonia Republic of Moldova Faroe Islands Romania Finland Russian Federation France San Marino Germany Serbia Greece Slovakia Hungary Slovenia Iceland Spain Ireland Sweden Isle of Man Switzerland Italy Ukraine United Kingdom of Great Britain and Northern Ireland

Oceania

Australia Papua New Guinea Fiji Solomon Islands New Caledonia Micronesia New Zealand Polynesia

32 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

B. LIST OF COUNTRIES USED AS UNDERLYING DATA FOR REGIONAL ESTIMATES

Child labour in global supply chains

• Central and Southern Asia: Afghanistan*, Bangladesh, Bhutan*, India, Kyrgyzstan, Nepal*, and Pakistan • Eastern and South-Eastern Asia: Cambodia, Indonesia, Lao People’s Democratic Republic, Mongolia, Philippines, Timor-Leste, and Viet Nam • Latin America and the Caribbean: Argentina*, Barbados*, Brazil, Chile, Colombia, Dominican Republic*, Ecuador, El Salvador, Haiti*, Jamaica*, Mexico, Nicaragua, Panama, Peru, Saint Lucia*, Suriname*, and Venezuela (Bolivarian Republic of) • Sub-Saharan Africa: Benin*, Burkina Faso*, Burundi*, Cabo Verde, Cameroon*, Central Africa Republic*, Chad*, Comoros*, Congo*, Côte d’Ivoire*, Democratic Republic of Congo*, Eswatini*, Ethiopia, Gabon*, , Ghana, Liberia, Malawi, Mali*, Mauritania*, the Niger*, Nigeria*, Senegal*, Sierra Leone*, South Sudan, Togo*, Uganda, United Republic of Tanzania • Northern Africa and Western Asia: Armenia, Egypt*, Georgia, Iraq*, Tunisia*, and Yemen*

Additional assumptions regarding sectoral distribution had to be made for the countries marked with * due to the lack of this information in the original micro datasets. Usually those are the cases in which MICS or DHS datasets were used.

Trafficking for forced labour in global supply chains

• Sub-Saharan Africa: Ethiopia, Ghana, Mali, Mozambique, Rwanda, Sierra Leone, and Uganda • Northern Africa and Western Asia: Armenia, Egypt, Georgia, Sudan, Turkey, United Arab Emirates, and Yemen • Eastern and South-Eastern Asia: Indonesia, Lao People’s Democratic Republic, Philippines, Thailand, and Timor-Leste • Northern America: United States of America • Europe: Austria, Bosnia and Herzegovina, Czechia, France, Greece, Italy, Portugal, Serbia, Switzerland, and United Kingdom of Great Britain and Northern Ireland

ANNEXES 33

C. SOURCES OF SURVEY DATA FOR CHILD LABOUR

Country Year Survey Afghanistan 2011 Multiple indicator cluster survey Argentina 2012 Multiple indicator cluster survey Armenia 2015 National child labour survey Bangladesh 2013 Labour force and child labour survey Barbados 2012 Multiple indicator cluster survey Benin 2011 Enquête modulaire intégrée sur les conditions de vie des ménages Bhutan 2010 Multiple indicator cluster survey Brazil 2015 Pesquisa nacional por amostra de domicilios Burkina Faso 2010 Enquête démographique et de santé et à indicateurs multiples Burundi 2010 Enquête démographique et de santé Cabo Verde 2012 Inquérito nacional sobre as actividades das criancas Cambodia 2012 Labour force and child labour survey Cameroon 2011 Enquête démographique et de santé et à indicateurs multiples Central African Republic 2010 Multiple indicator cluster survey Chad 2010 Multiple indicator cluster survey Chile 2012 Encuesta nacional de actividades de niños, niñas y adolescentes Colombia 2014 Gran encuesta integrada de hogares Comoros 2012 Enquête démographique et de santé et à indicateurs multiples Congo 2012 Enquête démographique et de santé Côte d’Ivoire 2012 Demographic and health survey Democratic Republic of the Congo 2014 Enquête démographique et de santé Dominican Republic 2011 Encuesta nacional de hogares de propósitos múltiples Ecuador 2012 Encuesta nacional de trabajo infantil Egypt 2014 Enquête démographique et de santé El Salvador 2015 Encuesta de hogares de propósitos múltiples Eswatini 2010 Multiple indicator cluster survey Ethiopia 2015 National child labour survey Gabon 2012 Enquête démographique et de santé Gambia, the 2012 Labour force survey Georgia 2015 Labour force survey Ghana 2013 Ghana living standards survey round 6 Haiti 2012 Enquête mortalité, morbidité et utilisation des services India 2012 National sample survey round 68 Indonesia 2009 Labour force and child labour survey Iraq 2011 Multiple indicator cluster survey Jamaica 2011 Multiple indicator cluster survey Kyrgyzstan 2014 National child labour survey Lao People’s Democratic Republic 2010 Labour force and child labour survey Liberia 2010 Labour force survey Malawi 2015 National child labour survey Mali 2013 Enquête démographique et de santé Mauritania 2011 Multiple indicator cluster survey Mexico 2015 Encuesta nacional de ocupación y empleo Mongolia 2012 Labour force and child labour survey Nepal 2014 Multiple indicator cluster survey Nicaragua 2012 Encuesta continua de hogares Niger, the 2012 Enquête démographique et de santé et à indicateurs multiples Nigeria 2011 Multiple indicator cluster survey Pakistan 2011 Labour force survey Panama 2014 Encuesta del mercado laboral Peru 2015 Encuesta sobre trabajo infantil Philippines 2011 Labour force and child labour survey

34 Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach

Saint Lucia 2012 Multiple indicator cluster survey Senegal 2014 Enquête démographique et de santé continue Sierra Leone 2013 Demographic and health survey South Sudan 2008 Population and housing census Suriname 2010 Multiple indicator cluster survey Timor-Leste 2016 National child labour survey Togo 2014 Enquête démographique et de santé Tunisia 2012 Multiple indicator cluster survey Uganda 2012 Labour force and child labour survey United Republic of Tanzania 2014 National child labour survey Venezuela (Bolivarian Republic of) 2012 Encuesta de hogares por muestreo Viet Nam 2012 Labour force and child labour survey Yemen 2013 National health and demographic survey

D. COVERAGE IN TERMS OF TARGET POPULATION

Child Labour

Region Population living Total in countries with population survey data in the region (5-17 years old, (5-17 years old, thousands) thousands) Coverage Central and Southern Asia 450 000 483 000 93% Eastern and South-Eastern Asia 113 000 391 000 29% Latin America and the Caribbean 123 000 140 000 88% Northern Africa and Western Asia 46 000 119 000 39% Sub-Saharan Africa 235 000 320 000 73% World (including other regions) 967 000 1 619 000 60%

Trafficking for Forced Labour

Region Population Total population coverage in the region (15-64 years old, (15-64 years old, thousands) thousands) Coverage Eastern and South-Eastern Asia 292 000 1 592 000 18% Europe 162 000 495 000 33% Northern America 212 000 237 000 89% Northern Africa and Western Asia 158 000 308 000 51% Sub-Saharan Africa 123 000 517 000 24% World (including other regions) 947 000 4 848 000 20%

Population based on UN Population Prospects 2019, for the year 2015. Measuring child labour, forced labour and human trafficking in global supply chains: A global Input-Output approach Technical Paper

Ali ALSAMAWI Tihana BULE www.Alliance87.org Claudia CAPPA Alliance8_7 Harry COOK Claire GALEZ-DAVIS #Achieve87 Gady SAIOVICI