Edited by Margaret Grosh and Paul Glewwe
The World Bank
Volume one
Lessons from 15 years of the
Living Standards Measurement Study
Developing Countries
Questionnaires
Household Survey
Designing
for
“Household surveys are essential for the analysis of most policy issues. This book has carefully assessed recent
experience and developed today’s best-practice technique for household surveys. Indeed, much of this technique
was developed and pioneered by the authors.This book is clear, systematic, and well structured. It is also wise and
scholarly. It will be indispensable to anyone involved in carrying out or analyzing household surveys, and thus it
is required reading for all those who wish to take evidence seriously when they think about policy.”
—Nicholas Stern, senior vice president, Development Economics
and chief economist, the World Bank
“This book is an ambitious undertaking, but it quickly exceeded my expectations. It has many strengths: …comprehensiveness,
…emphasis on practical application, ...and a sense of balance. For both my domestic and international
survey research, this volume will serve as a valued reference tool that I will consult regularly.”
—David R.Williams, professor of sociology and senior research scientist,
Survey Research Center, University of Michigan
“This is a comprehensive guide to planning household surveys on a range of socioeconomic topics in developing
countries. It is authoritative, clear, and balanced.The work is a valuable addition to the library of any survey
statistician or data analyst concerned with socioeconomic surveys in the developing world.”
—William Seltzer, former head,
United Nations Statistical Office
Household survey data are essential for assessing the impact of development policy on the lives of the poor.Yet
for many countries household survey data are incomplete, unreliable, or out of date.This handbook is a comprehensive
treatise on the design of multitopic household surveys in developing countries. It draws on 15 years of
experience from the World Bank’s Living Standards Measurement Study surveys and other household surveys
conducted in developing countries.
The handbook covers key topics in the design of household surveys, with many suggestions for customizing surveys
to local circumstances and improving data quality. Detailed draft questionnaires are provided in written and
electronic format to help users customize surveys.
This handbook serves several audiences:
• Survey planners from national statistical and planning agencies, universities, think tanks, consulting firms and
international organizations.
• Those working on either multitopic or topic-specific surveys.
• Data users, who will benefit from understanding the challenges, choices, and tradeoffs involved in data
collection.
Edited by Margaret Grosh and Paul Glewwe
Lessons from 15 years of the
Living Standards Measurement Study
Developing Countries
Questionnaires
Household Survey
Designing
for
Copyright © 2000
The International Bank for Reconstruction
and Development/THE WORLD BANK
1818 H Street, N.W.
Washington, D.C. 20433, U.S.A.
All rights reserved
Manufactured in the United States of America
First printing May 2000
The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any
manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent.The
World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence
of their use.The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the
part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries.
The material in this publication is copyrighted.The World Bank encourages dissemination of its work and will normally grant permission
promptly.
Permission to photocopy items for internal or personal use, for the internal or personal use of specific clients, or for educational classroom
use, is granted by the World Bank, provided that the appropriate fee is paid directly to Copyright Clearance Center, Inc., 222
Rosewood Drive, Danvers, MA 01923, U.S.A., telephone 978-750-8400, fax 978-750-4470. Please contact the Copyright Clearance Center
before photocopying items.
For permission to reprint individual articles or chapters, please fax your request with complete information to the Republication
Department, Copyright Clearance Center, fax 978-750-4470.
All other queries on rights and licenses should be addressed to the World Bank at the address above or faxed to 202-522-2422.
ISBN:0-19-521595-8
Library of Congress Cataloging-in-Publication Data has been applied for.
Part 1
Survey Design
5
1
Introduction
Margaret Grosh and Paul Glewwe
Accurate, up-to-date, and relevant data from household surveys are essential for governments to
make sound economic and social policy decisions. Governments need these data to measure and
monitor poverty, employment and unemployment, school enrollment, health and nutritional status,
housing conditions, and other dimensions of living standards.They need the data to determine
whether schools, health clinics, agriculture extension services, roads, electric power, and
other basic services are reaching the poor and other disadvantaged groups.And analysts need
household survey data to model economic behavior and thus provide answers to such important
policy questions as: How would changes in food subsidies affect the population’s nutritional status?
Would increasing fees for public schools reduce school enrollment, and how much revenue
would be raised by such fee increases? Who would participate in a new labor-intensive public
works program, and what would be the net benefit for participants? How would changes in the
price of fertilizer affect farmers’ production of different crops?
One way to collect the data needed to answer these
questions is to conduct separate household surveys on
each topic—that is, to conduct a labor force (employment)
survey, a health survey, a housing survey, and so
forth. Alternatively, data on many different topics can
be collected in a single survey. Such a “multi-topic”
household survey, which has many advantages, is the
type of survey considered in this book.
Household surveys are not a new invention.
Stigler (1954) points out that systematic collection of
data from households began over 200 years ago. The
first known efforts were the collection of family budgets
in England by Davies (1795) and Eden (1797). In
the 1800s similar data were collected in Saxony,
Prussia, Belgium, the United States, and undoubtedly
other places as well.The motivation for much of this
research was to focus public attention on the plight of
the poor. By the mid-1800s, generalizations about
household behavior were being drawn from these
data. For example, Ducpetiaux’s 1855 study of 200
Belgian households was used by Ernst Engel to derive
his classic law that the fraction of a household’s budget
devoted to food falls as income rises.
The statistical theory that supports modern survey
methods was developed in the 1920s.This led to the
establishment of high-caliber nationwide surveys in
many countries, especially after World War II.
Developing countries also participated in this phenomenon;for
example,India’s annual National Sample
Survey began in 1950. With the advent of modern
computing, and especially the appearance of powerful
personal computers, the collection and analysis of
household survey data has expanded rapidly in both
developed and developing countries. (See Deaton
MARGARET GROSH AND PAUL GLEWWE
6
1997 for a brief review of household surveys in the
20th
century.)
Since 1970 several major international programs
have been organized to support the collection of
household survey data in developing countries.
