Assessing the Effect of Complex Survey Design in the Analysis of Child Labour Prevalence Rate in Ghana

Abstract

Statistical surveys are conducted to estimate population parameters where there are reasons restricting the use of the total population. In practice, there are two different survey strategies (i.e. simple and complex survey designs) to be implemented and the choice of a strategy depends on several factors including the characteristics of the population, the nature of the research questions, etc. However, when the complex survey design is used, standard statistical methods that do not take into account the complex nature of the survey design may lead to inaccurate estimates. In Ghana, living standard surveys are conducted using complex survey design involving stratifications, clustering and estimation of survey weights. In this study, bootstrap resampling methods are used to explore the effect of complex survey design in the analysis of child labour prevalence rate. The relative efficiency of the complex survey design approach was determined by using design effect (deff). Data from the Ghana Living Standard Survey Round 6 (GLSS 6) conducted by the Ghana Statistical Service in 2012 was used for the analysis and the target population was children aged 5-17 years. The results from the simulation study shows that relative efficient estimates are obtained when the complex survey design characteristics are considered in the analysis. Thus, ignoring the characteristics of complex survey design could lead to unrealistic estimates.