Decision-making for heterogeneity : diversity in resources, farmers' objectives and livelihood strategies in northern Nigeria

As a result of increasing population pressure, the average farm sizes in the savannah regions of West Africa have reduced. By consequence, farmers can no longer rely on fallowing to maintain soil fertility. For long farmers have therefore resorted to other methods. The most common on-farm strategies to cope with reduced fallow lengths are rotation of cereals with nitrogen fixing legumes and crop-livestock integration. The most important component of crop-livestock integration is the feeding of crop residues to livestock and the subsequent use of manure as fertilizer. At the same time, many farmers can no longer rely on farming as their sole source of income and diversify into off-farm income sources such as petty trading; local manufacturing jobs; or migrate (seasonally) to large urban areas. Hence, the coping strategies in the wake of increased population pressure are manifold, and the rural population is far from homogenous. The aim of this study is to examine in detail three types of heterogeneity and their relationships with agricultural production. These three types of heterogeneity are: (1) heterogeneity in farmer goals and objectives, (2) heterogeneity in (on-farm) soil fertility resources, and (3) heterogeneity in crop-livestock integration. We thereby explore how differences in household characteristics and farming strategies relate to the three types of heterogeneity distinguished, and how this affects soil fertility levels. These types of heterogeneity affect production (decisions) and farmer soil fertility resources in different ways. First, developed with the purpose of analysing the ex ante impact of policies and technologies on farmers’ soil nutrient use, bioeconomic models frequently assume that farmers are homogenous in goals and preferences, i.e., their underlying utility function. Similarly, many studies on smallholder productivity and efficiency only include observable household characteristics and thereby implicitly assume that the relationship between household characteristics and farmer goals and objectives is homogenous. In neither type of study there is a clear reason to assume that such behavioural homogeneity holds. More importantly, ignoring farmer specific goals and objectives may lead to incorrect simulation outcomes from a bio-economic model, as well as biased estimates of efficiency or productivity. In both cases this could lead to ill-formulated policy recommendations. This is further investigated in this study. Second, most studies focusing on productivity and efficiency in agricultural production assume farm size to be homogenous with respect to its soil composition, an assumption refuted by numerous field studies. Again, ignoring such information may lead to biased estimates and policy recommendations. Third, livestock clearly plays an important role for production of manure, but manure production is not the main reason for households to keep livestock. Next to meat and other tangible benefits contributing to farm incomes in kind or in cash, several other non-tangible benefits, such as insurance and storing finances, play a role. The importance of these non-tangible preferences for keeping livestock may differ from one household to the other, giving rise to differences in the degree to which a farmer integrates crops and livestock. These types of heterogeneity are further analysed in Chapters 2 till 5 of this study. Thereby use is made from various data sources from northern Nigeria. The data used includes farmers from villages in different agro-ecological zones in northern Nigeria, as well as villages characterized by different levels of market access. The villages also differ in population density, but levels of agricultural intensification are high throughout the region of study, with fallowing non-existent in nearly all locations. The description of heterogeneity in farmer goals and objectives, and their effect on smallholder efficiency and on soil nutrient budgets is the subject of Chapters 2 and 3 respectively. Chapter 2 follows an explorative approach in documenting the various farmer goals and objectives. While arguable risk aversion and profit maximization are important attributes in farmer decision-making, other preferences and attributes may equally play a role. To capture such additional variables, a fuzzy pair-waise goal ranking is combined with a set of Likert scale questions. Principal component analysis is used to reduce these data into behavioural factors, i.e., the minimum set of underlying behavioural latent variables. We subsequently estimate technical and allocative efficiency levels by using Data Envelopment Analysis and analyse how these are related to farm characteristics and the identified behavioural factors. The models in which both intended behaviour and farmer characteristics are included give a significantly better fit over models in which only household characteristics are included. More importantly, next to expected effects of risk aversion, two other behavioural variables are identified that influence efficiency levels. These variables reflect the desire to be a successful farmer and the desire to fulfil subsistence demands from own production. On the other hand, the overall effects of these behavioural variables are small in relation to other observable household characteristics, and additional research should focus if and how agricultural policies should account for this heterogeneity. In Chapter 3, the relationship between differences in goals and objectives and on-farm soil nutrient budgets is explored in more detail, by using a combination of multi-objective programming, multi-attribute utility theory and bio-economic modelling. The first part of the analysis establishes trade-off curves between the most common production attributes included in smallholder studies, i.e., optimisation of gross margins, labour use, risk levels and sustainable use of soil resources. The estimated trade-off curves reveal that farm plans aimed at optimising gross margins and, arguably, sustainable use of soil resources are more favourable, considering the nutrient balances, than those aimed at minimising production risks. In the second part of the