When International Child Labour Definitions are Confronted with Reality

When International Child Labour Definitions are Confronted with Reality

Laterite has just recently completed a very large study on child labour in Rwanda’s tea growing areas … the analysis is currently under review and will hopefully be available online soon.

What we’ve learned throughout this process is that child labour is extremely difficult to measure and that there are fundamental limitations to international child labour definitions. These limitations imply that child labour estimates in different contexts are really not comparable. Small definitional and contextual differences can lead to large swings in child labour estimates.

Before we deal with some of these limitations and present some ideas on how to improve the way child labour is measured, here is a brief summary of how child labour is defined internationally.

What is child labour and hazardous child labour?

The international reference document for child labour statistics and related definitions is Resolution II of the seminal report of the 18th International Conference of Labor Statisticians (ICLS). Simplifying, this document sets out the following definitions for child labour and hazardous child labour:

Hazardous child labour (HCL), includes all children below the age of 17 who were engaged in a productive activity that is hazardous by nature or circumstance for at least one hour in the week prior to the interview (this includes hazardous chores).

Child labour (CL) includes all children below the age of 17 that are: (i) in employment below the minimum age (as per rules on the total number of hours worked); or (ii) in hazardous child labor

What is considered to be employment and what is considered to be hazardous?

A child is considered to be in employment when he or she engages in a “productive activity” for at least one hour per week based on an international reference called the “SNA production boundary”. Productive activities include: the production of all goods or services for the market (including for one’s own household), the production of all goods for self-consumption (including for one’s own household), the production of housing services to improve one’s dwelling (including for one’s own household), the production of knowledge-capturing products for self-consumption (e.g. information, news, reports, computer programs) and paid domestic work outside one’s own household.

In the Rwandan context, this would mean that a child that helps on the family farm or herds cattle for one hour per week is considered to be in “employment”. If this child is under the age of 13, then this child is not only considered to be a “working child”, but is also considered to be in “child labor”.

What is hazardous? A whole host of conditions would qualify a child as being in hazardous child labour. A child is considered to be in hazardous child labour for example if he or she works in hazardous locations (e.g. underground, in water, etc), in hazardous conditions (e.g. extreme heat or cold, carries heavy loads, is exposed to dust and fumes, chemicals, etc), has experienced injury at the work place (e.g. back-pain), uses dangerous equipment or products (e.g. automatic machinery, chemicals, fertilizer), works in particularly difficult conditions (long hours, work at night), in banned activities (e.g. mining, charcoal making, collecting scrap metal, etc) or if the child is exposed to physical, psychological or sexual abuse.

These definitions that make sense on paper sometimes struggle when confronted with the local context.

Putting one label on a very heterogeneous population

The first problem with these definitions is that they put one label on a very heterogeneous population. Based on the definitions presented above, children that work in very different conditions are put in the same bucket of “working children”, “child labor” or “hazardous child labor”. For example a child that uses pesticides for 1 hour per week while working on the family farm is put in the same hazardous category as a child that works 50 hours a week, has experienced injury as a result of the job, works at night, is exposed to dust and fumes, uses dangerous equipment and gets verbally abused by his/her employer. While both of these scenarios constitute hazardous work, they are inherently very different and probably require a very different set of policy responses. The question that then arises is how can one distinguish between what is more or less hazardous? For example, is using pesticides better or worse than getting verbally abused by one’s employer? While we could express an opinion on this matter, currently there is no way of objectively knowing what is better or what is worse. The answer to this question would also depend on the local context.

The local context matters

International conventions aim to harmonize the way child labor statistics are calculated to ensure the consistency and comparability of estimates, but the local context does matter. Countries have very different socio-economic contexts and different child labor regulations. Children in different countries engage in very different types of activities. In some countries the risk of children working in factories is high; in other countries, like Rwanda, the vast majority of working children work on the farm. In some places work is seasonal, in others it is permanent. Expectations and social perceptions of what is acceptable work and what is not differ. External weather conditions are also different – some locations have very harsh climates, extremely warm or cold, where just the act of working outside puts a child in a hazardous situation. Different conditions, expectations and regulations in different locations mean that international child labor and in particular hazardous child labor definitions need to be adjusted to fit the local reality. International child labor definitions allow for some flexibility to include local regulations into the equation, but even the smallest departures from definitions, can lead to large swings in the estimates.

Issues related to survey methodology

Another reason child labor statistics across different surveys are not directly comparable are potential differences in survey methodology, which can significantly impact results. Small differences between studies can lead to a large divergence in results. Below are a few dilemmas statisticians working on child labor statistics need to deal with:

Is it better to base child labor statistics on parent interviews or child interviews? Statistics based on the responses of parents or children, as we see saw in the study Laterite worked on, can be very different. The parents on the one hand might have a more mature understanding of the kind of activities children engage in and the risks involved; on the other hand, they might also have a greater incentive to under-report the amount of work their children engage in and might simply not know as much about the details of their children’s activities, especially in large families. While most child labor surveys rely on responses of the children themselves, and compare these to parental responses, there are valid reasons one might prefer to opt for either. In practice, some studies report child labor statistics using parental data, others using child level data. In the study we worked on the difference in the share of working children estimates between the two is of the order of 20 percentage points.

