Migration and nonfarm activities as income diversification strategies: the case of Northern Ghana

Abstract

ABSTRACT This article jointly analyses the determinants of participation in nonfarm activities and of migration in Northern Ghana, combining household data with community-level information and data on the evolution of the yields of cash and staple crops. Our results confirm the role of education as a key asset to pursue opportunities in the off-farm sector and the role of farm size in reducing the probability of participating in nonfarm activities. Poor households do not have enough resources to undertake nonfarm activities and they opt for migration as a diversification strategy. Community-level assets are found to play a crucial role for understanding off-farm diversification.

Résumé Cet article analyse les facteurs menant à la participation aux activités économiques au-delà du secteur agricole ainsi bien que le processus de migration au sein des communautés rurales dans le nord du Ghana. S'appuyant sur les données disponibles sur la revenue et la production des récoltes des ménages, nos résultats démontrent que l'éducation est clé dans la poursuite d'autres formes d'emploi dans les secteurs autre que le secteur agricole. La taille des fermes constituent aussi un facteur important dans cette dernière. Les ménages n'ayant pas assez de ressources pour trouver d'autres formes d'emploi doivent, par contre, recourir à la migration. Finalement, les biens communautaires jouent un rôle important dans le processus de diversification économique.

Keywords:

• income diversification
• nonfarm activities
• migration
• Ghana

Acknowledgements

The author is grateful to the editor, John Harriss, to three anonymous referees and to Simone Bertoli, Frank Ellis, Ornella Giambalvo, Leonardo Grilli and Donato Romano for their comments and suggestions. The usual disclaimers apply.

Notes

See Rimmer (1992) and Hutchful (2002) for the economic policies and their impact on growth adopted by Ghana since independence and on the structural adjustment policies respectively.

See Barrios, Bertinelli, and Strobl (2006, 2010) for empirical analysis on the impact of weather anomalies on internal migration and economic growth in Sub-Saharan African countries.

Lewis (1954) and Harris and Todaro (1970) laid the theoretical foundations for the analysis of migration out of rural areas.

See Hart (1973) for an early contribution on informal economic activities in Ghana.

We use the term off-farm diversification or activities when referring jointly to migration and local nonfarm activities.

This article refers implicitly to a unitary household model where the adopted livelihood strategy is the outcome of a household-level decision; see, for example, Newman and Gertler (1994), Yang (1997) and Kimhi (1998) for the analysis of the determinants of individual labour supply decisions in developing countries.

An overview of the relevant literature is provided when describing the setup of our empirical analysis.

The literature offers contrasting views on the consequences of these activities on income inequality –see Canagarajah, Newman, and Bhattamishra (2001) and Senadza (2011) on this.

The list of the 1984 population census enumeration areas, which contains population and household information, was used in the process of sample design. The enumeration areas were first stratified according to the three ecological zones – coastal, forest and savannah – and then within each zone further stratification was conducted with respect to the rural or urban location.

Each enumeration area can contain more than one rural community and the GSS does not disclose, to ensure confidentiality, the information that would allow connecting households to their community; luckily, in the three Northern regions, we have only one community for each sampled rural enumeration areas.

The GLSS4 reports non-negligible rental incomes, but this mostly reflects the imputed value of the rent for households who own their house, and in this view this income source is not included in our analysis; only 3.2 and 0.8 per cent of the households in our sample report income from renting livestock and equipment respectively.

The GSS includes (v) in the definition of agricultural income; still, the sale of processed crop products requires an additional, and often substantial, effort besides the one devoted to farming, and it may require the use of specific tools. The analytical choice to include the sale of processed crops among nonfarm income sources is in line with Assan et al. (2009) and Gordon and Craig (2001), who include among nonfarm activities petty trading in cooked food and drinks. Beer brewing, fish processing and rice parboiling are some examples of typical activities undertaken in Northern Ghana.

The extreme poverty line was set at 700,000 old Ghana cedis per adult equivalent per year (approximately 260 US dollars) by the GSS on the basis of the cost of purchasing a basket of goods required to achieve minimal nutritional requirements, estimated at the prices prevailing in Accra in 1999 (GSS 2000c); the GSS provides the factors required to account for variations in prices across space and time.

