A New Household Economy Method for Assessing Impact of Shocks

By Celia Petty and John Seaman¹

Celia Petty is Social Policy and Livelihoods Adviser at Save the Children UK. She has worked as an adviser on food security and livelihoods assessment methodologies and on policy and programmes for children affected by conflict.

John Seaman is currently a consultant working on operational approaches to measuring poverty. He was formerly Head of Policy, and Development Director of the Food Security Unit of Save the Children UK.

The work described in this report was funded by a grant from the UK Department for International Development.

This article describes the experiences of using an adapted household economy assessment approach in Uganda, Ethiopia, Swaziland and Mozambique. In Uganda and Ethiopia the approach was used to look at the impact of falling coffee prices on household poverty while in Mozambique and Swaziland the assessment estimated impact of HIV/AIDS on livelihoods and economic status. The main focus of the article is on the findings of the Mozambique and Swaziland assessment on impact of HIV/AIDS. This is particularly topical in the humanitarian sector as in spite of the common usage of terms such as 'new variant famine' and the increasingly accepted view that the HIV/AIDS pandemic is having a marked impact on food security, there is very little empirical evidence. The work described below is an almost unique snapshot of the impact of HIV/AIDS at community and household level. To some extent the findings counter the prevailing view that HIV/AIDS decimates local economies and livelihoods in high prevalence areas (Ed).

This article describes a new approach to poverty analysis and household level modelling. The method is based on the established Household Economy Approach (HEA)¹, and has been designed to assess the effects of policy changes and other defined shocks on disposable income and living standards. Four pilot studies were carried out during 2003 in Uganda, Ethiopia, Swaziland and Mozambique by Save the Children UK (SCUK). The aim was to test whether the approach was a practical methodology for field use, and whether output would provide a more rigorous basis for policy analysis, programme implementation and impact monitoring.

HEA was developed to predict the effect of a 'shock' (e.g. a crop failure or price change) on people's ability to maintain their income and to meet their survival and developmental needs. It is based on a quantitative and qualitative description of the economy of a defined population². A simple model is used to predict the likely impact of a particular event or events on the ability of households in different wealth categories to acquire sufficient food and meet defined nonfood needs.

The original purpose of HEA was to provide large-scale (e.g. national) predictions of food emergencies. Although it has been applied in other development contexts, its main use is still in the area of emergency assessment. This is largely due to the use of a simplified data set, with only one 'typical' household defined in each wealth group. The new model is based on a representative sample of individual households and is designed to handle a more complex data set and to predict the outcome of a more diverse set of problems. It has therefore been termed the 'Individual Household Method' (IHM)

The focus of the pilot studies was a) the household impact of falling coffee prices and b) the impact of HIV/AIDS on household economy. The effect of falling coffee prices was selected as the relationship between household poverty and internationally traded commodities is poorly understood and has attracted wider public interest. The impact of HIV/AIDS on household economy was selected as this subject presents major methodological problems (e.g. the difficulty of establishing control groups) which household methods are well suited to deal with. The debate around HIV/AIDS and food security also remains highly controversial. This article mainly presents the findings from the pilot studies in Mozambique and Swaziland, which examined HIV/AIDS and household economy.

It should be noted that the analysis from the case studies refers to the study sites only³.

Overview of methodology: the Individual Household Economy Approach

The individual household method differs from standard HEA in three ways:

1. Arandom sample of individual households is used (usually obtained by village mapping/ transect walks). Standard HEA is based on analysis of households that are 'typical' of a defined wealth group.
2. Results are expressed in terms of household disposable income (rather than the ability of a household to acquire food, given some level of non-food expenditure). The output produced shows the impact on household living standards across the population, given a defined change.
3. Because each individual household is described (rather than the 'typical' household), there is the possibility of extending the data set and model to include changes within the household e.g. in the case of HIV, changes to household demographic composition.