Is it more appropriate to base child labor statistics off filters or prompts? Take for example the issue of injuries at the work place, which would put a child in hazardous child labor. Is it better to base the statistics off a filter question, such as “have you fallen ill or been injured in past 12 months because of this activity?”, or is it more appropriate to prompt the respondent about their exposure to list of possible injuries, such as back-pain, extreme fatigue, wounds, eye problems, etc? Results obtained using the two different methods will yield radically different results. Using a filter question one would be likely to capture what children themselves perceive as a serious injury; prompting however, would also make it possible to capture data on issues that directly affect the health of a child, that are defined as serious health issues in international labor regulations, but that might not be perceived as an illness or injury worth reporting by the child in question. Prompting also takes significantly longer than using a filter. While it might increase precision, it also increases survey length significantly, especially if one needs to loop over a list of prompts for each activity for example. The longer the survey, the lower the quality of the data. There are also ethical issues involved as one should not interview a child for an excessively long period of time.

What time period should child labor rates be based on? It is generally accepted that whether a child is currently employed or not should be based on whether he or she has been engaged in a productive activity – under certain parameters – for more than one hour in the past week. It also makes sense to ask children about the details of this activity, such as how many hours they spend on this activity, within the framework of the past week. This is because people’s recall periods tend to be short. On some issues however, such as injury or potential abuse by the employer for example, it might make sense to ask about children’s experiences over a longer period of time in that same activity. A child might not have been abused by his/her employer over the past week, but might have been regularly abused in the past, thereby putting this child in a very hazardous working environment. A child might have been severely injured while doing a given task several months prior to the interview, thereby making that activity a dangerous activity for the child’s health, but not have experienced any specific ailments over the past week. Determining the most appropriate time limit to measure various aspects of what constitutes child labor is a difficult task.

Finally, how does one deal with parameters where questions of interpretation can make a big difference? Take for example a question relating to whether a child works in “extreme cold” conditions. Being exposed to “extreme cold” during work is something that would qualify a child as being in “hazardous child labor” based on international regulations. While it can get chilly in Rwanda, in particular at high altitudes, what children here would qualify as “extreme cold” when working on the household farm is not the same level of “extreme cold” that a child would experience in much colder climates or working in the cool room or cold storage room of a factory. An important question to ask is how to deal with questions like these that are very sensitive to interpretation and that can vary significantly depending on the local context.

Issues relating to the timing of the survey

It is key for child labour surveys to be conducted during an entire year to capture cyclical variation. The biggest limitation of the survey Laterite worked on was undoubtedly the timing thereof, as the survey was carried out over a short period of time – between mid October and December 2014. The survey overlapped with the start of Season A in Rwanda, which is one of the main harvest seasons, and the school holiday period, which started during data collection.

Taking a 2-month snap-shot of child labor dynamics in Rwanda’s tea growing areas therefore does not provide us with a full picture of child labor dynamics year round. Evidence from this report suggests that even within this 2-month period, seasonality can be very large. Seasonality is also compounded by the effect of holidays.

Finally, how should one deal with the holiday period? Is it more acceptable for a child to work during the holidays that during the school term? Should holiday work be discounted in a certain way?

Some suggestions for an improved way of measuring child labour

Laterite proposes a different approach to measuring child labour using complexity methods. The complexity-based metrics we propose to develop will complement existing measures of child labor and make the analysis of child labor much more specific, data-driven and better tailored to the local context. How would that work?

When studying child labor we work with an individual-CL factors matrix. For each child we have information, mostly binary, on whether the child is exposed to certain factors that qualify him or her as being in child labor (or hazardous child labor). There are many factors that would qualify a child as being in child labor.

We assume that children that are exposed to a diverse set of such factors are more “deeply” in child labor than others. Likewise we would assume that children that are exposed to non-ubiquitous factors – AND that are mostly experienced by children that are exposed to many different factors – are also more “deeply” into child labor. For example it would not be surprising to find that children that are exposed to sexual abuse at the work place (a non-ubiquitous factor), are also exposed to a whole host of other more ubiquitous factors that put a child into hazardous child labor (for example working for somebody outside the household, without pay, at night, etc). Early analysis from the recently completed study on child labor in Rwanda’s tea growing areas suggest that these assumptions hold.

Complexity methods that are based on the complex interaction of the ubiquity and diversity of child labor factors, would enable us to establish a measure of “how deeply in child labor” children are, but also “how serious” certain factors are from a child labor perspective. The individual – CL factors matrix, would also enable us to create a network of which factors are more closely related, thereby enabling policy makers to not only prioritize what issues to target but also to target inter-related issues in batches as opposed to individually.