Child Labor

OWID presents work from many different people and organizations. When citing this entry, please also cite the original data source. This entry can be cited as:

Esteban Ortiz-Ospina and Max Roser (2016) – ‘Child Labor’. Published online at Retrieved from: [Online Resource]

The International Labour Organisation states in its latest World Report on Child Labour (2013) that there are around 265 million working children in the world—almost 17 per cent of the worldwide child population. According to the publicly available data discussed in more detail below, Sub-Saharan Africa is the region where child labour is most prevalent. While absolute numbers are still high, particularly in those countries with the lowest standards of living, from a historical viewpoint there are concrete examples of countries that managed to virtually eliminate widespread child labour in the course of a century. The United Kingdom is a case in point. In terms of recent developments, on the aggregate trends show a significant reduction in child labour over the last two decades, albeit with wide dispersion in the progress that different countries have achieved.

# Empirical View

# Historical perspective

Historical studies suggest that child work was widespread in Europe in the 19th century, but declined very rapidly at the turn of the 20th century. The available historical evidence seems consistent with the fact that industrialisation in western European countries initially increased the demand for child labour, but then eventually contributed towards its elimination (see Cunningham, H., & Viazzo, P. P. (1996) 1 and the references therein). The following two visualizations show the share of children in employment for Italy and the UK at the turn of the 20th century century.

# Global overview today

A complete and up-to-date overview of recent global trends in child labour can be found in the ILO’s report Marking Progress Against Child Labour (2013)2 produced by the organization’s International Programme on the Elimination of Child Labour (IPEC). This report presents global estimates and trends for the period 2000-2012. The following visualization, based on this source, presents the recent changes in the world-wide share of children (ages 5-17) in employment.

child-labor-global-post-2000view the interactive version of the same chart (with access to the data)

Independently, the ILO Programme on Estimates and Projections of the Economically Active Population (EPEAP) has been producing statistics on labour force participation (for adults and children) since 1950, through the ILO’s cross-country database known as LABORSTA. Basu (1999)3 uses this source to produce global labour force participation rates for children (ages 10-14) in the period 1950-1995. The following visualization presents the corresponding trend using the data published in Basu (1999). While these estimates are informative about child labour, they cannot be linked directly to those of children in employment published by the ILO IPEC for the period 2000-2012 due to issues of comparability; specifically, the IPEC and EPEAP estimates discussed above rely on different survey instruments covering a different set of countries, and break up the relevant population in different age brackets.4 Nonetheless, regardless of discrepancies between these two sources, the trends tell a consistent story: the last fifty years have seen a marked improvement in terms of the share of economically active children in the world.

Contrary to popular perception, most working children in the world are unpaid family workers, rather than paid workers in manufacturing establishments or other forms of wage employment. The following visualization (Figure 9 in Marking Progress Against Child Labour (2013)) shows a breakdown of 2012 global estimates of child labour by employment status.

Breakdown of ILO-IPEC global estimates of child labour by employment status in 2012- Figure 9 in Marking Progress Against Child Labour (2013)

Breakdown of ILO's 2012 global estimates of child labour by employment status

Schultz and Strauss (2008)5 compile information from a number of different sources (mostly country-specific datasets from national statistics offices—see the original paper for detailed sources) to provide a picture of the industrial composition of economically active children. The following table (numbered as table 5 in Schultz and Strauss (2008)) presents their results. In almost every listed country, a majority of economically active children work in agriculture, forestry, or fishing.

# Industrial composition of economically active children – Table 5 in Schultz and Strauss (2008)6
Industrial composition of economically active children, compiled by Schultz and Strauss (2008)

# Recent developments

The visualization below shows the share of children (7-14 years) in employment for a number of countries (for the years in which data is publicly available from the World Bank consolidated dataset). As it can be appreciated, the prevalence of child labour varies widely by country; for instance, the share of children in employment (here defined in terms of being economically active for one hour a week) was fifteen times larger in Uganda than in Turkey according to 2005 estimates. While most countries exhibit a downward trend, many countries are lagging.

Switch to the map view in this chart to compare the level of child labour between countries.  Sub-Saharan Africa is the region where child labour is most prevalent, and also the region where progress has been slowest and least consistent.

