SUNScholar will be unavailable on 25 April 2018 from 08:00 to 10:00 South African Time for routine maintenance. Please contact scholar@sun.ac.za for queries.

The extension of existing data and methods to measure poverty and mobility in data-poor, agrarian Sub-Saharan Africa

Ferreira, Thomas (2017-12)

Thesis (DCom)--Stellenbosch University, 2017.

Thesis

ENGLISH SUMMARY : In sub-Saharan Africa (SSA), poverty rates have declined, but the absolute number of poor has increased (Beegle et al., 2016). The poor are disproportionately found in rural households which practise smallholder agriculture. Poor data availability and quality have no doubt curbed poverty alleviation efforts. Smallholder agriculture is hard to measure and the statistical capacity of many developing countries in SSA is low (Carletto et al., 2015). In this environment, this thesis shows through three separate studies, how existing data and methods can be extended to overcome usual data deciencies in SSA and enhance knowledge on welfare. In the rest study, econometric techniques used in analysis of the formal labour market are extended to understand the educational returns in agricultural productivity in Malawi. Poverty in Malawi ts the typical pattern in SSA. The majority of the poor live and work in smallholder agriculture. In settings like these, schooling is believed to be a valuable tool in lifting people out of poverty. Yet, little is known about how schooling affects agricultural productivity. The effect of education on smallholder agricultural production has been estimated for countries in SSA but no studies deal with the endogenous nature of education in the production process. This chapter contributes to the literature by estimating for the rst time, the causal effects of education on agricultural productivity in SSA. An instrumental variable (IV) approach is used, using the introduction of free primary education (FPE) and the age of paternal orphanhood as IV's for education. The instruments are shown to calculate local average treatment effects for individuals who only entered school due to FPE and only left school due to paternal orphanhood. It is found that there are large differences in returns to education between the subgroups. Returns are low and insignicant when FPE is used as an IV but they are larger and there is a signicant effect when age of paternal orphanhood is used. Thus, while education can have large effects on agricultural productivity, this is not so for most of the population, especially individuals specically targeted by large scale expansions in educational access. The second study shows how existing satellite data can be used to enhance knowledge of welfare in rural agrarian areas where data on poverty is infrequent and only statistically representative for large geographic areas. Satellite data overcomes both of these limitations as small geographic areas are captured at high frequency. Night lights satellite data have been shown to correlate well with GDP but are not in areas that are not electried, such as rural, agriculturally based communities in developing countries. This is why this study explores the use of daytime satellite data. Satellite data measuring land use and vegetation quality, have been used to model socioeconomic outcomes across regions, but no studies have explored whether daytime satellite data can be used to track welfare longitudinally. This study argues that indicators of vegetation quality, derived from satellite data, can be used to track welfare over time in agriculturally dominant areas. Such indicators are used extensively to predict agricultural yields and thus should correlate with welfare, as agriculture is an important source of income. This study explores whether this is the case using data from Namibia. Firstly, it is shown using classication of cropland, that daytime satellite data can identify areas of economic activity where night lights cannot. Secondly the relationship between vegetation quality and welfare is studied. Crosssectionally, increases in vegetation quality correlate negatively with welfare. This is expected as the poor are more likely to live in rural areas. Within rural areas, however, vegetation quality correlates positively with welfare. This shows that vegetation quality can be used to track welfare over time in areas where night lights are not present. It can also do this at higher frequencies and smaller areas than household survey data. The nal study shows how repeated cross-sectional data can be used to enhance knowledge on economic mobility, in a data decient setting such as Namibia, where panel data does not exist. Namibia has experienced sharp declines in poverty over the last two decades, but in 2009/10 the headcount poverty rate was still 28.7%. Nothing is known, however, about the household dynamics of poverty and poverty reduction over the period as there is no panel data available. To overcome this limitation, this study exploits a method developed by Dang et al. (2014) that uses repeated cross-sections to estimate bounds on the joint probability of staying in or out of poverty, or moving into or out of poverty over two periods. The results showed that Namibia deals with both chronic and transitive poverty. Poverty is found mostly in rural agriculturally dominant areas. It was shown that across these regions, the dominant nature of poverty varies. Particularly, chronic poverty was dominant in the poorest three regions. In general this thesis contributes by providing new information on the state of welfare and mobility of the poor in sub-Saharan Africa. This thesis also contributes by showing that even when ideal data does not exist, knowledge can be enhanced by the extensions of good methods, used in non-agricultural settings and data used for alternative purposes, such as commercial yield prediction. Finally, this thesis highlights the importance of acknowledging the links between poor households and smallholder agriculture. As is shown, survey data capturing these links is needed, and where it is available, it enhances knowledge on poverty and pathways out of it. Acknowledging the linkages also enables the use of alternative data sources, such as daytime satellites, to improve our understanding on welfare.

