Horst Wattenbach
Typically, field studies provide information from one or two sources of primary data: PRA surveys and household interviews. The collection of household data complements PRA results on the impact of globalization in light of policy reforms, but also provides the basis for household modelling.
For the sake of discussion, a spreadsheet-based model will be described. However, the same logical structure and data needs would apply in using other software for household modelling.
Framework for household modelling
Usually, a research framework establishes three organizational parameters which need to be considered during model specification: capacity of research team which will construct the model; the replicability of the model in other locations of the region; and the comparability of the structure of these models.
As a consequence, the software decision and model structure needs to find the appropriate balance between realistic representation of socio-economic factors at the household level, while remaining transparent enough to be developed and interpreted and the results successfully communicated to local decision makers. This suggests the use of readily available programs and not overly complex ones.
Normally, household models would be based on group averages for different household types in each location (e.g., food insecure, food secure and surplus producers).
Objective of household modelling
The contribution of household modelling in the overall policy analysis context would primarily be on the assessment of household level effects of policy change. Three impact areas are usually considered: income effects of policy change by household type; allocation decisions of production resources by household type and changes in consumption levels and possibilities by household type.
The comparative analysis of the results of the different household models allows the comparison of the impact of household characteristics as well as market access on the above three factors. To some extent, the household models are also useful to test the impact of compensatory measures for agricultural producers under risk of loosing from policy change.
Scope of household modelling
The specifications of household models often have the following scope: agricultural production, including the use of common property resources for livestock. Forest products and collected resources are generally omitted, except in cases they are traded by farmers; the model includes household consumption in the objective function of the household; to the extent possible, production risk factors are included in the models and the model is comparative static in nature, i.e. does not include herd dynamics or investment decisions at the household level.
Household and village modelling are closely related, and share a considerable part of field data. The main difference - from a household modelling perspective - lies in the fact that in the household models, typically: only agricultural producers are considered; the production process and production constraints are more disaggregated; consumption might be more aggregated than for village modelling and socio-economic factors are more readily considered than in village modelling.
The results of household modelling provide, at the same time, the possibility of some consistency checks with the village model results. The limitation lies in the fact that only agricultural producers are covered by household models, but their extrapolation to the producer population in the study villages is equivalent to the primary production sector of the village model.
The aggregated input demand of producing household likewise coincides with part of the resource transfers at the village level, and may also be used for consistency checks.
The following section specifies the data needs for household modelling, whether spreadsheet-based or a mathematical programming type agricultural household model. Crop-livestock integration (where applicable) can be modelled by the internal transfer of resources from crop to livestock gross margins. Farm level resources to be drawn upon include land of different categories and labour by gender and major seasons (or months with peak labour requirements).
Furthermore, capital requirements (including credit availability and working capital requirements for crop and livestock activities) are incorporated. However, modelling seasonal effects of capital inflows and outflows is complicated and data requirements are excessive. However, capital can be considered at the level of formal and informal credit requirements only.
It is advantageous if the models are based on template allowing simulation of household types in different locations. Restrictions on permitted activities in each of the locations can be imposed in the constraints fields of the model, in addition to the standard restrictions on resources available for these activities.
Available household resources
Land by different land categories, i.e. land under perennial crops (as it cannot be rapidly converted to other uses), land under annual crops (which could be allowed to shift rapidly between annual crops if the farmer prefers to do so) and land unsuitable for cultivation (permanent grazing land); grazing land without restrictions on conversion to agricultural land must be quantified, separately from marginal land (as it MUST remain grazing land).
Labour: number of family members, differentiated by the relevant age as well as gender categories. Relevant categories are those which determine and restrict production activities. Therefore, limits of school attendance, availability for executing light labour (such as livestock herding) and full labour. Restrictions on cultivation practices which limit the expansion of certain crops also fall into that category.
Especially in the case of West African households, resource management decisions are sometimes determined by complex resource access rights as well as social norms on contributing certain types of food to the kitchen of the extended family. Therefore, in some cases, e.g., Mali, household models might require additional specifications compared with East African household models. The dialogue on the trade-off between site specificity and use of model templates continues.
Information on production activities
Agricultural production is entered into the system with respective gross margins according to the textbook definitions. This includes standardized data, mostly on a per-hectare basis (for details, see below). as well as per head per year for milk cows, ewes, does and breeding males. For fattening of animals, the entire fattening period (after weaning) is considered as one unit. For most agricultural activities, different intensity levels will be defined. Intensity levels are generally understood in terms of yield, and corresponding combinations of inputs (new seed, fertilizer, better management etc.)
Crop production: This includes standardized data, i.e., per hectare data, for all crop production activities. Data needs include: yield (kg/ha) (main product as well as by-products, including straw/stover; physical inputs: seed, fertilizer, chemicals (kg/ha); labour inputs (days/ha) by month (periods without labour constraints can be omitted) (based on a cropping calendar, the conversion from labour for peak requirements can be converted from activity (such as land preparation, weeding etc) to calendar periods; and other inputs (ploughing animals, tractor services etc.)