Poverty is highest among the food crop farmers in Ghana; around 58 per cent of the poor belong to households primarily engaged in food crop cultivation, while around 24 per cent of the poor come from households whose main income source is represented by nonfarm self-employment (GSS 2000c).

The dependency ratio is defined as the number of dependent members, individuals aged below 14 or above 65 years, over household size; this variable is not correlated to the number of working age members, while it is positively correlated to the household size.

The same argument is suggested by some studies maintaining that households with fewer children under the age of five are more likely to participate in migration (Adams 1993; Lipton 1980).

According to Davis et al. (2007), female-headed households have a lower propensity to participate in nonfarm activities and they are more likely to receive both private and public transfers. Still, some studies have emphasised that the participation of women in nonfarm activities is on the rise in Sub-Saharan Africa: women take part in wholesale or retail trade and in manufacturing, in particular in the informal sector (Gordon and Craig 2001). Bryceson (1999) argues that gender barriers to participation in a wide set of activities are rapidly declining, and Marchetta (2011) suggests that a similar process is occurring in Northern Ghana. We have tried to include the gender of the household head in our model, but we did not find any significant result, so we decided not to include this variable in the final specification.

“The moderately poor group was found to migrate more than the other groups … and migration reduces the pressure on the resources of poor families, and is a way of building up investments. Remittances also help in land investments or building up entitlements by those who stay behind” (Yaro 2002, 18).

We define a household as moderately poor if its income is between the extreme poverty line and the moderate poverty line, as nonpoor if its income is above the moderate poverty line. The moderate poverty line was set as 900,000 old Ghana cedis per adult equivalent per year (approximately 334 US dollars).

Sixty-two per cent of the households with income from all sources receive transfers from urban areas, while the corresponding figure for households with income from agricultural activities and remittances is 52.5 per cent.

The Ministry for Agriculture provides information on the quantity of millet and groundnuts produced per hectare by year and district; using data on real prices of crops from GSS, we compute the real value of one hectare of cultivated land.

This variable is highly correlated with the presence of a junior secondary school and the availability of public transports.

It is important to observe the large variability, reflected in the standard deviation, of the measures of the evolution of the yields of the two crops.

We also included regional controls in x , to mop up any unobserved geographical factors which influence the choice of the preferred livelihood strategies, but these additional regressors proved to be non-significant.

The variable is equal to: 0 if all members have less than six years of education (primary education); 1 if at least one of the members has attended six years of schooling; 2 for some secondary education; and 3 if at least one of the members has completed secondary education.

Farm size refers to the size of the land owned or operated by household members.

As a robustness check, we have run the model using the ratio between the value of self-consumption and the poverty line, which probably represents a more precise measure of the ability of agricultural production to satisfy the basic consumption needs of the household. Results, which are available from the author upon request, are qualitatively unchanged, though we are aware that concerns about endogeneity still apply.

We also omitted this variable from the three specifications, obtaining similar results to those reported in the article; these estimates are available from the author upon request.

We include these variables in just one specification as we acknowledge that their informational content might be limited by measurement error.

For each specification, we tested the adequacy of the independence of irrelevant alternatives assumption (IIA) through the test proposed by Hausman and McFadden (1984), which entails re-estimating the model on a smaller subset of alternatives; the results of the tests are supportive of the distributional assumptions which underpin the multinomial logit model.

Loosely speaking, an effect is deemed to have an economic significance when its size is such to produce a nontrivial impact on the phenomenon of interest; a coefficient can be statistically significant while explaining only a limited part of the variability in the dependent variable (see McCloskey and Ziliak 1996).

If we also include age squared, we find an inverted U-shaped relationship between age and diversification in nonfarm activities, with the turning point around 40 years.

We consider as representative household a household whose variables take the median values of continuous variables and the value of the modal class for dummy and discrete variables. The representative household has two working age members, a dependency ratio equal to 0.5, with a 35-year-old household head, and with all illiterate members; it operates 5.1 acres of land with no agricultural equipment, it owns livestock, and its home production covers 23 per cent of its consumption, with its income falling below the extreme poverty line. It resides in a community with no market or hospital, and farmers are periodically visited by extension officers.

For example, if our representative household lived in a community having access to a market, then the predicted probability to be engaged in off-farm activities would rise to 0.63.