AFRIKAANSE OPSOMMING : Alhoewel die armoedekoers in Sub-Sahara Afrika afgeneem het, het die getal armes toegeneem (Beegle et al., 2016). Die meeste armes word disproporsioneel aangetref in landelike gebiede waar bestaanslandbou beoefen word. Die gebrek aan voldoende data bemoeilik ongetwyfeld pogings tot armoede-verligting. Opbrengste en inkomste in bestaanslandbou is moeilik meetbaar, en heelparty ontwikkelende lande in Sub-Sahara Afrika kort betroubare statistiese inligting (Carletto et al., 2015). Hierdie proefskrif spreek hierdie kwessies aan en toon, aan die hand van drie afsonderlike studies, hoe bestaande data en meetmetodes uitgebrei kan word om die gebrek aan goeie data te oorkom en in die proses ook kennis rakende welvaart uit te brei. In die eerste studie word ekonometriese tegnieke wat in die formele sekor gebruik word, uitgebrei om ‘n beter begrip van die invloed van onderwys op landouprodutiwiteit in Malawi te verkry. Armoede in Malawi volg die tipiese Sub-Sahara patroon, waar die meeste armes in bestaansboerdery is. In sulke omstandighede word onderwys dikwels beskou as instrument om mense uit armoede te lig. Maar min inligting is beskibaar oor hoe skolastiese ontwikkeling produtiwiteit in die landbousektor beïnvloed. Alhoewel die rol wat onderwys in bestaanslandbouproduktiwiteit speel al vantevore ondersoek is, probeer hierdie studie veral die endogene aard van onderwys in landbouproduktiwiteit aanspreek. Die studie dra tot die bestaande literatuur by deur vir die eerste keer die kousale rol van onderwys op landbouproduktiwiteit te probeer meet. ‘n Instrumentele-veranderlike (IV) benadering word gevolg, waar die inwerkstelling van gratis primêre onderwys en die ouderdom waarop kinders hulle vader verloor het as IV's vir skolastiese opvoeding geld. Die instrumente meet ‘n lokale gemiddelde behandelingseffek en die opbrengste op onderwys onderskeidelik vir individue wat slegs met skool begin het weens die inwerkingstelling van gratis primêre onderwys, en vir individue wat die skool vroeg verlaat het as gevolg van die dood van hul vader. Die resultate waartoe gekom word wys dat daar groot verskille in opbrenste tot onderwys is vir die twee groepe. Uitkomste is laag en onbeduidend waar gratis primêre onderwys as ‘n IV geld, maar is groter en meer beduidend waar vader se sterfte as IV geld. Dus, al hoewel opvoeding ‘n effek op landbouproduktiwiteit kan hê, geld dit hier nie ten opsigte van die grootste deel van die bevolking nie, veral nie individue wat spesiek deur die uitbreiding van toegang tot onderwys geraak word nie. Die tweede studie wys hoe bestaande satelliet-data gebruik kan word om kennis uit te brei oor die stand van welvaart in landelike boerderygebiede, waar gewone statistiese data oor armoede en die stand van welvaart nie geredelik beskikbaar is nie, of net statisties verteenwoordigend is oor groter geograese gebiede. Satelliet-data spreek beide hierdie tekortkominge gedeeltelik aan, omdat kleiner geograese areas teen hoë frekwensies gefotografeer kan word. Nagligte satellietdata korreleer goed met BBP, maar is nie verteenwoordigend van areas waar daar nie voldoende elektrisiteit is nie, soos landelike, landbou-gebaseerde gemeenskappe in ontwikkelende lande. Daarom ondersoek hierdie studie die gebruik van daglig satelliet-data. Daglig satelliet-data wat landgebruik en plant-gesondheid meet, word gebruik om sosio-ekonomiese uitkomste oor streke te modelleer, maar geen studie het tot nogtoe ondersoek of daglig satelliet-data gebruik kan word