Data collection and analysis

Villages are mapped and transect walks used to draw a representative sample of households. Information is then collected using standard methods. These include desk research, key informant interviews, focus group discussions and interviews with individual households.

The basic data set required from each household is:

1. Household demography, including gender and age.
2. The sources and amounts of household income from each income source.
3. Land and livestock holdings.

Additional extended interviews establish the local costs of food and basic items, which are used to establish a standard of living threshold.

Fieldwork for each of the case studies was carried out by teams of around six local staff; selection criteria included knowledge of the local language and prior experience of household level interviewing. Although the time required to cover each population varied according to the pattern of settlement and complexity of household demography, teams found that an average of around 90 household interviews could be completed over a 6 day period.

The Swaziland pilot study was carried out in a rural community in the Highvelt region, close to the South African border. In addition to standard household economy information, details of household demography were recorded covering a 5-year period. The previous employment of household members aged 21-50, who had died during this period was documented, as well as details of orphaned children who had joined the household.

For this study data collection and analysis of disposable income was conducted using the same techniques as in the Uganda and Ethiopia studies. Households were ranked according to disposable income per adult equivalent and those falling below the standard of living threshold were identified. In the Swaziland study, which used whole village enumeration rather than a representative sample⁴, additional information was collected on:

1. Household demography, the presence of orphans in the household, the year in which the parent/s of orphans died and their parent/s previous employment. The presence of orphans in the household was used as a proxy for HIV/AIDS⁵.
2. Employment histories of currently employed adults. This was undertaken to gain a better understanding of changes in the labour market.
3. Maize yields per hectare and the use of farm inputs over the past four years. These data were collected to assess returns on agricultural investments at different levels of input.

The Mozambique study was carried out in a semi-rural community, close to a district trading centre, with an HIV/AIDS prevalence of around 20%. In this study, a representative sample of households was used rather than a complete enumeration.

Presentation of the household data.

The output of the studies is presented as disposable income i.e. cash remaining to the household after basic food needs have been met. This is standardised in terms of the number of 'adult equivalents' in each household (i.e. gross household food energy requirement / an average adult male/ adult female energy requirement). A standard of living threshold was developed for each site, to identify the proportion of households with disposable income below this level. For our purposes, the cut off was based on the costs of primary education for all children in the household, and the cost of basic household and personal items required to meet minimum social norms.

Review of main findings and policy inferences

Ethiopia and Uganda: The economic impact of coffee price fluctuations

Individual HEA analysis showed that in the Ethiopia sites, the absence of alternative income sources either from agricultural or off-farm employment meant that households across the income range were extremely vulnerable to the effects of falling coffee prices. Sensitivity to coffee price changes was high (0.7%-1.5% increase in income for each 1% change in coffee price). This contrasts with low to negligible sensitivity in Uganda (see figure 1): (0.02% to 0.14% for each 1% change in coffee price).

At a macro policy level, this suggests that a change to coffee pricing that had a strong poverty impact in the Ethiopia sites would have a far weaker effect in the Uganda sites. The Uganda study showed that higher levels of wealth were only achieved by households that had access to salaried employment. Even if coffee prices were restored to pre slump levels, coffee would not provide a reliable route out of poverty. Moreover, niche market projects (marketing high quality organic beans etc) missed the poorest households, although they did benefit households in the middle income range.

As a methodological trial, the coffee studies were conducted in non-randomly sampled villages. To quantify the relationship between coffee prices and poverty at a national level, it would be necessary to scale up to include all coffee-producing districts. The practicality of implementing this approach at a national scale is discussed later.

 

Swaziland and Mozambique studies: The Economic Impact of HIV/AIDS

Although it has been widely debated for over a decade, surprisingly little is known about the impact HIV/AIDS on individual household economy. The complexities of analysis in this area are compounded by the fact that the HIV/AIDS epidemic has coincided with a period of economic restructuring in many of the worst affected countries, and with major climatic shocks. In assessing the impact of HIV/AIDS on household economy, studies in Mozambique and Swaziland were designed to take these other contextual events into consideration. In both countries, studies were conducted in areas of high HIV/AIDS prevalence.

Swaziland has one of the world's highest recorded rates of HIV/AIDS (around 38%) and has also experienced a series of economic shocks over the past decade. These include a substantial reduction in the quota of jobs available to Swazi nationals in the South African mining sector; private sector restructuring and job losses e.g. in the forestry sector; and structural adjustment of the national economy, including withdrawal of input subsidies, cost recovery and privatisation.

At both Sites households affected by HIV/AIDS were found across the income distribution. Both HIV/AIDS affected and non-affected households were found below the defined standard of living threshold. However, a disproportionate number of households below this threshold had suffered an AIDS death, or were supporting orphans from outside the community.

In Swaziland, employment histories allowed us to estimate the level of income lost through HIV/AIDS. Analysis showed that, omitting the costs associated with illness and funerals, extra adult mortality attributable to HIV/AIDS over the past 5 years has caused a fall of approximately 8%-12 % in total community disposable income. The economic impact on individual affected households is specific to that household, and ranges from a small improvement in income/adult equivalent (e.g. death of an unemployed adult) to devastating loss (e.g. loss of one or more salaried/public sector workers). Overall, in this relatively wealthy community, the net effect is to make very little change to the proportion of households falling below the defined poverty line.

To explore changes in the wider economic context, we looked at the effects of the recent withdrawal of farm input subsidies in Swaziland implemented as part of a structural adjustment programme. The main agricultural activity in the study community is maize production, and yields are highly input dependent. The vast majority of households, including the very poor, produce some maize. Assuming full use of agricultural inputs, the effect of the loss of subsidies was a 24% reduction in the net return on maize. However, it was also found that input use was low and largely independent of the wealth of the farmer (for reasons which are unclear, but probably include problems of pest control), and the actual loss of income due to the loss of subsidy would be less than this. The question of agricultural investment has clear implications for poverty mitigation and deserves further investigation.

As in the coffee studies, the wealthiest households in both communities derived most of their income from salaried, off farm employment.

At both sites the majority of households (both HIV/AIDS affected and non-affected) remained above the standard of living threshold (9% of households in the Mozambique site and 23% in the Swazi site fell below this threshold). In Mozambique, it was notable that teachers, health workers and policemen made up 26% of all employment in this community.

In both the Swaziland and Mozambique sites, households with orphans were found across the income distribution (see figs 2 and 3 below).

Further analysis of the characteristics of the poorest households in the two sites showed that there was no single 'cause' of poverty.

In the Mozambique site, although widows headed a disproportionate number of the poorest households, we were not able to ascertain on the available information, whether these households were poor before the husband died. A number of in depth interviews with widows suggested that wealthier households had been able to diversify their income base (e.g. to invest in profitable trading activities) on the death of a salaried male. Households that did not have the skills or capital to diversify remained poor.

In Swaziland, of the poorest 23 households (those falling below the standard of living threshold), 5 had suffered a 'catastrophic' fall in income as a result of adult deaths over the past 5 years. However, 10 of the poorest households were classified as 'not HIV/AIDS affected', and the remaining 8 HIV/AIDS affected households appear to have been poor for reasons that were not directly related to HIV/AIDS. The household death/s had not resulted in a significant loss of income, and the overall pattern of employment had not changed significantly in the past 5 years.

Practical inferences for policy and programmes

The study findings provide information that is directly relevant to policy and programming decisions, where the aim is to strengthen household economy and resilience to shocks. These include:

1. Measurement of the impact of HIV/AIDS on living standards.
The research provides a measure of the distribution of poverty in the study communities and casts doubt over the feasibility of deriving 'simple' HIV/AIDS related poverty indicators. This has implications for the design of social protection and welfare policies, as well as wider macro economic policy debates. For example, if only orphans were identified as eligible for free primary education or health care, many poor children would be excluded and some better off children included.

2. Asset protection and income generation
Individual HEA analysis provides a quantitative account of how a local economy works, a measure of potential demand and an objective assessment of productive capacities at household level. By providing realistic estimates of returns on investment for different enterprise alternatives, the approach could be incorporated in the design of micro credit and other economic interventions.

3. Targeting welfare and other forms of direct assistance.
HEA methods provide a basis for identifying, costing and targeting safety nets and other interventions at a community level. In rural contexts, these might range from 'one off' grants or loans to assist restocking or pay for agricultural inputs, to waiving of basic service fees and longer term social welfare support.

Akey inference from the HIV/AIDS studies is the importance of work at community level. Given the varied circumstances that face the poorest households, household specific needs must be identified and connected to relevant services or resources. Interventions, such as targeted food aid, distributed through school feeding projects, mother and child health (MCH), and community based orphan support programmes have been widely canvassed as a means of mitigating the effects of HIV/AIDS (e.g. FAO 2003). However, the view that the households that have lost labour and cash income through HIV/AIDS, could best be assisted by this mechanism, is not wholly supported by either study. There would be some scope for food distribution to selected poor households.

Conclusions

Better empirical data on the characteristics of the poorest households raises the prospect of a more effective set of responses. This can assist in selecting pro poor interventions, identifying typologies of households and targeting resources accordingly, as indicated in the pilot study results. At a more general level, it is clear that action in response to problems that emerge from the studies requires coordination and collaboration between a range of national and local government departments, major funding agencies and local and international NGOs.

By providing analysis based on representative samples of individual households, individual HEA methods allow decision makers to model the potential impact on living standards of different policy alternatives and ultimately, to measure actual impact against objectives.

Scaling up to larger populations and areas

The four pilot studies were all small scale. This may be appropriate for some applications (e.g. NGO project work). Other questions. e.g. monitoring the impact of coffee price changes or interventions to increase the value of coffee production (or any other internationally traded commodity, e.g. cotton, tea etc) would require a national system¹.

Scaling up does not seem to present any substantial technical difficulties. The only current question relates to the application of confidence intervals to the ranked values of disposable income, which would be necessary to establish a sample size². With that proviso, it does not appear that samples would need to be very large.

A problem that has been repeatedly encountered with HEA is that Governments are often not well organised to maintain information systems of this kind, as trained staff are moved to other tasks. A longer-term aim would therefore be to incorporate training and the operation of systems in universities. These methods complement other field research techniques routinely covered in many social science degrees.

For further details, contact: Celia Petty, Save the Children UK, 1 St. John's Lane, London, EC1M 4AR, Telephone +44 (0)20 7012 6400, Fax +44 (0)20 7012 6963, www.savethechildren.org.uk

 

 

---

¹HEA methodology is described in the 'Household Economy Approach' (Seaman J et al SCF 2000). HEA is based on Amartya Sen's theory of exchange entitlements. Standard HEA methodology is used extensively in sub Saharan Africa by governments, donors (e.g. USAID, DfID) and UN agencies, to assess food security across large geographical areas and to provide famine early warning.

²See The Household Economy Approach (Seaman J et al, op cit)

³They were designed to test a methodology, and cannot be generalised to a wider population. A discussion of the operational costs and human resource requirements for studies of individual populations and an estimate of the costs of taking the work to national scale is included in the final section.

⁴As this was the first time the method had been used to assess HIV/AIDS impacts across a community, whole village enumeration was carried out for completeness

⁵ The use of orphans as a 'best available' proxy, and the definition of orphan (loss of one or both parents) are discussed in the case study documents.

⁶17 years was chosen as the cut off, as it is the age at which young people in the study communities are likely to become economically active

⁷For example, in Uganda, this would cover the six coffee growing districts.

⁸This question has been discussed with a DfID based statistician.