Child Labour A Statistical Study Using Multistage Probability Proportional Strati ed Systematic Sampling

Abstract

Child labor is a widespread phenomenon in the world, occurring predominantly in develop- ing countries. Recently, there has been renewed concern about the presence and impact of child labor from politicians, activists and academics alike. Most of the popular discussion has centered on misleading statistics, harmful e ects of child labor and ways to curtail its incidence. Much of the recent theoretical literature has focused attention on the fact that the decision to send children to work is most likely made not by the children themselves, but by households who do so out of dire need. Poverty is considered to be the root cause of child labor. In fact, this is not true and literacy and household e ect are even bigger variables in the determination, and measurement of child labor in a society. This raises the issue of the impact of literacy and schooling on child labor and vice versa. Notwithstanding, a large and rapidly expanding literature on child labor, there is not much empirical evidence on this issue since much of this literature has concentrated on socially, anthropologically, or somewhat psychologically, analyzing the causes of child labor rather than studying its consequences, especially for the impact of learning and household. The present study seeks to ll this signi cant gap in the literature on child labor. Broadly, the study can be divided into three parts; developing a reliable estimate to calculate number of children doing work, identifying important factors for child labour, and thirdly, developing a pro le of today's child labourer. The exercise is conducted on a primary data set involving 5-14 years old children from Lahore & Bahawalpur divisions, Pakistan, using a multi-stage probability proportional strati ed systematic sampling scheme. Close ended questionnaire was specially developed keeping in view the eld and data processing requirements of the project. To avoid possible biases, proper interviewer's training and practice sessions were conducted. Information was collected on family demographics, place of origin and current living status, personal information, current work history and conditions, personal behaviour, health, perceptions and knowledge and literacy level on a household basis from the house- hold head. The estimator is developed using Sampford (1967) extension to Brewer (1963) approach for calculating internal selection probabilities. The numerical strength of child labour in these two divisions turns out to be 3,440,411 chil- dren which happens to be 32% of total children living in these two divisions. Monte-Carlo simulation is carried out to develop its probability distribution which turns out a bi-modal distribution. This bi-modalness is probably because of di erent boys and girls labourers or due to di erent sizes of districts and tehsils included in the sample. This distribution is then used to develop con dence intervals associated with the total number of child labourers in these two divisions. E ect of household, literacy and poverty are quantitatively investigated and these turns out to be the biggest instrumental variables in the dynamics of child labour- ers. Speci c generalized Poisson regression models are developed for various situations to ascertain and gauge the veracity of associations and relationships between child labour dy- namics and its causes like household demography, household poverty and household literacy. It turns out that household demography, including its physical and familial structure, plays a statistically signi cant role in the dynamics of child labour. Household poverty, on the second hand, turns out to be promotive for child labour. While, increasing household literacy turns out to be negatively associated with the dissemination of child labour. Multivariate cluster analysis is also conducted to develop a household characteristics based segmentation in the child labour community which results in three clearly separated clusters of labouring kids; mechanics, chotta, and girls. A multiple discriminant analysis is also conducted to develop a household characteristics based yard stick to index households for their propensity towards child labour. It also helps in identifying the potential entrants in this labour. In the end, a pro le is developed for a typical child labourer on the basis of accumulated data envisaging di erent facets of his life. Such a pro le is useful in understanding the life and miseries of a child labourer and his household.