om welvaartsverandering oor tyd te meet nie. Hierdie studie voer aan dat aanwysers (indikatore) van plant-gesondheid, afgelei vanaf satelliet-data, gebruik kan word om welvaart oor tyd te meet in areas waar landboubedrywighede dominant is. Sulke aanwysers dien reeds algemeen om verwagte landbouopbrengste te voorspel en behoort dus met welvaart te korreleer. Die studie vors na of dit wel so is, deur data oor Namibië te gebruik. Eerstens is landerye waarop gewasverbouing plaasvind, geklassiseer. Die bevindinge wys dat daglig satelliet-data areas van ekonomiese bedrywighede in die landbou kan aandui waar nagbeligting-satelliete dit nie kan doen nie. Tweedens is die verwantskap tussen welvaart en aanduiding van plant-kwaliteit ondersoek. In ‘n algemeen-ekonomiese deursnit van ‘n land of gebied, sal plant-gesondheid nie positief met welvaart korreleer nie, wat verstaanbaar is omdat meer arm mense in landelike gebiede woon. Maar in die landelike gebiede self korreleer plantkwaliteit positief met welvaart oor tyd. Dus kan daglig satelliet-data gebruik work om welvaartsveranderings te monitor in gebiede waar daar min of geen lig in die nag aanwesig is. Die nale studie toon aan hoe opeenvolgende deursnit-data aangewend kan word om kennis oor ekonomiese mobiliteit uit te brei, veral in ‘n data-arm gebied soos Namibië, waar longitudinale of paneeldata nie bestaan nie. Alhoewel daar oor die afgelope twee dekades ‘n skerp afname in armoede in Namibië was, was die armoedekoers in 2009/10 steeds 28,7%. Geen inligting bestaan egter oor die dinamiek teenwoordig in huishoudelike armoede asook dit wat armoede oor die tydperk heen laat afneem het nie, omdat geen paneeldata beskikbaar is nie. Om hiedie tekortkoming aan te spreek, ontgin die studie ‘n metode ontwikkel deur Dang et al. (2014), wat herhalende deursnit-data gebruik om die betroubaarheidsgrense te bepaal van die proporsie van die bevolking wat oor tyd arm gebly het of buite armoede gebly het, asook van die wat oor die armoedelyn beweeg het. Die resultate toon aan dat Namibië sowel chroniese as kortstondige armoede beleef en dat armoede meestal veral in landelike landbou-dominante gebiede voorkom. Die studie toon ook dat die aard van armoede in hierdie gebiede varieer, maar dat chroniese armoede veral in die drie armste streke dominant is. Oorkoepelend beskou, dra hierdie proefskrif by om nuwe inligting te verskaf oor die stand van welvaart en inkomste-mobiliteit onder armes in Sub-Sahara Afrika. Die proefskrif lewer ook ‘n bydrae deur aan te dui, ook in gevalle waar voldoende goeie data nie bestaan nie, hoe kennis vermeerder kan word deur bestaande analitiese metodes uit te brei, insluitende metodes van buite landbou-omgewings of data wat vir alternatiewe doelstellings, soos byvoorbeeld kommersiële opbrengs-skattings, gebruik word. Laastens beklemtoon die proefskrif die belangrikheid van die verband tussen arm huishoudings en bestaansboerdery-landbou. Soos getoon is, is daar ‘n behoefte aan meer meting en opnames waardeur meer data ingesamel word, om hierdie verband verder te ontgin, want waar sulke inligting beskikbaar is, bied dit sowel inligting oor armoede as moontlikhede om armoede te verminder. Die erkenning van hierdie sterk verband tussen armoede en landbou skep ook moontlikhede vir die gebruik van alternatiewe data-bronne, soos byvoorbeeld daglig-satelliete, om ons begrip van ekonomiese welvaart uit te brei.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/102939
This item appears in the following collections: