Chapter 17: Consumption

Designing Household Survey Questionnaires for Developing
Countries: Lessons from Ten Years of LSMS Experience

Edited by Margaret Grosh and Paul Glewwe
Chapter 17: Consumption
Angus Deaton
Princeton University
Margaret Grosh
Development Research Group
The World Bank
First Version, May, 1997

This Version, July 1998
Fiona’s 1st Version: October 23, 1998
Angus and Margaret: November 12, 1998
This research paper is funded by the Research Support Grant 679-61 "Increasing the Policy Relevance of LSMS Surveys."
The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They do not necessarily
represent the views of the World Bank, its Executive Directors, or the countries that they represent.
The authors are grateful for useful comments to Richard Blundell, Martin Browning, Paul Glewwe, John Hoddinott, Alberto Martini,
Andrew McKay, Raylynn Oliver, and, especially, Christopher Scott. The Statistical Institutes of Armenia, Ecuador, El Salvador,
Jamaica, and Latvia and the Kiev Institute of Sociology in Ukraine conducted experiments in the collection of consumption data that
are summarized. Eric Edmonds, Dean Jolliffe, Kinnon Scott, Diane Steele, Tilahun Temesgen, and Wlodek Okrasa contributed to the
analysis that underlies some of the discussion.
1
Introduction
The measurement and understanding of living standards are overarching goals of the
living standards surveys. Much of the focus is on poverty or deprivation, the lack of adequate
living standards. Standard economic measure of deprivation are concerned with the lack of
goods, or the lack of resources—income, expenditure, or assets—with which to obtain goods.
But it is always important to keep in mind that many of the most important aspects of deprivation
go beyond purely material deprivation. Deprivation of health, deprivation of education,
deprivation of freedom from crime, and deprivation of political liberty are all important—and
often more important than deprivation of material living standards. The role of development in
freeing people from deprivation in a wide sense has been forcefully argued by Amartya Sen, see
Sen (1999) for a recent and comprehensive account. Data from the living standards surveys
frequently help us take a broad view of poverty, particularly data from the modules on health and
education. Other important aspects of broadly construed living standards, such as life-
expectancy, infant mortality, or the threat of crime, must be constructed in other ways.
Nevertheless, measuring the material basis of living standards will always play an important role
in the assessment of levels of living, and how to collect data for a consumption-based measure is
the topic of this chapter.
The measurement of consumption has been a central objective of the LSMS program
since its inception in 1980 and has remained so throughout some 50 surveys, all of which have
been used to document living standards and poverty. Although the program has always
acknowledged that living standards have many dimensions and has taken care to measure them in
its surveys, the narrowly economic aspect of living standards in the program title was taken to
mean not income, as had been the case in many previous surveys, but consumption. This
consumption focus differentiated the LSMS surveys from some surveys in developing countries
that immediately preceded the establishment of the LSMS, such as RAND’s Malaysian Family
Life Survey. However, many earlier surveys, including the Indian National Sample Survey
(NSS), had long used per capita household expenditure as its measure of living standards and its
basis for counting poverty.
The basic ideas underlying the measurement of consumption are straightforward.

Nevertheless, there are various practical complexities, many of which are discussed in this
chapter. Although income and wealth are what enable people to obtain goods and services, it is
those goods and services themselves that directly generate economic well-being. The
consumption module of the LSMS survey is designed to measure the consumption of these items
in some detail and in the aggregate (with the aggregate being the total value of consumption at
suitable prices). At its simplest, the module collects data on how much people spend on various
goods and services. How best to gather such information and in how much detail, how to deal
with goods that are not obtained through the market, and how to obtain accurate data on prices
are among the topics discussed in this chapter.
The LSMS surveys differ from many other household surveys in that their primary
concern is not the estimation of means or of totals. Instead, the single most important concern of
the LSMS surveys is documenting the distribution of living standards -- measuring poverty
(often but not exclusively the fraction of the population in the left tail of the distribution) and, to
2
a lesser extent, inequality. The LSMS data are also used to illuminate a wide range of policy
issues from descriptive tabulations to econometric modeling.
This emphasis on poverty and distribution must constantly be kept in mind because it has
implications for the design of any LSMS survey. A survey that yields accurate estimates of
average levels of income or consumption may nevertheless do a poor job of documenting income
and consumption among the poor or of estimating the inequality of incomes. For example, if
people have difficulty remembering high-frequency purchases (for example, food) after a day or
two, then asking respondents about the purchases that they made on the previous day will yield
more accurate data than asking them about the purchases that they made over the previous week
or month. If the main concern is to estimate mean expenditure for the population, it may be
sufficient to collect data on the average of the previous day’s consumption since this figure
would include all purchases both for those who purchased nothing the previous day and those
who purchased several days’ supply. In contrast, the average of the previous week or month’s
purchases will be biased downward if the longer recall period implies that some purchases will
be forgotten and thus not reported. However, for measuring poverty, the previous day’s measure
will not be sufficient because all those who did not purchase anything would be counted as poor.
This means that it might be better to ask them about purchases that they made over the previous
week or month in spite of its downward bias. Much of the literature on the design of surveys in
general is concerned with how best to estimate means and totals and can be seriously misleading
when applied to LSMS-type surveys that have different concerns. This is particularly true of most
existing consumer expenditure surveys around the world, which are designed to collect weights
for consumer price indexes.
Part I of this chapter briefly reviews the arguments for using consumption rather than
income as a measure of living standards and for using it to measure poverty and inequality. It
goes on to discuss the principal uses to which consumption data have been put; while the docu-
mentation of living standards remains the central aim of LSMS surveys, there are a number of
other important policy issues that can be illuminated using consumption data. Thereafter, Part I
reviews some of the experience of more than 10 years of LSMS surveys in collecting
consumption data. Part II discusses the data that are needed to construct a consumption-based
measure of living standards and reviews the design issues that affect the cost of collecting data as
well as its eventual accuracy. Part III presents a draft consumption module, while Part IV
provides explanatory notes on that draft module.
Part I: Policy Issues
There are several good arguments for using consumption rather than income as a measure
of living standards and to use it in calculations of poverty and inequality.
Why Use Consumption to Measure Living Standards?
Although the LSMS surveys, in common with many surveys in developing countries, give
primary emphasis to consumption rather than income, there are many other surveys concerned
with well-being that do not attempt to collect consumption data. Many of these are in
industrialized countries, but the income focus is also standard in most surveys in Latin America.
3
There are both theoretical and practical considerations that affect the choice of income or
consumption, and the balance in favor of one or the other may be different in different
circumstances. Thus, it is useful to start by rehearsing the main arguments for and against each
measure.
Theoretical Issues and Implications for Measurement. Income and consumption are dif-
ferent concepts, not just two different ways of measuring the same concept. Some economists
prefer income as a measure of living standards, perhaps following a “rights” approach according
to which income together with assets measures a person’s or family’s potential claims on the
economy. Other economists prefer to use consumption because it measures what people actually
acquire so that, if the level of living is a measure of economic input, consumption is the
appropriate concept. Both can be defended as approximations to utility; the “indirect” utility
function expresses welfare in terms of resources (positively) and of prices (negatively), which in
practice usually means income or resources deflated by a price index -- real consumption or
income, not money consumption or income. Whether consumption or income is measured,
measures of prices are going to be needed whenever analysts wish to compare people who face
different prices, which will be whenever they make comparisons over time or space.
Another consideration about whether to use income (including income from assets) or
consumption is the time-period over which living standards are to be measured. At one extreme,
there is a lifetime living standard measured either by average consumption over a person’s
lifetime or by their total lifetime resources, which will be the same apart from any bequests. The
issue here is that some poverty is only temporary (for example, students are poor in the short
term but not over their lifetimes, while the elderly may be poor but have not been poor
throughout their lives) so that short-term measures of inequality can overstate lifetime inequality.
One influential theory of consumption and saving is the “life-cycle hypothesis,” which asserts
that a person’s consumption at any age is proportional to his or her lifetime resources. If this is
true, measuring consumption is not only useful in its own right but also provides an indication of
lifetime resources. However, the evidence for this hypothesis is controversial to say the least; for
many people, the promise of resources in the future may do little to pay the bills today.
Policymakers have to take account of current poverty regardless of the long-term prospects of the
poor, because saying “Don’t worry, they will be OK later” about poor children or “Don’t worry,
they’ve had their turn” about the elderly are not regarded as acceptable responses.
If a lifetime is too long a reference period, a day, a week, or a month are all clearly too
short. Arguments can be made in favor of using a “season” as a reference period, and there is a
substantial literature on seasonal poverty (see, for example, Sahn, 1989). However, there seems
to be a general consensus that a year is a sensible reference period over which to judge people’s
living standards, even if this is inevitably a compromise that is too long for some purposes and
too short for others. There is also a good deal of empirical evidence that, even in poor
agricultural societies and even without the ability to borrow much, people can smooth their
incomes within a particular year and perhaps over a series of years, so that consumption will
reflect at least living standards throughout the year and perhaps even over a series of years (see
Bhalla, 1979 and 1980; Musgrove, 1978 and 1979; Paxson, 1992 and 1993; Wolpin, 1982; and
Chapter 6 of Deaton, 1997 for a review).
4
If a year is chosen as the standard for assessing living standards but the survey in question
can only hope to measure flows over a shorter period, it is better to collect consumption data than
income data in order to measure living standards. Most people do not receive income every day,
and many do not receive income every season or at least not an equal amount in each season. So
while consumption over a week, a fortnight, or a month is likely to be a reasonable indicator of
living standards over the year or over a few years, income will not be. If analysts are interested
in measuring averages, this will not matter much if the survey itself is spread over a year, since
some people’s zero incomes would be averaged with others’ high seasonal incomes. However,
they are usually not only interested in means -- in fact LSMS surveys are rarely the instrument of
choice for estimating mean income or mean consumption -- but in inequality and in poverty,
which are sensitive to the tails of the distribution, especially the lower tail. Gathering data on the
previous month’s income will overestimate inequality in annual living standards and, provided
the poverty line is below the mode of the distribution, will overstate the fraction of people below
the line. Although there are also random irregularities and seasonal patterns in consumption, they
are typically smaller than those in income, because consumption is less tied to seasonal and
weather-related patterns in agriculture than is income. Even so, consumption measured over a
reference period of less than a year is still likely to overstate poverty and inequality on an annual
basis. Also, the overstatement may not be constant over time if seasonal patterns change with
time because one year is different from another or over the long run because agriculture accounts
for a shrinking share of household income as economies become richer.
These arguments provide a persuasive case that, given the choice, (perfectly measured)
consumption is a more useful and accurate measure of living standards than is (perfectly
measured) income. These theoretical advantages of consumption are likely to decrease as the
period over which it is feasible to gather data gets longer. If it is feasible to visit households on
many occasions throughout the year this will clearly capture any seasonality in the household’s
income. Moreover, if the survey has a panel element so that income can be averaged over a series
of years, there would be little to choose between income and consumption if one can be
measured as accurately or as cheaply as the other.
Practical Issues. The choice between income and consumption is often determined more
by practical than by theoretical considerations. In the United States, poverty is assessed by income
not consumption, which cannot be used because the US does not have a consumption survey of
adequate size and quality to permit the estimation of the poverty numbers. In general, however,
whenever a new or reformed survey is being planned, the designers will have to choose whether to
collect data on household income or consumption, and much will hinge on the relative costs and
relative precision of the data collection required.
Neither consumption nor income data are easy to collect. For consumption, the need is for
data on total household expenditure on goods and services and, as will be discussed in the next
section, these usually have to be gathered item by item. In some cases, a substantial fraction of
consumption does not come through the market, so those imputations have to be made. In
industrialized countries such as the US and Britain, the detail and the associated time and effort of
asking dozens or sometimes hundreds of questions will often make it seem relatively more
attractive to collect income data, especially in situations where income comes from one or two
sources (for example, wages and pensions) that are easily recalled or for which independent
5
documentation exists. By contrast, consumer expenditure surveys are seen as among the most
“difficult and expensive surveys” to field in the statistical system (McWhinney and Champion,
1974). In the US, the Consumer Expenditure Survey (CEX) costs about five times as much per
household as the current population survey (CPS), which is the main source for data on income,
earnings, and employment.
Even so, the concept of expenditure (in other words, giving money in exchange for a good
or service) is clear both to interviewers and interviewees whereas the concept of income,
especially income from self-employment or own-business activity, is not. For own-account
workers in agriculture and small businesses, their personal and their business accounts are often
hopelessly entangled. Thus, in agriculture and elsewhere, the only practical way to estimate
income is to gather data on all transactions, business as well as personal, and then to impose an
accounting framework on the resulting information. This process is extraordinarily time-
consuming, and the results are subject to large margins of error. Note that these difficulties are not
specific to developing countries; even in the US and Britain, the various surveys (CPS and CEX
in the US and the Family Expenditure Survey in Britain) do relatively poorly in gathering data on
income from self-employment (see Coder, 1991 for the CPS; Branch, 1994 for a comparison of
CPS and CEX; and Atkinson and Micklewright, 1983 for the FES). The difference between
developing and developed countries is the fact that formal-sector wages and salaries are much less
common in the developing countries.
Note that the income of many households -- particularly but not exclusively agricultural
households -- varies seasonally throughout the year. In such circumstances, measuring
households’ annual income (which is the minimum amount of data that would be needed to get
adequate measures of poverty and distribution) would require many visits to the household or
reliance on the ability of the household respondents to remember their income from many months
earlier. However, if consumption is smoothed over the seasons (much of the already cited
literature suggests that this is done in most households), then consumption will vary less
seasonally than income. It may also be possible to collect useful data on annual consumption
without making multiple visits, an issue that will be further discussed below.
It is also generally thought that respondents are more reluctant to share information about
their income and (to an even greater degree) their assets than about their consumption, which
means that they are more likely to give deliberately inaccurate answers to questions about their
income than to give the same kind of answers to questions about their consumption. In many
countries, income is taxable, at least in principle, and it may be hard for the survey interviewers to
persuade respondents that the information that they give will not be passed to the tax authorities.
Rich households may even refuse to grant interviews to the survey team and, even if they do, the
respondent, who may be a family member or even a servant, may be more knowledgeable about
the household’s consumption than about its sources and levels of income. Income from assets is
likely to be particularly hard to capture because the ownership of assets is highly unequal and the
wealthy who own most such assets are typically thought to be the least likely to cooperate. Given
the fact that most of these survey interviews in developing countries are frequently conducted of
necessity in a semi-public place, respondents are often reluctant to state their wealth in the
presence of relatives and friends, even when their consumption levels are in part supported by that
wealth. These problems of measuring assets and asset income are likely more severe for
6
measuring inequality than poverty, since those households who are below the poverty line
typically have few assets.
What Analyses are Consumption Data Good For?
The consumption data that can be gathered in an LSMS survey have a number of
important analytical uses.
Measuring Welfare. The policy importance of measuring living standards is indisputable.

Household budget analysis has been used to document and to publicize poverty since the late 18th
century. Although there are many non-economic components of living standards, such as health,
access to education, and political freedom, consumption is the best measure of the economic
component of living standards. Formally, the real value of consumption can be thought of as an
approximation to utility, or “money-metric” utility, according to which an indifference curve is
labeled by the amount of money at constant prices that is required to reach it (see Chapter 5 of
Deaton and Muellbauer, 1980). Total household expenditure adjusted by a price index and divided
by the number of people in the household (or by some more sophisticated count such as the
number of equivalent adults) is a measure of the living standard of each member of the household
and is the measure recommended in this book for analyzing poverty and inequality. See Deaton
and Zaidi (1999) for a more comprehensive dicussion.
LSMS surveys also collect data on a wide range of other household and community
variables that help describe other dimensions of living standards. For example, they collect data,
on individual and community health outcomes and facilities and on the educational attainments of
individuals and the educational facilities that exist in the community. These measures are
frequently used not only to document living standards but also to explore their determinants in
studies, for example, of the relationship between income, assets, and consumption, between
earnings and schooling, or between health status, income, and consumption.
In addition to being used to construct a single summary measure of the economic welfare
of households, the consumption data that can be collected in LSMS surveys have other important
uses, some of which are discussed briefly below. For a much longer account with applications, see
Deaton (1997).
Evaluating the Impact of Price, Subsidy, and Taxation Policies and the Provision of
Public Goods. Analysts are often concerned with the effects of the price changes caused by
changes in policy – taxes or subsidies - or by fluctuations in world prices. Consumption data are
invaluable for assessing these effects, in particular who gets hurt by a price increase and by how
much. Many developing country governments collect a large share of their revenue through tariffs
or through taxes on consumption while simultaneously subsidizing the provision of many goods
and services ranging from basic foods (such as bread, wheat, or rice) to transportation, health, and
education. To a first approximation, a price increase hurts consumers in proportion to the amount
of the good that they purchase, so, in order to know the distributional effects of a price change,
analysts need to know who consumes what and where consumers are in the overall welfare
distribution. For example, do transport subsidies really benefit the poor as is often claimed, or are
the beneficiaries in reality people who are much better off? Improving the quality of clinics or
7
increasing the number of teachers in schools will not help the poor if the poor do not use these
clinics or attend the kind of schools where these teachers are employed. Even simple cross-
tabulations can establish results that, if not necessarily surprising, can resolve major policy
controversies, see for example Grosh (1997) on kerosene pricing in Ghana and health care use in
Guyana and Deaton (1988) on rice pricing in Thailand.
More complex modeling of price reform requires estimates of how consumers respond to
price changes, so analysts can calculate dead-weight loss and the tradeoffs made between equity
and efficiency. Once again, data on consumption, income, and prices that are needed to estimate
these responses (see Newbery and Stern, 1987; Ahmad and Stern, 1991; and Chapter 5 of Deaton,
1997).
Nutrition and Poverty Lines. There is a long tradition in development economics of

counting calories and of defining poverty in terms of malnutrition, for example by counting
people whose caloric intake falls below some recommended standard. The authors of this chapter
do not view nutrition as an adequate measure of welfare because people consume more items than
food and clearly make tradeoffs between food and other goods, so that collecting data on calories
consumed is no substitute for estimating consumption. Nevertheless, the documentation of
nutrition is clearly of considerable interest in its own right.
In some surveys, a household’s calorie consumption is estimated directly by nutritionists

