Per Capita Household Expenditures in HIV-affected Households
Export Indicator
The indicator measures the extent to which expenditures in HIV-
affected households are changing. Household expenditures are a common proxy used in lieu of
direct measures of household income. Household income determines the household’s ability to
purchase food in the marketplace, which is a critical determinant of food security. Research
indicates that many poor and vulnerable households in developing countries (including rural
households) are net purchasers of food. As their incomes rise, these households spend more
on food, purchase a more diverse variety of foods, and shift to higher quality foods with greater nutritional value. Rising incomes also increase the ability of poor and vulnerable households to
manage risks, cope with stresses and shocks, and build or replenish assets, which are important
determinants of household food security.
Interpretation. This indicator is interpreted to measure the extent to which changes in
household expenditures are occurring in the program area from which measurements are
taken. Because an average value is used, the indicator does not measure changes in
expenditures at the individual household level. Rather, it measures the percentage change in
average household expenditures in the relevant program area.
Changes in average household expenditures measured by this indicator may not be entirely
attributable to program interventions. External conditions such as weather, changes in the
economic environment, conflict, or government policies may also influence household incomes.
These factors need to be considered and accounted for when interpreting results for this
indicator.
Household expenditures are difficult to measure, and any indicator used to measure them will
include some degree of measurement error. The data collection methods proposed here,
however, follow established practice and should produce credible expenditure estimates if
implemented properly.
The definition of this indicator allows for cross-program comparisons. Situations may arise,
however, in which different programs use different methods to measure expenditures or draw
their expenditures estimates from different target populations (e.g., all households in the
program area vs. HIV-affected households in the program area). In such situations, any
differences in measurement methods or target populations should be noted and their
implications for the interpretation of results explained. In all cases, interpretation of the
indicator should describe the target population and the data collection method used.
Because this indicator measures the change in household expenditures over time, it is
necessary to establish a baseline value for the indicator before or early in the program
implementation phase.
Uses. The information provided by this indicator can be used for a variety of purposes. At the
global level, it can be used by donors and international organizations to track the extent to
which efforts to increase income among HIV-affected households are effective, compare results
achieved by different programs, and identify countries or regions where greater efforts may be
required. Similarly, national governments can use this indicator to track the success of its policies or interventions and to prioritize needs within their countries. Programs can use the
information to assess the impact of their interventions, inform resource allocation and program
management, and report to donors.
To compare per capita expenditures across households and over time, all expenditure values
must be adjusted for inflation and normalized to a single reference period, typically a day,
month, or year. Information on price changes is needed to adjust for inflation and ensure that
expenditure values from different periods are in the same constant terms. The estimated
expenditures are then divided by the number of people in the household to obtain per capita
expenditures. The average per capita household expenditures is then calculated by adding all
the household per capita expenditures together and dividing by the number of households
whose expenditures are included. Finally, the percentage change is calculated by subtracting
the initial level (baseline) from the later level (endline) and dividing the difference by the
baseline value.
This indicator measures the percentage change in average per capita
household expenditures among HIV-affected households. Generally, economists prefer to
measure household income indirectly using household expenditures rather than to measure
household income directly. Direct measures of household income are unlikely to be reliable for
a few reasons. First, households are often reluctant to reveal income information to a stranger.
Second, household members are unlikely to know total income flows where there are multiple
income generating activities undertaken by different household members, especially when
household members have incentives to hide income from each other. Third, income flows in
poor and vulnerable households tend to fluctuate significantly over the course of the year,
particularly where households are involved in seasonal income generating activities, as is
common in rural areas. By comparison, expenditures tend to be easier to measure than
income, yield more reliable and accurate results, and fluctuate less.
Household expenditures are measured using estimates of expenditure totals over the relevant
reference period for food items, non-food items, household durables, non-durables, and other
expenditures based on the recall of a household head. The relevant reference period for each
expenditure category is determined by the frequency of the expenditure (e.g., weekly, monthly,
yearly). Where feasible, household expenditures should also include an estimate of ‘imputed
expenditures’ as measured by the monetary value of home-produced and gifted food and non-
food items.
To compare per capita expenditures across households and over time, all expenditure values
must be adjusted for inflation and normalized to a single reference period, typically a day,
month, or year. Information on price changes is needed to adjust for inflation and ensure that
expenditure values from different periods are in the same constant terms. The estimated
expenditures are then divided by the number of people in the household to obtain per capita
expenditures. The average per capita household expenditures is then calculated by adding all
the household per capita expenditures together and dividing by the number of households
whose expenditures are included. Finally, the percentage change is calculated by subtracting
the initial level (baseline) from the later level (endline) and dividing the difference by the
baseline value.
Information to calculate this indicator comes from a minimum of two rounds of a household
survey administered to a sample of HIV-affected households, or a broader survey of households
from which data are drawn from households identified as HIV-affected. Survey rounds
typically, but not necessarily, occur at the beginning (baseline) and end (endline) of a program,
although they can occur more frequently as need, capacity, and budget allow. After data from
households in a program area are collected, the average per capita household expenditures are
calculated for each survey round. The two values are then compared to each other and the
percentage change calculated. Data collection method. Data to measure this indicator are collected directly from members of
HIV-affected households using a formal survey instrument. Or if the survey is carried out
among a sample of all households, only the households determined to be HIV-affected are
included when calculating the indicator. Data collection is expected to occur primarily at the
homestead. Where resource or logistical constraints or stigma do not allow collection of data
at the homestead, data may be collected from clients at facility-based HIV care delivery sites, at
community service delivery points, or through other locations.
Economists and national statistical agencies often attempt to measure household expenditures
using very long and detailed expenditure modules including dozens of disaggregated
expenditure categories. This approach, however, is infeasible for most development programs.
As an alternative, programs may use an abbreviated set of expenditure categories capturing the
largest and most common household expenditures at higher levels of aggregation. While this
approach probably yields less accurate expenditure estimates, it compensates with
substantially increased feasibility. Academic researchers also frequently use this approach. An
example of this approach is shown below. This is only one example, and several alternative
approaches may also be used. However, the same approach should be used for multiple
measures of the same population so that the measurement of change uses like values.
Over the past 7 days approximately how much have
you spent for each of the following items? Purchased
Home
Produced
Or Received As
Gift
1. Food and non-alcoholic beverages (e.g., meat,
vegetables, fruits, dairy, legumes, grains,
starches, water, juice, soda, etc.)
2. Alcoholic beverages and tobacco
Over the past 30 days, approximately, how much
have you spent for each of the following items? Purchased
Home
Produced
Or Received As
Gift
1. Payment for housing (rent, maintenance and
repair, water, electrical power, fuel)
2. Non-Durable and Personal Goods (e.g., toiletries,
personal grooming, handbags, travel bags,
newspapers and magazines)
3. Transport and Communication (e.g., tires, tubes,
taxi/bus fares, mobile phone airtime, fuel)
4. Health and Medical Care (e.g., consultations,
medicines, hospital/clinic charges)
5. Supporting relative/friends, religious donations,
6. Other (e.g., entertainment, laundry, barber and
beauty shops, domestic servants, hotels and other
lodging)
Over the past 12 months, approximately, how much
have you spent for each of the following items? Purchased
Home
Produced
Or Received As
Gift
1. Clothing and Footwear
2. Furniture, Furnishing, etc.
3. Household Appliances and Equipment (e.g.,
refrigerator, iron, stove, TV, radio, cassette,
bicycle, motorcycle, computers, mobile phone,
jewelry, watches)
4. Glass/Table Ware, Utensils, etc. (e.g., basins,
plates, tumblers, buckets, enamel and metallic
utensils)
5. Education (e.g., school fees, boarding and lodging,
uniforms, books, supplies)
6. Livestock
7. Other (funerals, bride price, festivals/events)
8. Land
Because the average values for the program area are used for this indicator, it is not necessary
to collect data from a panel of the same households over time, although a panel of households
is also acceptable. If at all possible, the data should be collected from a representative sample
of HIV-affected households in the program area. If a representative sample is not possible, then
the implications for the results in terms of bias, accuracy, and other factors should be explicitly
acknowledged. In all cases, programs should report the sample size and how the sample was
constructed. The size of the sample will also affect the precision with which the indicator
values reflect the status of the larger population of HIV-affected households.
Frequency of measurement and reporting. While household expenditures are expected to
fluctuate less over the course of the year than actual household income, they too can show
significant temporal variation. Significant and permanent changes in household expenditures,
moreover, can take years to emerge. Additionally, information on household expenditures can
be time and resource intensive to collect. For these reasons, it is recommended that the
collection of household expenditure data takes place no more than once every 12 months and that it take place at the same time of the year to account for seasonal differences in
expenditures.
Disaggregation. Because this indicator measures average changes in per capita household
expenditures, disaggregation at the individual level is not possible. Programs may choose to
disaggregate the indicator based on categories that are relevant to their target groups and
services, such as type of household, geographic region, or type of intervention. In this case, the
sample of HIV-affected households may need to be stratified by the relevant disaggregation
categories.
Annual
Strengths: The primary strength of this indicator is its wide-spread relevance. Increasing
household income is an objective of many food security programs, and thus this indicator will
be of interest to a broad range of stakeholders. A second strength is that it follows
international best practice by measuring expenditures as a proxy for income, thereby increasing
the accuracy, credibility, and comparability of resulting indicator values. Measuring
expenditures instead of income directly also potentially allows users to measure changes in the
average poverty status of sample households by comparing expenditure values to national
poverty and extreme poverty lines, which are also often measured in terms of household
expenditures.
