Cash transfer programmes have been proposed as a means of providing benefits to targeted individuals or households in poorer African countries. Pilots have been run in some poorer countries with limited administrative capacity, but have not so far been expanded to a national scale.

A large programme aimed at poverty relief would require information to establish targeting criteria, and to estimate the proportion and number of poor households/people in different locations. These questions are not specific to the design of cash transfer programmes, but raise a general and longstanding problem of poverty measurement. Surveys are often dated, data is often difficult to obtain and the reliability of much of the income data is questionable.

A simplified method of obtaining household budget data, the 'household economy approach' (HEA), is widely used at a national scale in southern Africa for crisis prediction. The HEA model is used to simulate the impact of changes in the economic context, for example the impact of a crop failure and/or a price change on the (reference) income established by survey.

The HEA is practical and economical in use and appears to give reliable income estimates. However, the technique uses a simplified data set which does not allow the level of discrimination between households necessary for the design of social protection programmes.

The Individual Household Model (IHM) was developed to overcome the limitations of HEA witha view to modelling a wider range of changes to individual households. IHM is based on household demographic, asset, and income data obtained from individual household interviews. The limitation of IHM is that so far it has not been applied on a large scale and is relatively expensive in use and demanding of skills and organisation.

An extended HEA model (HEA+) was designed to combine the practical advantages of HEA and at least some of the detail supplied by data from individual household surveys. The question is whether the collection of a small amount of data additional to the standard HEA data set provides a sufficient increment in information to potentially extend the use of HEA to cash transfer programmes.

The study was conducted in Kazangula District in Southern Zambia from a base in Livingstone. The initial intention was to work in four villages in two livelihood zones, for one of which (Zambezi Valley West) HEA data already existed. However, unforeseen difficulties occurred and data was only obtained from a single village. The economy of the study village is primarily agricultural, and food aid was distributed in the village through a number of channels.

Household income estimates were made using HEA+ and IHM. A fairly good fit is obtained between the two methods. The differences between the income estimates obtained using the straight line model and the individual household data are 0.2%, 8.5%, 13.5% and 3.9% for the 'very poorest', 'very poor', 'poor' and 'middle' groups respectively. There is a close correspondence between the actual household income estimate in the reference year and the HEA+ model. The findings also tend to support the reliability of the HEA data.

HEA+ can be used to obtain estimates of:

• the proportion of poor households/people in each livelihood zone,
• he cost of bringing this population up to the standard of living threshold,
• changes in poverty rates following changes in production, assistance and the price of traded goods,
• information which may be useful to establish targeting criteria.

Assuming that an HEA data set was being gathered or an existing data set was being updated, the additional cost of using the HEA+ model would be very low. On the experience of the pilot the HEA+ data set would add approximately 5-10% to the work required to gather a 'standard'. HEA data set.

A single small study is obviously insufficient to establish the validity of the proposed method and further experimentation is required. Further testing of the method would be most simply and economically done in a location where existing HEA data sets were already being updated. This would give a much larger HEA sample than was available in this study. Individual household income data could be obtained from an appropriate sample of households from the HEA sample sites, rather than from a single village.

However, most countries already have a poverty measure, and if HEA+ were used as a poverty measure there are outstanding questions about sampling. It is therefore important for further development of the method to be done in agreement and discussion with the relevant national agencies.