A Bayesian Hierarchical Model to Estimate and Predict Child Labour in India

Abstract

Child labour - work identified as harmful to children aged 5-17 years - is not yet eradicated in India. Consequently, there is a strong demand for the number of child labourers to be measured with more accuracy. Furthermore, understanding the diverse routes to and causes of child labour is essential for the improvement of its estimation and prediction. In India, child labour is a combined result of social structure, institutions, and norms. Notably, the gendered aspects of child labour add complications: largely based on gender roles and attitudes, and norms in India. Accordingly, this study aims to offer a new way to gauge the number of child labourers in India by using advanced modelling and revealing the relationship between its socioeconomic and cultural attributes. This study mainly uses the Indian Human Development Survey 2011/12 and the National Sample Survey on Employment and Unemployment 2011/12 for the analysis of the child-labour problem in India. This thesis is composed of three journal articles that address key conceptual and methodological questions regarding child labour. The first article investigates how to improve the estimation of child labour using a Bayesian hierarchical model within a combined data approach. The second examines the interactive effects of social group, class and gender on child-labour participation and child labourers' working hours. The third article reveals the impact of gender norms on child-labour risks in terms of household occupation and state. All of these are based on the gender and development approach, in which social and gender relations explain agents' decisions regarding child labour. Gender is constructed within social and cultural circumstances which decide whether and where boys and girls work. In the first article, a Bayesian statistical method was used for analysis with a combination of the two representative datasets in India from 2011/12. A combined-data approach enables us to accumulate knowledge from multiple data sources, as it incorporates diverse priors and assumptions as a probability distribution. Second, a Bayesian hurdle Poisson model provided a consistent and efficient way to estimate the prevalence and magnitude of child labour and to reveal its gender-based trends. The third article suggested a child-labour risk model that could be incorporated together with the norms and socioeconomic variables used in the estimation of child labour, thereby providing a way to predict risks of child labour according to households' socioeconomic backgrounds and locations. The main findings include a more accurate estimation of the number of child labourers across the whole of India. The best estimate for child labourers aged 5-17 years old in 2011/12 is 13.2 million (95-percent predictive intervals are 11.4-15.2 million). The suggested Bayesian model can be used to project the potential numbers of child labourers in the near future, through applying more accurate population data and priors. This study confirms the importance of including unpaid household services in the definition of child labour as a way to make girls' labour more visible. Moreover, this research has revealed new evidence of the relationship of gender, class, and social groups with child labour, such as the low levels of participation in the work force among female children in contrast with their long working hours as informal labourers. Lastly, the findings demonstrate that group-based norms and institutions, along with household structural positions, play clear roles in decisions regarding child labour. Norms supporting women in working outside the home are found to help reduce child labour. This research accounts for diverse aspects of child labour - social structure, institutions/norms and gender - advocating therefore for an integrated and long-term approach towards the child-labour problem. As the findings imply, the distribution of labour across children according to gender should be taken into accou

Date of Award	31 Dec 2021
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Wendy Kay Olsen (Supervisor) & Arkadiusz Wisniowski (Supervisor)