Among the largest such programs have been the
United Nations Household Survey Capability
Program, the World Fertility Surveys (which later
became the Demographic and Health Surveys), and
the World Bank’s Living Standards Measurement
Study (LSMS) survey program. Other organizations,
including the International Food Policy Research
Institute, the RAND Corporation, and Cornell
University, have also carried out household surveys in
developing countries. Some U.N. organizations regularly
participate in single-topic household surveys in
developing countries, such as employment surveys
done in collaboration with the International Labour
Office. And two regional survey programs have been
strongly influenced by, and indeed have grown directly
out of,theWorld Bank’s LSMS program.The first of
these, the Social Dimensions of Adjustment (SDA)
program for Sub-Saharan Africa, was supported by a
consortium of agencies and administered by theWorld
Bank. The second and more recent regional survey
program, the Improving Surveys of Living Conditions
program for Latin America, is sponsored jointly by the
Inter-American Development Bank, the World Bank,
and the Economic Commission for Latin America.
(The Spanish name for this program is Mejoramiento
de las Enquestas de Condiciones de Vida; it is often
referred to by its Spanish acronym, MECOVI.) The
surveys done under the LSMS, SDA, and MECOVI
programs are all multi-topic surveys.
Because of these and other efforts, household survey
data are now much more widely available than
they were 10 or 20 years ago.World Bank statistics on
the extent of poverty in 1985 were based on data from
only 22 of 86 developing countries.Although these 22
countries accounted for 76 percent of the population
of the 86 countries (Ravallion, Datt, and van de Walle
1991), it is significant that at that time no reliable data
existed for three-fourths of the developing countries.
Similar calculations currently underway are based on
data from about 70 of 100 developing and transition
countries, covering about 88 percent of the total population
of the countries. Data for more than one point
in time are now available for 50 countries. Coverage
has grown the most in the region where it was lowest,
Sub-Saharan Africa. In the 1985 calculation only 6
percent of Sub-Saharan Africa’s population was represented,
while recent estimates cover 66 percent of this
population (Ravallion and Chen 1998). Finally, the
time lag between collection and dissemination of the
data is getting smaller. In the 1985 World
Development Report the average lag was 11 years, so
the average survey date was 1974. Now the lag is only
five years (Ravallion and Chen 1997).
The surge in the collection of household survey
data in developing countries has greatly increased the
demand for knowledge on how best to design and
implement such surveys. Moreover, the growing number
of surveys provides a vast amount of experience
from which to draw lessons.Yet until now it has often
been difficult for those planning a new survey—especially
one in a developing country—to find out about
the experiences of previous household surveys: what
was tried, the factors that influenced decisionmaking,
what worked, and, most importantly, what did not
work. The formal literature is scattered across disciplines—statistics,
economics, sociology, psychology—
and often contained in conference proceedings or
government document series that are not widely
indexed and are seldom available outside the country
where they were written. An additional limitation is
that a considerable amount of the formal literature
pertains to surveys in industrialized countries. While
much can be learned from such literature, it is still
unclear how well the literature applies to settings with
lower literacy rates, different income levels and
employment and consumption patterns, and differing
factors that affect the social interaction of the inter-
view.
Much of the experience of surveying in developing
countries is poorly documented. Statistical institutes
in developing countries have little money or staff
to devote to experimentation or research; their mandate
is production and their resources are few.Articles
published in the formal academic literature that use
the data from these surveys typically provide only a
brief description of the data used. They may contain
some hints about whether the data collection methods
worked, but almost by definition, data collection
efforts that failed usually do not lead to academic publications.
Household survey questionnaires and their
associated statistical abstracts (reports) contain both
implicit and explicit information, but they are sometimes
available only in the country in which they were
CHAPTER 1 INTRODUCTION
7
administered. Moreover, statistical abstracts tend to
minimize any problems that may have been associated
with a survey because the statistical agencies that produce
these abstracts do not want to publicize a survey’s
shortcomings. In principle, the most useful information
for the designers of future surveys would be the
internal memoranda and informal notes of the agencies
and people involved in designing and implementing
past surveys. However, these are rarely filed and
seldom preserved after a survey is completed, much
less systematically made available to people outside the
agency.
The Objective and Audience for this Book
The objective of this book is to provide detailed
advice on how to design multi-topic household surveys,
based on the experience of past household surveys.This
book will help individuals and organizations
that are planning a comprehensive, multi-topic survey
to define the objectives of their survey, identify the
data needed to analyze those objectives, and draft
questionnaires that will collect such data. These tasks
are not easy, because designing such a survey for a
given country (or an area within a country) usually
involves a host of tradeoffs among different objectives.
This book aims to help survey designers evaluate these
tradeoffs, set realistic objectives, and design a survey
that best fulfills those objectives.
This book was written with several target audiences
in mind. The primary audience consists of the
people most likely to carry out household surveys
similar to the ones discussed in the book—the staff of
the national statistical agencies and planning agencies
responsible for their countries’ household surveys. A
second audience consists of individuals or groups in
consulting firms or international aid agencies that
advise governments on the design of household surveys.
A third audience is composed of researchers or
research agencies that plan to field a survey to pursue
their own research objectives.A fourth audience consists
of individuals or groups working on a survey
intended to evaluate or monitor the impact of a development
project in a particular country—either a
nationwide project or a project limited to a small part
of the country.A fifth audience is composed of people
working on a single-topic survey, because the book
can provide them with guidance on how to collect
“background” information from households—
information on, for example, a household’s composition,
basic characteristics, and level of welfare. Finally,
this book will assist researchers who use household
survey data produced by others, because it will help
them understand the challenges, possible options, and
tradeoffs involved in data collection. Such an understanding
will allow these researchers to interpret
household survey data more accurately and use these
data more fully.