As shall be discussed in more detail, child labour is by definition problematic whenever it interferes with the children’s development. Because of this it is informative to study child labour specifically when it is coupled with absence from school. The following visualization shows the share of children in employment who work only (i.e. those children who are economically active and do not attend school). Again, there is wide variation across countries; while in Latin America the majority of children who are economically active also attend school, in sub-Saharan Africa this is not the case. However, trends are encouraging on the whole, particularly in sub-Saharan Africa where the problem is most acute. The next section exploring correlates, determinants and consequences of child labour, provides more information about the link between work and school attendance.

# Current gender gaps

In the majority of countries boys are more likely than girls to be engaged in economic activity. The following visualization presents the incidence of child labour for boys vs. girls by country, according to the most recent estimates available from the data published by the World Bank. Here, an imaginary diagonal line with slope of one would mark equal incidence for boys and girls; as it can be appreciated, most countries lie below such line (with a couple of exceptions, such as Pakistan and Cambodia).

# Correlates, Determinants, & Consequences

This section discusses the relationship between children’s work and two crucial variables: schooling and income. Low school attendance is typically considered one of the most important consequences of child labour, while low household income is considered an important determinant. See Schultz and Strauss (2008) for a discussion of academic evidence informing other potential determinants and consequences.

# Child labour and schooling

As it has already been mentioned, child labour is particularly problematic to the extent that it hinders the children’s development, notably by interfering with schooling. Since time is a scarce resource, the extent to which children’s employment is linked to school attendance depends on the type and number of hours worked. The following visualization taken from Schultz and Strauss (2008) presents evidence of this link using data from UNICEF’s Multiple Indicator Cluster Surveys (2000 and 2001); it plots school attendance rates for children 10–14 against total hours worked in the last week (by type of work) with 95 percent confidence intervals (labeled CI and plotted in lighter shades). The steepest part of the pictured curves are in the range 20-45 hours, suggesting—as one would naturally expect—that it is most difficult for a child to attend school when approaching full-time work. This evidence also shows that there are no significant difference by domestic or marketed work.

School Attendance vs. Hours Worked – Schultz and Strauss (2008) 7
School Attendance vs. Hours Worked, from Eric V. Edmonds "Child Labor", Chapter 57 in T. Paul Schultz, John Strauss (2008), Handbook of Development Economics, Volume 4. North Holland.

The above relationship between work and schooling is informative about the impact of children’s work on schooling, but is not sufficient to establish causality; there are many potential economic and cultural factors that simultaneously influence both schooling and work decisions; and in any case, the direction of the relationships is not obvious—do children work because they are not attending school, or do they fail to attend school because they are working? A number of academic studies have tried to establish causality by attempting to find a factor (an ‘instrumental variable’) that only affects whether a child works without affecting how the family values other uses of the child’s time (e.g. Rosati, F., Rossi, M. (2003)8; Gunnarsson, V., Orazem, P., Sanchez, M. (2006)9)While these studies can be criticized on the grounds of the validity of the instrumental variables used, they seem to agree on the fact that there is a stronger association between child labour and schooling than the raw data would suggest.

# Child labour and income

Cross-country data on child labour and economic growth suggests a strong negative correlation between economic status and child labour. The following visualization depicts the cross-country incidence of child labour (share of children ages 7-14 involved in economic activity) against GDP per capita (PPP adjusted GDP per capita in international dollars). To provide some context regarding the absolute number of children, each country’s observation is pictured as a circle where the size of the circle represents population aged 5-14.

This evidence cannot be interpreted causally; as before, countries differ in many aspects that may be associated with child labour choices and income. But there are a number of reasons why, conceptually, child labour might be indeed caused by poor living conditions. For example, children might only work if the parents are unable to meet subsistence conditions; or it could be the case that parents allocate more of the children’s time to schooling as they afford the necessary inputs for schooling (text-books, uniforms, etc).

