Please use this identifier to cite or link to this item:
Title: Modelling poverty in Namibia using beta distribution.
Authors: Mafale, Ndubano
Keywords: Modelling poverty
Poverty Namibia
Beta distribution
Issue Date: 17-May-2022
Publisher: Namibia University of Science and Technology
Citation: Mafale, N. (2022). Modelling poverty in Namibia using beta distribution. (Unpublished master's thesis). Namibia University of Science and Technology, Windhoek.
Abstract: Modelling poverty is important as it helps to pinpoint human development areas that are most affected by poverty. Also, modelling poverty helps in understanding the patterns and levels of poverty, which helps policy makers to plan and make targeted interventions to reduce poverty. The traditional methods of estimating poverty such as the cost of basic needs approach or the poverty line approach are surrounded by a lot of controversies as they are said to underestimate or overestimate poverty. These methods are uni-dimensional as they only estimate poverty in one dimension (e.g consumption, income and expenditure) which neglects the humanistic needs side of poverty such as access to health or education. On the other hand, methods that include the Alkire and Santos (2011) method measure poverty in more than one dimension (e.g living standards, health, and education) but are faced with prejudice as the weighting method used is based on experts’ opinion or the consensus of different stakeholders. Thus, this type of weighting method may result in biased weights and consequently result in inaccurate estimates of Multidimensional Poverty Index (MPI) values. This study focused on developing a multidimensional poverty model using beta distribution capable of estimating poverty for Namibia on regional and national levels. In addition, the study aimed at assessing the impact of weighting methods on MPI. The first specific objective was to develop a multidimensional poverty model using beta distribution that could be used to model poverty for Namibia. The developed model showed that the MPI is equivalent to the expected value of the left-truncated beta distribution. The uncertainty surrounding the MPI was measured through the specification of the variance. The second specific objective was to assess the impact of weighting methods on MPI. Two weighting methods (equal and entropy weighting) were adopted and their effect was assessed on the MPI obtained using the Alkire and Santos (2011) and the beta distribution approaches. The results revealed that the MPI values obtained when using entropy were iii slightly bigger than the MPI values obtained using equal weighting under the Alkire and Santos (2011) approach compared to the beta distribution approach where the MPI values obtained when using equal weighting were bigger than the ones obtained using the entropy weighting method. Moreover, the entropy weighting method was found to be better than equal weighting as it is a mathematical based approach and is not affected by a change in the number of indicators compared to equal weighting which is subjective and sensible to the number of indicators. The third specific objective was to identify more potential indicators that could be used in computing MPI which were not used in the initial computation of MPI by fitting a beta regression model. Using the NHIES 2015/2016 data, we fitted a beta regression model and identified the indicators that were left out in the initial computation of MPI. In conclusion, the results revealed that the beta distribution model can be used to estimate regional and national poverty. The results also revealed that the entropy weighting method is useful in allocating weights when computing MPI as it eliminates the bias that comes with allocating weights. Moreover, the model can be used to identify areas that are highly affected by poverty and thus helping to come up with ways to alleviate the poverty. Finally, the beta regression model can help to identify indicators to be included in computing MPI.
Description: A thesis submitted in fulfilment of the requirements for the degree of Master of Science in Applied Statistics in the Department of Mathematics and Statistics Faculty of Health and Applied Sciences Namibia University of Science and Technology.
Appears in Collections:Theses and Dessertations

Files in This Item:
File Description SizeFormat 
Ndubano_Mafale_Thesis_May 2022.pdf1.38 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.