A Research Report submitted in partial fulfilment of the Degree of Master of Commerce (Economic Science) in the School of Economic and Business Sciences, University of the Witwatersrand, 28 February 2019 / This paper investigates the impact of the South African Child Support Grant (CSG) on the beneficiary’s height-for-age z-score (HAZ). We make use of data from the National Income Dynamics Study (NIDS), the wave 3 dataset. Using Propensity Score Matching approach the effect of the CSG is positive but statistically insignificant and relatively small. This popular technique rely on assumptions that often do no hold for observational studies. Furthermore, it is susceptible to misspecification of the propensity score equation which could bias the results. This paper therefore look into a technique that address these limitations and can assess the treatment effect robustly and with more precision.
We apply genetic matching algorithm, namely GenMatch. GenMatch is an iterative search algorithm that uses distance metrics to optimize covariate balance in the process of estimating the treatment effect. It automates the search process without the need of manual intervention to achieve the best balance. This algorithm is applied using two balance measures namely, the entropic distance metric and the standardized difference in means. The former compares distributions while the latter compares the first two moments (means and variances) of distributions. The results showcase the significance of utilizing a method that automates the process of optimizing balance and the influence of balance measures on the resulting treatment effect estimate. Specifically, we found that the estimate of the effect of the CSG is larger and more precise than the one reported in the literature. / PH2020
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:wits/oai:wiredspace.wits.ac.za:10539/29817 |
| Date | 28 February 2019 |
| Creators | Mthembu, Lerato Eunice |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | Online resource (40 leaves), application/pdf |
Page generated in 0.0025 seconds