analysis, by using multi-attribute utility theory, farmer specific weights for each of these attributes are identified. Risk aversion, operationalised through variance minimization, appears an important attribute in this study for many farm households with smaller land holdings. Subsistence production of cereals is dominant in such farm plans that lead to negative soil nutrient balances, especially for potassium. Farmers who place a large importance on gross margins in their utility function are likely to benefit most from policies aimed at enhancing profitability through improving the functioning of markets. The large group of risk averse farmers will have the largest immediate gain in utility from policies and technologies aimed at lowering production risk in high-value crops. Additional policies aimed at creating a stronger market–oriented production by the least-endowed farm households could play a role in reducing intensity of soil fertility mining. Then, the efficient cropping pattern shifts (partially) from cereal cropping to high value crops, associated with higher input use. In Chapter 4 it is analysed how heterogeneity in soil fertility resources at farm level affects maize and sorghum production, and measures of technical efficiency for these crops. While arguably crop production is dependent on natural soil fertility levels, this is not always taken into consideration in production function estimations. Two variables that can easily be derived from household production surveys are introduced as proxies for on-farm heterogeneity in soil fertility. Next to these proxies, detailed soil fertility data at village level is included to account for differences in soil fertility levels between villages. The results show that the used soil fertility variables have significant effects on production, although not always of the expected sign. Secondly, the inclusion or omission of such soil fertility variables plays a critical role in testing for the presence of inefficiency. In the case of maize production, inefficiency is no longer observed after inclusion of the soil fertility variables. Finally, variation in labour availability is an important determinant of the inefficiency found in sorghum production. The findings highlight the need to further develop and include proxies for on-farm soil fertility heterogeneity in smallholder efficiency and productivity studies. In Chapter 5 it is investigated how preferences in non-tangible benefits of keeping livestock relate to differences in herd size and crop choice at different types of farms. Integrating crops and livestock is widely advocated as a method to maintain soil fertility levels through increased use of manure. On the other hand, there are many other benefits of keeping livestock, such as insurance and storage of finances, in addition to manure production. The role of such non-tangible benefits could differ across farms, thereby driving apparent differences in observed levels of crop-livestock integration. First, a bio-economic simulation model is used to identify, at different farm types, the relationships between preferences for non-tangible benefits, optimal herd size and crop choice. The simulation outcomes show that optimal herd size increases for non-tangible benefits, though herd size decreases again for increased importance of tangible benefits, i.e., liveweight production. Furthermore, the results from the model suggest that for increasing labour supply, herd size decreases due to a shift into vegetable cultivation and consequent reduction of on-farm fodder supply. Second, a novel method to measure non-tangible benefits empirically is introduced in this chapter. This measurement is done by calculating the difference between simulated herd size at maintenance levels, given on-farm fodder supply, and actual herd size observations. A regression analysis shows that farm households wellendowed with farm and labour are more likely to maintain (too) large herds, possibly as a mechanism to store finances. Consequently, these farmers also use more manure. It shows that herd size increases demand for fodder products, while there is additional evidence that manure use benefits cereal production, but does not benefit other crops. Hence, similar to the results in Chapter 3, these results suggest that specific policies and technologies are needed to enhance use of manure at the least-endowed farm households. Finally, in Chapter 6 the wider implications of the research findings and the methods used are discussed. More specifically, three topics are discussed in more detail. First, it is discussed how research should further address heterogeneity in goals and objectives in various types of study. It is thereby argued that experimental field research methods could potentially further improve the accuracy of latent variables, including the newly identified ones in Chapter 2. Furthermore, such variables could further shed new light in other agricultural or development studies at smallholder level, such as (dis-)adoption studies. Second, it is discussed how simulation models can be improved for more accurate design of policies to promote growth at smallholder levels. Both the inclusion of heterogeneity in farmer goals and objectives, as done in this study, and the use of robust optimisation methods to account for data uncertainty thereby play an important role. Finally, the implications of this research for the enhancement of sustainable use of soil resources in the savannah regions in Nigeria, and Africa in general, are discussed. Most importantly, the results in Chapter 3 and 5 suggest that mostly wellendowed farmers tap into markets for high-value crops, and thereby use more organic and inorganic inputs. Hence, research should focus on how production, and sustainable use of soil resources, at the least-endowed farmers can be enhanced further. This can be partially achieved by developing technologies that reduce the risk in the cultivation of high-value crops and policies aimed at bringing the leastendowed to the market. Potentially this can be achieved through cooperative agreements between farmers, but little is yet known if and how such agreements can play a role. This can be further investigated by combining bio-economic simulation models and methods from cooperative game theory.