The recommendations in this book apply to a
broad range of multi-topic household surveys, reflecting
the authors’ expectation that future surveys in
developing countries will be increasingly diverse in
their purposes and content.The book provides survey
designers with a wide range of options from which
they can pick and choose according to both the purpose
of their survey and the prevailing circumstances in
the country studied.Future household surveys will,and
should,evolve in ways that are hard to foresee.Thus this
book should be regarded as a starting point for planning
new surveys rather than as an exhaustive treatise
on the way to design all future household surveys.
This book assumes that the survey designer has
already decided to implement a multi-topic household
survey, as opposed to a census, a qualitative study, or a
single-purpose survey. Nevertheless, several chapters in
this book compare the advantages and disadvantages of
different data sources for studying certain topics. In
addition,Chapter 25 provides a thorough discussion of
qualitative data collection methods.
The Experience on Which This Book Is Based
Much of this book is based on the experience of the
World Bank’s Living Standards Measurement Study
(LSMS) program (Box 1.1), one of several recent
international efforts to expand the pool of data on
poverty and living standards in developing countries.
The World Bank established the LSMS program in
1980 to explore ways of increasing the accuracy, timeliness,
and policy relevance of household survey data
collected in developing countries. Because the first
LSMS surveys were designed by the World Bank for
research purposes, there was little variation in these
surveys’ design and implementation. However, by the
late 1990s LSMS surveys had been carried out in a
wide range of low- and middle-income countries,
with the involvement of many different national agencies
and international organizations. Over time LSMS
MARGARET GROSH AND PAUL GLEWWE
8
surveys have become increasingly customized to fit
specific country circumstances,including policy issues,
social and economic characteristics, and local household
survey traditions. Each survey has also inevitably
reflected the interests (and prejudices) of the individuals
planning it.
The LSMS program has had its share of successes.
Most importantly it has shown the feasibility of collecting
comprehensive household survey data in developing
countries. Since the first LSMS survey in 1985,
LSMS surveys have been implemented in about 30
developing countries (Table 1.1). In some of these
countries the original LSMS survey prototype was
implemented in its entirety.In other countries this prototype
was significantly altered to suit local circumstances.
In still other countries it was used as a guide to
redesign surveys that already existed. LSMS surveys
were also the starting point for SDA surveys, which
have been implemented in about 20 Sub-Saharan
African countries, and for the MECOVI program now
in progress in eight Latin American countries.
The increase in the number of LSMS surveys and
other household surveys has substantially expanded
the stock of data that can be used to study poverty
and, more broadly, economic and social development
in developing countries. In every country where an
LSMS survey has been done, the data have been used
to measure and analyze poverty by the government, an
international development agency, or both working
together. In several countries LSMS data have directly
influenced specific government policy decisions (see
Box 1.2). Data from LSMS and similar surveys have
also been used in hundreds of studies of developing
countries, helping to extend what is known about
poverty, household decisionmaking, and the impact of
economic and social policy changes on household
The overall objective of LSMS surveys is to measure and
study the determinants of living standards in developing countries,
especially the living standards of the poor.To accomplish
this objective, LSMS surveys must collect data on many
aspects of living standards, on the choices that households
make, and on the economic and social environment in which
household members live. Much of the analysis undertaken
using LSMS surveys attempts to investigate the determinants
of living standards—which requires more sophisticated analytical
methods than simple descriptive tables.
LSMS surveys have several characteristics that distinguish
them from other surveys. One of the most important is
that they use several questionnaires to collect information
about many different aspects of household welfare and
behavior.These consist of a household questionnaire, a community
questionnaire, a price questionnaire, and, in some
cases, a facilities questionnaire. (For more details on the questionnaires
see Box 1.4.)
Another characteristic of LSMS surveys is that they typically
have nationally representative, but relatively small, samples—usually
between 2000 and 5000 households. This will
yield fairly accurate descriptive statistics for the country as a
whole and for large subareas (such as rural and urban areas
or a few agroclimatic zones), but usually not for political jurisdictions
(such as states or provinces). The surveys’ sample
sizes are generally adequate for the regression methods often
used for policy analysis of LSMS survey data.
Because of the complexity of most LSMS surveys, these
surveys have rigorous quality control procedures to ensure
that the data they gather are of high quality. These procedures,
which are generally difficult to implement on larger
samples, usually include several key elements. Both the survey’s
fieldwork and its data entry are decentralized, and the
people who carry out these tasks are strictly supervised.
Interviewers receive extensive training (usually for about four
weeks) prior to the survey. In the field, information is gathered
not by asking one person all the questions about the household
and its members but through a series of “mini-interviews,”
with each adult responding for himself or herself.This
procedure minimizes any errors caused by respondent fatigue
or by the use of proxy respondents .The interviewers make
multiple visits to households to find any members who were
not home during the interviewer’s earlier visits—which also
reduces the need to use proxy respondents.
There is one supervisor for every two or three interviewers.The
supervisors must revisit a significant percentage (often
25 percent) of the sampled households to check on the accuracy
of the interviewer’s data.They must directly observe some
interviews, and they must review each questionnaire in detail.
Supervisors’ performance of these procedures is documented,
and the supervisors are in turn supervised by staff from the central
office of the statistical agency. Data entry and editing are
done as soon as each interview is over, either in the local field
office or by a data entry operator who travels to households
with the team of interviewers. As data are entered into the
computer, a data entry program carries out a large number of
quality checks to detect responses that are out of range or
inconsistent with the other data from the questionnaire. Any
problems this program detects can be verified or corrected in a
subsequent visit to the household by the interviewer.
Box 1.1 An Introduction to LSMS Surveys
CHAPTER 1 INTRODUCTION
9
welfare. Many of these studies have been presented at
conferences and published in books or academic journals,
and have thereby shaped thinking about these
issues far beyond the countries in which the data were
collected.
Despite these successes, several challenges remain
for LSMS surveys and other multi-topic household
surveys. First and most obviously, many developing
countries still have inadequate household survey data.
This is true even for some of the countries that have
recently fielded new surveys, including LSMS surveys.
Ideally, all governments should collect data on a regular,
ongoing basis in order to monitor poverty trends
over time. However, survey efforts are still sporadic in
many developing countries today, and many surveys
have serious deficiencies such as limited questionnaires,
samples that exclude rural areas, and long delays
in processing the data after completing the fieldwork.
