Child labour measurement: Whom should we ask?

AuthorSarah A. JANZEN
Published date01 June 2018
DOIhttp://doi.org/10.1111/ilr.12041
Date01 June 2018
International Labour Review, Vol. 157 (2018), No. 2
Copyright © The author 2018
Journal compilation © International Labour Organization 2018
* Assistant Professor of Economics, Department of Agricultural Economics and Econom-
ics, Montana State University, email: sarah.janzen@montana.edu. This research was made possible
using the WEKEZA (Wezesha Ustawi, Endeleza, Kiwongo cha Elimu Kuzuia Ajira kea Watoto)
baseline survey data collected by Savannas Forever Tanzania for the International Rescue Com-
mittee, World Vision, Kiota Women’s Health and Development, the Tanga Youth Development As-
sociation and the Institute of Development Studies, with funding support from the United States
Department of Labor and assistance from the National Institute for Medical Research-Muhimbili
Medical Research Centre, United Republic of Tanzania. The author would like to specically ac-
knowledge the contributions of Susan James, Jovit Felix, Nai-Nancy Laiser, David Lawson, Andrew
Ferdinands, Deborah Levison, Bernard Ngowi and Nike Start for research assistance and support
during the survey implementation process. Mark Anderson, Paul Glewwe, David Lawson, Carly
Urban and audiences at the University of Montana and the Economics of Global Poverty Confer-
ence provided helpful feedback. The author accepts responsibility for any errors that may subsist.
Responsibility for opinions expressed in signed articles rests solely with their authors, and
publication does not constitute an endorsement by the ILO.
Child labour measurement:
Whom should we ask?
Sarah A. JANZEN*
Abstract. Funding decisions to support vulnerable children are tied to child labour
statistics, hence the importance of an accurate measurement. The author analyses
whether the type of respondent plays an important role in explaining variations in
child labour statistics. Using data from two sites in Tanzania, the analysis shows
that “whom we ask” matters considerably when estimating the prevalence of
child labour. The results suggest that prevalence increases by approximately 35 to
65 per cent when using child self-reports rather than proxy reports. This bias af-
fects 14 to 31 per cent of the sample, depending on the indicator. Discrepancies
decrease as the child ages and increase if proxy attitudes demonstrate opposition
to child labour.
Recent estimates from the International Labour Organization (ILO) sug-
gest that approximately 168 million children are actively participating in
the global economy through activities that are potentially harmful – mentally,
physically, socially or morally – or that commonly interfere with a child’s edu-
cation (ILO, 2013). Udry (2006) describes child labour as “an insidious evil”
because it frequently (though not always) limits a child’s future opportunities.
This is precisely why the determinants and consequences of child labour have
been analysed extensively in literature. As highlighted by Anker (2000), pol-
icy-makers often use child labour statistics to set priorities, allocate resources,
International Labour Review170
monitor progress and evaluate programmes. Yet measuring child labour re-
mains a complex process, leading to widespread inconsistency across studies.
In this article I investigate the role of one critical factor – the survey respond-
ent – in explaining the variation in child labour statistics. The ndings suggest
that “whom we ask” (the selected respondent) matters considerably when es-
timating the prevalence of child labour.
When measuring child labour, the respondent can be either the actual
child or a proxy respondent who reports on a child’s behaviour and time use. In
2004 the ILO recommended that questions be answered directly by the child,
provided the child is at least 10 years old (ILO, 2004). More recent guidelines
from 2008 suggest that “sections of the questionnaire may be addressed to the
children themselves, particularly on hazards at the workplace, and the main
underlying reason for working” (ILO, 200 9, pp. 61–62). These guidelines are
widely acknowledged as the “gold standard”, yet the evidence supporting the
exclusive or even partial use of child reports over proxy reports is weak. The
cognitive development of children is likely to affect the quality of the infor-
mation provided, and it is not entirely clear that child reports are more accu-
rate than proxy reports. According to the Resolution II concerning statistics
of child labour adopted by the 18th International Conference of Labour Stat-
isticians in 2008 , “the general practice is to address survey questions to the
most knowledgeable adult member of the household (or sometimes the head
of household, who is often also the parent or guardian of the working child)”
(ibid., p. 61). Indeed, it is often easier and cheaper to interview an adult than
it is to interview a child. When children are not directly accessible for inter-
views, as is frequently the case, a self-response requires multiple visits and
raises the cost of a survey.
With these issues in mind, I analyse whether the respondent matters
when measuring the prevalence of child labour. Specically, I pose the follow-
ing two questions: (1) Do the answers provided by proxies differ signicantly
from those of the child on whose behalf they are answering?, and (2) Do child
or proxy characteristics explain some of the difference between child and proxy
reports of child labour participation? The answers to these questions have
important implications regarding the cost of collecting accurate child labour
statistics that are comparable both geographically and intertemporally. Such
statistics are regularly used to set targets and monitor progress, and they often
have important implications for the allocation of resources.
Two recent studies seeking to answer the aforementioned questions have
produced conicting results. Dillon et al. (2012) use a randomized survey ex-
periment in the United Republic of Tanzania (hereinafter “Tanzania”) and
conclude that respondent selection, whether a child respondent or a proxy,
does not matter when determining rates of child labour. Conversely, Dam-
mert and Galdo (2013) nd signicant disagreement between child and proxy
responses. Despite the fact that the data used in this study are most similar to
the data used by Dillon et al. (2012), with both data sets coming from Tanzania,
the results reported here provide further evidence in support of the ndings

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT