Structural Transformation and African Smallholders: Drivers of Mobility within and between the Farm and Non-farm Sectors for Eight Countries

Abstract

Using longitudinal data from 2354 smallholder households in 103 villages in eight African countries, three processes of agrarian transformation are analysed for the period 2002 to 2008: intensification of grain production, commercial diversification from staple crops and income diversification out of agriculture. Methodologically, three multi-level, binary logistic models are used. The trends observed provide grounds for some optimism: despite an overall picture of stagnation, intensification in grains (yield per hectare) seems to be increasing. Farmers have, however, raised productivity through the more intense use of labour resources rather than through technological change, while political commitments to agriculture have not improved the production environment. Rather, economic growth and commercialization emerge as strong drivers of intensification, both at country and household levels. Tendencies towards distress-driven income diversification out of agriculture appear to have abated somewhat in the face of more dynamism in the grain sector, with households moving between the farm and non-farm sectors in response to shifts in producer incentives and non-farm opportunities. Diversification processes within agriculture, meanwhile, point to both push- and pull-driven diversification occurring simultaneously. Grain markets, crop diversification and non-farm opportunities complement one another over time. There is little evidence of even incipient processes of structural transformation among the smallholders surveyed.

Notes

¹ For the purpose of this paper, we define intensification as an increase in production due to increasing yields per unit of area for a given crop. This category hence includes households that have increased production on a given area and also households that have increased yields while also expanding the total area devoted to grains. In earlier work on maize (see Djurfeldt et al., 2008), we separated the two trajectories, defining the latter as expansion, but here the two patterns are discussed collectively as grain intensification. The reason for using agronomic definitions rather than conventional economic definitions based on value of production is that the quality of our agronomic data is higher than that of the economic data.

² Here, we are referring not to diversification of crop patterns per se, but to an extension of the range of crops sold, from staple crops to non-staples and non-food cash crops.

³ That is, diversification of income sources or allocation of household labour away from agriculture/animal husbandry to the non-farm sector.

⁴ We also use the alternative term “pluriactivity”.

⁵ The variable “descendant household” is used as a control variable in the models provided next.

⁶ This could be a panel effect, but a comparison of the two cross sections supports the interpretation of less diversification: in the 2002 cross section, 55% sold other crops, while in the 2008 round 52% (Table A1, rows 23 and 24).

⁷ For reasons of space, we do not report these figures in tabular form, only in the text.

⁸ The adoption of animal traction could have contributed to intensified grain production, but this variable suggests no significant association and was not included in the final version of the model.

⁹ The connection between cotton production (0) and grain intensification has been convincingly demonstrated in the context of Mali by Tefft (2010), who shows how technological improvements in cotton production, primarily in terms of animal traction, spill over into intensified grain production.

¹⁰ Given that we have 74 villages in the models, they can accommodate up to seven village-level variables, but the low number of cases makes us willing to accept a 10% level of significance. For household-level variables, we accept a 5% level or lower only.

¹¹ The dependent variables in the intensification and diversification models have been tested for inclusion in the straddling model. As they do not contribute to the explanatory power of the model, they were not retained in the final version.

¹² Tanzania is an outlier and is excluded here.

¹³ Note that the economic growth variable refers to the economy as a whole, including the grain sector, as sectorwise GDP data were not available for all countries.

¹⁴ We have used maize as a proxy for grain imports, as maize is the most commonly grown and imported grain in all the countries in question.

Additional information

Notes on contributors

Göran Djurfeldt

This article draws on data from the Afrint II project which features collaboration between researchers in nine African countries. The team was led by Göran Djurfeldt, Lund University, Sweden, and involved a team of researchers from Lund and Linköping Universities. The country teams were: for Ethiopia, Dr Wolday Amha, Ethiopian Economic Association, Dr Teketel Abebe, Addis Ababa University, Dr Mulat Demeke, Addis Ababa University; for Ghana, Professor Ernest Aryeetey, Institute of Statistical, Social and Economic Research (ISSER), Legon-Accra, Dr Daniel Bruce Sarpong, Department of Agricultural Economics and Agribusiness, University of Ghana, Mr Fred Danku, Institute of Statistical, Social and Economic Research (ISSER), Legon-Accra; for Kenya, Professor Willis Oluoch-Kosura, African Economic Research Consortium (AERC), Dr Stephen K. Wambugu, Department of Geography, Kenyatta University, Dr Joseph Karugia, the same department; for Malawi, Mr John Kadzandira, Centre for Social Research, University of Malawi, Zomba, and Dr Wapulumuka O. Mulwafu, Faculty of Social Science, University of Malawi, Zomba; for Mozambique, Dr Peter Coughlin, EconPolicy Research Group, Ltd, Maputo; for Nigeria, Professor Olatunji Akande, Nigerian Institute for Social and Economic Research (NISER), Ibadan, and Dr Olorunfemi Oladapo Ogujndele, the same Institute; for Tanzania, Professor Aida Isinika, Institute of Continuing Education, Sokoine Agricultural University; for Uganda, Dr Bernard Bashaasha, Department of Agricultural Economics and Agribusiness, Makerere University, Kampala; and for Zambia, Mr Mukata Wamulume, Institute of Economic and Social Research (INESOR) and Ms Charlotte Wonani, Development Studies Department, University of Zambia. The Swedish team consisted of the following team from Lund University: Professor Göran Djurfeldt, Department of Sociology, Associate Professor Magnus Jirström, Dr Agnes Andersson, Ms Johanna Bergman Lodin, Ms Cheryl Sjöström, Department of Human Geography, Professor Björn Holmquist, Ms Sultana Nasrin, Department of Statistics. Associate Professor Hans Holmén, Linköping University, was also a member of the team. We had a distinguished group of advisors, including Professor Göran Hydén (now Emeritus), University of Florida and Prof. Oliver Saasa, Institute of Economic and Social Research, University of Zambia. The authors wish to thank Martin Andersson, Christer Gunnarsson, Ellen Hillbom and Tobias Axelsson, whose inputs in various ways have contributed to the following analysis. Thanks are due to two anonymous reviewers from Oxford Development Studies. The research was funded by the Swedish International Development Authority (Sida) and the Swedish Research Council (Vetenskapsrådet).