Second, improvements are needed in the process
of adapting the LSMS approach to countries that have
not yet implemented LSMS-type surveys. It has been
difficult for people working on a survey in one country
to learn from the experience of other countries
that have carried out LSMS and other multi-topic surveys.
Mid-level staff in government statistical agencies
know the details of why particular choices were made
and know how well the choices worked, but they
rarely meet with their counterparts in other countries.
A small pool of World Bank staff and consultants also
know many of these details and have been in contact
with many of the people developing new surveys in
particular countries. However, until now, they have
Table 1.1 LSMS Surveys
Has the survey been repeated,
Country Year of first survey or will it be repeated? Number of households in sample
Albania 1996 No 1,500
Algeria 1995 No 5,900
Armenia 1996 No 4,920
Azerbaijan 1995 No 2,016
Bolivia 1989 Yes 4,330–9,160
Brazil 1996 No 5,000
Bulgaria 1995 Yes 2,000
Cambodia 1997 Yes 6,010
China (Hebei and Liaoning only) 1995 No 800
Côte d’Ivoire 1985 Yes 1,600
Ecuador 1994 Yes 4,500
Ghana 1987/88 Yes 3,200
Guyana 1992/93 No 1,800
Jamaica 1988 Yes 2,000–4,400
Kazakhstan 1996 No 2,000
Krygyz Republic 1994 Yes 2,100
Mauritania 1988 Yes 1,600
Morocco 1991 Yes 3,360–4,800
Nepal 1996 No 3,373
Nicaragua 1993 Yes 4,454
Pakistan 1991 Yes 4,800
Panama 1997 Yes 4,945
Paraguay 1997/98 Yes 5,000
Peru 1985 Yes 1,500–3,623
Romania 1994/95 Yes 31,200
South Africa 1993 No 8,850
Tajikistan 1999 No 2,000
Tanzania—Kagera 1991 No 800
Tanzania—Human Resource Development Survey 1993 No 5,200
Tunisia 1995/96 No 3,800
Turkmenistan 1997 No 2,350
Vietnam 1992/93 Yes 4,800–6,000
Source: LSMS data bank.
MARGARET GROSH AND PAUL GLEWWE
10
shared their knowledge of past surveys mostly on an
informal basis, one person or one country at a time.
And since the teams that are assigned to work on each
specific survey are usually small, these teams start the
survey development process with detailed knowledge
of some of the topics to be covered by the survey
questionnaires but less detailed knowledge of others.
Third,the data gathered from some parts of LSMS
survey questionnaires have been disappointing. Two
particularly difficult problems are how to measure
household income from agriculture and nonagricultural
self-employment and how to measure savings and
financial assets.
Fourth, new issues have emerged since the first
LSMS surveys were implemented. The economics
profession has increasingly discounted the notion of
the household as a unified decisionmaking body, trying
instead to understand how goods, services, and
power are allocated among the different members of a
given household. In addition, there is growing interest
in using qualitative and quantitative techniques in
complementary ways, or even combining these techniques.And
analysts increasingly use household survey
data to address environmental issues.
How this Book Came to Be
In recognition of the continuing challenges for LSMS
surveys, in the mid-1990s theWorld Bank initiated the
multiyear research project that developed this book.
(See Box 1.3 for a brief description of related initiatives.)
The project was assigned three goals: to extend
the range of policy issues that can be analyzed with
LSMS data; to increase the reliability and accuracy of
LSMS household surveys are designed to collect data that can
be used to study living standards and how living standards are
affected by government policies.The following examples illustrate
how some governments and donor agencies have used
LSMS data to help make policy choices.
In 1989 the Jamaican government was considering
whether it should eliminate subsidies for basic food items and
use the funds saved to expand its food stamp program.While
the government was making this decision, data from the
Jamaican LSMS survey became available.Analysis of these data
showed that most of the benefits from general price subsidies
went to nonpoor households, while most of the benefits of the
food stamp program went to the poor.This information helped
the government decide to remove the subsidies on basic foodstuffs
and expand its food stamp program. The government
then commissioned further analysis of the LSMS data to find
out how many families needed help in purchasing a minimum
food basket, and how much help these families needed. The
government used this information to choose new eligibility
thresholds and benefit levels for the food stamp program.
The Jamaican government has used its LSMS data in
making many other decisions, such as whether to change
kerosene subsidies and whether and how to subsidize medicines
distributed through public health clinics. In addition, the
government has used LSMS data to study the effects of raising
user fees for public health care services. The Jamaican
LSMS survey is conducted annually; the incidence of poverty
is measured in each survey.
In South Africa the 1993 LSMS survey provided the first
comprehensive, credible data set for the entire territory of
South Africa, including the homelands.The survey was completed
just before the first democratic elections were held in
the country.The data were quickly put to extensive use both
by the new government and by academic researchers. The
first product, an extensive statistical abstract, was followed by
a poverty profile prepared jointly by the World Bank and the
government’s Ministry of Reconstruction and Development,
then by other studies and reports. This body of work has
helped to shift the national debate about poverty away from
the nature and extent of poverty toward policy options for
reducing poverty. For example, young women in rural areas
were made eligible for public works employment programs
after the data showed that these women were often needy
and that they would be able to participate in such schemes
since they had access to childcare. The survey data also
revealed that the old age pension program was well targeted,
which convinced the government not to modify that program
but instead to consider reforming other programs that
appeared to be less well targeted.
In 1998 the Government of the Kyrgyz Republic undertook
a thorough assessment of the current and projected
impact of its state pension reform.With the help of a World
Bank team, the government analyzed data from the 1993 and
1996 Kyrgyz LSMS surveys to examine a range of policy alternatives.The
survey data were used to show rates of participation
in, contributions to, and receipts from pension programs
by age cohort and by level of welfare.The data were
particularly helpful to the government when it worked on
setting a new level for the minimum pension, based on average
earnings in the poorest quintile of the population.
Forthcoming analytical work will include an assessment of
household consumption and demand for utility services, and
the formulation of a strategy to compensate the poorest for
increases in utility prices.
Box 1.2 Using LSMS Data to Inform Government Policy Choices
CHAPTER 1 INTRODUCTION
11
the surveys; and to make it easier to implement LSMS
surveys, either by simplifying survey design or by providing
more and better instructional materials on survey
design and implementation.This book contributes
to the achievement of all three goals and thus addresses
the four challenges facing the LSMS that were
described above.
Past LSMS surveys have typically consisted of a
household questionnaire, a community questionnaire,
and a price questionnaire; sometimes they have also
included a school or health facility questionnaire.The
household and community questionnaires are each
composed of separate modules, sections of the questionnaire
that focus on different topics (Box 1.4).This
book reviews each module that has typically been a
part of past LSMS surveys, and offers some interesting
new additions.
The author or authors of each chapter of this
book were chosen according to the following criteria:
extensive research experience on the topic in question;
experience in analyzing data on that topic using
data from LSMS and non-LSMS surveys (both multitopic
and single-topic); and experience in collecting
data in developing countries. In order to ensure that
experiences and perspectives from both LSMS and
non-LSMS surveys were included, a concerted effort
was made to include not only people who have long
been associated with LSMS surveys but also people
associated with other survey traditions.
The authors of the chapters that focus on specific
modules have reviewed the relevant literature (both analytical
literature and literature on survey experience),
analyzed existing survey data,and,in the case of the consumption
module,experimented with different methods
of collecting data. Many authors have drawn lessons not
only from LSMS experience but also from experience of
other surveys,including the RAND Family Life surveys,
the World Bank’s Social Dimensions of Adjustment
(SDA) surveys, and the Demographic and Health
Surveys (DHS). The authors of many chapters have
reviewed a large number of single-topic surveys, including
ones on housing, agriculture, water and sanitation,
time use, and household income and expenditure.
While this book was being written,two workshops
were held that brought together all of the authors, as
well as representatives from the various organizations
that constitute the main audiences for this book. The
participants in the first workshop were primarily data
users—researchers and policy advisors.They were invitBox
1.3 Other LSMS Products
A manual for planning and implementing the LSMS survey.
When work began on this book about questionnaire
design, work also began on a companion volume about
planning and implementation: “A Manual for Planning and
Implementing the Living Standards Measurement Study
Survey” by Margaret Grosh and Juan Muñoz.The manual,
completed in 1996, is intended for all people involved in
planning and implementing an LSMS survey, including staff
in planning agencies, statistical agencies, line ministries, academic
institutions, and development agencies.The manual
discusses such issues as sampling, fieldwork, data management,
initial analysis, dissemination, and a host of planning
and budgeting issues—in each case explaining the technical
procedures and standards used in LSMS surveys. The
manual is available in English, Spanish, and Russian.
The LSMS data bank. Data from LSMS surveys are now
much more accessible than they were in the early years of
the LSMS program in the late-1980s. The LSMS website,
http://www.worldbank.org/lsms/lsmshome.html, contains a
catalogue of the data sets that are available, the documentation
for most surveys, and the data from some of the surveys.
Data sets and documentation not yet available from
the website are available by mail.Three factors have made
it possible to increase the accessibility of LSMS data. First,
a growing number of countries have adopted more open
data access policies, and some have even given the World
Bank permission to place their data on the LSMS website.
Second, the LSMS team at the World Bank has thoroughly
documented most of the surveys, whether working alone,
working with managers of survey projects, or commissioning
documentation work. Good documentation preserves
institutional memory, lowers the cost to the Bank of disseminating
data, and reduces startup costs for new users
of LSMS data sets.Third, the LSMS team now has a full-time
data manager, good technical support, and adequate space
to stock an inventory of questionnaires, manuals for field
staff, abstracts, and other useful documents from each
country’s survey.
Other tools.The LSMS program periodically produces other
tools for survey planners and analysts.The best way to keep
abreast of these tools is to look on the “tools for managers
of new surveys” and “tools for using household survey
data” pages of the LSMS website. Readers of this book may
be particularly interested in a paper by Deaton and Zaidi
(1999) on how to construct consumption aggregates,
which complements Chapter 5 in this book, and in a recent
book by Deaton (1997) on analyzing household survey
data, which brings together a large amount of statistical and
econometric material relevant for policy analysis.
MARGARET GROSH AND PAUL GLEWWE
12
ed to ensure that the book had correctly identified the
research and information needs of potential data users.
A larger share of the participants in the second workshop
were data producers—staff from national statistical
agencies and representatives of organizations that provide
technical assistance or funding to national statistical
agencies.They were invited to ensure that the book
addressed their concerns—a requirement for any successful
survey. Prior to the first workshop each chapter
was reviewed by an expert in the relevant field. Before
the second workshop the draft manuscript as a whole
was reviewed by several experts in analyzing and producing
household survey data in developing countries.
After all of these people’s advice had been incorporated
and a polished draft produced, the book was subject to
another round of (anonymous) peer review and revisions.In
addition,many of the draft chapters were given
to people who were in the process of advising governments
or survey institutions on the design of a multitopic
household survey.This served as a limited field test
for the book,and also confirmed that government planning
and statistical agencies—and advisors of these
agencies—had a genuine and pressing need for this
book.
Nevertheless it must be recognized that because
the draft modules presented in each chapter are based
primarily on lessons from past surveys, few of them
have been rigorously field tested in the exact form
presented here. Thus extensive field testing must be
done in each country implementing a new survey.
Survey designers should consider this testing a vital
part of their job after they have chosen a set of modules,
modified these modules, and combined the modules
into survey questionnaires.
Each chapter contains a “cautionary advice” box
that specifies how much the draft module has been
changed from its design in previous LSMS surveys,
how well similar modules have worked in the past, and
which parts of the modules most need to be customized
to fit specific country circumstances.
This book represents a major advance on three
fronts. First, the book makes it easier for those workOne
distinguishing characteristic of LSMS surveys is that they
are both multi-topic and multi-level: they use several questionnaires
to study many different aspects of household welfare
and behavior. The largest LSMS questionnaire is the
household questionnaire.The LSMS household questionnaire
always collects detailed information to measure household
consumption, which is the best monetary indicator of household
welfare (see Chapter 5 for further discussion). The
household questionnaire also collects information on income;
transfer income and income from wage employment are collected
in almost every LSMS survey, and many LSMS surveys
also collect data on income from agriculture, household
enterprises, and miscellaneous sources.
LSMS household questionnaires always record information
on a variety of other dimensions of welfare and on the
use of social services; housing and related amenities such as
water and sanitation; the level of education of adults, grade
attainment and current enrollment rate of school-aged children;
and vaccination histories and anthropometric (height
and weight) measurements for children. A typical household
questionnaire collects more information than this, in order to
expand the range of living standards indicators that can be
studied and to allow researchers to model the choices
households make.The traditional list of modules included in
a prototype LSMS survey includes: household roster, education,
health, employment, migration, anthropometry, fertility,
consumption, housing, agriculture, household enterprises,
miscellaneous income, and savings and credit. Some of the
information (consumption, housing quality, agricultural production)
is collected only at the household level, but much of
it (employment, education, health) is collected at the individual
level.
The community questionnaire gathers information on
local conditions common to all households living in the same
community. Many of these conditions recorded can be directly
influenced by government actions.The information covered
typically includes the basic characteristics (including distance
from the community) of nearby schools and health facilities,
the existence and condition of local infrastructure (such as
roads and public transportation), sources of fuel and water,
availability of electricity, means of communication, and local
agricultural conditions and practices.
A separate price questionnaire is used to record the prevailing
prices of commonly purchased items in local shops and
markets. In almost all countries prices vary considerably
among regions—in order to compare the welfare levels of
households that live in different regions one needs information
on the prices that they face when purchasing goods and
services. The community and price questionnaires are discussed
in Chapter 13.
Finally, in some LSMS surveys special facility questionnaires
are used to gather detailed information on schools or
health facilities. These questionnaires are discussed in
Chapters 7 and 8.
Box 1.4 Components of a Typical LSMS Survey
CHAPTER 1 INTRODUCTION
13
ing on new household surveys to learn from the wide
range of LSMS and other survey experience. Second,
the book makes it much easier to customize the design
of questionnaires for new multi-topic surveys. Third,
the material presented in the book deals with new
policy questions, presents new analytical methods to
address both new and long-standing policy issues, and
provides new ways to reduce or avoid measurement
problems.
How to Use this Book
The process of designing a comprehensive,multi-topic
survey can be divided into five steps.First,survey planners
must define the fundamental objectives of the
survey and decide on the overall design of the survey
in light of these objectives. Second, within this general
framework the survey planners must choose which
modules to include in the questionnaires, the objectives
of each of these modules, and the approximate
length of each module.Third, the planners must work
out the precise design of each module, question by
question, in light of the module’s specific objectives
and approximate length. Fourth, the modules must be
integrated with each other and combined into a complete
set of draft questionnaires (household,community,
price, and in some cases, facility). Fifth, the draft
questionnaire should be translated (if applicable) and
field tested. Ideally, the five steps should be completed
in chronological order. However, in practice, implementing
any given step may reveal information that
requires survey designers to rethink a previous step.
This book consists of three volumes. Volumes 1
and 2 contain all 26 chapters of this book.Volume 3
provides the draft questionnaires introduced by the
chapters in Volumes 1 and 2. Volumes 1 and 2 are
organized into four parts. The first three chapters of
Volume 1 constitute Part 1, which discusses the “big
picture.” This includes decisions that must be made
about the overall design of the survey and the modules
to be used, as well as procedures for combining modules
into questionnaires and questionnaires into a survey
(or sequence of surveys). Chapter 2 starts by
describing how to choose from among the three “classic”
survey designs and how to select the modules to
be included in the survey. Chapter 3 describes general
procedures for designing each module, combining the
modules into a well-integrated set of questionnaires,
and translating and field testing the questionnaires.
The remaining chapters ofVolume 1 form Part 2
of the book, and the first nine chapters of Volume 2
comprise Part 3.The chapters in Parts 2 and 3 discuss,
in great detail, the individual modules that are the
building blocks of any multi-topic household survey.
Each chapter reviews the main policy issues pertinent
to the subject matter of the module, identifies the data
needed to analyze these issues, introduces one or more
draft modules (which are presented inVolume 3), and
provides annotated notes that explain the reasoning
behind many of the details of each draft module. For
most modules, two or three different versions are
introduced, each of different length.Which module to
use depends on the level of interest in the particular
topic. In addition, many of the chapters in Parts 2 and
III discuss how to add or delete submodules within
each module in order to provide a better fit with local
circumstances and the specific focus of the survey.
Part 2 (Chapters 4–13), in Volume 1, introduces
“core” modules that must be included in virtually all
LSMS-type surveys: metadata, consumption, roster,
education, health, employment, anthropometry, transfers
and other nonlabor income, housing, and the
community questionnaire.The modules on health and
education come with draft questionnaires that can be
used for gathering data from local schools and health
care facilities.The collection of community-level data,
including data on local prices, is discussed in Chapter
13, the final chapter of Part 2.
Part 3 (Chapters 14–22), in Volume 2, introduces
modules that are optional: environmental issues, fertility,
migration, total income, household enterprises,
agriculture, savings, credit, and time use.
The last four chapters ofVolume 2 constitutes Part
4 of the book.These chapters contain material that is
more methodological in nature. Chapter 23 discusses
when and how to collect panel data—in other words,
whether to interview the same households when
doing a sequence of surveys, and how best to do so
when this option is chosen. Chapter 24 reviews the
issues involved in analyzing the allocation of resources
and power within households, and summarizes the
implications of this analysis for data collection.
Chapter 25 summarizes how qualitative research
methods can be used to complement the quantitative
methods typically used in the design and analysis of
multi-topic household surveys. This chapter stresses
that qualitative methods can play a useful role in the
design of multi-topic household surveys, especially in
MARGARET GROSH AND PAUL GLEWWE
14
formulating questions and developing hypotheses for
data analysis. It is unfortunate that these methods have
been neglected by most survey designers, who usually
have quantitative backgrounds.
Chapter 26 reviews the basic economic and econometric
concepts that underpin many of the chapters in
this book. Although many survey designers have an
economics background, many others do not, and even
some economists may benefit from a review of this
material. The chapter begins by presenting the basic
economic model of the household and goes on to discuss
standard econometric techniques that have often
been used in policy research on developing countries.
The reader should understand that the questionnaires
provided in Volume 3 and discussed in Parts 2
and III are not polished or completed, and cannot be
used immediately in any developing country. Final
versions of the questionnaires for any country must be
developed by the survey designers themselves. Survey
designers must combine their own experience and
expertise with the information in this book to design
a country-specific questionnaire that will elicit the
information needed to answer the most important
policy questions of that country. This book is just a
starting point. It provides survey designers with the
lessons learned from past experience and with advice
from experts who are familiar with both LSMS and
other household surveys.
Part 1
All readers of this book should read all of Part 1,which
in addition to this chapter includes Chapters 2 and 3.
CHAPTER 2: MAKING DECISIONS ON THE OVERALL
DESIGN OF THE SURVEY. This chapter leads survey
designers through the factors that need to be considered
when determining the basic scope of the survey.
It sketches three alternative designs for an LSMS-type
survey—a full LSMS survey, a “core” (scaled-down)
LSMS survey, and a core and rotating module
design—and suggests rules to help survey designers
choose the design most appropriate for the circumstances
that they face. Chapter 2 also explicitly defines
the “core” components that should be included in any
LSMS-type survey.
CHAPTER 3: DESIGNING QUESTIONNAIRE MODULES AND
ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES. This
chapter moves to a finer level of detail, providing general
guidance on how to design individual modules and
combine them into an integrated set of survey questionnaires.
First each module must be customized to
meet the specific objectives set out for it. Then the
modules must be compared with each other to check
for gaps and overlaps and to harmonize wording,codes,
and recall periods. Next, survey designers must decide
on the order of the modules and combine the modules
into draft questionnaires. Finally, the questionnaires
must be translated and field tested. Throughout this
process, issues of questionnaire formatting will arise;
Chapter 3 explains the principles and conventions used
in formatting LSMS questionnaires.
Part 2
After survey designers have read Part 1 of this book
and decided on the broad issues concerning the scope
and design of the survey, they can begin the painstaking
but crucially important task of designing the individual
modules. Detailed advice on designing individual
modules is given in Parts 2 and 3 of the book. Part
2 includes all modules that should be included in
almost any LSMS-type survey, while the modules in
Part 3 are optional.Almost all of the chapters in Parts
2 and 3 follow a similar outline. The first section
reviews the current policy issues in developing countries
for the topic or sector covered by the chapter.
The second section explains what kinds of data are
needed from household surveys to address these policy
questions and also discusses any measurement
issues.The third section introduces one or more versions
of a draft module (the modules themselves are
presented in Volume 3), and the fourth section provides
notes that explain the reasoning behind, and
important details of, each version of the draft module.
Chapters 4 on metadata and 5 on consumption
are the first chapters in Part 2 because they contain a
good deal of information on survey methods and
issues of validity and measurement, information with
broad implications for the subsequent chapters. Like
Chapters 4 and 5, Chapters 6–13 cover “core” modules.
Each of these topics does not have to be covered
in great detail, but it is recommended that at least the
essential parts of each of these modules be included.
For example, the fullest version of the health module
introducted by Chapter 8 is so extensive that it should
be included only in a survey specializing in health
issues.Yet questions 10–38 of the short health module
are an essential part of the core.
CHAPTER 1 INTRODUCTION
15
CHAPTER 4: METADATA—INFORMATION ABOUT EACH
INTERVIEW AND QUESTIONNAIRE. “Metadata” are data
about the survey itself, such as dates of interviews,
identities of respondents, and time required for each
interview.This topic has frequently been neglected in
LSMS and other household surveys. Metadata are
needed to guide the implementation of a survey, to
help analysts interpret survey data, and to allow a
quantitative evaluation of different survey procedures.
Chapter 4 reviews the different kinds of metadata that
should be collected as part of any multi-topic household
survey, and provides guidance on how to collect
them. If the recommendations made in this chapter
had been adopted at the beginning of the LSMS program,
the rest of this book and its companion manual
(Grosh and Muñoz 1996) would have had a firmer
empirical foundation for discussing the tradeoffs in
survey design and implementation. For example,
information on the time required to complete specific
modules of varying lengths would have been very
useful for deciding costs of expanding a module in
terms of interview time.
CHAPTER 5: CONSUMPTION. This chapter differs from
the others by focusing most of its attention on measurement
issues, specifically on how to measure household
consumption.The chapter draws on the literature
on data collection, and also on data collection experiments
that were part of the research for this book.One
important conclusion of Chapter 5 is that measurement
of consumption is highly sensitive to differences
in methods, and that consumption measurement techniques
within a given country must therefore be standardized
over time. Researchers and consumers of
household survey data need to be aware of the stringent
comparability requirements that consumption
data must satisfy before comparisons can be made
across different surveys.
CHAPTER 6: HOUSEHOLD ROSTER. One of the fundamental
decisions in any household survey is deciding
who is and who is not a household member. Chapter
6 provides a basis for identifying all members of the
household, thus selecting from which individuals the
survey will collect information.The recommendations
in this chapter do not differ greatly from procedures
used in past LSMS surveys.The chapter offers no new
proposals on how to define a “household,” which is a
difficult issue in almost any household survey. LSMS
surveys have not made any significant contribution to
this issue.The chapter confirms the usefulness of gathering
household roster information on any children
and parents of household members who do not live in
the household, as well as of linking parents and children
to each other when both are household mem-
bers.