It may be possible, moreover, to integrate this indicator into Household Expenditure Surveys,
Core Welfare Indicators Questionnaires (CWIQs), or other multipurpose surveys administered
by national statistical agencies or other research institutions, which are common in developing
countries. Alternatively, it may be possible to calculate this indicator with data in existing
national or regional data bases. The popularity of expenditure modules in national and regional
surveys increases the opportunities for piggybacking the measurement of this indicator on
other data collection efforts. However, this approach requires identifying which households are
HIV-affected, which may be challenging unless such questions are part of the survey.
Another strength of this indicator is that there has been extensive experience using the
indicator in field settings. For example, the World Bank has collected household expenditure
data in 32 countries through its living standards measurement survey (LSMS).
A final strength of this indictor is that measuring the percentage change in average household
expenditures can help to compensate for any inaccuracies or biases that may exist in making
specific numeric estimates of household expenditures.
Weaknesses: The principal weakness of this indicator is the challenge involved in collecting
accurate data on household expenditures. This challenge stems from two related sources. The
first is the challenge involved in capturing accurate estimates of household expenditures. The
second challenge is related to the financial and technical demands of capturing household-level
information.
The use of simplified expenditure survey modules is one way to address the challenges related
to data collection, although the tradeoff is a likely loss in accuracy. Any loss in accuracy,
however, is of less concern if a purpose of using the indicator is to measure the change in
household expenditures, provided the abbreviated expenditure model includes the household’s
primary expenditure categories. If, however, a purpose of the indicator is to measure household poverty status by comparing results to national poverty or extreme poverty lines
that were derived using more detailed expenditure modules, then the loss in accuracy may be
more important. If a country has existing national or regional data bases with expenditure
modules, programs can determine the most important expenditure categories by consulting the
existing databases, as well as determine the extent to which the program’s abbreviated
expenditure modules produce results at variance with the more detailed expenditure modules.
With regard to the second challenge, data collection from individual households can be costly,
time intensive, and technically challenging, making it a potentially significant burden for many
programs. Collecting information on imputed household expenditures adds to the difficulty
and cost of collecting information for the indicator. The failure to collect information on
imputed expenditures likely leads to a systematic underestimation of household expenditures,
particularly in rural areas.
In some situations, moreover, stigma may pose a challenge to data collection. This challenge,
however, exists for all indicators measured at the household level. If the program is already
doing household surveys for other purposes, the marginal cost for adding an expenditure
module to the survey is relatively moderate. In certain circumstances, programs may be able to
reduce the costs and burden of data collection by collecting expenditure data from clients at
facility-based HIV delivery sites, at community service delivery points, or through other means.
Another weakness of the indicator is that it may be challenging to identify HIV-affected
households. If questions about HIV are not already part of the household survey used to collect
expenditure data, such questions could possibly be added, though there may be sensitivities to
asking such questions through household surveys. Alternatively, if programs already have
information about which households in a program area are HIV-affected, this information could
be used.
A final weakness is that when programs measure household expenditures using random
samples in each survey round (as opposed to a panel of households), the results do not
measure and track changes in household expenditures among specific HIV-affected households
over time. Rather, the results only allow the programs to reach conclusions about average, or
general, changes over time. This approach does not allow the same type of in-depth analysis of
factors driving observed changes at the household level that would be possible with a panel
survey. The loss in analytical power may be accepted as a necessary trade-off to increase the
practical feasibility of the indicator.
Resources required. The primary resources required to use this indicator are those associated
with carrying out a household survey: enumerators, training, transportation, survey forms. As
discussed above, in some cases household expenditure data may already be collected through
existing household surveys, in which case additional resources would not be required for data
collection. Staff time will also be needed for compilation and analysis of the data, as well as
training of staff in analysis. In cases where household surveys are already being conducted but
expenditure data are not already being collected, adding survey questions on expenditures will require information on which expenditure categories to include and prices, as well as training in
the questions and in analysis. In some cases household expenditure data will be available, but
information about which households are HIV-affected will not be readily available. In such
cases, resources will be needed to collect information about HIV among households (e.g.,
adding questions to the household survey, or collecting additional data); or if program
information about HIV-affected households exists and can be matched with household
expenditure data, staff time will be needed to match these datasets.
Deaton, Angus and Margaret Grosh. “Chapter 17: Consumption.” Designing Household Survey
Questionnaires for Developing Countries: Lessons from Ten Years of LSMS Experience. Margaret
Grosh and Paul Glewwe, eds. World Bank, 1998.
Morris, Saul, Calogero Carletto, John Hoddinott, and Luc J. M. Christiaensen. Validity of Rapid
Estimates of Household Wealth and Income for Health Surveys in Rural Africa. FCND Discussion
Paper No. 72. Washington D.C.: International Food Policy Research Institute, 1999.
Smith, Lisa C., and Ali Subandoro. Measuring Food Security Using Household Expenditure
Surveys. Food Security in Practice technical guide series. Washington, D.C.: International Food
Policy Research Institute, 2007.