Most developing countries lack the capacity to analyze their agricultural sector and in particular its poverty dimensions, formulate strategies, and propose programs for international and national funding. The need for modelization of the agricultural sector – including crops, livestock, forestry and other sub-sectors – is acknowledged by many development agencies. A “Multipurpose Agricultural Data System” (MADS) is thus proposed. The objective is to develop a system able to deal not only with Poverty Reduction Strategy Paper-type sector analysis but that could also be used as a decision tool for testing policy scenarios, for sector monitoring, and for regional and project level analysis.

The core model would represent a production system i.e. a production unit / farm model defined as a linear combination of crop and/or livestock systems and other farm or non-farm activities. A region, a development program or the entire agricultural sector of a country is then represented as a linear combination of production units.

The system is based on simple deterministic cost-benefit analysis models, i.e. it is the user responsibility to determine the value of the parameters (e.g. milk production per lactation of a cow, yield of a crop), and the software would do no more than compiling the data to produce results.

The above means that – at least in this version – MADS is not designed to handle other forms of analysis such as optimization models (linear programming), probabilistic models (stochastic variables) or rule-based decision models (expert systems).

The MADS approach wants to be a synthesis between the manipulation of non-typed variables and the detailed structuring of data found in other software. The idea is to structure the data in order to allow for maximum pre-definition of calculations, as well as to provide for mechanisms that would let the user specify additional calculations and define the format of customized reports (tables). The number of data types is kept limited by combining generic and specific data structures in order to provide maximum power and flexibility.

MADS Data Structures

MADS offers two primary generic data structures (data object types), called Commodity and Plan, that are meant to be general enough to accommodate most basic analysis situations, as well as a specialized data structures to handle livestock demographic models (Herd). At the lowest level, plans can be thought of as representing activities as a combination of commodities produced or consumed and/or herds. These activities can in turn be combined linearly into higher-level plans. For example, one can define a farm model as a plan made of crop plans and herds, and a given sector, region or project as a plan made of farm plans. The PLAN is thus the structure that represents a particular model.

Calculations will be performed and reports produced for a given plan. Calculations corresponding to commodities and herds within a plan are predefined. Additional variables can be specified and assigned values through user-defined calculations – that can make use of commodity or herd results within a given plan – by defining scripts. A Script would be executed for a given Plan by adding it to its list of components.

Finally, MADS would offer a fifth generic data structure, called Table, to handle the production of user-defined table reports. The production of a report for a given Plan is triggered by a command included in a Script.

Within all the above data structures, information would often take the form of time series. For the sake of simplicity, MADS offers only one time step, assumed to be the year (the entity Herd is a special case – see details in the Herd reference section). One could decide to use a different time step and set the different parameters accordingly. Future versions might offer more ways to specify series of values over time, including options for mixing different time steps (year, semester, quarter, month or even week).

Using MADS

The system would thus be able to deal with the following:

In practice, the first step in the analysis of the agricultural sector of a country, or the crop sector, the livestock sector, a given region or an investment operation, is to identify all stakeholders, and in particular to identify representative production or farm models, ideally by drawing on recent surveys and national statistics that indicate the typology of production systems. Each model would eventually be represented in MADS by a Plan. Although MADS, as currently proposed, does not include any poverty-specific feature, poverty focus would come from the selection of the production models, and of their unit size. Selected indicators of wealth (net cash income, financial return per day of labor) can be calculated for each model and compared to thresholds defining poverty for the area.