who enter the household and observe what is eaten by each household member, either by
weighing and measuring foods as they are consumed or by asking the members questions about
their dietary intake during the previous 24 hours. It is possible to imagine a module of this sort
that could be added to a multi-topic survey, although, because it would be a lengthy module, it
may displace other modules in the questionnaire as a whole. Swindale (forthcoming) provides
guidelines on how to collect dietary intake data. Many writers believe that such methods are
necessary to obtain accurate estimates of calorie intake -- see, for example, Bouis (1994) and
Bouis and Haddad (1992). However, such dietary surveys also involve a number of difficulties.
The survey techniques are invasive and may cause people to alter their behavior. There may also
be enough variation in a household’s consumption from one day to another to make a 24-hour
recall period too short to yield accurate data with which to estimate poverty, yet longer periods
may be too expensive or too invasive. A more common (albeit probably less accurate) way to
count calories is the “indirect” method, which is usually used in an expenditure survey. Figures
for the quantity of each good that the household has consumed can be obtained either by having
the survey interviewers ask direct questions about the physical quantity consumed by the
household as well as about the household’s expenditure on the good or by collecting data only on
expenditures and then deflating these expenditure figures by the prices of the commodities in
question as obtained in the community or price questionnaires. Standard conversion tables are
then used to convert quantities into a count of the number of calories contained in the food
purchased, a measure known as “caloric availability.”
Data on caloric availability has been used together with data on household income or
expenditure to calculate calorie Engel curves, which plot the average household calorie
consumption at each level of income or expenditure. Following work that was done in India over
25 years ago (see Dandekar and Rath, 1971a and 1971b and Government of India, 1993 for a
8
review), income or total expenditure poverty lines are obtained by calculating the income or total
expenditure level at which the calorie Engel curve gives the recommended calorie intake. If the
calorie Engel curve has a relatively high slope, then increasing household income will eliminate
hunger relatively rapidly. If, as some recent writers have suggested, the elasticity of calorie
consumption with respect to income is close to zero, economic growth alone will not eliminate
hunger. This means that poverty can only be reduced by direct intervention, an approach which is
closer to the basic needs philosophy. (See Behrman and Deolalikar, 1987 and Bouis and Haddad,
1992 -- who also argue that estimates are biased when caloric availability is used rather than direct
dietary surveys -- as well as a contrary position by Subramanian and Deaton, 1996 and a review by
Strauss and Thomas, 1995.)
The demand analysis discussed in the previous subsection can also be applied to calories,
so as to calculate the effect on calorie intake of changes in prices, for example, of the elimination
of subsidies on basic foods (see Laraki, 1989).
Intrahousehold Allocation and Gender Bias. Expenditure data are an important tool for
researching the allocation of resources within the household and for testing different models of
how that allocation might work. In recent years, many studies have found different outcomes for
males and females, particularly boys and girls, within the same household. In some countries,
infant mortality is higher among girls than among boys, and, in even more countries, educational
outcomes are worse for girls than for boys. Several scholars have explored the possibility of using
data on household expenditures to cast light on these different outcomes for boys and girls, as well
as other groups, for example adult women compared with adult men or the elderly compared with
prime-aged adults or widows compared with other household members. (See Chapter 24 on
intrahousehold issues for a fuller discussion, as well as Chapter 4 in Deaton, 1997.)
It is costly and time-consuming for surveys to collect complete data on the consumption of
every item by each family member. In fact, this may be impossible for the many joint (or
household public) goods that are shared by all household members. As a consequence, most
multi-purpose surveys, including the LSMS surveys, have collected household-level data on
consumption and have made little effort to collect individual data. Nevertheless, there are some
cases where consumption at the individual level can be inferred from household data such as
health expenditures that are linked to an identified episode of illness on the part of one member or
expenditures on men’s clothing when there is only one man in the household. In some surveys,
data on expenditures have been collected using the diary method, in which each adult family
member has been asked to keep a diary about who spends what in the household, even if not on
who consumes what. Even when data are to be collected by interviewers, it is probably possible to
collect more individual data than has typically been done in the past if the interviewer can find out
who consumes how much of such obviously private goods as tobacco, transportation, clothing, or
entertainment.
Even when individual-specific data are not collected, it is possible to examine the effect of
household characteristics, including the composition of the household, on the way that households
allocate their budgets. For example, it may be that household expenditures on food and children’s
clothing are higher when there are relatively more women in the household or when a large share
of household resources are earned by, and thus putatively controlled by, women. There is also a
9
developing literature (Bourguignon and Chiappori, 1992; Bourguignon et al, 1993; and Browning
et al, 1994) that has identified sharing rules within the household. Provided that some goods can
be identified that are consumed exclusively by one group within the household or if analysts have
some data on who consumed what of each good, then it is possible to infer whether or not income
is shared equally across the groups. Related to this is the examination of expenditures on “adult
goods” (usually alcohol, tobacco, and adult clothing) for signs of gender bias in the treatment of
children. Since the total household budget is not increased by the presence of children, parents
typically reduce their expenditures on adult goods to make room for the costs of the children. If
the parents cut back on their own consumption more for their sons than for their daughters, then
this is evidence of discrimination against the girls in the household. Surprisingly, analysts have
consistently failed to find such differences, even in places where there is other evidence (such as
differential infant mortality) of bias against girls (see Chapter 4 in Deaton, 1997).
Family Structure, Child Costs, and Economies of Scale. The most commonly used
measure of living standards is total household expenditure divided by the total number of
household members, in other words, household total expenditure per capita. This measure, while
convenient, ignores the fact that the needs of one household member differ from those of another
household member, particularly between adults and children, and that there are likely to be some
economies of scale to household size. Larger households are usually those with many children and
those who would benefit most from economies of scale, so using the per capita total household
expenditure measure almost certainly overstates the number of large households that are poor and
understates the number of small households that are poor. In some countries -- most notably the
US -- there is a different (official) poverty line for each type of household, and these lines embody
both economies of scale and the different needs of adults and children.
There is a long history of studies in economics that have attempted to use consumption
data to derive the cost of living for families of different types by inferring equivalence scales
across age groups and estimating the extent of economies of scale. If such calculations were
feasible and credible, they would have the key advantage over dividing resources by the number
of people in the household that they take into account country-specific and local differences in the
costs faced by different types of families. For example, it is often argued that children are
relatively more expensive in rich countries than in poor agricultural societies. Unfortunately, all
procedures for estimating equivalence scales are controversial, and many economists would argue
that the task is a misguided and even impossible one (see Chapter 4 of Deaton, 1997 for a
discussion of both sides of the argument).
Nevertheless, the consumption data have a more limited but less controversial role to play
in helping analysts to check the implication of various models. Although all methods for
measuring economies of scale or estimating equivalence scales must contain untestable
identifying assumptions, most have stronger implications that can be tested using the data. The
results can reveal a great deal about the plausibility of the models (see, for example, Deaton and
Paxson, 1998, who used a number of LSMS data sets to show that the relationship between food
expenditures and household size contradicts most of the obvious notions about how economies of
scale might operate). Without consumption data, it is impossible to make any progress on the
extremely important policy issue of how to factor differences in household size or structure into
the assessment of household welfare. Until some agreed basis is established for correcting for
10
differences in the costs of living faced by households of different sizes and composition, there is
no way to address such issues as the relationship between poverty and fertility or whether children
are more likely to be poor than adults or the elderly.
Credit and Saving. A traditional use of expenditure data in analysis is to combine it with
income data to derive estimates of saving at the household level. The role that saving plays in
economic development has always been an important intellectual issue, and both public and
private saving are rarely absent from the policy debate. Unfortunately, the generally poor quality
of data on saving collected through household surveys has limited their contribution to this
debate, except perhaps in countries like Taiwan where household saving rates are very high.
Microeconomic income data have typically been poorly measured, and, even if consumption
measures tend to be more accurate, the estimate of saving is the relatively small difference
between two large and inaccurately measured numbers, and as such, may be mostly measurement
error. It is not clear that having such measures of saving is worth the effort of obtaining them. To
the extent that it is the owners of small-scale, household-based activities who are doing the
saving, it is even more difficult to measure saving because data on income from these activities
are extremely hard to measure accurately.
Using data from an expenditure module to contribute to a measure of use of credit may not
be such a daunting prospect. Supplier credit constitutes a large share of households’ total use of
credit. A convenient way of eliciting information about supplier credit is to add questions in the
consumption module about purchases on credit (see Chapter 22 on credit in this volume.)
Do LSMS Surveys Collect Accurate Consumption Data?
In most developing countries, there are unlikely to be any independent estimates of poverty
and inequality against which LSMS data can be checked. However, it is possible to compare
estimates of per capita consumption from the surveys with similar estimates from the National
Income and Product Accounts (NIPA). Although the main purpose of the LSMS is not to measure
means, if the data are enormously different from the NIPA estimates, this is likely to erode public
confidence in the survey itself and, particularly, in its estimates of consumption. While it is
important to make these comparisons, it should not be automatically assumed that the NIPA
estimates are necessarily correct and that discrepancies are wholly due to errors in the survey data.
The quality of NIPA accounts varies widely across the world and, while some items of
consumption are likely to be well estimated (for example, when consumption is from imports and
there is good record keeping at the border), the data on others are often no more than educated
guesses (see Srinivasan, 1994). Even when this is not the case, there are often important
differences in the definition of consumption between the NIPA and the household survey that, if
not corrected for, can invalidate the comparisons (see, for example, Gieseman, 1987 and Branch,
1994 for evidence from the US).
11
Table 1: LSMS and NIPA estimates of Average Per Capita Consumption, Selected Surveys
Country Dates Currency LSMS Annual NIPA Annual Per Ratio LSMS/ Sources
Mean PCE Capita Consumption NIPA
Bulgaria, 95,5 to 95,7 Levas 50,436 90,021 .56 Authors’ calculations from data on LSMS
1995 website; staff estimates
Côte d’Ivoire, 85,2 to 86,1 CFA 237,853 184,935 1.29 Grootaert, 1993, p. 30; IMF 1995
1985
1986
1987
Côte d’Ivoire, 88,5 to 89,4 CFA 173,072 190,203 .91 Grootaert, 1993, p. 30; IMF 1995
1988
Ecuador, 94,6 to 94,9 Sucres 2,032,560 2,230,392 .91 Lanjouw and Lanjouw, 1996, Table 3; IMF,
1994 1996.
Ghana, 1987 87,9 to 88,8 Cedi 56,645 45,568 1.24 Glewwe and Twum-Baah, 1991, p. 17; IMF,
1995
Guyana, 1993 93,1 to 93, 11 Guyana 91,602 53,750 1.70 World Bank, 1994a, p.7; Baker, 1996
Jamaica, 1988 88,8 to 88,9 J$ 4,700 5,210 .90 World Bank, 1996, p.28
Jamaica, 89,11 to 90,3 J$ 6,304 6,568 .96 World Bank, 1996, p.28
1989-2
Kyrgyz, 1993 93,10 to Som 2,273 907 2.50 World Bank, 1995a, p.60
93,11
Morocco, 90,10 to Dh 6,870 6,384 1.08 World Bank, 1994b, Vol II, Annex 1,
1990-91 91,11 Table 2
Nicaragua, 93,2 to 93,6 Cordobas 4,079 2,312 1.76 World Bank, 1995b, Vol II, p.46; IMF,
1993 1996
Pakistan, 91,1 to 91,12 Rupees 6,835 6,037 1.13 Lanjouw and Lanjouw, 1996, Table 4; IMF,
1991 1995
Peru, 1985 85,7 to 86,7 Intis 4,616 6,359 .73 Glewwe, 1987, p. 9; IMF, 1995
2
Table 1 (Continued)
Peru, 1991 91,10 to New soles 750 1,178 .64 Webb and Baca, 1993, p.266; IMF, 1995
91,11
Peru, 1994 94,7 to 94,8 New soles 2,190 3,539 .62 Authors’ calculations from data on LSMS
website, 1996; IMF, 1996
Romania, 93,4 to 94,12 Lei 1,126,558 1,348,055 .84 World Bank, 1987, Annex 1, Table 4;
1995 National Commission for Statistics.
Russia, 1993 93,10 to 94,2 Rubles 1,071,312 497,512 2.15 Foley, 1996; IMF, 1996
Tanzania, 92,9 to 93,11 Schillings 129,708 37,718 3.44 World Bank, 1995c, p. 30
1992/93
Venezuela, 92,1 to 92,12 Bolivares 69,684 142,104 .49 Scott, 1994, p.15; IMF, 1995
1992
Note: Adjustments were made for inflation as follows. If the country has less than 15 percent inflation, and the survey period covered the whole year, no
adjustment was made. If the country has less than 15 percent inflation and the survey period covered only part of the year, the survey figures were adjusted to
correspond to midyear prices. If the country had greater than 15 percent inflation, the monthly CPI was used to adjust the NIPA numbers to the month for which
the survey’s data are priced.
3
Table 1 presents a number of LSMS survey estimates together with their NIPA
equivalents. (As far as the authors of this chapter are aware, LSMS survey data were not used to
construct any of these national accounts.) However, this comparison should not be taken too
seriously for two reasons. First, no detailed investigation of NIPA practices for the countries in
the table has been undertaken, so there is no rigorous information about the accuracy of their
estimates. Second, the survey numbers were taken from the various survey reports rather than the
original microeconomic data, which would have been prohibitively expensive. As a result, there
may be some incomparabilities in their calculation.
The table shows average per capita consumption for 6 LSMS surveys. The ratio of the
LSMS to the NIPA estimate has a median of .96. Though these summary measures indicate an
impressive consistency between the survey and NIPA estimates, there are large discrepancies for
some countries and years. Nevertheless, the survey estimates provide no evidence of the general
understatement of expenditures, which defies the common belief among survey experts and
differs from substantial literature demonstrating that the understatement of expenditures is the
major problem with consumer expenditure surveys in industrialized countries.
Looking at some of the cases with large discrepancies between the household survey data
and the NIPA estimates illustrates the difficulties involved in making comparisons between these
surveys. Several of the countries surveyed were undergoing major economic changes at the time.
For example, when the Kyrgyz survey was carried out in October 1993, the som had been
introduced as a national currency just five months before. While the ruble had lost its status as
legal tender in the Kyrgyz Republic, it was still used in the Republic’s neighboring countries with
which it had substantial formal and informal trading relationships. In this rather chaotic situation,
inflation rose to 772 percent per annum, and the dollar became something of a de facto unit of
account -- sometimes the unit of transaction for large purchases or purchases of imports and also
the most reliable store of value. The planners of the Kyrgyz survey had been considerably
perplexed about whether to use the som, the ruble, or the dollar as the survey’s unit of account
and about what would constitute reasonable recall periods, especially for non-food items. The
same changes that were complicating surveying were also complicating national accounting in
the Kyrgyz Republic. The national statistical office had just begun to calculate the international
NIPA numbers rather than the Gross Social Product numbers that used to be calculated in the
Soviet Union. With such complications in the measurement of both numbers, it is not surprising,
nor even particularly alarming, that the survey data are not very close to the NIPA calculations
for the Kyrgyz Republic. Similar, though somewhat less dramatic, issues plague the comparison
for Russia, Romania, and Bulgaria, which also have very high discrepancies between the LSMS
and the NIPA estimates.
An anecdote about Guyana further illustrates the problems of comparing household

survey data and NIPA estimates. When the first tabulations were becoming available from the
4
1
survey, they yielded a mean per capita expenditure that was about twice that estimated in the
official NIPA. This caused some consternation among the team working on the survey analysis.
However, within hours, they discovered that there was a concurrent effort being made to adjust
the NIPA for various flaws and biases. The new estimate for NIPA was within $30 per capita of
the survey data (van der Gaag, 1994).
The teams working on the surveys in both Tanzania and Nicaragua likewise noticed that
the estimates of consumption from these respective surveys were much higher than those in the
national accounts. However, they were not disturbed by this, given the fact that the national
accounts in those countries had a reputation for being inaccurate (Tsoflias, 1996 and Scott,
1997).
As can be seen in Table 1, the survey estimates for Venezuela are lower than the NIPA
estimates. However, in this case, it appears to be the survey that is wrong. A very short list of
consumption items was used in the Venezuelan survey, and, as a result, consumption was
probably underestimated (see Part II below). On the other hand, why the Peruvian surveys
substantially underestimate consumption compared to the NIPA is a puzzle, although it may have
been that the recall periods that were used were too long given prevailing rates of inflation. Yet
the sample for the 1991 LSMS survey in Peru omitted some sizable rural areas, which means that
the survey estimate should be higher, rather than lower, than the NIPA estimate.
Part II: The Data Requirements of a Consumption Survey and How to Meet Them
Part II discusses the data needed to obtain a consumption-based measure of living stand-
ards as well as to analyze the other policy and research issues that were outlined in Part I. To the
extent that it focuses on the main purpose of this chapter -- the measurement of a single
aggregate for consumption -- this discussion is driven less by what to measure than by how to
measure it. In this respect, this chapter differs from many of the other chapters in this book. This
section discusses conceptual issues, the measurement of prices, and the design issues that have
loomed large in analyses of LSMS data and in the previous literature. Thereafter, the section
presents the current best practice on each of the issues, sometimes based on the literature on
consumption but also on some of the LSMS experience, and results from special experimental
surveys that were specifically conducted to study some of the points discussed here.
This task has been a difficult one, and there are several important issues on which there is
little to report. Even for consumer expenditure surveys in industrialized countries, which have
been extensively documented and where there has been a good deal of experimentation, the
literature has not produced a satisfactory synthesis between theory and practice. The literature
itself is difficult to find and is scattered across various disciplines, including economics, market-
ing, psychology, sociology, and statistics. Much of it is contained in poorly catalogued
1
In Guyana, the “LSMS” survey consisted of the union of two surveys. A traditional income and consumption survey was
carried out with a full year’s sample. Interviewers then revisited one quarter’s sample households to administer the Survey of
Living Conditions, which consisted of modules on health, education, and anthropometrics. When the two data sets were merged,
the number of topics that were covered was close to the number that have been covered in most previous LSMS surveys,
although the amount of detail varied greatly among the topics.
5
government reports and conference proceedings rather than in academic journals. Even so, there
has been a good deal of recent progress in understanding how to measure consumption,
particularly from interactions between survey statisticians and cognitive psychologists. There are
now several topics that are reasonably well understood by analysts and where conclusions can be
extrapolated from previous experience with some confidence (see in particular Sudman et al,
1996 and the volume edited by Biemer et al, 1991). The discussion presented here will draw on
this work.
Even so, it should be kept in mind that there are many design issues where there is
evidence of problems but little understanding of their causes nor of solid recommendations for
solutions. As Sudman et al (1996) point out, “The theoretical basis of interviewing is still less
rigorously developed than the theoretical basis of sampling,” so that the wording and design of
questionnaires has largely remained an art that is governed by in-house tradition and personal
experience. In his useful discussion of the sources of measurement error in expenditure surveys,
Neter (1970) gives the following typology of (non-sampling) errors that remains relevant today
(and whose applicability goes well beyond consumption data): (i) recall errors associated with the
fading of people’s memories; (ii) the “telescoping” of reported events by incorrect dating; (iii)
reporting errors associated with respondents being overwhelmed either by the length of the
survey or by the number of items covered; (iv) “prestige” errors, in other words, misreporting due
to various social pressures; (v) conditioning effects from being in the survey; (vi) respondent
effects where the identity of the respondent affects the answers that they give; (vii) interviewer
effects; and (viii) effects associated with the design of the instrument. Another non-sampling
error that could be added to this list is biases in the data due to non-responses or to the use of an
inadequate sampling frame. All of these issues have been extensively discussed in the subsequent
literature, but for only (i) and (ii) are the causes and treatments reasonably well understood 30
years later.
The literature on expenditure surveys in developing countries is even thinner. While

findings from rich countries often carry over to the poor countries, there are always reasons to be
cautious. For example, response rates in LSMS surveys are very high, nearly always higher than
80 percent and often closer to 100 percent. This compares with 85 percent in the US Consumer
Expenditure Survey, about 70 percent for the British Family Expenditure Survey, and typically
lower figures for surveys in much of western Europe. As a result, if the difficulty of surveying
wealthy households is one factor that causes average consumption to be understated in
industrialized countries, then there is likely to be less underestimation from surveys in poor
countries. While the high response rates are good in themselves, they may (or may not) also
indicate high levels of attention and cooperation among those who choose to respond, for
example, in terms of tolerating long questionnaires or being prepared to take more time to try to
remember the information that they are being asked to provide.
The LSMS program of surveys itself has made little contribution to the methodology of
measuring consumption. Systematic experiments have been conducted only recently and are still
largely not written up in places that are easily available to the community of survey researchers.
The emphasis on collecting data rather than on furthering methodology is, perhaps, inevitable
given that the countries and country departments within the World Bank that fund the surveys are
more interested in increasing their understanding of development policies than in increasing
6
survey know-how. Nonetheless, it is lamentable that the LSMS program has not done more in
this area and desirable that it should do more. Good survey practice should include the continual
evaluation of methods, just as good social policymaking includes evaluating the impact of
government programs. In not supporting investigations into survey methodology more strongly,
the LSMS program has been remiss.
Even within its focus principally on data production, there are four areas in which the
LSMS could make useful contributions to survey methodology. First, it would be helpful and not
too costly for the LSMS survey teams to produce fuller documentation of their pre-tests of
questionnaires, including the different options that were tested, the process and lessons of the
field test, and the reasons for making the various choices that were made about the questionnaire.
A great deal is learned in this process and never made systematically available to others. The
record might include taping of interviews and debriefing of respondents about the interviews.
Second, doing a more rigorous job of collecting and disseminating metadata on the process of
interviewing (as described in Chapter 6 on metadata) will enable analysts to do more systematic
studies of approaches, costs, and quality. Third, the controlled experiments with alternative
modules that were set up or analyzed as part of the background research for this chapter paid off,
and more of these should be done. These will be particularly appropriate in those countries where
the new survey will use a different consumption module than has been used in past surveys. In
these cases, the experiments would not only increase knowledge about how to measure
consumption but would allow the host country to make adjustments in comparisons between its
older surveys and the new one, so that the differences in results due to differences in method are
identified. Fourth, although the consumption modules used in most surveys are increasingly
standard, there is much to be learned from talking through design issues in “cognitive
laboratories” in survey organizations (see Sudman et al, 1996). In these sessions, potential
respondents are brought into a laboratory, asked sample questions, and then debriefed on how
they interpreted the questions, and on how they went about answering them. The results of these
sessions are then used to modify the questions in the module as necessary. Sudman et al (1996)
wrote “In questionnaire design, we strongly recommend the use of think-aloud interviews for
determining what respondents think the questions mean and how they retrieve information to
form a judgment.”
What Consumption Data are Needed?
The measure of living standards that analysts wish to construct is a real value of total
household consumption on a per capita or per equivalent basis. Thus, they need to have available
information on three things: (i) consumption; (ii) household size (and, for the equivalence scales,
the age and sex of the household members); and (iii) prices. Data on item (ii) can be gathered in
the household roster (see Chapter 7),while items (i) and (iii) are discussed here.
A measure of total household consumption is built up from several components. First, it

is necessary to add up all reported expenditures on individual goods and services or on groups of
goods and services (for a fuller discussion, see Deaton and Zaidi, forthcoming). Then a value for
consumption that does not go through the market must be added in (in other words, consumption
out of home production or in kind received from employers). In those countries where
households hold significant stocks or inventories of goods and, particularly, of expensive durable
7
goods, it is necessary to correct for the difference between consumption and expenditures. Then
the estimates must be converted to real terms by adjusting them by a price index to account for
differences in prices among different regions or interview dates. It is important to note that this
accounting should be done after the data are collected because it is neither necessary nor
advisable for the respondent to understand the economic concept of consumption (nor of
income.) The questionnaire should be designed around items that are familiar to the respondent --
typically cash flows or flows of goods -- provided that enough information is gathered to allow
total consumption to be calculated. However, consumption questionnaires often collect
additional data on cash flows that are not part of the economist’s definition of consumption -- for
example, taxes, contributions to savings accounts, or loan repayments -- but that are of interest in
themselves and that the respondent sees as outlays of cash similar to purchases of goods.
To be a good measure of welfare, the consumption concept must be comprehensive. All

goods and services that contribute to people's standard of living need to be included in the
measure, which can be thought of as a practical approximation to an indirect utility function or
money metric measure of welfare. While it is often tempting -- and economical -- to collect data
on only a subset of consumption (sometimes even a single good or group of goods, such as
housing or food), the relationship between the part and the whole can vary a great deal from one
household to another and from one place or time to another, so that rankings or living standards
by the shortcut measure may not be universally valid. A good example comes from the spatial
differences in relative prices that cause people to substitute cheaper goods for relatively more
expensive goods. Poor urban dwellers must often live in poor housing in order to have access to
income-earning opportunities in the city, but the standard of their housing will understate their
overall standard of living.
As will be seen in the following sections of this chapter, there is not a clearly “right” or
“wrong” way of doing things for many issues about how to measure consumption. Rather, there
is a range of good practice techniques and not enough empirical evidence about which is best.
However, it is true that the estimates of consumption are sensitive (sometimes only slightly,
sometimes markedly) to which method is used to reach the estimate. In addition, survey
designers will often wish to ensure that the new survey data are comparable with previous survey
data in that country, which is a powerful argument in favor of whatever method was used in all
(or most) of those previous surveys. However, this may sometimes conflict not only with the
interests of accuracy and best practice but also with standards that would permit comparing the
survey data with equivalent data from other countries. Comparability over time within a country
is most useful for monitoring poverty, which often takes precedence over other considerations.
International agencies and researchers value international comparability, and, because they are
often the sources of funds or technical assistance for surveys in developing countries, this
consideration is often influential. Even so, the past survey practices in a given country may be far
from sensible or standard, in which case, when a new survey is planned, it would be better to be
the first in a new series of consistent and potentially comparable surveys than to replicate its
previous method of measuring consumption. At the very least, it would be worth conducting
experiments to test whether the previous surveys actually collected the data that they purported to
collect. Comparability of nonsense is no great virtue.
For measuring welfare, consumption is ultimately a more useful measure than
8
expenditures (purchases). For most, although not all, non-durable goods, it is safe to assume that
a person’s or household’s consumption is closely tied to their purchases. A kilo of tortillas or a
bunch of bananas must be consumed soon after they are purchased, so expenditure and con-
sumption will approximate each other for specific goods over a short period. Even for less
perishable commodities, some averaging may occur across goods in a fairly short period of time.
A person may buy a pound of coffee one week and consume it over a month, but the next week
he or she may purchase a pound of sugar that will, likewise, last for a while, and the next week
he or she may purchase a bag of flour and so on.
However, in the case of major durable goods, expenditures and consumption are not
closely related in the short run and household expenditures on durable goods will be a poor guide
to their consumption of durable goods. (In some cases, where grains can be stored for substantial
periods of time, the same may be true for goods that are not conventionally classed as durable.
Rice or dried pasta may be stored for a long period of time and, thus, their consumption and the
expenditures made to purchase them may deviate a good deal). For major durable goods (and in
some cases for stocks of grain or of fuel), consumption should be linked to stocks not purchases,
so that the sub-module that deals with durable goods needs to collect data on a list of durable
goods possessed by the household. From these, some sort of consumption flow needs to be
imputed. To do this sensibly, analysts need to have information on the age of the good and on its
original (and perhaps current) value or, in the case of housing where there is no adequate rental
market, on any characteristics of the good that can be used to impute its rental value. Of course,
such imputation is at best a hazardous undertaking in countries where there are few rental units to
judge by, and the quality of the resulting data may not be worth the effort put into collecting
them. Great care must also be taken to avoid erroneous interpretations of the results in cases
where such imputations have an important effect on the total consumption measure or on the
welfare rankings of households. If there is no rental market and possibly only a limited housing
market, an imputed rental value may overstate the value of the housing to its inhabitants.
Particularly in an emergency, it may be hard if not impossible for them to turn this imputed value
into urgently needed cash. It is unwise to let policy decisions rest on often arbitrary and
contentious imputations.
The policy issues discussed in Part I can mostly be analyzed with data that are the by-
product of the need to construct of an estimate of total household expenditure. There are,
however, some exceptions where more data will need to be collected to allow analysis of a policy
issue.
First, the decision about the level of disaggregation at which to collect the data must be
guided by the needs of analyst to have data on specific items of expenditure. For example, they
might wish to have data on goods of particular nutritional significance, such as rice or milk.
Likewise, when different goods are taxed or subsidized at different rates and analysts want to use
the survey data to investigate tax reform, then the different goods must be distinguishable from
each other. If flour is subsidized in the country where the survey is to be fielded, then it would be
a good idea to include a separate question on the consumption of flour rather than including it in
a broader question about “staples” or “flour, rice, and cornmeal.” Where analysts are concerned
with the relationship between consumption and the environment, it would be necessary to
distinguish consumption items that were gathered or hunted from consumption items that were
9
grown at home or bought in the market.
Second, analysts may wish to collect data that are disaggregated in a manner that will
yield information about intrahousehold allocations, even if such disaggregation is not required to
estimate the total of all expenditures. Items that are exclusively consumed by different groups are
an obvious example (for example, collecting data on men’s and women’s clothing and footwear
separately from data on children’s clothing, rather than having a single item for clothing). It is
also possible to include questions about who consumed what, at least for some goods. Tobacco
and alcohol are individually, not jointly, consumed, and it is probably possible to obtain
reasonable estimates of the individual-level consumption of these products. Other examples of
goods that are consumed individually are tickets for entertainment and for transportation. One
difficulty is that these may be the items for which the household member who is the best-
informed respondent for the household overall gives the worst answers. For example, the
homemaker may know a lot about the household’s food consumption but relatively little about
the expenditures on alcohol, tobacco, or entertainment by individual members. This problem can
be dealt with by conducting individual interviews on consumption of some goods, a feasible but
costly solution.
Third, some of the research topics -- such as the calculation of calorie availability or the
estimation of price elasticities -- ideally require data on quantities of individual items at the
household level, whereas the welfare measure makes do with expenditures, deflating when
necessary by a price index. The issue of quantities will be addressed in the next subsection.
Collecting Price Data
The conversion of money values to real expenditures requires the construction of a price
index, for which price information must be available. This price information must capture not
only temporal variations in prices but must also accurately represent the price level faced by each
of the households in the survey sample. In some countries, adequate price indices may already be
available, but this is rare, because many price surveys exclude rural areas. Urban prices are only
useful for nationwide analysis when spatial price variation is limited, for example, where there is
a good transportation network and markets are well integrated (although even in this case there
may be marked regional variations in the costs of housing).
When there is no other adequate information on prices, the data must be collected in the
household survey. This may be done at the household level or at the community level. At the
household level, the survey may be designed to ask each household how much they paid for each
unit of an item that they purchased or about the quantity of the good that they purchased as well
as their total expenditures on the good. When households report physical quantities (such as
kilos, sacks, or numbers), it is then possible to divide their reported expenditure by the reported
quantity to yield a price -- or more precisely a unit value -- for each good, and these can be
weighted together to give household-specific price indexes.
In the case of most previous LSMS surveys, the designers have opted for collecting data
from local markets in a price questionnaire at the community level, with few LSMS surveys
having collected information on quantities from households. However, many other surveys
10
around the world do collect information on quantities, including the Indian NSS, the Pakistan
Household Income and Expenditure Surveys, and the Indonesian SUSENAS surveys. While the
LSMS survey in Viet Nam did collect price data in local markets, it also collected quantity data
at the household level, as did the Brazilian, Ecuadorian, Nicaraguan, Kyrgyz, and Russian
surveys. The LSMS surveys in Pakistan, Bulgaria, and Ecuador included questions about
expenditures per unit. Apart from the Pakistan case where the unit cost data had serious problems
(probably for local reasons, almost certainly inadequate interviewer training), there appears to
have been no systematic evaluation of this particular option so far.
There are several advantages to collecting data on prices by asking household respondents
about both their expenditures on and the quantities of their purchases. This procedure yields
measures of physical quantities that are useful in their own right, for example, for computing
calorie availability or for estimating the elasticity of quantities to changes in taxes or subsidies. It
also yields the raw material for a price index for each household, without requiring assumptions
about where the household buys its good. A price index constructed in this way, however, covers
only those goods (typically but not exclusively foods) for which quantity data can be well-
defined in the questionnaire. Such price indexes are automatically tailored to the consumption
patterns of the households in the survey, so there is no discrepancy between the price data and the
goods that people buy. Having data on prices for individual goods at the household level is also
useful for analyzing demand patterns and policy issues, such as price reform, that depend on the
results of demand analysis. One of the authors of this chapter has satisfactorily matched
household data on unit costs from the Indian NSS to the prices that the Government of India
regularly collects from local markets around the country, at least in cases where the local markets
are located near the survey households. Not only do these data match across districts, but the unit
values from the survey reflect the appropriate seasonal patterns of agricultural prices (see Chapter
5 in Deaton, 1997).
There are also disadvantages to collecting data on prices in this way. Unit values are not
prices, and they even vary among households that purchase from the same sources, because
better-off households typically buy higher qualities, even of fairly narrowly defined commodities
like rice or sorghum, let alone for more broadly defined commodities such as meat. This problem
can be dealt with, for example, by averaging the unit values over all of the households in a
primary sampling unit (PSU), and the Indian evidence quoted above suggests that the averaged
unit values are not likely to be misleading as indicators of price. A more serious problem is that,
with a few exceptions such as fuels, it is not easy to define physical units for goods other than
foods. Gathering data on the price of food may be enough, particularly in very poor economies
where food consists of two-thirds or three-quarters of the budget of most households, but this is
clearly not true in general. In many past LSMS surveys, even the definition of physical units of
some food items was unclear or subject to error. For example, the respondents may have been
unclear whether to report the price that they paid per egg or the price that they paid for a dozen
eggs. Moreover, goods are often sold locally in amounts or units that are often not very precise
and, thus, can be hard to interpret at the analytical stage; for example, a “bunch” (or any other
unit) of “vegetables” is much less clearly defined than a kilogram of rice.
LSMS surveys have had less difficulty in defining useful units for consumption than for
production, where respondents often cannot provide any precise quantities (“I sold three “sheets”
11
of this and five “baskets” of that"). Since most of these difficulties were experienced in Africa
and since the successful quantity experiences of statistical offices cited are largely Asian (though
there is also positive experience in Latin American and in countries of the former Soviet Union),
there may be a “continent effect” here, perhaps reflecting the degree to which the economy is
monetized or to which non-standard prices or units prevail.
The alternative to collecting prices from households is to collect prices at the community
level at village and local markets in the PSU. This option is cheaper, because prices are collected
only for each PSU, not for each household. It also has the advantage that, in principle, the prices
in the market are the prices that consumers actually face, not the unit values that they choose to
purchase. The fact that observed prices are the same for everyone in the PSU is thus an
advantage, not a disadvantage. Because most countries have some sort of regular method for
collecting data on consumer prices and aggregating them into a price index, at least in urban
areas, the survey designers may be able to make use of or at least adapt these well-established
procedures.
However, there are also problems in collecting data on prices at the community level.
First, common sense suggests that, in some circumstances, it is difficult for a survey team to
replicate the sort of transactions that are engaged in by locals; haggling is often an important
factor in defining the prices actually paid by local consumers, which may mean that the prices
that vendors quote to survey enumerators may be different from those actually paid by long-
standing or regular customers.
Second, the price questionnaire can only collect prices on items that are available in the
local markets, which may exclude many non-food items as well as those food items only
consumed seasonally or regionally rather than nationwide. To solve this problem as well as the
problem of defining a suitable unit, the survey may be biased in favor of manufactured or
processed items that can easily be defined such as a can of standard-brand tomato paste, a two-
gallon plastic bucket, or a two- pound packet of sugar from the national refinery. Some of the
problems here are more than procedural. In countries where consumption patterns differ radically
across regions (for example, between northern and southern areas of India), there are serious
conceptual difficulties in defining price indexes that are as bad as those involved in comparing
prices across countries.
Third, in some (but not all) countries, it can be hard to know what is meant by a “local
market.” The image of a PSU as an isolated rural village with a single market is an appealing
one, but it is not accurate in all parts of the world. In urban areas, people may buy goods far away
from where they live, although the obvious solution to this problem would be to use the urban
price indices that are often available and that are sometimes of acceptable quality. However, even
in rural areas, PSUs are defined by statistical and often ultimately administrative criteria that may
not accurately represent actual villages or village markets. So there is no guarantee that the prices
in any given local market are the relevant prices actually faced by the households in the survey.
This problem can be exacerbated when there are several different types of outlet in a given
community (such as markets, corner stores, supermarkets, and government ration shops selling
subsidized products), all with different prices and clienteles.
12
Fourth, at least in a few LSMS surveys in the past, the procedures for entering data from
the community questionnaire into the computer were not well enough established, and the data
from many questionnaires or (in extreme cases) the whole community survey were lost to
analysts. The recommendations in the new LSMS implementation manual by Grosh and Munoz
(1996) should overcome some of the problems, but it will remain the case that community
questionnaires are more novel to most survey agencies than household questionnaires and,
therefore, may be managed less well. In the LSMS surveys for Pakistan (1991) and Viet Nam
(1992-93), price data were collected at both the household level (on expenditures and quantities
in Viet Nam and on quantities and unit values in Pakistan) and the community level, and the
results for various goods compared. The comparison is clouded by the fact that, for the reasons
given above, the two data collection procedures measured different things, so that it is unclear
exactly how close the unit values from the household questionnaires could be expected to be to
the market prices from the community questionnaires. For both countries, the two sets of
estimates are similar at a sufficiently high level of aggregation; for example, there is little
difference between the unit values and prices at the all-province level in Pakistan. However, in
both countries’ data sets, the unit values and prices differ markedly at the PSU level, comparing
the estimate from the price questionnaire with the average of households in the PSU. In Viet
Nam, the correlations between the reported unit values from market purchases and the directly
observed market price vary from 0.77 (for noodles) and 0.76 (for pork) to -0.07 (for cassava) and
-0.34 (for mangoes). For the 16 foods in the comparison, these are the only two negative
correlations, and the median correlation is 0.34. The correlations are similar, if somewhat lower,
for Pakistan, perhaps reflecting the problems with the unit value data.
Since any household survey will always require some estimate of price variations and
since households cannot usually provide price (or unit value) data for most non-foods or even for
some foods, the survey will always have to include a price (community) questionnaire of some
kind. If the household questionnaire does not collect data on quantities, meaning that it will yield
no unit values, then the community questionnaire must gather data on the price of food as well as
of non-food items. When both approaches are available for many foods, it is unclear from these
two examples or from the literature that one method has any clear advantage over the other for
any given survey. Of course, duplicating the collection of some data is an insurance policy, and
there has been at least one LSMS survey where respondents were (mistakenly) given the option
of reporting either quantities or expenditures, so that, without the community questionnaire, it
would have been difficult to construct the essential consumption expenditure aggregate for each
household.
It is useful for analysts to have data on quantities for reasons other than the construction
of the price indices. Thus, most surveys with objectives similar to those of an LSMS survey will
want to collect data on quantities if it is feasible and economical to do so in the country in
question. Thus, if previous surveys in the country have yielded good data on quantities or if it
proves possible to gather data on quantities in an exploratory field test, questions about quantities
should be included in the consumption module in the household questionnaires, at least for food.
Level of Disaggregation
13
The number of items about which data are to be collected is one of the central issues in
designing the questionnaire for a consumption module. On the one hand, it is an important
determinant of the cost of the consumption module, with longer modules being more costly or
crowding information out of other modules of the questionnaire. On the other hand, asking about
items in detail is generally assumed to yield fuller reporting and greater accuracy from the detail
and detailed prompting than would the use of shorter lists of items. There have, however, been
suggestions that a survey can try to gather data in too much detail. Respondents may become
bored, despondent, or believe that they are being uncooperative or showing themselves to be
inadequate consumers if they have nothing to report in response to a long list of questions and,
thus, may invent purchases to be “helpful” or to enhance their prestige. However, there has been
a validation using crop and trade data of food estimates from the Indian National Sample Survey
which uses a very long list of items (Minhas, 1988 and Minhas and Kansal, 1989), and, although
there is some slight evidence of overestimation for cereals (about 3 percent), this seems well
within reasonable bounds.
Traditional expenditure surveys in developing countries (for example, the NSS surveys in
India) use very long lists of consumption items with each food being named with great
specificity, and lists of 200 to 300 items are not unknown; the Brazilian budget survey uses a list
of 1300 items; LSMS surveys have been less detailed; the 33 foods and 20 non-food items listed
in the Pakistan survey in 1991 and the 45 foods and 46 non-food items listed in the Viet Nam
survey in 1993/94 are typical. As has already been discussed, some disaggregation is necessary to
obtain information on some specific items of interest, but the precise level of disaggregation that
survey designers choose for any given survey will depend on their views about the tradeoff
between costs and accuracy.
There is a good deal of debate about whether short (or at least shorter than those
commonly used) consumption questionnaires can save time and money and still deliver accurate
estimates of total consumption, but the issue does not seem to have been settled in the literature.
One set of results suggests that short lists of items will yield reasonably accurate data. A study of
farm operators in the US by Reagan (1954) found that total consumption was only modestly
lower -- about 10 percent overall -- for a condensed list of 15 items than for a list of over 200
items. For developing countries, Bhattacharya (1963) reported on a small-scale experiment on 44
households in two villages in West Bengal, who were presented with the usual detailed NSS
questionnaire as well as with a questionnaire covering broad commodity groups, and with a
single question about their total expenditure during the previous month. The consumption totals
from the questionnaire with the broad groups were slightly lower than those from the detailed
questionnaire but not significantly so, but the single-question method gave estimates that were 25
to 30 percent lower, although still highly correlated (0.98) with the estimates from the full NSS
list. There is also some different evidence from the US Consumer Expenditure Survey where
some households are asked to keep detailed product diaries of their purchases of food, while
others, who are in the interview section of the survey where food is not the main focus, are asked
to report their total expenditures on food at home and away from home for each of the previous
three months. According to Gieseman (1987), the amount that the respondents reported that they
spent on food at home was actually larger in the interview part of the survey than in the diary part
of the survey and was closer to the NIPA estimate. For the estimates of the amount spent on food
away from home, the diary appears to be quite accurate relative to the NIPA’s estimate, but there
14
was substantial underestimation of this expenditure in the interview part of the survey. Even
more positive results were reported by World Bank (1993) from a test survey in Indonesia. Both
a short and a long questionnaire was administered to 8,000 households. In the short
questionnaire, the number of food items was reduced from 218 (in the long questionnaire) to 15
and the number of non-food items from 102 (in the long questionnaire) to eight. Total measured
food expenditures differed little between the questionnaires, either in terms of means or of
distribution, although the long questionnaire yielded about 15 percent more non-food
expenditure.
These results have not been replicated elsewhere. A similar experiment in El Salvador
with 72 versus 18 food and 25 versus six non-food items gave ratios (long-to-short) of 1.27 for
food and 1.40 overall (Jolliffe and Scott, 1995). A 1994 experiment in Jamaica compared
modules with a total of 119 items to modules with a total of 37 items and produced similar
results, with a long-to-short ratio of 1.26 for both food and non-food items (Statistical Institute
and Planning Institute of Jamaica 1996, Appendix 3). In Ecuador in 1993 two versions of the
sub-module on food items were piloted, one with 122 food items, the other with 72. The ratio of
total food expenditures reported on the long module to those on the shorter module expenditures
was 1.67 (see Steele, 1998). Although the shorter questionnaires sometimes dramatically reduce
both survey costs and times compared to the longer questionnaires (for example, in West Bengal
from 180 minutes to 90, and in Indonesia, from 80 minutes to ten), it seems that such savings are
gained at the expense of accuracy.
However, there are alternatives to having either a long questionnaire or a short

questionnaire. One alternative, which appears never to have been used in an LSMS survey,
would be a hierarchical scheme in which respondents are first asked if they have purchased
anything within a broad class of goods and are only asked the detailed questions about items in
that class if they say they have bought a good in that class. For example, if a household
responded that it had not bought any dairy items, it would not be asked specific questions about
milk, yogurt, butter, and ice cream. In fact, it is likely that something like this often happens in
practice when interviewers and respondents are faced with lists of several hundred items that are
inevitably grouped into broad categories. This kind of approach has obvious advantages when
interviewers use computers to administer questionnaires to respondents, in which case the
detailed questions about each item would never come up in the interview unless the respondent
first indicated that they were relevant. The obvious risk is that consumption will be
underestimated if going through the list in detail prompts people to remember purchases. There
is evidence from diary surveys in industrialized countries that preprinted diaries that identify
more categories of consumption cause respondents to report more consumption, presumably due
to this prompting effect (see Tucker, 1992 and Tucker and Bennett, 1988).
Note that there are many different ways of turning a list of several hundred items into a
list of only a few dozen. The traditional procedure might be referred to as the “botanical”
method, whereby cereals are grouped together, as are pulses, root vegetables, or leafy greens.
Since botanically similar foods often contain similar amounts of calories per kilo, this way of
aggregating the list of items ensures that analysts can calculate calorie counts when necessary.
However, other criteria can be used for aggregating these items, such as where the consumer
typically buys the goods and consumers may be able to remember how much they spent in
15
individual stores more accurately than individual goods, let alone on groups of goods. There is
some evidence from an experiment in Jamaica where botanical versus “point-of-purchase”
aggregates were compared, (Statistical Institute and Planning Institute of Jamaica, 1996,
Appendix 3). Here there was little difference in the two sets of means, and the variances were
also similar. There is also related evidence from the literature on diaries. Sudman and Ferber
(1971) tested diaries that used (i) itemization by purchase, (ii) groupings by product type, and
(iii) groupings by outlets, and found that respondents were more likely to agree to cooperate with
the survey and maintained their diaries for longer if a product diary was used. The relevance of
this evidence to LSMS interview surveys is a matter for conjecture.
Finally, there are implications for the degree of aggregation if the survey is attempting to
collect data on physical quantities. Some foods, such as many cereals, can be grouped together,
and a meaningful total weight devised. However, this is not the case, for example, for canned
goods, for vegetables, or for many processed goods, where considerable disaggregation is
required in order to obtain appropriate units.
In summary, it seems that using drastically short questionnaires is likely to be risky and to
lead to the underestimation of total consumption. Although it is difficult to be more precise, 300
items are probably too many and 10 is probably too few. The draft module presented in Part III
includes the approximately 70 to 100 items that have commonly been used in past LSMS
surveys, on the grounds that many more items would increase costs noticeably and that the
comparisons to NIPA have not shown huge, systematic biases. On the other hand, using a much
smaller number of items would both increase the risk of underestimating total consumption and
certainly decrease analysts’ ability to calculate rough estimates of caloric content, which are
sometimes used to calculate the poverty line.
Recall Period
For each consumption item in the consumption module, there must be a recall or
reference period. For example, the questionnaire may ask how much rice the household
purchased during the previous week, two weeks, or month or it may ask about the household’s
expenditure on clothing during the previous two weeks, month, or year. The recall period is
sometimes tied to a particular event, most commonly the interviewer’s last visit. Another option
is to have respondents report how much they “usually spend” over a month or a year. While it
makes obvious sense to use longer reference periods for items that are rarely purchased and
whose purchase is a significant event that is easy to remember and to use short periods for high
frequency purchases, this guideline still leaves a great deal still to be decided.
Introduction. The choice between recall periods is one of the most important and difficult
design issues for the consumption module. It is also an issue that cannot be dealt with in isolation
because it interacts with other elements of the module (such as whether expenditures are
collected by diary or by interview) and with the survey’s design more generally (in particular
whether the design permits multiple visits at least a week apart). The ultimate objective is to
obtain a reasonably accurate estimate of the rate of each household’s total consumption
expenditure over the previous year. There are many different ways to fulfill this aim.
16
One possibility is for the interviewer to make a single visit to the household during which
the respondent is asked to recall how much the household spent during the previous 12 months,
either in total or on a list of items. However, this is likely to lead to either an underestimation of
household expenditure, because it is difficult for people to remember their expenditure from so
long ago, or to educated guesses, where respondents estimate their expenditure over the whole
year from their current rate of expenditure. Another alternative is for the interviewer to visit the
household many times throughout the year and to ask the respondent for details of the
household’s expenditure over shorter periods. However, if people’s memories of their
expenditure fade quickly, many visits may be required to ensure that accurate data are collected
on high frequency purchases, and such visits can be prohibitively costly. The diary method was
designed to minimize the need to rely on respondents’ memories because they are supposed to be
filled out at or near the time when the purchase is made. However, diaries clearly pose special
problems when a substantial fraction of the population is illiterate, problems that will be
discussed more fully below.
Note that one of the special issues for LSMS surveys is the requirement that each survey
provide an estimate of annual expenditures at the household level. Most consumption surveys do
not make this demand and are content with estimating aggregates or averages over households,
for example, for weights for a consumer price index. If the only problem with reporting were
progressive forgetting (the fact that respondents’ memories of their expenditures fade as the time
since the purchase grows longer) and if there were no other systematic biases (but there are, as
will be discussed below), averages could be obtained accurately with short reporting periods.
However, this will not do for measuring welfare at the level of the individual household. Longer
recall periods are better than shorter ones for measuring the distribution of consumption because
averaging consumption over many days eliminates the randomness of some of the household’s
day-to-day purchases that have nothing to do with its standard of living. However, if people find
it harder to remember more distant events, longer recall periods will miss more consumption, and
lead to downward bias. If short recall periods are used, and if people report accurately, not
everyone will purchase every item every day or every week, because many goods can be stored,
because many goods that the household consumes regularly do not have to be consumed every
day, and because some items of consumption have seasonal patterns. Provided that the survey’s
fieldwork is spread throughout the year and provided that the respondents’ reports are accurate,
short reference periods will yield unbiased estimates of the mean for the population. Those
households that do not purchase anything during the reference period will be averaged with those
who happen to make purchases for several periods, or those interviewed about their consumption
during a festival are averaged with those interviewed about non-festival consumption. However,
such data (for periods when some households spend nothing while others spend a lot) do not give
an adequate picture of the annual consumption of individual households nor of the distribution of
consumption across households.
Whenever the recall period used in the consumption module is shorter than the period
over which living standards are defined in analysis and even given perfect recall, the data will
include variance that does not reflect the true distribution of living standards. All the standard
measures of inequality have the property that they are increased by adding variance -- effectively
a “mean-preserving” increase in spread” -- so that inequality will be exaggerated by the
measurement error, and, provided the poverty line is below the mode of the distribution,
17
estimates of poverty will also be exaggerated. In the extreme case, a single day could be used as
the recall period to eliminate bias in the mean, with the consequence that anyone who did not go
shopping the previous day would be classified as poor. Since the LSMS surveys are as much or
more concerned with the dispersion of households across the distribution than with means,
single-visit consumption modules with very short recall periods should be avoided, except when
purchases are known to be evenly spread or when there is rapid inflation, which will itself cause
people to make frequent, regular purchases. This is not to say that short recall periods and
multiple visits can never work well; for example, several African expenditure surveys use a daily
recall period coupled with seven daily visits, and this kind of design is used in conjunction with
diaries in consumer expenditure surveys in Singapore (see Silberstein and Scott 1991).
The difficulties that can arise with recall periods that are too short are not confined to the
estimation of poverty and inequality. If the recall periods for reporting consumption are shorter
than the period over which living standards are defined in the analysis, the measurement error in
each individual expenditure will be transmitted into the total expenditure estimate, which is the
sum of all of the individual expenditures. As a result, there will be a (non-standard) measurement
error bias in the estimation of Engel curves, including calorie Engel curves (see Cramer, 1969
and Bouis and Haddad, 1992). If this is not corrected for, estimated Engel elasticities will vary
with the length of the recall period (see Ghose and Bhattacharya, 1993 and 1995 for evidence
from India).
Theory and Evidence in the Literature. It is a well-known phenomenon that people forget
events as they recede into the past. There has been experimental evidence for more than a century
on “forgetting curves” that plot the degree of memory errors against the length of the period over
which people are trying to remember. People do not forget all events at the same rate, and they
can remember some “flashbulb” events (for example, the death of President Kennedy in the US)
in great detail for many years. Purchases of consumer goods are no exception to these rules, and
there is a large body of evidence in industrialized countries on “recall bias” or increasing
underestimation as the recall period is increased. For reviews and general discussions, see Neter
(1970), Eisenhower et al (1991), Silberstein and Scott (1991), and Sudman et al (1996). Some of
the cited evidence comes from developing countries, for example, Mahalanobis and Sen (1954)
and Ghosh (1953). Most relevant to our current concerns, is the study by Scott and Amenuvegbe
(1990) who ran experiments with households from the Ghanaian Living Standards Survey, which
showed that, for 13 items that were frequently purchased by these households, reported
expenditures fell at an average of 2.9 percent for every day that was added to the recall period up
to a week, so, for a seven-day recall, expenditures were 87 percent of what they were for a single
day. A two-week recall period yielded a total that was 5 percentage points lower than that
yielded by a one-week period. Annual recall based on explicitly normative questions (“How
much do you usually spend on xx?”) gave a total 91 percent of that of the one-day recall, while
annual recall based on ostensibly factual questions (“How much did you spend on xx?”) gave a
total 113 percent of the one-day recall figure.
Progressive forgetting is only one of the tricks that memory plays. There is a second kind
of important bias that affects expenditure estimates. This is “telescoping” whereby respondents
include in their reports events that happened earlier than the limit of the recall period. Asked
about their expenditures during the previous year, respondents may include a car that was bought
18
13 months ago (see Neter and Waksburg, 1964 who identified such effects in the US Consumer
Expenditure Survey for home-owners’ alterations and repairs). This effect is closely related to the
“boundary effect” identified in crop surveys in Bengal by Mahalanobis (1946) where neighboring
farmers each included their common boundary in reporting estimates of their crop acreage.
Telescoping has been formally modeled by Rubin and Baddeley (1989) and by Bradburn,
Huttenlocher, and Hedges (1994). According to these models, people do not remember dates very
well, so, even though they may remember the event (or expenditure), they may be unsure about
the date of the event. If uncertainty about dates increases as the event recedes, such telescoping
errors will cause a net upward bias in the resulting data. The further in the past the event is, the
larger will be the uncertainty about its date, and thus the probability that it will be misplaced into
the “present” is higher than the probability that a “present” event will be misplaced into the
“past.” Moreover, by definition, “present” events cannot be misplaced into a time that is
“future” relative to the interview date, so that on average, there is greater intrusion of past events
into the reference period than there is loss of relevant events out of it.
The fact that people increasingly forget events as they recede into the past can explain
why people remember and report items that they purchase frequently (such as food) less as the
recall period increases and why telescoping can cause some upward bias in the data on
expenditures in large or important items that people remember purchasing but do not remember
when they did so. Short recall periods are likely to lead to the overstatement of purchases of
those items that are subject to both telescoping and recall bias, with telescoping dominating. On
the other hand, as the recall period is lengthened, recall bias will become more prevalent, which
will result in a downward bias (see Eisenhower et al, 1991). These effects will work differently
for different goods, with purchases that are highly salient to the respondent, such as purchases of
durable goods, expenses for wedding or funerals, or for stocking up on grain for the year being
more subject to telescoping and with high frequency smaller purchases on such items as food and
household supplies being more likely to be forgotten. Only those purchases that people remember
making can be telescoped.
Neter and Waksberg (1964) proposed a procedure for dealing with telescoping that
appears to be reasonably effective. This is the method of “bounded recall,” whereby a
preliminary interview is conducted in which respondents are asked about the household’s
expenditures in (say) the previous month. These data are not be used because they are subject to
telescoping bias, but are nevertheless important because the process of noting them will provide a
record that prevents their being reported again in the first “real” interview and thus eliminates
telescoping. In the second interview, which is the first to yield usable data, respondents are asked
about their purchases since the first interview. Note that this option is only possible when the
interviewer makes at least two well-separated visits to the household. Recall bias is harder to deal
with, although, if respondents are capable and willing to keep an accurate diary, it becomes
unnecessary to rely on their memories, in which case both recall bias and telescoping are
eliminated.
Cognitive psychology and questionnaire “think aloud” interviews are yielding insights
into how people answer questions. It seems that, as long as the interview is well-conducted,
respondents typically do the best they can to answer to provide accurate and truthful answers. At
the same time, they try to minimize the effort they have to make to respond to the interviewer’s
19
questions, thus acting as “cognitive misers.” As a result, they may switch their tactics for
answering expenditure questions as the difficulty of the task increases. Over short recall periods
or for rare but important events in their lives, their answers are based on counting; they recollect
individual events and then add them up. Over long periods, or for unimportant events that happen
frequently, or for aggregates containing large numbers of items, they resort to more approximate
ways of estimating their answer by estimating the frequency of the occurrence in question and
then multiplying this number by the length of the reference period. The frequency that they
choose may or may not be accurate and may, for example, be overweighted towards their current
or recent behavior, with the respondent ignoring or giving inadequate weight to exceptional
events. Sudman et al (1996) stated that their “tentative finding is that estimation is unbiased but
counting methods, although reducing variance, may be biased either up, for short time periods, or
down, for long periods.” In this sense, respondents’ estimation strategies can be thought of as an
alternative to diaries as a means of dealing with recall and telescoping biases. Indeed, there is
some evidence from consumer expenditure surveys in Canada (McWhinney and Champion,
1974) that diaries and interviews with annual recall periods gave closely similar results.
It is worth recording how one survey, the US Consumer Expenditure Survey (CEX), has
changed its design to accommodate the findings in the literature (see Jacobs and Shipp, 1993).
Although the CEX is fielded in an industrialized country rather than a developing one and
although its primary function is to collect weights for the price index rather than to monitor
welfare, it collects detailed and aggregated consumption measures, and its experience is relevant
to designers of LSMS surveys. The CEX is a survey that could be seen by an ambitious statistical
office in a developing country as a model to emulate and there is good documentation on its
experience with telescoping, recall bias, seasonality, and the advantages and disadvantages of
diaries compared to interviews. While it is wise to be cautious about generalizing to other
countries, all of this experience is relevant for LSMS designers.
In the earliest CEX surveys up to the 1960–61, expenditures were obtained by interview
with an annual recall period. An interesting feature of this methodology was a “balancing” pro-
cedure whereby, if a household’s reported expenditures and reported income differed by more
than a pre-specified limit, the household was revisited for cross-checks. However, in the decade
before the next survey in 1972–73 came the work on telescoping by Neter and Waksburg (1964),
as well as the influential experiments on diaries and interviews by Sudman and Ferber (1971).
The annual recall was abandoned, as was the “balancing” procedure, which was considered
arbitrary and unworkable in the absence of an annual recall period. It is worth noting that the
available literature does not provide any evidence that the annual recall procedures were
unsatisfactory or that the balancing procedure failed to work, only that such techniques had been
superseded.
In the current design, which was introduced in 1980, one set of households keep diaries
that cover their expenditures on food and minor household items (mostly grocery items), while a
different set of households is interviewed on five separate occasions. The data from the first
interview, to which a recall period of the previous month applies, are used to eliminate
telescoping bias, and are discarded. At each of four subsequent quarterly interviews, households
are asked to recall their expenditures in each of the previous three months. The evidence from
these surveys is entirely consistent with previous evidence about the significance of telescoping
20
bias. According to Silberstein (1990), the discarded rates of expenditure from the first interviews
are much higher than those at the subsequent interviews; for clothing, the totals are 40 percent
higher than the average of the subsequent four interviews. However, there is also “internal” teles-
coping or recall bias in the data from the subsequent interviews, with respondents consistently
reporting higher expenditures in the most recent (third) month than in the earlier (first two, and
furthest back) months in the three-month reports. There are also pronounced seasonal effects in
the reported expenditures in the CEX, especially associated with the year-end holidays
(Silberstein and Scott, 1992). The US Bureau of Labor statistics doubles the size of the sample
over the holiday period to deal with these and other (largely non-response) effects, although the
effectiveness of this measure is in some doubt.
Experience from Past LSMS Surveys. Past LSMS surveys have used a range of recall
periods for consumption items, depending on the item and on the survey. For example, for food
purchases, the on-going Jamaican survey uses a seven-day and a 30-day recall periods. In South
Africa, respondents were asked whether they bought each food item on a weekly basis or a
monthly basis and were then asked to report their purchases during the last such period. The
survey in Ecuador took a similar approach in that the respondent chose the recall period. In the
Kyrgyz Republic, Nicaragua, and Russian surveys, the recall period was one week, in Brazil, it
was two weeks, and in China (the Hebei and Liaiong provinces only), it was specified simply as
“1994.” In many surveys, non-food items have often been separated into high frequency or
“daily” categories, for which the recall period is shorter (a week or two) than it is for
“occasional” items, which have tended to be given monthly, quarterly, six monthly, or annual
recall periods. For non-food items, some surveys have had two recall periods, the Jamaica survey
uses a month and a year, and other surveys have sorted different items into different single recall
periods. For example, expenditures on soap may be reported on a monthly basis, clothing on
quarterly basis, and vacations on an annual basis. In spite of this variation, there is one design
that is frequently thought of as an LSMS standard, partly because it is not used widely in other
surveys and partly because it was used in several of the earliest and most widely analyzed LSMS
surveys. In this protocol, food expenditures are collected in a two-stage process. Respondents are
asked whether the household has consumed a particular food item during the last year. Those
who answer in the affirmative are then asked a series of follow-up questions. First, the
respondent is asked whether the household purchased any of the foodstuffs on the list and, if the
answer is yes, a further question about the value of any purchases that they made since the
interviewer’s last visit (which in the prototypical fieldwork plan for LSMS surveys is 14 days).
Second, the respondent is asked in how many months of the year the household purchased the
food item, how often it purchased the item in each of those months, and how much it usually
spent each time. Data on the value of home-produced food is collected in a separate set of
questions that ask how often the home-produced food is consumed with a recall period that has
varied from country to country in previous surveys ranging from “each time the home-produced
food is consumed” to each day to a typical month.
This design makes it possible to compute two different estimates of the monthly rate of
expenditure for each food item. The “last-visit” measure would be zero if no purchases were
reported; otherwise it would be the amount reported since the last visit divided by the number of
days since the last visit and multiplied by the number of days in an average month (365 over 12).
The “usual month” measure would be zero if nothing were purchased in the previous year;
21
otherwise it is the reported usual monthly expenditure multiplied by the number of months in
which purchases were made and divided by 12. Most analysts of the data, whether constructing
poverty profiles or conducting research, appear to have used the “usual month” figures.
In some surveys, the respondent was offered only a single recall period for some non-food
expenditures -- a week, since the interviewer’s last visit, a month, or a year depending on the
supposed frequency of expenditure. For example, frequently purchased non-food items, such as
newspapers and tobacco, are usually collected with a recall period of the previous seven days or
since the interviewer’s last visit. However, for a substantial number of items, a dual procedure
similar to that used for food has often been followed with a last visit measure constructed as
before, although the alternative is now expenditures “in the last year” rather than usual monthly
expenditure. In this case, a monthly estimate can be constructed by dividing the response by 12.
In terms of the literature discussed in the previous subsection, the “last- visit” measure
can be thought of as an explicitly bounded measure that elicits from the respondent an answer
based on his or her recall and counting of events, while the “usual month” question is an attempt
to elicit from the respondent an answer based on a rate or frequency. Of course, these are
generous interpretations. In the standard protocol, no consumption questions are asked at the first
(in other words, previous) visit. Thus, although the visit itself may be fixed in the respondent’s
memory, it is not clear that this will help to reduce the errors in the dating of purchases that
underlie telescoping. The respondent may remember the previous visit of the interviewer very
well but still be unable to recall whether his or her trip to the market occurred before or after that
visit. There is also a serious issue about whether the “usual month” recall is likely to be
independent of the “last visit” recall. If respondents are “cognitive misers” who are casting about
for a basis for estimating the “usual” rate, the questionnaire’s wording will to give them one, in
other words, the answer to the question that they have just been asked -- the “last-visit” question.
Nor would omitting the “last visit” question necessarily solve the problem if the respondent’s
answer to the frequency question is unduly influenced by his or her recent behavior.
In background work for this chapter, data from the LSMS surveys in Côte d’Ivoire (from
1986), Ghana (from 1988), Pakistan (from 1991), and Viet Nam (from 1992–93) were used to
compare the consumption estimates from the different recall periods used in the consumption
modules. The aim was to look for evidence that reported mean expenditure rates decline with the
length of the recall period and to check whether the same is true for the dispersion of the
estimates.
For purchases of food (excluding the value of home production), the estimates are very
similar between the “last visit” and “usual month” methods of calculating monthly food
expenditures. Where there are differences, they do not conform to the expected pattern in which
both mean and dispersion fall from the last visit to the usual month figure. Indeed, the differences
are most marked in Côte d’Ivoire, where the last visit measures have lower means and medians
by 5 and 8 percent respectively, and the measures of dispersion are very close. For Ghana, the
two sets of numbers are effectively identical, a finding which extends to the complete
distributions so that no poverty or inequality measure would be different if one kind of data were
used to calculate it rather than the other. In Viet Nam, the last visit measures are slightly lower
than the usual month measures, and the dispersion in the latter is perceptibly lower. A close
22
inspection of the details revealed that there are fewer very low reports in the usual month data
than in the last visit data. (The Pakistan data cannot be used for this comparison due to the
problems discussed above.)
For non-food items, for which the frequency of purchases is generally lower than for
food, the differences between the two measures are more marked. For the means, the usual
month measures are lower than the last visit measures for Côte d’Ivoire, Pakistan, and Ghana but
not for Viet Nam. The dispersions are lower for the usual month measures than for the last visit
measures in all four surveys. Except for Viet Nam, the decline in means is consistent with the
syndrome of increasing forgetfulness over the longer period, and all the dispersions are consistent
with the view that the last visit (two week) measure of these items is too short to give an accurate
measure of households’ annual standards of living.
When food and non-food are put together with other items to give a consumption
aggregate, dispersion of total per capita household expenditure from the usual month data is less
than for the last visit data, but the difference is not very marked, given the fact that the item with
the biggest drop in dispersion -- non-foods -- is a relatively small share of most households’
budget. Nevertheless, the differences are enough so that the headcount ratio measures of poverty
will all be lower when the usual month data are used rather than the last visit data.
These results can be seen as encouraging, since they show that the results are only slightly
sensitive to the choice between two of the most obvious reporting periods. An optimistic
interpretation is that food expenditures are made with sufficient frequency and are sufficiently
stable so that the last visit and usual month estimates are similar. While this is not true for
infrequent expenditures for which the choice of reporting period affects both bias and variance,
the total of such items is usually too small for their net effect on total consumption to be very
large.
Of course, there are some caveats to this finding. If respondents forget about their food
purchases at the sort of rates suggested by Scott and Amenuvegbe (1990), the last visit measures
may themselves be substantially underestimated. Also, the consistency between the usual month
and last visit measures for food may simply reflect households using their responses to the last
visit questions to guide their answers to the usual month questions, which would not then
constitute any sort of independent check on the validity of either measure. Note also that there are
some non-food items in these surveys (mostly frequently purchased items) for which there was
only a single recall period. While it might be hoped that these non-food items would not be
sensitive to the choice of recall period, the inclusion of these items in the consumption totals
certainly mutes the effects of the different recall periods on the overall totals.
Options for Future Surveys. From the discussion to date, it is clear that there are no
definitive answers about the optimal recall period. In the meantime, however, surveys must be
designed so there follows a discussion of some suggestions and options.
Note first that, if the survey being planned is meant to be comparable to another survey, it
makes sense to use the same recall periods as were used in the other survey, provided that the
previous survey conformed with best practice standards as outlined in this section of the chapter.
23
Beyond this consideration, there are two main routes to go. The first is essentially the
“status quo” of the design of most recent LSMS surveys, while the second is a more extensive
revision premised on the supposition that the current design is unsatisfactory. The preference of
the authors of this chapter (which is reflected in the draft questionnaire) is to make only minor
modifications to the status quo and to experiment with the components of any revision before
putting it into practice.
The “status quo” design is to use two recall periods -- one being since the interviewer’s
last visit and the other being for “usual” expenditures. As already argued, the former yields
estimates that have a minimal amount of telescoping, while the latter is a calculated, unbiased
estimate. The evidence of the comparisons with NIPA data can also be cited; these do not show
the gross and systematic underestimation of expenditures that might be expected if the recall data
missed a very large fraction of expenditures. Finally, an important characteristic of LSMS
surveys has been their coverage of many topics, which would be threatened by including a vastly
more extensive consumption module. Consumption is already one of the longest and most
expensive modules of any LSMS questionnaire, and to extend it further (even if the resources
were available to do so) would inevitably crowd out other important topics. Nevertheless, these
are clearly not strong arguments. The NIPA comparisons are weak, the consistency checks in the
surveys are capable of many interpretations, not all of which support the validity of the data, the
“last visit” recall period contains no bounding questions, and arguments in terms of expediency
are much weaker if the data are of uniformly poor quality. Moreover, the two-visit structure is
being used increasingly less frequently (for reasons not related to the design of the consumption
module), in which case the last visit question would have to be replaced by a question with an
unbounded recall, which will increase the potential for telescoping.
The other point of view is that the LSMS standard questions are far from being the best
practice and that, to quote one of our reviewers, the late Chris Scott, “the use of the question,
`How much do you usually spend on mangoes in one of the months that you purchase mangoes?’
appears to fall far outside of reasonable best practice.” Scott, who had extensive experience of
carrying out consumption surveys in Africa (and elsewhere), believed that only a more extensive
approach would yield adequate data. In particular, he argued that respondents should be
interviewed several times, separated by the most accurate recall period, perhaps as little as a day.
The number of interviews should be as many as are needed to cover the reference period, say a
week or a fortnight, with a bounding interview at the outset. Some or all of these interviews can
be replaced by diary keeping by the respondents themselves, by proxy record keepers in the
household, or by interviewers who complete the diaries with help from the respondents, an
alternative that blurs the line between diary and interview procedures (see below). There is some
evidence reviewed below that diaries can capture some expenditures that may be missed in
interviews, and with a sufficient number of such interviews, the substantial additional cost of
conducting interviews can be translated into high-quality data. Nevertheless, even with a two-
week reference period (requiring 10 or more interviews), seasonality is still not captured nor are
other fluctuations in consumption over the year. It is conceivable that “usual month” responses
do capture some of this variation, but this is probably little more than a hope.
It is recommended that every survey have a budget for experimentation. Questions should
24
occasionally (if not regularly) be subjected to cognitive laboratory techniques and revised and
updated in the light of the results. Even more importantly, the extensive interviewing and diary
techniques discussed above urgently need to be compared against the standard “last visit” and
“usual month” responses, preferably on randomly selected subsets of households within the same
survey.
Finally, it cannot be overemphasized that the wordings of the recall periods must be
unambiguous and well understood by interviewers. Wordings such as “since my last visit, two
weeks ago” are obviously ambiguous if the visit did not occur exactly 14 days ago. There are also
possibilities for confusion in recording the units of a purchase. Where quantity or price data are
sought, it must be clear what unit they refer to. Obtaining the units for quantities purchased may
work better than obtaining the units for prices or unit values, so this is reflected in the draft
consumption module presented in Part III.
Multiple Visits Throughout the Year
In most past LSMS surveys, primary sampling units and the households within them were
visited on two occasions two weeks apart, but the consumption data are collected only during the
second visit. Thus, there is a single record of consumption over whatever recall period is
selected. This has also necessarily been the case in the increasing number of LSMS surveys
where only a single visit is made to the primary sampling unit. Even if the surveys were to adopt
a more intensive program of multiple interviews, consumption data would still be collected over
only a relatively brief period, say a week or a month. No proposed design of the LSMS
consumption module would capture variations in household consumption over a whole year, so
that the consumption data that are currently collected may not reflect the annual consumption
flows in which analysts are fundamentally interested. To collect better data, it would be necessary
to revisit households on several occasions throughout the year and to collect consumption data
during each visit. Such data could be used to increase the accuracy of the consumption
aggregates, allowing for variations in households’ consumption over time, and they could also be
used for a number of analytical exercises.
One issue is seasonality. The collection of agricultural data and agricultural income
typically requires that the survey interviewers visit households (farms) in different seasons of the
year. Farm incomes are seasonal, and it may not be possible for a respondent to remember all of
the transactions that go into a calculation of net income many months afterwards. In
consequence, it is widely believed -- and suggested in Chapter 19 in this volume -- that accurate
farm surveys require multiple seasonal visits. The concern in the current chapter is with con-
sumption, but it is conceivable that consumption expenditures -- like incomes -- vary
systematically with the seasons, so that if respondents’ memories are a problem, accurate
estimation of annual consumption flows will require multiple, seasonal visits just as does the
accurate estimation of income.
While the seasonality of consumption is a fact in most countries, it does not necessarily
follow from the seasonality of income, and it may have a quite different seasonal pattern. At the
theoretical level, farm households have strong incentives to untie their consumption from the
seasonal patterns of their incomes. In extreme cases, harvests generate income only during a few
25
weeks in the year, while consumption has to be maintained throughout the year. Note that it is
not necessary to accept the permanent income or life-cycle hypothesis of consumption to believe
that households can smooth their consumption over the year. Even without access to credit
markets, farmers can store some of their output or can save some of their income from the
harvest to support their consumption throughout the rest of the year. Indeed, given the
predictability of the seasons in agriculture, it would be extraordinary if farmers had not learned to
do so. Of course, there will be still be seasonal patterns in their consumption. Festivals such as
Tet or Christmas are associated with higher than normal expenditures. Furthermore, storage is
costly, so that prices will generally be higher just before the harvest than just after it, and there
will be some resulting effect on consumption. However, the point is that seasonal variation in
consumption is (i) not closely tied to, and (ii) less than, seasonal variations in income. Reliable
evidence on the relative seasonal smoothness of consumption comes from Paxson (1993). Rice
farmers in Thailand who double crop (with irrigation) have quite different seasonal patterns of
income than farmers with only one crop, yet their consumption patterns are almost identical and
exhibit little seasonal variation. The same is true of a comparison between farm and non-farm
households and among farmers in different agroclimatic zones.
An attempt has been made for this chapter to look for seasonal patterns in the
consumption data from three LSMS surveys. Cross-sectional surveys are not well-suited to this
task, since what is required is not a large number of households observed throughout a single
year but rather a large number of years over which the seasonal patterns can be established.
Nevertheless, the consumption totals for Ghana, Côte d’Ivoire, and Viet Nam were examined for
differences across months. In any one survey for a single year, there tend to be significant
differences in the consumption total from one month to another. Some of these -- such as Tet in
Viet Nam -- are easy to explain, but others are not, even though their effects are sometimes large.
However, many of the differences that can be seen are probably not driven by seasonal patterns in
consumption. Evidence from this comes from Côte d’Ivoire and Ghana, for which there are data
for more than one year. In both cases, the monthly patterns in the consumption total are quite
different across the survey years, which does not support any simple unvarying seasonal pattern.
Even so, the source of the monthly differences remains something of a puzzle. The progression
of the survey teams through the country may generate some variation as they move from poorer
to richer villages. Random measurement error is also likely to be an important factor. Never-
theless, this limited analysis provides no ground for supposing that consumption would be better
measured by including multiple, seasonal visits to households in LSMS surveys in the future.
However, there are other reasons why multiple visits might be useful. The average of two
consumption totals, each for a two week recall, will give a better, lower-variance estimate of
longer-term consumption. More radically, Scott (1992) and Central Statistical Office of Zambia
(1995) advocate using the correlation between consumption across multiple visits to correct
measures of inequality to bring them closer to what would have been measured had it been
possible to collect consumption data over a full year for each household. The idea is as follows.
Suppose that analysts are interested in annual consumption measured at a monthly rate but that
the only observations that they have are of consumption over the previous month. At one
extreme, each household may consume the same in each month so that the monthly totals are
correct and can be correctly used to give a measure of dispersion over households. At the other
extreme, suppose that each household’s consumption is uncorrelated from one month to the next.
26
The “last month” totals are then correct on average, but their dispersion over households would
be larger than the dispersion in which analysts are interested because it has a “within-household”
component in addition to the “between-household” component with which analysts are
concerned. With multiple observations for at least some households and under some reasonable
assumptions, it is possible to estimate the size of the within-household dispersion and to correct
the measures of the total. The Central Statistical Office of Zambia (1995) uses such a procedure
to correct dispersion measures for a Zambian survey, and the technique could usefully be applied
elsewhere.
Note also that, as always, measurement error will add dispersion to measured
consumption, so that, if measurement error is random, the dispersion that it causes will be in
addition to the genuine dispersion that comes from the within and between household
components of dispersion. Although this is not explicitly allowed for in Scott’s formulation, his
corrections should still lead to a better estimate of dispersion. With repeated observations of the
same households, there are a number of techniques that would make it possible to assess the size
of measurement errors (see particularly Griliches and Hausman, 1986).
Finally, note that, in some circumstances, multiple consumption measures can be useful
at the analytical stage. Multiple visits generate a type of longitudinal or panel data that can be
useful for studying changes over time and for sorting out cross-sectional as opposed to time-
series variation as in Scott’s work above. However, if the visits are separated by only a few
months, the changes that may have occurred may not be large enough to be interesting, and the
measured changes are likely to be dominated by measurement error.
The recommendations made here are once again tempered by the increased costs of
extending the consumption module of a typical LSMS survey, both in terms of money and of the
consequences for the rest of the survey. Because consumption is smoothed within the year,
measuring it over a fortnight or a month may yield a sufficiently accurate picture of annual
consumption to make it not worth incurring the cost of adding yet more visits. On the other hand,
if the detailed agricultural module is included in the survey with multiple visits at different
seasons, there would be little additional costs involved in collecting at least some consumption
data at each visit. Such cases apart, multiple visits throughout the year are probably not the
highest priority for improving the typical LSMS survey.
Imputing Values
In nearly all LSMS surveys, calculating a comprehensive measure of consumption will

require at least some imputations. Not all consumption is obtained through market purchases, so
that if analysts want to calculate consumption in monetary units, they must find some way of
pricing its unmarketed components. In many of the poorest countries, and especially for the
poorest people, a large share of food comes from home production or from collecting wild
foodstuffs, hunting or fishing. These imputations for food are likely to be those that are most
important for the totals. However, there are a number of other commodities that are obtained by
the household’s non-market labor such as home-made clothes or wood and water fetched by
children or women. Household members often receive gifts or payments in kind that need to be
priced before being added in to the consumption total. Finally, it is necessary to obtain
27
information on the consumption that comes from using durable goods. For non-durable goods --
even those that are partly durable -- it is probably safe to assume that consumption and purchases
are one and the same. However, for large durable goods such as houses, cars, or bicycles that are
expensive and that last for many years, it is important to try to make some adjustments. This
subsection reviews the data needs that are required to make those imputations.
Before doing so, it is worth noting that imputation is an inherently difficult and error-
ridden process. It is likely to work best where there is relatively little need for it, which is when
the economy is highly monetized but where there is a relatively small amount of own production
(such as vegetable gardens) involving goods that have clear market equivalents. It works badly in
economies where a large share of transactions do not pass through the market. LSMS procedures
for estimating welfare stem from a theory of a consumer with well-defined preferences operating
in a market where prices are well-defined and unaffected by the agent’s behavior. Where these
markets do not exist, analysts are effectively imposing an accounting framework on the physical
data that is of dubious relevance to the lives of the people that are being studied.
Home-produced food, or food received as gifts or payment in kind, has been much the
most important imputed item in most of the LSMS surveys to date. In principle, the calculations
are straightforward. The respondent is asked to report the value of any home-produced food
consumed by the household during the reference period, and then the sum of these items is added
in to the consumption total. Given the seasonality of production, the recall period probably has to
be a year, or at least a typical month over the last year. It may be possible to do better than this
when there is a multiple-visit agricultural module in the survey. However, the major difficulties
are with valuation, since the respondent is being asked a purely hypothetical question about the
sale or purchase of an item that is only rarely traded or that may have only been traded some time
ago.
The draft module presented in Part III recommends collecting data on physical quantities
of goods consumed since these are magnitudes that the respondent observes, at least in principle.
The value or price of these quantities can be obtained in several ways. Farm-gate prices, defined
as what the household could get for its production, set a lower bound on valuation, since it is
usually presumed that consumption is evidence that the good is valued beyond what it would
fetch. Market prices, by contrast, are likely to be too high since they include transport and
distribution margins and because the commodity that is traded is often of a higher quality than
the home-grown variety. However, once the quantity has been obtained, the respondent could be
asked to report one or both of these two prices or simply to estimate the value of the commodity
directly. Some degree of cross-checking is possible from the quantities and prices of purchases
reported in the agricultural module or from the prices gathered in the community questionnaire.
In some circumstances, it may be possible (or important) to carry out a similar exercise
for non-food items. Clothes and furniture are often made at home, and household labor is used to
collect firewood, dung, or water. These items have usually been omitted from past LSMS
surveys, probably because of difficulties in valuation. One danger is that the welfare of poor
households might be overstated by using inappropriate prices or wages to value their production
or their labor. For someone with no other employment opportunities who ekes out a living
gathering firewood or coal, it is adding insult to injury to impute a high standard of living to them
28
by valuing their time in terms of the market wage in a formal sector that is inaccessible to them.
In some circumstances, the wage data from the community questionnaire may be a better basis
for imputations based on the value of time, but they do not eliminate the dangers inherent in the
procedure.
Note also that it is incorrect to compute an extended concept of consumption by adding

“expenditure” on leisure (in other words, the value of leisure at the market wage) to total
consumption. This “full-income” concept has its uses, but it is a nominal measure and, like other
nominal measures, must be converted to real terms before being compared across households or
individuals. The problem here is that, even if everyone in the comparison faces the same (or
similar) prices for goods, they do not face the same price of leisure because wage rates differ. As
a result, full income must be deflated by some price index that includes the price of leisure before
it can be used as an indicator of welfare. Alternatively, if the value of leisure is to be added into
the value of total expenditure, the same wage rate should be used to value everyone’s leisure.
For durable goods, the consumption flow is best thought of as a rental equivalent or “user
cost.” This has two components: (i) the opportunity cost of the funds tied up in the good that
could be realized through its resale and (ii) the value of the physical depreciation of the good that
comes through using it or simply through the passage of time. To estimate these magnitudes,
some measures of depreciation and of current value are needed. Perhaps the simplest way to
obtain the information, at least for goods purchased in the previous five to ten years, is to ask
respondents to report when they purchased the good and how much it then cost. These are both
factual matters and will often be well-remembered, at least in the case of large, important items.
Provided that the good has been available for some time and provided that purchases have been
made relatively evenly over time, an estimate of the average lifetime of the good can be obtained
by doubling the average age over all similar goods for all households in the survey. Once this is
known, the depreciated value of the good can be estimated given its age and original value,
which is then used to calculate the first component of the user cost. It is also possible to ask
respondents for direct reports of the current market value of the used durable, though there is no
evidence on the likely accuracy of such (hypothetical) reports, and field tests are likely to be
useful.
For housing, the largest of the durable goods, the imputation approach again starts from
the rental equivalent. Unlike the value of most other durable goods, rents can sometimes be
observed directly, and these are the correct numbers to add into the consumption aggregate. For
households that do not report rents, the standard procedure is to impute a rent based on the
characteristics of the house, as reported in the housing module. This is typically done by
“hedonic” regressions in which reported rent is regressed on the house’s characteristics (such as
is size, number of rooms, construction material, and location) and the results used to calculate
rents for other properties where rents are not reported. The credibility of these regressions is
compromised if only a small fraction of the sample report rents and, more generally, if those who
report rents are unrepresentative of the population as a whole. While it is possible to make
mechanical corrections for the selection, these usually require arbitrary and untestable
assumptions that further compromise the credibility of the process. This is a difficult area. In
general, survey analysts should make sure that indefensible imputations are not dominating
welfare comparisons. The data required for such imputations are gathered in the housing module
29
(and, occasionally, in the community questionnaire) and so are not discussed further in this
chapter.
A number of other imputations come from other modules in the survey. In particular, the
employment module gathers information on income in kind provided by employers, including
(free or subsidized) transport to and from work, food at work, or housing.
Respondents for the Consumption Module
Most LSMS surveys have interviewed a single respondent for the whole of the
consumption module or for each different part of it. The household is asked to determine who the
“best informed individual” is and to have that person respond. This has the appealing feature of
not pre-judging the division of labor in a household, either by gender or by age, as would be the
case if it were assumed that the wife does the shopping or if any similar assumption were made.
In many, perhaps most countries, the single respondent approach works well. In particular, it will
be satisfactory where food is a large share of the budget, where there is a common cooking pot,
and where most of the household resources are pooled.
Even within resource-pooling households, it may be useful to have multiple respondents

for some expenditures or different household members reporting on different categories of
expenditure. While the person who does most of the shopping for food will know about this large
share of the household’s budget, another large share could be most accurately reported by the
person who pays the housing and utility bills (who may or may not be the shopper). Then there
are other expenditures of which no one single person may have a very accurate picture.
Individuals may not know even how much “walking around” money the other household
members have, much less how they spend it -- whether on bus fares, meals away from home,
newspapers, tobacco, alcohol, or entertainment. Also, there may be larger items, such as clothing,
that individuals purchase without any other household member knowing how much was spent.
This is particularly true where several adults live in a household, each contributing something
less than all of their income to the household’s joint expenses and reserving the remainder for
their individual use. For example, in a household with a mother and grown sons, the mother may
handle all the household expenses, including food and utilities, paid for from her pension and
from regular monthly sums given to her by her sons. However, she may have little or no idea
about her sons’ incomes or other expenditures, which could account for most of the household’s
total income and outlay. In this situation, no one person can give an accurate report of the
household’s income and expenditure, nor is the “household” really the relevant unit for analysis.
The LSMS has little experience in procedures for dealing with these situations. In the
literature discussed in the next subsection, there have been cases where each adult member of the
household kept a diary of at least some categories of expenditures. It is possible in principle in
interview situations to interview each member of the household about at least some expenditures,
such as those paid for by “walking-around” money. A more ambitious prospect would be to try to
record incomings and outgoings for each member of the household who spends money. This is
likely to be prohibitively expensive for a general multi-topic survey, although, as always, there
are potential benefits from conducting experimental work either on a few households or within a
special survey. Multi-person accounts are likely to provide a fascinating picture of how
30
intrahousehold transfers of resources take place, on who gets what, and on what it is that makes a
group of people function as a household. This is clearly an important research area but not a
prime candidate for incorporation into standard LSMS surveys at this time.
Diaries versus Oral Interviews
The use of consumption diaries, in which households are asked to record their purchases
as soon as they make them, is common in full-fledged single-purpose consumption surveys. The
ideal diary would yield a record of each purchase immediately after it takes place, so eliminating
the need for respondents to rely on their memories and removing any associated errors, including
telescoping. In principle, the diaries may be kept by a single respondent or by several or all
members of the household, so they can also help to resolve the question of who makes the best
respondent for the whole household while simultaneously yielding information on intrahousehold
allocations. Diaries can be organized on a product basis, on an outlet basis, or on a purchase
basis, and the forms can be designed to allow a large degree of prompting (for example, by listing
many types of products) without the associated tedium of a long interview.
There are some practical implications to the use of diaries. The most obvious is that the
person filling them out must be literate. Nevertheless, the diary approach has been used in
household expenditure surveys in a large number of countries where literacy is not universal.
This has been achieved by having the most literate member of the household (sometimes a child)
help the one who does the purchasing to fill out the diary or having the interviewer visit the
household frequently to help the household to fill it out, perhaps even on a daily basis, see
Blaizeau (1998). In such cases, the distinction between a diary and an oral interview becomes
blurred. This blurring occurs even in literate households in which the members of the household
either forget to fill out the diary or get tired of doing so. In the US CEX, substantial numbers of
diaries are completed by the interviewer at the time of collection based on the respondent’s
memory. When collecting the diary, the interviewer examines it briefly and, if it appears to be
incomplete, he or she will try to prompt the respondent to fill it out more completely, which
becomes essentially an interview. To the extent that these “diaries” rely on respondents’
memories, telescoping and recall bias once again become potential problems.
The second logistical issue is that the diary must be explained and left with the household
and picked up after its completion. If the diary period is relatively short (a week or two), this
would not necessarily pose a problem in an LSMS survey, since the completion of the whole
questionnaire typically involves multiple visits by the interviewer to the household so that each
member can be interviewed and so that the length of each interview can be kept reasonable.
However, leaving a diary with a household for a long period of time, such as a month or a
quarter, would be more difficult and would not be possible within the current design of most
LSMS surveys. Thus, diaries will only be pertinent for items for which a relatively short recall is
appropriate and, thus, not for a large proportion of non-food expenditure.
Third, the use of a diary changes the burden of interviewing. Most directly and measur-
ably, it reduces the amount of time that the interviewer has to spend interviewing households that
fill out the diary completely, though it may increase the time that the interviewer must spend to
travel to the household to collect it and to help illiterate households to fill it out. Moreover, it
31
shifts the burden of response onto respondents. The effect of this is unclear. Some survey
statisticians speculate that households may enjoy the novelty of filling out the diaries, while
others think that it makes it possible for the household to fill out the diary at the time of day or in
the place that is most convenient to them. If either of these speculations are true, then the burden
on the households would be limited, and they would not be deterred by the prospect of
participating in the survey or of keeping accurate records.
Nevertheless, evidence from industrialized countries shows that keeping diaries is likely
to deter households from participating in surveys and that the burden of keeping the diary causes
respondents to drop out of the survey over time. There is also evidence that the rate of reporting
declines with time, so that, in two-week diaries, more consumption is recorded in the first week
than in the second. This is also true for the CEX in the US, and is documented for seven West
African countries by Blaizeau, 1998. In Belarus’ 1995 and 1996 income and expenditure surveys,
the expenditures recorded in the second week were about 15 percent lower than those in the first
week of diary keeping (Martini and Ivanova, 1996). In Armenia, the diary was kept for four
weeks, and the downward trend continued over this longer span. The second week’s expenditures
on food were 26 percent lower than the first, the third week’s were 35 percent lower than the
first, and the fourth week’s were 40 percent lower than the first (calculations done for this
chapter). This could be due to respondent fatigue or to the fact that the novelty of diary keeping
wears off. Also, the very fact that they are keeping a diary may cause people to spend more or to
shift their expenditures forward into the diary keeping period. Keeping a diary may cause people
to think more about their consumption and perhaps to take the opportunity to buy some items that
they needed in any case. To the extent that diaries are not filled out every day, there is also scope
for telescoping and recall errors within the diary period.
Having several members of the same household each keep a diary is also an attractive
option in some circumstances, and there is considerable variation in practice among different
kinds of surveys. For example, in the US CEX, there is a single household diary, while in the
British FES, all adults in the sample keep individual diaries. There is also evidence from Hong
Kong on the use of multiple diaries, reported by Grootaert (1986), from Papua New Guinea,
reported by Gibson (1998), and from surveys in Benin, Burkina Faso, Côte d’Ivoire, Mali, Niger,
Senegal, and Togo reported in Blaizeau (1997). This literature finds that multiple diaries can be
useful for obtaining records of expenditures that would otherwise be missed, but that it is
difficult to get all household members to cooperate, so that response rates can be reduced in the
attempt. It is also clear that for some household members, certainly children and perhaps some of
the elderly, proxy or household-level reporting can be more accurate.
To the existing literature can now be added evidence from experiments in three former
Soviet republics -- Latvia, Armenia, and Ukraine. In the Latvian experiment (analyzed by Scott
and Okrasa, 1998), a nationally representative set of 300 households were given oral interviews
and asked to keep diaries covering a comprehensive lists of foodstuffs and a selection of non-
food items. In half of the sample, the diaries were administered first, and, in the other half, the
interviews were done first. The order of the instruments had virtually no effect on the results.
Overall, food expenditures were about 46 percent higher for the diary than for the interview; the
coefficients of variation were quite similar. This pattern also held for 13 of the 15 subgroups of
food and generally for the quantities as well as for the expenditures. For non-food, the results
32
were much more mixed. For one of the four categories, the expenditures reported in the diaries
were significantly greater than those reported in the interviews, for two of the categories, the
opposite was the case, and in the fourth category, the differences were not significant.
In Armenia, as part of nationwide survey, a quarter of the households in each cluster were
given diaries for 30 days and interviews with 30-day recall periods were conducted with the
remaining three-quarters of the households. The only diary data available to the authors in usable
form are from the food sections of the diary, so the comparisons done for this chapter had to be
be limited to those items. Total expenditures for food recorded in the diary were about one-third
higher than those resulting from the interview, a result similar to that in Latvia. However, this
pattern did not hold as strongly for the sub-groups of food. Of the 15 groups (which each
included from one to 19 items), the diary yielded significantly higher expenditures than the
interviews in eight cases, significantly lower expenditures in two, and not significantly different
expenditures in the other four cases. There is no noticeable pattern to these results in relation to
the mean expenditure, the number of items in the sub-group, or the average frequency of
purchases. The coefficients of variation for the diary were twice as large as for the oral
interviews. The same pattern of results generally held when the whole sample was divided into
rural and urban areas, although the differences were slightly larger in rural areas. The diary
estimates for total food expenditure were 43 percent higher than for the interviews in the rural
areas and 31 percent higher in the urban areas.
In Ukraine, a diary was administered to about 500 households in selected locations as an

experiment in conjunction with a national survey that used interviews. The diary used a recall
period of four weeks for all items. The oral interview used a recall period of two weeks for food
and four weeks for non-food items. The lists of items in the two instruments were not the same.
The comparisons made here were limited to those categories that were either identical between
the diaries and the interviews or that could be clearly mapped into each other. For instance,
“butter” was an item listed as such on both lists, but the interview list contained a single question
about “smoked sausage and other smoked meats” while the diary had separate items for “smoked
sausage” and “other smoked meats.” Thus, in this experiment, there are two possible effects that
will work in opposite directions. On the one hand, the diaries might yield higher expenditures
than the interviews, partly because of the greater disaggregation of the items on the diary list.
However, the shorter recall period in the interview may have meant that the interviews yielded
higher expenditure numbers due to the greater proportionate influence of any telescoping.
Moreover, if diaries are kept less rigorously as time goes on, the shorter recall period in the
interview may mean that higher numbers are reported in the interviews than in the diaries. These
effects cannot be disentangled because data available to the authors contain only the sub-totals
for each item recorded in the diary rather than each individual purchase of it.
In fact, in the Ukrainian experiment, the expenditures for the subtotal of food items that
could be matched between the two instruments are 10 percent lower in the diaries than in the
interviews, a difference which is significant only at the 10 percent level of confidence. In only
three of the 11 categories are expenditures significantly different between the diary and the
interview. The category that covers bread and flour accounts for about two-thirds of the total
differences found in the food sub-totals. For the sub-total of non-food items that could be
matched between the two questionnaires, the diary sub-totals were 7 percent lower than those for
33
the interviews, a difference that was not significant.
There are several features of these three experiments that limit the extent to which they
can be generalized to other contexts and countries. The populations of Latvia, Armenia, and
Ukraine are literate, so these experiments cannot indicate what would happen if diaries were used
in largely illiterate survey populations. All three countries have long traditions of household
expenditure surveys that use diaries. In Latvia and Armenia, the experiments were carried out by
the statistical office using regular interviews, so that the experiments used experienced staff who
were thoroughly familiar with the procedure. This was not the case in Ukraine, where the
experiments were done by a private survey organization, which may account for some of the
differences in the two sets of results.
Past LSMS surveys have made less use of diaries than might have been the case, and
perhaps more use should be made of them in the future. It is certainly true that an ambitious
statistical office in a developing country, looking to the industrialized countries for inspiration,
would probably use diaries. What is less obviously true is whether, even in rich countries, there is
any indisputable evidence of the superior performance of diaries over interviews to justify such a
decision. While diaries would produce better results if they were used in ideal circumstances, it is
unclear whether their practical superiority over interviews has ever been convincingly
demonstrated. Given the literacy issue in many poorer countries, the argument for a change
becomes even less overwhelming. While there is no doubt that diaries can be used in situations
where interviewers make many visits to the household to help them remember their purchases
and complete the diary, this is closer to diary keeping by the interviewer than diary keeping by
the respondent. In countries such as Armenia and Latvia where diaries have been routinely used
in the past, there is every reason to incorporate them into LSMS-type surveys. However, there
seems to be no compelling case for introducing the use of diaries as standard in LSMS surveys at
this stage.
Data from Other Parts of the Questionnaire
While much of the data required to calculate consumption aggregates come from the
consumption module of the household questionnaire, other important data are usually collected
in other modules of the survey. These are reprised briefly here, to serve as a checklist for the
overall survey design. If survey designers decide not to collect data in the other modules on the
items mentioned here, then they should ensure that they are collected in the consumption module.
Roster. Analysts need to know the number of members that the household has in order to
compute per capita expenditure measures. If they intend to calculate equivalence scales, then they
will need data on the age and sex of the members as well. These are never omitted from rosters,
so this should not be a problem.
Housing, Water, Sanitation, and Fuel Modules. The housing module will collect most of
the information needed to impute the use value of owner-occupied housing. It is also the usual
place to gather information on utilities (such as electricity, piped gas, and telephone service) and
on expenses for water, sanitation, and some kinds of fuel although it is now being suggested that,
in some surveys, these may be moved to separate extensive water, sanitation and fuel modules
34
(these are discussed in Chapter 16).
Education Module. Detailed data on household expenditures on school fees, uniforms,

textbooks, supplies, bus fares and so on are usually collected for each student in the education
module with recall periods that are considered to be appropriate to the category of expenditure.
These expenditures are used to compute consumption aggregates. Collecting this information in
the education module makes it easier to make the questions refer to specific individuals, which is
necessary for much education analysis. The fact that detailed information is collected also means
that the information is more likely to be complete. Conceptually, there may potentially be some
overlap between some of the items recorded in the education module and those in the
consumption module. For example, does the category on children’s clothing in the consumption
module include or exclude school uniforms? Few LSMS questionnaires have been careful to
specify this in the questionnaire itself, although it may have been addressed during the
interviewer training process.
Health Module. Detailed data on health care expenditures such as payments to doctors or
other medical professionals, for prescription medicines, and for lab tests are usually collected in
the health module for each person who incurred such expenses during the recall period, which is
usually the previous four weeks. Data on expenditures on health insurance and over the counter
medicines are usually collected in the consumption module. Some surveys have also included a
general question or two in the consumption module about expenditures on the items covered in
more detail in the health module. In the consumption module, the coverage is for the whole
household and for a reference period of up to a year. The detailed data that are usually collected
in the health module will generally give higher means than the aggregate questions in the
consumption module, but the shorter recall period will result in higher variances. Analysts may
then chose which method best suits their particular analysis. However, survey designers must
ensure that one or the other is included; the detail provided in the health module is probably
preferable if only one is to be included because it will yield more accurate means and support
many health sector analyses.
Employment Module. The employment module is usually the place to gather information
on households’ consumption of goods provided in kind as a part of wages and on their members’
commuting expenses. Different analysts tend to handle commuting and other job-related
expenses (such as childcare, uniforms, and fees for professional associations) differently. Some
even exclude them altogether from the consumption total on the grounds that they do not increase
the household’s welfare. At any rate, it is useful to gather data on commuting expenses to give
analysts the choice.
Environment Module. Households sometimes obtain important resources from the

environment. Water for household use and firewood are perhaps the most familiar examples,
although a wide range of plants, animals, and minerals can be gathered from wild or common
property to be used as food or fodder or as inputs into the household’s enterprises, agriculture, or
housing. To gather a full range of data on households’ use of such resources would probably
require the inclusion in the survey of a special purpose module and the modification of the
consumption, agriculture, household enterprise, and housing modules to ensure full accounting
while avoiding double counting. This has yet to be done in an LSMS survey, though it has been
35
done in some interesting single purpose surveys (see Cavendish, 1998). However, several past
LSMS surveys have gathered information on at least a few of these resources, especially water
and firewood. Such questions have typically covered the quantities used by the household,
sometimes the time spent in collecting or carrying the goods, and sometimes the distance traveled
to find them. Survey designers need to decide how much emphasis to give to this issue in the
survey being planned and to review all pertinent modules to ensure that modifications have been
made where needed.
Community and Price Questionnaires. The community and price questionnaires are the
places to gather data on prices at the community level. If neither a community nor a price
questionnaire is to be included in the survey, then it is of paramount importance that either: (i)
adequate regional price indices are available from some other source or, more likely, (ii) data on
quantities of food and fuel consumed by households will be obtained in the consumption module.
It is also vital that the survey designers are willing to base regional price adjustments only on
food and not to use non-food in the price index.
Saving Module. Data on any regular saving that households may make can be gathered in
either the saving module or the consumption module.
Inter-household Transfers. Many surveys have included a separate module that elicited
data on flows of transfers in and out of the household. In general, it makes sense to put the
questions about inflows and outflows of transfers together and to elicit parallel information about
them. In this book, the inflows are discussed in the miscellaneous incomes chapter (Chapter 20)
and a page is included in the draft consumption module presented in Part I of this chapter on
outgoing transfers.
Part III: Draft Modules
This section contains a set of draft sub-modules to make up a consumption module to be

included in an LSMS-type household survey. The draft module is something of a middle ground
in terms of length. In certain circumstances, some questions could be deleted, and some areas
could be explored in much more detail. When such adjustments are warranted and how to make
them is described in Part IV.
Survey designers need not follow the order in which the sub-modules are presented here
since the same information could have been organized differently. There seems to be no rigorous
evidence about how different organizations would affect the answers given by respondents.
However, there are a few principles that survey designers should bear in mind, which are
discussed below, but the particular layout that best accomplishes these principles may vary.
36
Organization
There are different ways of grouping items together -- by the type of the item (such as
food, fuel, or clothing), by the place of purchase, by recall period, or by the kind of follow-up
questions to be asked (about quantities, prices, or home production.). The exact layout will depend
upon the circumstances in the country where the survey is to be fielded (including what people
consume and how they acquire it) and on the objectives of the survey (is it important that the
survey be able to make a rough estimate of the number of calories consumed by the household or
that it should be able to support intrahousehold analysis or the analysis of home-produced
services?). The survey designers must draft one or more versions of a questionnaire and field test
them rigorously to determine if the proposed module actually works.
Comprehensiveness
The questionnaire should cover all types of consumption. It will often be pragmatic to
define some categories that contain groups of items (for example, “canned goods”) in order to get
a comprehensive list of expenditures without making the list unduly long.
Box 1: Cautionary Advice
How Much of the Draft Module is New and Unproven? The draft consumption module presented here
closely follows the approach taken in many previous LSMS surveys.
How Well has the Module Worked in the Past? The consumption modules in most previous LSMS surveys
have produced data that have appeared to many analysts to have reasonable magnitudes and to show expected
patterns of consumption by categories of items and over types of household. However, field tests of the module in
some countries in Africa showed that respondents had difficulty reporting data on quantities of goods purchased, so
those questions were deleted in those countries. In some surveys, respondents have seemed to be confused about
whether to report a total value and a quantity or a value per unit of purchase and a quantity. Where this confusion
prevailed, it was necessary to delete some observations from the resulting data set or even to discard all of the data
with a short period recall. To minimize this risk, the questions should be clearly formulated, the order of the
questions should be thoroughly field tested, and the interviewers should be fully trained. However, the most
effective insurance would be to use two recall periods, at least for food, to ensure that there is always a backup
measure of consumption in the data set
Which Parts of the Module Most Need to be Customized? Survey designers should base their decisions
about which items to include in the different sub-modules on local consumption patterns. If a comparable survey
was previously fielded in the country in question and if the consumption module in that survey was designed
broadly in accordance with the best practices described in this chapter, then the designers of the new survey should
choose to use the same recall periods and the same degree of item disaggregation as in that previous survey.
Specificity
It is important to list certain items individually. In general, this will be of interest for items
that are important sources of calories or that are particularly interesting to analysts in their own
right, such as subsidized fuels and those goods whose consumption is particularly favored (or
disfavored) by policymakers (“merit” goods or “sin” goods).
37
Double-counting
It may sometimes to be useful to create a cross-check by gathering data on the same items
in more than one place in the questionnaire. In these cases, the questions should be worded
carefully so that the analyst knows how to exclude one or the other to avoid double-counting these
data in the estimate of total consumption. For example, if purchases of “major kitchen appliances”
are listed among the consumption expenditure items, then the inventory of durable goods should
not list “stoves” since it would not be clear whether stoves were included in “major kitchen
appliances” or not. In this case, the durable goods inventory should either use the same category --
“major kitchen appliances” -- or omit all such items.
Customization
Each questionnaire must be customized to reflect the circumstances that prevail in the
country or area where the survey is to be fielded. The lists of consumption items must be
customized to reflect local consumption patterns and local terminology. Which items are listed for
which recall period will depend on the shopping patterns of the local people. A few examples of
this are given in Part IV.
38
Part IV: Annotations on Questionnaire
This section consists of a series of notes and annotations to each of the draft sub-modules
that were presented in Part III. This is followed by several general guidelines for dealing with
some special circumstances within which a survey may have to operate. The section ends with
some advice about how to keep the module as short as possible consistent with the need to gather
sufficient data to support the analysis of the most important policy issues in the country in
question.
Part A: Daily Expenditures
The list of items to be included in this sub-module will vary from country to country. The
idea is to capture (with a short recall period) the many small, repetitive miscellaneous transactions
that many people engage in on an almost daily basis. Individuals often purchase the kind of items
to be listed here with their “walking around money.” As such, this is one of the places in the
consumption module where it would be feasible to collect individual-specific data.
It is important to check that all items are covered in one part of the questionnaire or
another, and that any double-reporting is intentional as a cross-check rather than as a the result of
an ambiguity. For example, if fuel sources are included in the housing module, then they need not
be included here. In some countries, gasoline for cars is included in the daily expenses page,
whereas in others, it is in the non-food consumption page.
The list of items used in the draft sub-module is a mixture of the lists from various
previous LSMS surveys. Other items that could be included are: khat, flowers, gasoline, firewood,
haircuts, shaves, baths, and tips.
In this sub-module, there is a special grid for meals consumed away from home.
Experience has shown that the more detail that appears in the grid, the higher will be the numbers
reported. In Jamaica, in 1993, the questions were individual-specific, and the estimate of
expenditures on food purchased away from home accounted for 5 percentage points more of total
food consumption than in 1992 when a single question on the daily expenditures grid was used
(Table B-6 in Statistical Institute and Planning Institute of Jamaica, 1994 and Table B-4 in
Statistical Institute and Planning Institute of Jamaica, 1995). If the grid is individual-specific, then
it would probably be better to have each person respond for him or herself rather than to use a
single respondent as specified in the draft sub-module. In that case, the page should be adjacent to
other individual-specific modules. Care may be needed to avoid double-reporting subsidized
lunches in the factory canteen or the schoolyard, as these may also be captured in the employment
or education modules.
This draft sub-module is modeled after the one used in the Kazakstan questionnaire
because most of the answers obtained from it made intuitive sense. The number of meals reported
was mostly divisible by five, indicating regular patterns associated with the work week, and the
unit value for the meals seemed to be plausible.
39
Part B: Food and Fuel
This sub-module gathers information on the food and fuel consumed by the sample
households.
List of Items. The list of items that should be included in this sub-module will vary from
country to country. The full-scale expenditure survey done every five or ten years in most
countries to provide weights for the consumer price index has served as the basis for the
development of the lists of foods in past LSMS surveys. Special care should be taken to itemize
those goods that are substantial contributors to the total number of calories consumed and to
expenditures, as well as those that are (or are most likely to be) subsidized. Thus, in Central
America, rice, beans, and tortillas would each be listed separately rather than in a group of
“starches” or “basic grains.”
To prevent the list of items from becoming unwieldy (in the order of hundreds), there will
inevitably have to be some grouping of items, such as “canned foods” or “vegetables.” This
grouping is necessary to ensure completeness and a brevity that is acceptable to respondents. Such
groupings, however, can pose problems for interviewers in eliciting answers about quantities of
the item that were consumed and for analysts in trying to establish the nutrient content of the
items. There is no perfect solution to the problem, but survey designers should bear it in mind. For
some items, it may be pragmatic to black out the quantities question, thereby eliminating a source
of confusion in the interview at the small cost of losing a piece of ambiguous data. For other
items, taking care to keep the sub-grouping reasonably homogenous may help to resolve the issues
both of establishing quantities and of estimating caloric content. “Leafy green vegetables” and
“potatoes, sweet potatoes, and other tubers” are more internally homogenous categories than
“vegetables.”
It is usual, and probably helps respondents to remember the necessary information, to list
the items so that “similar” items are together. In past surveys, this has usually meant that
“botanically” similar items were placed side by side (in other words, all meats together and all
fruits and vegetables together) and that, within each group, those items that are more commonly
consumed were placed first. This probably also reflects people’s shopping patterns in most
countries if vegetables are in one section of the market and meats in another of if they buy their
vegetables from the greengrocer and their meat from the butcher. If a very long list of items is to
be used, it might make the interview more manageable to divide the list into sub-groups in a more
explicit way than is shown in the draft sub-module, with sub-headings for each sub-group. Then
each sub-group could have a filter question such as “did your household purchase any meat since
my last visit?” After that, separate questions on each kind of meat would follow. Some surveys
supplement this general filter question by showing the respondent a card that either lists or
illustrates (depending on the degree of literacy in the country) the various items in the sub-
category as a prompt. However, no past LSMS survey has tried this.
The list of foods used in the draft sub-module as an illustration came from the Pakistan
LSMS questionnaire, with the addition of beer and other alcoholic beverages, which are important
items of consumption in many countries, though not in Pakistan. Thus, the list is specific to
Pakistan and includes items (dal, gur, and ghee) that may be unfamiliar and, thus, inappropriate to
40
list in other countries. As a result, the list also omits some other items that should be included in
other countries (such as pork, cassava, yams, tomatoes, papayas, and bananas). The key point is
that the list must be customized to reflect local patterns of food consumption.
Recall Period. The recall period used in this sub-module is the time since the interviewer’s
last visit to the household. This question presumes that the interviewer’s last visit to the household
was about two weeks before the interview on consumption, as was the case in the old prototypical
fieldwork plan where each primary sampling unit was visited by the survey teams twice, two
weeks apart, and different sections of the questionnaire administered in the two visits. In this plan,
the two visit routine served a number of functions other than serving as the recall period, but, now
that data entry operators often travel with the field teams rather than remaining in a regional
office, the logistical reasons for the two separate visits are being eroded. Moreover, in many
surveys, especially those that have shorter questionnaires than the full LSMS, only one visit is
ever made to the primary sampling unit (PSU) and the entire questionnaire is canvassed in that
visit. In this case, the recall should be changed from “since my last visit” to “in the last two
weeks.” In countries with high inflation rates or where the other expenditure surveys use a shorter
reference period, survey designers may shorten the recall period to a week or to whatever period
matches that of the other survey.
Barter. If barter is common in the country where the survey is being planned, it can be
included in the wording of the purchase question. If it is very important, survey designers may
wish to give it a question of its own. It will be particularly important to pilot test the wording of
the question as this can often be awkward.
Quantity. The quantity question applies to the short recall period question so that, when a
unit value is derived from this data, it refers to a specific time period. To have a unit value (quasi
price) for some indeterminate month during the year is not very helpful where there is even
modest inflation or seasonal variation in prices.
Ordering of Quantity and Expenditure. There is some debate about the proper ordering of
these two questions. Essentially, the field test should be the guide. Expenditure has been placed
first in the draft sub-module for two reasons. First, it is the most important piece of data to be
collected. Second, collecting it first is likely to reduce the risk of ambiguity or misunderstanding
on the part of the respondent. It is possible to imagine that, if questions were in the reverse order,
a respondent might give the interviewer an answer that referred to the household’s expenditure per
unit of quantity rather than for the total purchase. Instead of “I spent 50 pesos on meat. I bought
two kilos,” the informant might answer “I bought two kilos” and “I paid 25 pesos.” The
respondent would mean 25 pesos for each kilo, not in total, and the true response would be
misrecorded. In Pakistan, this kind of problem occurred often enough to call into question the
accuracy of all of the short period data. However, in the Panamanian field test, the questions
seemed to work better with the quantity question placed before the expenditure question.
Fuel. Fuel is placed in the same sub-module as food because it is the main category of
non-food that is purchased in convenient, standard units. This follows the principle that, wherever
quantities can be collected easily and accurately, they should be. Hence it is convenient, though
41
not usual, to put fuel in the food sub-module so that the follow-on questions on quantities will
apply.
Home Production. There are several options about how to arrange the questions on home
production. The option that has most commonly been used in previous LSMS surveys has been to
place them either together with food purchases as shown here or in a separate sub-module on
home production. Occasionally, the consumption of home-produced food has appeared in the
agricultural module as part of the means by which a crop is disposed of. Since agriculture modules
are becoming increasingly infrequent in LSMS surveys and since they may not always include
crop disposal questions anyway, the following discussion concentrates on the first two options.
Whether the questions on households’ consumption of home-produced food should be

placed adjacent to the questions on their purchases of food or in a separate sub-module is an issue
that will be affected by both the percentage of households that produce their own food and the
range of goods produced in this way. If questions on home-produced goods are placed in a
separate sub-module, it is simple to add a filter question, “has your household consumed any food
produced at home?” Then it is possible to avoid asking a long list of inapplicable questions to
those households that do not produce any food at home. This will be most appropriate when home
production is not very common. Ultimately what works best should be determined by the field
test. Do respondents find it easier to think about all of the sources of a single foodstuff together?
(For example, I bought some tomatoes, I grew some tomatoes in my garden, or my co-worker
gave me some tomatoes?) Or is it easier for them to think about their food budget in terms of the
source of that food? (For example, when I shop I buy beans, tortillas and milk, but in my garden, I
grow tomatoes, papaya, and bananas.) The decision about whether or not to put these questions in
a separate sub-module will not materially affect the length of the interview, although it may take
up more paper and make the physical questionnaire longer to put home production separately from
purchased food.
It is not necessary to have a home production question for every food item in the list, since
some are industrially manufactured items that cannot be made in people’s homes. Even some
items that are not industrial are not commonly produced in people’s homes and can likewise be
omitted from the list of home-produced items. In the draft of the sub-module in Part III, the boxes
are blacked out in the home-production columns for items that are unlikely to be produced at
home. If a separate page is used for home-produced foods, then the list on that page should apply
only to those likely to be produced at home.
It is recommended that information about the quantity of food produced at home should be
regarded as the most important piece of data to be gathered about home production of food. The
quantity of a good is a factual, observable piece of information, although there are anecdotes
about how difficult it has sometimes been for respondents to answer quantity questions about
items that they harvest only one meal’s worth at a time such as cassava. Because they never see
the total cassava harvest at one time, they may have difficulty estimating its size. Questions about
value are more hypothetical, since, by definition, the home-produced item did not pass through the
market. Theoretically, a farmgate price would be too low and a market price too high. There
seems to be no evidence as to which price the consumer of home-produced is likely to know more
often or which answer he or she would give in response to a general question on the “value of
42
home production.” The draft sub-module includes a general (ill-defined) question on the “value of
home production” on the grounds that, for goods for which there are markets, the food purchases
questions and/or community questionnaire gathers information on prices or unit values from food
purchases and may gather data on farmgate prices in the agricultural module if there is one. Thus,
the “value of home production” contributes a different set of information or at least on the aspect
that the respondent feels is most pertinent.
Gifts. In some past LSMS surveys, the gifts section included the value of food received as
payment in kind from employers. This is not included in this draft sub-module, because it already
appears in the various employment sub-modules.
Use of Environmentally Provided Goods. If survey designers wish to try to put an explicit
value on the use of all environmentally provided resources, the list of food items should include
any specific food items that are likely to be gathered, fished, or hunted. In additional, the wording
of the questions may be changed or additional questions added to clarify that these items are being
encompassed.
Part C: Non-Food Consumption
Again, the list of items to be included in this sub-module will be country-specific and can
usually be derived from the survey used to weight the CPI. The list is likely to contain more
groups of items (for example, “clothing”) than was the case for the food list (for example, “rice”).
The non-food list should be designed with the following objectives in mind: (i) to cover all
aspects of the household’s budget that are not covered elsewhere in the household questionnaire;
(ii) to do so in a logical manner that respondents find congenial; (iii) to gather information on
specific items that may receive heavy subsidies or attract heavy taxes (for example, kerosene or
gasoline or tobacco or alcohol); and, sometimes, (iv) to gather the data in a manner that will
enable analysts to study intrahousehold issues.
For most items, it will not be practical to collect data on quantity (and, implicitly, unit
values) due to the difficulties of establishing meaningful units, especially for groups of goods. For
items where meaningful units can be established (usually fuel), data on the quantity consumed
should be gathered.
Note that the boundary between consumption items and durable goods is somewhat fuzzy.
Many of the items usually listed in the consumption module (and included in this draft sub-
module) may actually last for more than a year. Kitchen equipment (including cups, forks, plates,
and saucepans), furniture (including beds, tables, cupboards, chairs, and rugs), and linens
(including sheets, towels, and blankets) all appeared in the consumption module of the Cote
d’Ivoire questionnaire rather than in the durable goods module, even though all of these items
were likely to last for more than a year. Nonetheless, there are so many different items that could
be listed, that it seems pragmatic to measure their consumption via the flow into the household
43
rather than trying to enumerate the whole stock item by item and then to compute a use value for
it.
The list of items in the draft sub-module is derived from several different questionnaires
from previous LSMS surveys, principally the surveys in Jamaica and Nepal.
The same considerations as apply for food apply to the barter and gift of non-food goods
and to non-food goods (such as firewood) gathered from the environment. Home production is
theoretically of interest here as well but has rarely been specifically enumerated in past LSMS
surveys. If survey designers decide that this is important, it can be handled in a manner similar to
that for home-produced food.
Part D: Expenditures on Private Inter-household Transfers
Respondents may find it most logical to be asked questions about their income from and
expenditures on inter-household transfers in the same place in the questionnaire. In this book,
however, questions on household expenditures on private inter-household transfers are included
here and those on income from inter-household transfers are included in the miscellaneous income
in Chapter 20 where the annotations on the module are included. The two sub-modules were
developed together and can be positioned adjacent to one another in the household questionnaire
in a separate module.
Two versions of the sub-module are included. The short one is designed to get basic
information when the consumption module is being shortened as much as possible and the study
of private safety nets is not deemed important. The longer sub-module includes questions about
the recipients and their relationship to the donor or head of the household and on the amounts,
regularity, and purpose of the transfers. The answers to these questions supply analysts with a lot
of information for studying transfers.
On question 12, it would be useful to get as much detail on the destination as can easily be
coded, for example, to the district or county level. For both questions 11 and 12, codes should
coincide with those used on the migration module. Questions 6 and 7 use the same codes as on the
roster, which is why code 1 is missing (on the roster module it is for the head of household).
Part E: Durable Goods
The items to be included in the durable goods list should be those that last for substantially
longer than a year and that are so large in relation to the household’s standard of living that they
can be separately enumerated and that respondents can accurately remember information about
their purchase after several years have gone by. Thus, a car would meet this definition of a durable
good, whereas a shirt would not, even though each may last for several years.
The durable goods page in this sub-module is divided into two blocks so that households
can report on the value of two of the same kind of item (for example, two bicycles).
Most past LSMS questionnaires have included a list of a dozen or two different kinds of
44
durable goods, though the lists have usually been longer in the surveys fielded in the countries of
the former Soviet Union. The most appropriate items to include in the durable goods list will vary
from country to country. What is considered expensive enough to be a durable good will also vary
from country to country. For example, cooking pots were on the durable list in the survey in
Kagera, Tanzania, whereas the list in the Jamaican survey included satellite dishes. The most
common durable goods may also vary by climate. In tropical countries, air conditioners or fans
will often be on the list, whereas, in cold climates, various sorts of heaters will be listed. Also,
culture matters. Jewelry, carpets, and guns were listed in Pakistan but are unlikely to be listed in
the durable goods sub-module in many other countries.
Special Circumstances
This subsection provides survey designers with advice on how to deal with various special
circumstances that they may face in planning and designing a survey in their particular country.
Dealing with Inflation. In those countries and during those times when inflation is high, it
may be necessary to modify the questionnaire. (However, if inflation is very high, it is not clear
how well these suggested modifications will work.)
First, the recall period should be shortened to the shortest period that is reasonable for the
type of item in question. Thus, the recall period on food might change from the previous two
weeks to the previous week. The “usual month” questions on food might be dropped altogether,
since it would be unclear to which time period they referred, which would make it impossible to
deflate the expenditures appropriately. The recall period on non-food items might be shortened
from a year to three or six months. Shortening the recall periods may increase the variance of the
estimates, but in any case, it would be impossible to interpret any means for data collected when
prices were very different. Moreover, inflation itself will cause people to make more frequent
purchases (so that the real value of their money does not diminish), so the tradeoff between biases
in the mean and variance may be less than in those places where inflation is low.
The second modification that can be made to take inflation into account involves asking
respondents about the dates when they made certain large purchases so that analysts can deflate
these figures appropriately. Third, if a currency other than the national currency has become a de
facto unit of account, then survey designers might allow respondents to give their answers in
either the local or the international currency.
Credit. If the use of consumer credit is of particular analytical and policy interest in the
country in question, then a short module on households’ use of credit (particularly for purchasing
food) could be inserted after the food sub-module, and questions on purchases on credit added to
the durable goods sub-module as is suggested in Chapter 22 on credit.
Supporting Intrahousehold Analysis. If supporting intrahousehold analysis is a special goal

of the survey, the consumption module can be re-organized so that at least some of the
information is individual-specific. The main principle is to classify all items according to two
criteria. The first criterion is whether they are consumed individually (like a taxi fare or a shirt) or
jointly (like washing powder or the use of a television). The second criterion is whether it is easy
45
to distinguish whether the item was consumed by one household member rather than another.
Food eaten at home is individually consumed, but, since it is usually purchased and prepared for
the household as a whole, it is difficult to distinguish how much is consumed by each individual.
On the other hand, it is easier to assign the consumption of food consumed in restaurants or from
food stalls to particular individuals. After the items that are individually consumed and easily
distinguished as being consumed by a given individual have been determined, these items can
then be listed on a separate grid that the interviewer would fill out for each individual.
Keeping the Module as Short as Possible
It is important to keep the module to a reasonable length to avoid the risk of tiring or
annoying the respondents with an overly long interview. However, there are relatively few options
for shortening the consumption module, aside from using the shorter of the two versions of the
sub-module on inter-household transfers. As explained in Part II, reducing the number of items or
categories for which information is collected can easily lead to an underestimation of
consumption. Also, getting a comprehensive measure of consumption requires inquiring about
purchases, home production, and gifts of all commodities consumed by the household, so it is not
easy to reduce the number of questions about each item.
There are a few questions that could be cut in the draft sub-modules, though this would
result in a loss of important data in each case (which is why the questions are included in the draft
sub-modules in the first place). Probably the most expendable question is the last of the series for
durable goods -- question 7. Removing this question would mean that the valuation of durable
goods would rest solely on the assumption that the average life of the good is twice the average
age reported across all households.
Another possibility would be to reduce the number of questions on inter-household

transfers. The questions that essential for calculating outflows are 1, 2, and 13-15. The answers to
the other questions contribute to the understanding of why and to whom the transfers are made,
but some of them could be dropped if studying such issues is of relatively low priority.
It would be possible to drop question 9 in the food sub-module, and instead to value home
production using data from the expenditure questions and/or the community questionnaire. There
is an element of risk in this, since households would no longer have the option to express their
answers in value terms instead or as well as quantity terms, which they may have preferred found
easier to do. Of course, this option is only available if there are questions on quantities in the food
expenditure sub-module of consumption or if data on prices are gathered in the community
questionnaire for all items on the home-produced list. Alternatively, survey designers could drop
the quantity questions about purchases and home production (note that the quantity and the value
questions on home production cannot both be dropped simultaneously). However, quantity is an
important piece of data in its own right, so dropping these questions reduces analytical potential,
most importantly it means that prices from some other sources are required to derive a calorie-
based poverty line from the data set.
A final option would be to drop the dual recall periods in the food or the non-food sub-
module. Doing so would be risky as, if something were to go wrong in the fieldwork, analysts
46
would be unable to use the data set to calculate what is arguably the single most important
variable in the whole enterprise -- total household consumption. Because food is more important
in the total than non-food, it is preferable to drop the dual recall periods for non-food items before
dropping those for food items.
47
References
Ahmad, S. Ehtisham, and Nicholas H. Stern. 1991. The Theory and Practice of Tax Reform in
Developing Countries Cambridge: Cambridge University Press
Atkinson, Anthony B. and Mickelwright. 1983. “On the Reliability of Income Data in the Family
Expenditure Survey 1970–77,” Journal of the Royal Statistical Society, Series A 146
Baker, Judy, 1996. Personal correspondence
Behrman, Jere R., and Anil B. Deolalikar. 1987. “Will Developing Country Nutrition Improve
with Income? A Case Study for Rural South India” Journal of Political Economy 95
Bhalla, Surjit S. 1979. “Measurement Errors and the Permanent Income Hypothesis: Evidence
from Rural India,” American Economic Review, 69
Bhalla, Surjit S. 1980. “The Measurement of Permanent Income and its Application to Saving
Behavior” Journal of Political Economy 88
Bhattacharyya, Nikhilesh. 1963. “On the Effects of Itemization in the Family Budget Schedule”
Processed. Indian Statistical Institute: Calcutta
Biemer, Paul P., Robert M. Groves, Lars E. Lyberg, Nancy A. Mathiowetz, and Seymour
Sudmanr (eds). 1991. Measurement Errors in Surveys Wiley: New York
Blaizeau, Didier, 1998, “Household Expenditure Surveys in the Seven UEMOA Countries,”
Processed. INSEE: Paris
Bouis, Howarth E. 1994. “The Effect of Income on Demand for Food in Poor Countries: Are Our
Databases Giving us Reliable Estimates?” Journal of Development Economics, 44
Bouis, Howarth E. and Lawrence J. Haddad. 1992. “Are Estimates of Calorie-income Elasticities
too High? A Recalibration of the Plausible Range” Journal of Development Economics,
39
Bourguignon, François, and Pierre-André Chiappori. 1992. “Collective Models of Household

Behavior: an Introduction” European Economic Review, 36
Bourguignon, François, Martin J. Browning, Pierre-André Chiappori, and Valérie Lechene. 1993.
“Intrahousehold Allocation of Consumption: a Model and Some Evidence from French
Data” Annales d’Économie et de Statistique, 29
Bradburn, Norman M., Janellen Huttenlocher, and Larry Hedges. 1994. “Telescoping and Temp-
oral memory” in Norbert Schwartz and Seymour Sudman (eds) Autobiographical
Memory and the Validity of Retrospective Reports New York: Springer-Verlag
48
Branch, E. Raphael. 1994. “The Consumer Expenditure Survey: a Comparative Analysis”
Monthly Labor Review, 47–55
Browning, Martin, François Bourguignon, Pierre-André Chiappori, and Valérie Lechene. 1994.
“Income and Outcomes: a Structural Model of Intrahousehold Allocation” Journal of
Political Economy, 102
Cavendish, William. 1998. “The Complexity of the Commons: Environmental Resource

Demands in Rural Zimbabwe” Mimeo. Centre for the Study of African Economies,
University of Oxford: UK
Coder, John. 1991. “Exploring Non-sampling Errors in the Wage and Salary Income Data from
the March Current Population Survey” Processed. Household and Household Economic
Statistics Division, Bureau of the Census: Washington DC
Cramer, Jan S. 1969. Empirical Econometrics Amsterdam: North-Holland
Dandekar, V. M., and N. Rath. 1971a. “Poverty in India: Dimensions and Trends” Economic and
Political Weekly 6
Dandekar, V. M., and N. Rath. 1971b. Poverty in India Pune: Indian School of Political
Economy
Deaton, Angus, 1997. The Analysis of Household Surveys: a Microeconometric Approach to

Development Policy Baltimore, MD: Johns Hopkins University Press for the World Bank
Deaton, Angus S. 1988. “Rice Prices and Income Distribution in Thailand: a Non-parametric
Analysis” Economic Journal 99 (Supplement)
Deaton, Angus S. and John Muellbauer. 1980. Economics and Consumer Behavior, Cambridge:
Cambridge University Press
Deaton, Angus S. and Christina H. Paxson, 1998. “Economies of Scale, Household Size, and the
Demand for Food” Journal of Political Economy 106
Deaton, Angus S. and Salman Zaidi. 1999. “Guidelines for Constructing Consumption
Aggregates for Welfare Analysis” Processed. Princeton University and Development
Research Group, World Bank: Washington DC
49
Eisenhower, Donna, Nancy A. Mathiowetz, and David Morganstein. 1991. “Recall Error:
Sources and Bias Reduction Techniques,” in Biemer et al Measurement Errors in Surveys
Wiley: New York
Foley, M., 1996. Personal correspondence
Ghose, Suchismita and Nikhilesh Bhattacharyya. 1993. “Engel elasticities of clothing and other
items,” Sarvekshana, 17
Ghose, Suchismita and Nikhilesh Bhattacharyya. 1995. “Effect of Reference Period on Engel
Elasticities of Clothing and Other Items: Further Results” Sankhya: The Indian Journal of
Statistics 57, Series B
Ghosh, A. 1953. “A Note on the Accuracy of Family Budget Data with Reference to the Period
of Recall.” Calcutta Statistical Association Bulletin
Gibson, John. 1998. “How Robust are Poverty Comparisons to Changes in Household Survey
Methods? A Test Using Papua New Guinean Data” Department of Economics, University
of Waikato: New Zealand
Gieseman, Raymond. 1987. “The Consumer Expenditure Survey: Quality Control by

Comparative Analysis” Monthly Labor Review, 8–13
Glewwe, P., 1987. “The Distribution of Welfare in Peru in 1985-86” Living Standards
Measurement Study Working Paper No. 42. World Bank: Washington DC
Glewwe, P., and K.A. Twum-Baah, 1991. “The Distribution of Welfare in Ghana, 1987-88”
Living Standards Measurement StudyWorking Paper No. 75. World Bank: Washington
DC
Griliches, Zvi and Jerry A Hausman. 1986. “Errors in Variables in Panel Data” Journal of
Econometrics, 31
Grootaert, Christiaan. 1986. “The Use of Multiple Diaries in a Household Expenditure Survey in
Hong Kong” Journal of the American Statistical Association, 81
Grootaert, C., 1993. “The Evolution of Welfare and Poverty Under Structural Change and
Economic Recession in Cote d’Ivoire, 1985-88” Policy Research Working Paper No.
1078. World Bank: Washington DC
Grosh, Margaret. 1997. “The Policymaking Uses of Multi-topic Household Survey Data: A
Primer” World Bank Research Observer, 12
50
Grosh, Margaret and Juan Muñoz, 1996. “A Manual for Planning and Implementing the Living
Standards Measurement Study Surveys” Living Standard Measurement Study Working
Paper 126, World Bank: Washington DC
India, Government of. 1993. “ Report of the Expert Group on Estimation of Proportion and
Number of Poor” Perspective Planning Division, Planning Commission: New Delhi
International Monetary Fund. 1995. International Financial Statistics Yearbook. IMF:

Washington DC
International Monetary Fund. 1996. International Financial Statistics Yearbook. IMF:

Washington DC
Jacobs, Eva and Stephanie Shipp. 1993. “A History of the US Consumer Expenditure Survey:
1935–36 to 1988–89” Journal of Economic and Social Measurement, 19
Jolliffe, Dean and Kinnon Scott. 1995. “The Sensitivity of Measures of Household Consumption
to Survey Design: Results from an Experiment in El Salvador” Processed. Policy Re-
search Department, World Bank: Washington DC
Lanjouw, J. and P. Lanjouw, 1996. Draft. “Aggregation Consistent Poverty Comparison:

Theory and Illustrations” Yale University and World Bank
Laraki, Karim. 1989. “Ending Food Subsidies: Nutritional, Welfare, and Budgetary Effects”
World Bank Economic Review, 3
Mahalanobis, P.C. 1946. “Recent Experiments in Statistical Sampling in the Indian Statistical
Institute” Journal of the Royal Statistical Society, 109
Mahalanobis, P. C. and S. B. Sen. 1954. “On Some Aspects of the Indian National Sample
Survey” Bulletin of the International Statistical Institute, 34, 5
Martini, Alberto and Anna Ivanova. 1996. “The Design and Implementation of the Income and
Expenditure Survey of Belarus” Processed. Urban Institute: Washington DC
McWhinney, Isabel and Harold E. Champion. 1974. “The Canadian Experience with Recall and
Diary Methods in Consumer Expenditure Surveys” Annals of Economic and Social
Measurement, 3/2
Minhas, B. S. 1988. “Validation of Large-scale Sample Survey Data: Case of NSS Household
Consumption Expenditure” Sankhya Series B, 50, (Supplement)
Minhas, B. S. and S. M. Kansal. 1989. “Comparison of NSS and CSO Estimates of Private
Consumption: Some Observations based on 1983 Data” The Journal of Income and
Wealth 11
51
Musgrove, Philip. 1978. Consumer Behavior in Latin America Brookings Institution:
Washington DC
Musgrove, Philip. 1979. “Permanent Household Income and Consumption in Urban South
America” American Economic Review, 69
Neter, John and Joseph Waksburg. 1964. “A Study of Response Errors in Expenditure Data from
Household Interviews” Journal of the American Statistical Association
Neter, John. 1970. “Measurement Errors in Reports of Consumer Expenditures” Journal of

Marketing Research, 7
Newbery, David M. G. and Nicholas H. Stern (eds). 1987. The Theory of Taxation for
Developing Countries Oxford University Press: New York
Paxson, Christina H., 1992, “Using Weather Variability to Estimate the Response of Savings to
Transitory Income in Thailand” American Economic Review, 82
Paxson, Christina H. 1993. “Consumption and Income Seasonality in Thailand” Journal of

Political Economy, 101
Reagan, Barbara B. 1954. “Condensed versus Detailed Schedule for Collection of Family
Expenditure Data” Agricultural Research Service, US Department of Agriculture
Rubin, David C. and Alan D. Baddeley. 1989. “Telescoping is not Time Compression: a Model
of the Dating of Autobiographical Events” Memory and Cognition, 17
Sahn, David. 1989. Seasonal Variation in Third World Agriculture: the Consequences for Food
Security Baltimore: Johns Hopkins University Press for the International Food Policy
Research Institute
Scott, Christopher and Ben Amenuvegbe. 1990. “Effect of Recall Duration on Reporting of
Household Expenditures: an Experimental Study in Ghana" Social Dimensions of
Adjustment in Sub-Saharan Africa Working Paper 6. World Bank: Washington DC
Scott, Christopher, Mary Strode, and Oliver Chinganya. 1998. “Estimation of Annual Expenditure
from Monthly Survey Data” Processed
Scott, Kinnon. 1994. “Venezuela: Poverty Measurement with Multiple Data Sets.” Draft. World
Bank: Washington DC
Scott, Kinnon. 1997. Personal communication
Scott, Kinnon and Wlodzimierz Okrasa, 1998. "Analysis of Latvia Diary Experiment" Internal
Memorandum. Development Research Group, World Bank: Washington, DC
52
Sen, Amartya K., 1999, Development as Freedom, New York. Knopf.
Silberstein, Adriana R. 1990. “First-wave Effects in the US Consumer Expenditure Interview

Survey” Survey Methodology, 16
Silberstein, Adriana R., and Stuart Scott, 1991. “Expenditure Diary Surveys and their Associated
Errors” in Biemer et al (eds) Measurement Errors in Surveys Wiley: New York
Silberstein, Adriana R. and Stuart Scott. 1992. “Seasonal Effects in the Reporting of Consumer
Expenditure” Proceedings of the American Statistical Association Survey Research
Section
Srinivasan, T. N. 1994. “Database for Development Analysis” Journal of Development

Economics, 44
Statistical Institute and Planning Institute of Jamaica. 1994. Jamaica Survey of Living Conditions
1992, Kingston
1993. Kingston
1994, Kingston
Steele, Diane, 1998. "Ecuador Consumption Items" Internal memorandum. Development

Research Group, World Bank: Washington, DC
Strauss, John and Duncan Thomas. 1995. “Human Resources: Empirical Modeling of Household
and Family Decisions” in Jere Behrman and T. N. Srinivasan (eds) Handbook of Development
Economics, 3a. Elsevier: Amsterdam
Subramanian, Shankar and Angus S. Deaton. 1996. “The Demand for Food and Calories”
Journal of Political Economy, 104
Sudman, Seymour, and Robert Ferber. 1971. “Experiments in Obtaining Consumer Expenditures
by Diary Methods” Journal of the American Statistical Association, 66
Sudman, Seymour, Norman N. Bradburn, and Norbert Schwarz. 1996. Thinking About Answers,
Jossey-Bass: San Francisco
Swindale, Anne, and Punam Ohri-Vachaspati, 1997. "Household Food Consumption Indicator
Guide" Draft. IMPACT Food Security and Nutrition Monitoring Project, International
Science and Technology Institute: Washington DC
Tsoflias, Lynn. 1996. Personal communication
53
Tucker, Clyde. 1992. “The Estimation of Instrument Effects on Data Quality in the Consumer
Expenditure Survey” Journal of Official Statistics, 8
Tucker, Clyde, and Claudette Bennett. 1988. “Procedural Effects in the Collection of Consumer
Expenditure Information” Proceedings of the Section on Survey Research Methods,
American Statistical Association
van der Gaag, Jacques. 1994. Personal communication
Webb, R., and G. Fernandez Baca. 1993. Peru en Numeros 1993. Anuario Estadistico. Cuanto.
Lima, Peru
Wolpin, Kenneth I. 1982. “A New Test of the Permanent Income Hypothesis: the Impact of
Weather on the Income and Consumption of Farm Households in India” International
Economic Review, 23
World Bank. 1993 “Indonesia: Public Expenditures, Prices, and the Poor” Processed Report #
11293–IND, Indonesia Resident Mission Jakarta
World Bank, 1994a. “Guyana: Strategies for Reducing Poverty” Report 12861-GUA, Country
Operations Division 2, Country Department III, Latin America and the Caribbean
Region. Washington DC
World Bank, 1994b. “Kingdom of Morocco: Poverty, Adjustment & Growth. Volume II”
Report 11918-MOR. Middle East and North Africa Region. Country Department I.
Country Operations Division. Washington DC
World Bank, 1995a. “The Kyrgyz Republic. Poverty Assessment and Strategy” Report 14380-
KG. Human Resources Division, Europe and Central Asia Country Department III
Washington DC
World Bank, 1995b. “Republic of Nicaragua. Poverty Assessment. Volume II” Report 14038-
NI. Country Operations Division. Country Department II. Latin America and the
Caribbean Regional Office. Washington DC
World Bank, 1995c. “Tanzania Social Sector Review” Report 14039-TA. Poverty and Human
Resources Division. Country Department II, Africa Region, World Bank Washington
DC
World Bank, 1996. "Jamaica Survey of Living Conditions (JSLS) 1988-94 Basic Information."
Poverty and Human Resources Division, Policy Research Department, World Bank:
Washington DC
World Bank, 1997. “Romania, Poverty and Social Policy” Report No. 16467, Human Resources
54
Sector Operations Division, Country Department II, Europe and Central Asia region,
World Bank: Washington DC
Zambia, Central Statistical Office. 1995. Household Budget Survey 1993/94 Lusaka
55
PART A: DAILY EXPENSES MOST KNOWLEDGEABLE MEMBER
1. 2. 3. 4.
In the past 7 days, has any member of your household How much did How many [MEALS/SNACKS] were What was the value
spent money on any of the following items? your eaten by household members of these [MEALS]
household outside of the home during the past eaten outside of the
U
PUT A CHECK ( ) IN THE APPROPRIATE BOX FOR spend for 7 days? home in the last 7
EACH ITEM. ASK QUESTION 1 FOR ALL ITEMS [ITEM]? days?
BEFORE GOING TO 2.
NO YES CURRENCY NUMBER CURRENCY
1
Tobacco, cigarettes, cigars Breakfast
2
Newspapers or magazines Lunch
3
Lottery tickets Dinner/supper
4 Snack or beverages
Fares for busses, trams, taxis, etc. (including alcohol)
5
Parking
6
Regular worship
7
Alms
8
Shoeshines
Consqu05.xls11/13/99
PART B: FOOD AND FUEL MOST KNOWLEDGEABLE MEMBER
PURCHASES SINCE LAST VISIT PURCHASES TYPICAL MONTH HOME PRODUCTION GIFTS UNIT CODES:
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. USE CODES
WITH STAR
In the following questions, I want to ask Have the How much How much did How many How much do you How many How much did you What was the What is the WHENEVER
about all purchases made for your members of did you pay you buy? months in the usually spend on months in the consume in a typical value of the total value of POSSIBLE
household, regardless of which person made your in total? past 12 months [FOOD] in one of past 12 months month? [FOOD] you the [FOOD]
them. household did your the months that did your consumed in a consumed KILO*....1
bought any household you purchase household typical month that you GRAM*....2
Has your household consumed [FOOD] [FOOD] since purchase [FOOD]? consume from your own received as a POUND*...3
during the past 12 months? Please exclude my last visit, [FOOD]? [FOOD] that you production? gift over the OUNCE*...4
from your answer any [ITEM] purchased for that is since grew or past 12 LITER*...5
processing or resale in a household [DAY/DATE] ? produced at months? CUP*.....6
enterprise. home? PINT*....7
QUART*...8
U
PUT A CHECK ( ) IN THE APPROPRIATE IF NONE IF NONE IF NONE, GALLON*..9
BOX FOR EACH FOOD ITEM. IF THE WRITE WRITE WRITE BUNCH...10
ANSWER TO Q.1 IS YES, ASK Q.2-13. ZERO, » 7 ZERO, » 10 ZERO PECK....11
BUSHEL..12
YES.1 TIN.....13
NO..2 PIECES..14
NO YES CODE (»5) CURRENCY AMOUNT UNIT MONTHS CURRENCY MONTHS AMOUNT UNIT CURRENCY CURRENCY DOZENS..15
BOTTLES.16
Wheat (grain) 1
Wheat (flour or maida) 2
Maize (flour or grain) 3
Jawar/Bajra 4
FIne rice (basmati) 5
Coarse rice 6
Other grains/cereals 7
Gram 8
Dal 9
Groundnuts 10
Liquid vegetable oils
(dalda) 11
Ghee, Desi ghee 12
Fresh milk 13
Yogurt and Lassi 14
Milk Powder 15
Baby Formula 16
Sugar (refined) 17
Gur/Desi sugar 18
Mutton/lamb/goat 19
Beef/Buffalo 20
Consqu05.xls 11/13/99
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. USE CODES
WITH STAR
QUART*...8
U
BUSHEL..12
YES.1 TIN.....13
NO..2 PIECES..14
BOTTLES.16
Chicken and other
poultry 21
Eggs 22
Fish 23
Vegetables 24
Melon 25
Bananas 26
Citrus Fruits 27
Mango 28
Other Fruits 29
Canned Foods 30
Bottled Beverages
(Coca cola, roohafza, 31
Beer 32
Other alcoholic
beverages 33
Chapatti, Nan, other
breads 34
Fried items such as
somosas, pakoras 35
Biscuits and cakes 36
Spices and condiments 37
Tea 38
Coffee 39
Misc. other food
expenses 40
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. USE CODES
WITH STAR
QUART*...8
U
BUSHEL..12
YES.1 TIN.....13
NO..2 PIECES..14
BOTTLES.16
Firewood 41
Charcoal 42
Paraffin 43
Cooking gas 44
PART C: NON-FOOD MOST KNOWLEDGEABLE MEMBER
PURCHASES 30 DAYS 12 MONTHS GIFTS 12 MONTHS

1. 2. 3. 4. 5. 6.
In the following questions, I want to ask about all purchases Have the How much How much did Did you receive What is the value
made for your household, regardless of which person made members of did you your household any [ITEM] as a of all the [ITEM]
them. your spend? spend for gift during the past that you received
household [ITEM] during 12 months? as a gift during the
Has your household bought, spent money on or received bought any the past 12 past 12 months?
gifts of any [ITEM] during the past 12 months? Please [ITEM] in the months?
exclude from your answer any [ITEM] purchased for last 30 days?
processing or resale in a household enterprise.
U
PUT A CHECK ( ) IN THE APPROPRIATE BOX FOR
EACH FOOD ITEM. IF THE ANSWER TO Q.1 IS YES,
ASK Q.2-13
YES.1 YES..1
NO..2 NO...2
NO YES CODE (»4) CURRENCY MONTHS (»NEXT ITEM) MONTHS
Personal care items (soap, shampoo,

toothpaste, etc.) 1
Cosmetics 2
Women’s clothing 3
Men’s clothing 4
Children’s clothing 5
Women’s footwear 6
Men’s footwear 7
Children’s footwear 8
Cloth and sewing supplies 9
Tailoring expenses 10
Laundry and dry cleaning 11

Personal services (haircuts, shaving,
manicures, etc.) 12
Traditional remedies and over the
counter remedies 13

1. 2. 3. 4. 5. 6.
U
ASK Q.2-13
YES.1 YES..1
NO..2 NO...2
Modern medicines and health services

(doctor fees, hospital charges, etc.) 14
Books, stationery (excluding textbooks) 15
Postal expenses, telegrams, 16

Entertainment (cinema, cassette
rentals, cultural and sporting events, 17
Household cleaning articles (soap,
washing powder, bleach, etc.) 18
Kitchen supplies (napkins, matches,
garbage bags, etc.) 19
Toilet supplies (toilet paper, cleanser,
etc.) 20
Electrical items (light bulbs, cords,
plugs, batteries, etc.) 21
Repair and maintenance of household
articles 22
Household linens (sheets. blankets,
towels, etc.) 23
Small kitchen appliances (blender,
mixer, etc.) 24
Dishes (crockery, cutlery, glassware,
etc.) 25
Kitchen utensils (pots, pans, buckets,
tools, etc.) 26

1. 2. 3. 4. 5. 6.
U
ASK Q.2-13
YES.1 YES..1
NO..2 NO...2
Small electrical items (radio, walkman,

watch, clock, etc.) 27
Household tools 28
Sports and hobby equipment 29
Toys 30
Musical instruments 31
Vehicle repair, maintenance, parts and
licenses (do not include gasoline) 32
Repair and maintenance of the house 33
Home improvements and additions 34
Insurance (auto, property) 35
Health insurance 36
Membership fees 37
Excursion, holiday (including travel and
lodging) 38
Charity, donations 39

1. 2. 3. 4. 5. 6.
U
ASK Q.2-13
YES.1 YES..1
NO..2 NO...2
Income tax 40
Land tax 41
Housing and property taxes 42
Gambling losses 43
Cash losses 44
Contributions to ROSCAs, tontins,
Christmas clubs, etc. 45
Deposits to savings accounts 46
Legal or notary services 47
Marriages, births, and other ceremonies 48
Dowry or bride price 49
Funeral expenses 50
PART D: EXPENDITURES ON PRIVATE INTER-HOUSEHOLD REMITTANCES MOST KNOWLEDGEABLE PERSON
1. During the past 12 months has any member of your household provided money or goods to persons YES..1
who are not members of your household? For example for relatives living elsewhere, child support or NO...2(»NEXT SECTION)
alimony, or to friends or neighbors?
2. 3. 4. 5. 6. 7.
What are the names of the persons Is this Which Is [NAME OF How is [NAME OF RECIPIENT] related How is [NAME OF RECIPIENT] related to
to whom household members have assistance sent household RECIPIENT] to [NAME OF DONOR]? the head of the household?
sent assistance during the past 12 to [NAME OF member sends related to
RECIPIENT IS… OF DONOR RECIPIENT IS… OF DONOR
months? RECIPIENT] [NAME OF [NAME OF SPOUSE...............2(»10) SPOUSE.................2(»10)
given by a RECIPIENT]? DONOR]? CHILD................3(»8) CHILD..................3(»8)
specific member GRANDCHILD...........4(»10) GRANDCHILD.............4(»10)
of this NIECE OR NEPHEW......5(»10) NIECE OR NEPHEW........5(»10)
household? PARENT...............6(»9) PARENT.................6(»9)
LIST ALL NAMES
SISTER/BROTHER.......7(»10) SISTER/BROTHER.........7(»10)
BEFORE GOING
COPY ID SON/DAUGHTER-IN-LAW..8(»10) SON/DAUGHTER-IN-LAW....8(»10)
TO 3-20
CODE OF BROTHER/SISTER-IN BROTHER/SISTER-IN LAW..9(»10)
PERSON LAW..................9(»10) GRAND
FROM GRAND FATHER/MOTHER.........10(»10)
ROSTER FATHER/MOTHER.......10(»10) FATHER/MOTHER-IN-
FATHER/MOTHER-IN- LAW...................11(»10)
YES..1 YES..1 LAW.................11(»10) OTHER RELATIVE........12(»10)
NO...2 ID NO...2 OTHER RELATIVE......12(»10) NO RELATION...........13(»10)
NAME (»7) CODE (»9)
10
11
12
8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Is [RECIPIENT] Does What kind of place In what region is How much Has the What is the Is the
male or [RECIPIENT] does [RECIPIENT] this place? money have household approximate assistance
female? live in this live in? members of also given value in cash sent to
place? the household any of the [RECIPIENT]
sent to assistance to assistance provided for a
COPY ID COPY ID [RECIPIENT] [RECIPIENT] given in food specific
CODE OF CODE OF in the past 12 in the form of or other reason?
RECIPIENT RECIPIENT CODE LIST months? food or other goods?
FROM FROM goods?
OF AREAS
ROSTER C ROSTER B
USED ON
MIGRATION
PAGE
CAPITAL
CITY......1
LARGE
MALE.....1 YES..1 CITY......2 YES..1 YES..1
ID ID FEMALE...2 NO...2 TOWN.......3 NO...2 NO...2
CODE CODE (»14) VILLAGE....4 AMOUNT (»17) AMOUNT (»19)
10
11
12
18. 19. 20.

What is the main reason How many times Do you expect
why [RECIPIENT] was per year does [RECIPIENT] to
sent this assistance? [RECIPIENT] repay the
receive this assistance sent
CHILD assistance? by this
SUPPORT......1 household?
EDUCATIONAL
EXPENSES.....2
MEDICAL
EXPENSES.....3
WEDDING.......4
FUNERAL.......5
INVESTMENT IN
HOUSEHOLD
ENTERPRISE....6
PURCHASE OF A
YES..1
DURABLE GOOD.7
NO...2
OTHER ........8
(»NEXT)
10
11
12
PART D: EXPENDITURES ON PRIVATE INTER-HOUSEHOLD REMITTANCES - SHORT MOST KNOWLEDGEABLE PERSON
1. During the past 12 months has any member of your household provided money or goods to persons YES..1
who are not members of your household? For example for relatives living elsewhere, child support or NO...2(»NEXT SECTION)
alimony, or to friends or neighbors?
2. 3. 4. 5. 6. 7.
What are the names of the persons Is this Which Is [NAME OF How much What is the
to whom household members have assistance sent household RECIPIENT] money have approximate
sent assistance during the past 12 to [NAME OF member sends related to members of value in cash
months? RECIPIENT] [NAME OF [NAME OF the household of the
given by a RECIPIENT]? DONOR]? sent to assistance
specific member [RECIPIENT] given in food
of this in the past 12 or other
household? months? goods?
LIST ALL NAMES
BEFORE GOING
COPY ID
TO 3-7
CODE OF
PERSON
FROM
ROSTER
YES..1 YES..1
NO...2 ID NO...2
NAME (»6) CODE AMOUNT AMOUNT
10
11
12
Page 12
PART E: DURABLE GOODS MOST KNOWLEDGEABLE PERSON
2. 3. 4. 5. 6. 7.
LIST ALL THE ITEMS OWNED BY How many Did you How much did How much If you wanted
1. Does your household own any of the
THE HOUSEHOLD, THEN years ago purchase it or you pay for it? was it worth to sell this
following items?
PROCEED TO ASK Q.3-7. did you receive it as a when you [ITEM] today,
acquire this gift or payment received it? how much
I [ITEM]? for services? would you
DETERMINE WHICH DURABLES THE
T receive?
HOUSEHOLD OWNS BY ASKING Q.1.
E
FOR EACH DURABLE OWNED, WRITE
M
THE DESCRIPTION AND CODE IN THE
SPACE PROVIDED UNDER Q.2, AND »»7
PROCEED TO ASK Q.3-7 FOR EACH
ITEM.
PURCHASE..1
GIFT OR
PAYMENT..2
ITEM CODE YES NO DESCRIPTION CODE YEARS (»6) CURRENCY CURRENCY CURRENCY
Stove 201 1
Refrigerator 202 2
Washing Machine 203 3
Sewing/knitting machine 204 4
Fan 205 5
Television 206 6
Video player 207 7
Tape player/CD player 208 8
Camera, video camera 209 9
Bicycle 210 10
Motorcycle/scooter 211 11
Car or truck 212 12

13
14
15
16

Chapter 17: Consumption

Uploaded by

Copyright:

Available Formats

Chapter 17: Consumption

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 17: Consumption

Uploaded by

Copyright:

Available Formats

Designing Household Survey Questionnaires for Developing

Countries: Lessons from Ten Years of LSMS Experience

Chapter 17: Consumption

First Version, May, 1997

The basic ideas underlying the measurement of consumption are straightforward.

Part I: Policy Issues

Why Use Consumption to Measure Living Standards?

What Analyses are Consumption Data Good For?

Measuring Welfare. The policy importance of measuring living standards is indisputable.

Nutrition and Poverty Lines. There is a long tradition in development economics of

In some surveys, a household’s calorie consumption is estimated directly by nutritionists

Do LSMS Surveys Collect Accurate Consumption Data?

An anecdote about Guyana further illustrates the problems of comparing household

The literature on expenditure surveys in developing countries is even thinner. While

What Consumption Data are Needed?

A measure of total household consumption is built up from several components. First, it

To be a good measure of welfare, the consumption concept must be comprehensive. All

For measuring welfare, consumption is ultimately a more useful measure than

Collecting Price Data

However, there are alternatives to having either a long questionnaire or a short

Multiple Visits Throughout the Year

In nearly all LSMS surveys, calculating a comprehensive measure of consumption will

Note also that it is incorrect to compute an extended concept of consumption by adding

Respondents for the Consumption Module

Even within resource-pooling households, it may be useful to have multiple respondents

Diaries versus Oral Interviews

In Ukraine, a diary was administered to about 500 households in selected locations as an

Data from Other Parts of the Questionnaire

Education Module. Detailed data on household expenditures on school fees, uniforms,

Environment Module. Households sometimes obtain important resources from the

Part III: Draft Modules

This section contains a set of draft sub-modules to make up a consumption module to be

Box 1: Cautionary Advice

Part A: Daily Expenditures

Whether the questions on households’ consumption of home-produced food should be

Part C: Non-Food Consumption

Part D: Expenditures on Private Inter-household Transfers

Part E: Durable Goods

Supporting Intrahousehold Analysis. If supporting intrahousehold analysis is a special goal

Keeping the Module as Short as Possible

Another possibility would be to reduce the number of questions on inter-household

Baker, Judy, 1996. Personal correspondence

Bourguignon, François, and Pierre-André Chiappori. 1992. “Collective Models of Household

Cavendish, William. 1998. “The Complexity of the Commons: Environmental Resource

Cramer, Jan S. 1969. Empirical Econometrics Amsterdam: North-Holland

Deaton, Angus, 1997. The Analysis of Household Surveys: a Microeconometric Approach to

Foley, M., 1996. Personal correspondence

Gieseman, Raymond. 1987. “The Consumer Expenditure Survey: Quality Control by

International Monetary Fund. 1995. International Financial Statistics Yearbook. IMF:

International Monetary Fund. 1996. International Financial Statistics Yearbook. IMF:

Lanjouw, J. and P. Lanjouw, 1996. Draft. “Aggregation Consistent Poverty Comparison:

Neter, John. 1970. “Measurement Errors in Reports of Consumer Expenditures” Journal of

Paxson, Christina H. 1993. “Consumption and Income Seasonality in Thailand” Journal of

Scott, Kinnon. 1997. Personal communication

Silberstein, Adriana R. 1990. “First-wave Effects in the US Consumer Expenditure Interview

Srinivasan, T. N. 1994. “Database for Development Analysis” Journal of Development

Steele, Diane, 1998. "Ecuador Consumption Items" Internal memorandum. Development

Tsoflias, Lynn. 1996. Personal communication