A particular difficulty lies in the definition of mixed cropping activities. In these cases, area based definition of cropping will be carried out, hence resulting in the production of an output mix (e.g. maize and beans) from one unit of land with the associated inputs. Some standardization of mixed cropping elements by household type and farming systems zone is required. Details are to be defined with the local teams.
Livestock production (ruminants only): Standardized data for livestock are on a per-head-per-year basis for milk cows, ewes, does, breeding males. For fattening of animals, the entire fattening period (after weaning) is the unit.
The raising period is defined as the period from weaning to first calving/kidding/lambing. In most cases, no separate management will be carried out, so all input use will simply be linked to that of production animals. However, the reference periods (length of raising period etc.) for keeping the animals MUST be specified (as it affects the variable costs).
Alternatively: if gross margins for all animal groups (breeding females, raising of replacement animals and breeding males) will be unavailable, aggregated livestock units could be used (based on the existing herd composition). Fattening enterprises should always be kept separately, due to their positive impact on reducing pressure on grazing land and income generating possibilities.
Labour requirements
Labour requirements for livestock management need to consider the existing requirements as well as the capacity for expansion of herd sizes with the existing labour force. Particular in grazing management, there is a large difference between existing labour use per head of livestock and maximum management capacity. Labour requirements for livestock management need to consider social norms on who performs certain activities, such as: men do not herd sheep or goats in some societies, therefore small ruminant flocks are not restricted by male labour constraints. Please consider similar restrictions in your case.
Feeding regime: periods with exclusive grazing, periods with complementary as well as exclusive feeding with farm or purchased feed (including type of feed)
Information on consumption activities
For the main agricultural products, the use of output at present is required. This would serve as a minimum production constraint on basic foodstuffs to meet consumption requirements. Models will allow excess production to be sold freely. Additional consumption is (for the level of household modelling) mostly required as cash requirements for consumption purposes.
Collection/Gathering activities
So far, only sheanut gathering (and processing) in Ghana has been identified as a relevant activity which might require inclusion in the household models.
Working capital and credit
Working capital and liquidity constraints frequently limit the farmers’ choice of activities. This situation has become more serious where crop finance or agricultural credit has been reduced through the political adjustment processes. At the same time, the capital constraint can only be modelled to a limited degree.
Credit available for crop finance will be the key component, especially fertilizer, seed etc needs to be specified (interest rate, credit limits)
Linear programming models are strict maximizations of the objective function. Therefore, considerable care needs to be taken to specify the objective function of the households, as we intend to integration production and consumption decisions and to a certain degree also attitudes to (perceived/real?) risk. Furthermore, some restrictions on allowing certain activities to occur need to be imposed (upper bounds/minimum activity level). For the latter, information on reasonable maximum activities needs to be provided. Otherwise, the models will yield extreme results, or become unsolvable.
Typically, constraints are set for one or more of the following: marketable quantities; total hired labour availability; reasonable limits to family members working off-farm at a given wage rate; shift from one production activity to another one; restrictions on technology levels available for some household types and credit/working capital constraints?
The following farm level effects of policy modelling are often analysed: price changes of inputs and outputs; changes in constraints (capital, resource availability); and availability of new production technologies (different levels of existing or new crop or livestock enterprises).
Typical policy scenarios for household modelling are: staple food price change, e.g., the effect of 10 percent increase in major staple food prices; cash crop export price change, e.g., the effect of 10 percent increase in cash crop prices; agricultural export price, e.g., the effect of 10 percent increase in the prices of agricultural exports; change in agricultural subsidies, e.g., the effect of 10 percent increase in the subsidies on agricultural commodity productions, or the 10 percent reduction of value-added tax on agricultural commodities; productivity gain: impacts of technological progress in staple food production; income transfers, e.g., household receives 10 percent higher transfers from government or the rest of the world; wage rate change: increase in hired labour cost in staple food production; migration remittances change, e.g., increase in migration remittances to the village; or devaluation, e.g., 10 percent devaluation of currency exchange rate.
The decision about the software to be used needs to consider the following factors: capacity of the local teams; a standard template as the basis for all country teams; models transparency in structure and output tables; and software costs.
Potential candidates for those purposes are: Excel solver; XA83; GULP; Lindo (including Lingo, WhatsBest!, Lindo) and GAMS.
Hardaker, J.B., Pandey, S., & Pattan, L.H. 1991. Farm Planning under Uncertainty: A review of alternative programming models. Review of Marketing and Agricultural Economics 59: 9-22.
Hazell, P.B. & Norton, R.D. 1986. Mathematical Programming for Economic Analysis in Agriculture.. Macmillan Publishing Company, New York.
McCarl, B.A. 1984. Model Validation: An overview with some emphasis on risk models. Review of marketing and Agricultural Economics 52(3): 153-173.
McCarl, B.A. & Apland, J. 1986. Validation of linear programming models. Southern Journal of Agricultural Economics 1986: 155-164.
Pannell, D.J. 1997. Introduction to practical linear programming. John Wiley and Sons, New York.