Partly following this logic, several countries have implemented cash transfer programmes in an attempt to discourage child labour and increase schooling. The idea behind these programmes is that the cash transfers are conditioned on a number of desirable actions, including sending children to school; and in doing so, they lower the relative costs of schooling and raise family income.  Schultz (2004)10 evaluates one such program in Mexico (the so-called ‘Progresa’ program) and  finds a significant reduction in wage and market work associated with eligibility for Progresa. Similar findings have been found in other countries as well. The literature often refers to these programmes as the prime example of “collaborative measures” against child labour: non-coercive interventions that alter the economic environment of decision makers in order to make them more willing to let children stay out of work.

Definitions, Measurement & Data Quality

# Definitions and measurement

Broadly speaking, the term “child labour” is defined as the employment of children in any work that deprives them of their childhood and dignity, and that is harmful to their physical and mental development.

The ILO defines child labour as work that is mentally, physically, socially or morally dangerous and harmful to children; and that interferes with the children’s schooling by depriving them of the opportunity to attend school, either by obliging them to leave school prematurely, or by requiring them to attempt to combine school attendance with excessively long and heavy work (a general definition along these lines can be found in the ILO’s Child Labour website ). The ILO’s Convention No. 138, adopted in 1973 and ratified by most countries of the world, stipulates the relevant ages that different countries use to define child labour.

Strictly speaking, whether or not a given job falls within this definition of “child labour” depends on the details of the actual context—the child’s age, the type and hours of tasks performed, the conditions under which they are performed. In fact, in the strictest sense, a counterfactual (i.e. what the child would be doing in the absence of work) would be necessary to determine precisely whether an activity is harmful to a child’s development. Since accounting for all of these factors is extremely challenging in practice, it is common for studies to concentrate on employment in general—through indicators constructed from surveys of economic activity by age groups—, while leaving room for exploring the consequences of such employment (e.g. school attendance). This is the approach followed here:  we have been referring to “children in employment” as those engaged in any economic activity (for at least one hour during the last week), hence including work in market production as well as certain types of non-market production (e.g. the production of goods and services for own use).

It’s important to bear in mind that many studies distinguish between “children in child labour” and “children in employment”, while using the terms “working children”, “children in economic activity” and “children in employment” interchangeably. In such cases, the former (“children in child labour”) are considered a subset of the latter (“children in employment” or any of the aforementioned interchangeable terms). More specifically, children in child labour include those in worst forms of child labour and children in employment below the minimum age, excluding children in permissible light work—where “permissible light work” is defined as any non-hazardous work by children (ages 12 to 14) of less than 14 hours during the reference week (for more details see ILO-IPEC, Diallo, Y., et al. (2013)11). While standardized cross-country data is not publicly available for “children in child labour”—or other definitions using other working hour thresholds—, it is possible to obtain raw data on children’s time allocation from UNICEF and ILO (see Data Sources below). Studying how many hours children work is important considering that one hour of work in a week, without any further context, constitutes a rather slack measure of child labour.

# Data quality

Schultz and Strauss (2008) provide a complete account of the particular challenges that arise from measuring children employment through household surveys. The authors highlight difficulties arising from coverage (i.e. capturing the most vulnerable children through random sampling) and accuracy (i.e. misreported hours worked and sensitivity to the recall period used).

Specifically regarding the information published by the World Bank in their World Development Indicators, it is important to highlight that, while definitions are standardized (children in employment are always defined as those children aged 7-14 involved in economic activity for at least one hour in the reference week of the corresponding survey), the data-collection instruments are not standardized across the different sub-sources feeding the consolidated dataset. It is because of this that many policy reports (such as the much-referenced report Marking Progress Against Child Labour (2013) ) ‘homogenize’ the data before reporting estimates, by correcting for discrepancies in the underlying survey instruments. Those visualizations presented here that use the consolidated data published by the World Bank have not been corrected.

Issues of consistency across different  survey instruments in the World Bank consolidated data  can help us explain country-specific patterns that are otherwise difficult to interpret. Consider the case of India. As it can be appreciated in the following visualization, the incidence of child labour in India seems to jump up in 2006, only to go back in 2010 to the levels that would have been predicted with the observations from 2000 and 2005. After checking the survey catalogue, it becomes clear that the estimates for 2006 come from the country’s Demographic and Health Survey, while those for the other years come from consecutive rounds of the National Sample Survey. Cases such as this illustrate why current academic studies typically rely on data stemming from a single survey instrument, such as UNICEF’s Multiple Indicator Cluster Surveys.


# Data Sources

# Children in employment (country-specific historical data)

# Toniolo and Vecchi (2008)
  • Data Source: Toniolo, Gianni, and Giovanni Vecchi. Italian Children at Work. 1881—1961. Giornale degli economisti e Annali di economia (2007): 401-427.
  • Description of available measures: Share of children ages 10-14 (total, and by gender) who are economically active
  • Time span: 1981-1961
  • Geographical coverage: Italy

# Cunningham and Viazzo (1996)

# Children in employment (consolidated cross-country data)

# World Development Indicators published by the World Bank (based on data from ILO, UNICEF and the World Bank)

The main source of consolidated data on child labour is the inter-agency research cooperation programme Understanding Children’s Work. The principal source for this programme is the ILO’s Statistical Information and Monitoring Programme on Child Labour (SIMPOC), which is the statistical arm of the International Programme on the Elimination of Child Labour (IPEC). Understanding Children’s Work links the SIMPOC data with data produced by the World Bank (specifically the Bank’s Living Standards Measurement Study datasets) and UNICEF (specifically datasets produced with the organization’s so-called Multiple Indicator Cluster Surveys); as well as data from direct partnerships with national statistical offices. Details about the corresponding household surveys used to produce these datasets, including information about sample size, sample units and coverage, can be found in survey catalog of Understanding Children’s Work.

  • Data Source: Understanding Children’s Work project based on data from ILO, UNICEF and the World Bank. This data is published by the World Bank.
  • Description of available measures:
    • Share of children aged 7-14 (total, and by gender) involved in economic activity for at least one hour in the reference week of the corresponding survey (irrespective of school attendance)
    • Share of children aged 7-14 (total, and by gender) involved in economic activity for at least one hour in the reference week of the corresponding survey (not attending school)
    • Share of children aged 7-14 (total, and by gender) involved in economic activity for at least one hour in the reference week of the corresponding survey (working while attending school)
  • Time span: 1994-2015
  • Geographical coverage: Global by country (predominantly low and middle income countries)
  • Link:


The ILO Programme on Estimates and Projections of the Economically Active Population (EPEAP) has been producing statistics on labour force participation (for adults and children) since 1945, through the database known as ILOSTAT (formerly LABORSTA). Many studies rely on the LABORSTA/ILOSTAT data to shed light on the extent of child labour in the 20th century, before ILO started producing specialized child labour data. LABORSTA/ILOSTAT data is however problematic as a source to measure child labour, since data on work inside the household (even market work) are often not collected. Schultz and Strauss (2008) say that this source is not reliably useful for analyzing changes in child labor over time, given that the survey instruments, coverage, and estimation methodologies are not designed for this purpose.12

  • Data Source: ILO Programme on Estimates and Projections of the Economically Active Population (EPEAP), through the
  • Description of available measures: Labour force participation rates by age groups
  • Time span: Since 1950
  • Geographical coverage: Global by country
  • Link:

# Children’s time allocation (cross-country data)


The UNICEF’s Multiple Indicator Cluster Surveys, which can be obtained upon request, contain rich information about children’s time allocation in 108 countries (essentially the same set of countries for which the World Bank publishes the data referenced above). They include a child labour module which asks children 5–14 whether they work outside of their household in the last week and the last year as well as how many hours they worked outside the household in the last week. The surveys also collect hours in the last week for work in domestic chores and in the household business.

  • Data Source: UNICEF MICS
  • Description of available measures: Children’s time allocation outside their household and in domestic chores in the household
  • Time span: 1994-2015
  • Geographical coverage: Global by country (predominantly low and middle income countries)
  • Link:

# Composite indices

# Maplecroft’s Child Labour Index

Maplecroft’s Child Labour Index evaluates the frequency and severity of reported child labour incidents in 197 countries. The private risk analysis firm producing the data does not provide details about its methodology, but it does produce periodic analysis reports that are publicly available. Further information can be found on Maplecroft’s website.