CHAPTER 7: EDUCATION. This chapter recommends
only modest changes to the design of the education
module that has been used in most previous LSMS surveys,
because this design has worked quite well in the
past. The education module (presented in Volume 3)
collects information about the schooling of all household
members, including highest grade attained,
degrees obtained, and grades repeated. Individuals currently
in school are also asked about the type of school
they attend (public or private), their recent attendance,
and the amount of money the household spends on
their schooling. Chapter 7 also introduces an expanded
version of the education module,which is useful for
the design of surveys that focus on education issues.
This expanded module requires administering relatively
short tests of cognitive skills to members of the
household as well as collecting information about local
schools through school and teacher questionnaires.
CHAPTER 8: HEALTH. The health data collected in previous
LSMS surveys have been of limited usefulness
for policy analysis.The draft health module introduced
in Chapter 8 (and provided inVolume 3) has been dramatically
revised from previous surveys. The new
module consists of a series of submodules on selfreported
health status, health-related behavior, child
immunization, insurance coverage, health service utilization
and cost, and knowledge of health providers.
One key change is that information is collected on all
visits made by household members to medical facilities
during the reference period—not just the most
recent visit.Another improvement is the collection of
self-reported data on “activities of daily living,” data
which cover the ability to climb stairs, carry heavy
loads, or walk long distances.The chapter also presents
an expanded version of the module that includes
questions related to mental health and very detailed
questions on the household’s health expenditures and
utilization of health facilities.The expanded version of
the module also collects data on observed activities of
daily living and the cognitive functioning of house-
MARGARET GROSH AND PAUL GLEWWE
16
hold members that are observed by the interviewer.
CHAPTER 9: EMPLOYMENT. The collection of data on
employment and labor force participation in past
LSMS surveys has been fairly successful. However,
given the large size of this module, there is ample
room for many small improvements. Several modifications
are suggested to past designs. First, detailed information
on household members’ work for household
businesses or on the household farm is collected in the
household business and agriculture modules, not in
the employment module. Second, job history information
is collected in a way that focuses on each individual’s
employment five years before the time of the
survey. Third, summary employment information is
collected in a way that can accommodate individuals
who have done many different kinds of work in the
past 12 months.
CHAPTER 10: ANTHROPOMETRY. Nutritional status has
always been one of the key nonmonetary indicators of
welfare in LSMS surveys,especially for children;future
surveys should continue to collect such data. Chapter
10 discusses the tradeoffs involved when anthropometric
information is collected only for young children,
rather than for all household members. The
chapter recommends that in general anthropometric
data should be collected for all household members,
both adults and children. Chapter 10 also discusses the
merits of collecting data on mid-upper arm circumference,
which until now has been done in only one
LSMS survey.
CHAPTER 11: TRANSFERS AND OTHER NONLABOR
INCOME. Many households receive income unrelated to
any of their members’ current work activities. Past
LSMS surveys have usually collected data on this
income in a module on transfers and other nonlabor
income. Chapter 11 introduces an improved version of
that module,capturing more information on both public
and private transfers. For private transfers, the new
module also collects more information about the donor
household and about the purpose of the transfers.
CHAPTER 12: HOUSING. Past LSMS surveys collected
data on housing to serve as indicators of “basic needs”
and to derive the implicit consumption value (imputed
rent) associated with owner-occupied housing.
Chapter 12 introduces two draft housing modules, a
short module and a longer module.The short module
collects data similar to those collected by the housing
module used in previous LSMS surveys, albeit with
several improvements. The longer module collects
information that can be used to study a wide range of
housing policy issues.Both modules are flexible in that
they can gather data on complex water supply systems
and on many different rental arrangements. Finally, in
the new housing module questions are added that are
appropriate to places with cold climates and well
developed infrastructure;such questions will be particularly
useful for surveys in the transition economies of
Eastern Europe and the former Soviet Union.
CHAPTER 13: COMMUNITY AND PRICE DATA. Community
questionnaires have been used in many past
LSMS surveys to gather information on the economic
environment in which households operate.
Community characteristics that affect households’
economic environment can often be directly changed
by government interventions.This chapter provides a
much-needed discussion of how to define the “community”
for which the information is to be collected
and how to gather community information from a
group of respondents. Finally, it introduces a longer
and much more comprehensive community questionnaire
(presented in Volume 3) than has been used in
previous LSMS surveys.The design of this draft questionnaire
is based on the experience of both LSMS
surveys and the RAND Corporation’s Family Life
Surveys, as well as on suggestions from many of the
authors of the other chapters in this book.
Part 3
This part, consisting of Chapters 14–22, is in Volume
2. The chapters in Part 3 follow the same format as
those in Part 2. Part 3 covers topics that are likely to
be of interest, but one would never include all of them
in any one survey. None of the topics in Part 3 are
required for an LSMS-type survey.
CHAPTER 14: ENVIRONMENTAL ISSUES. To date, very
few LSMS surveys have collected data that can be used
to examine environmental issues. The environmental
module introduced in Chapter 14 (and presented in
Volume 3) offers a series of submodules that can be
used, as appropriate, in different settings. There are
very brief submodules on environmental priorities in
urban and rural areas, on attitudes and perceptions
CHAPTER 1 INTRODUCTION
17
about urban air quality, and on discount rates. All of
these submodules could be used in many surveys, even
surveys that do not focus on the environment. The
environmental module also includes lengthy submodules
on water, sanitation, and fuel, to be included in
surveys for which the use of these resources is of particular
interest. There are also contingent valuation
submodules that attempt to measure the extent to
which households are willing to pay for improvements
in urban air quality, the urban water supply, urban sanitation,
or the rural water supply.The design of all of
the above submodules is based on extensive experience
from single-purpose surveys.
CHAPTER 15: FERTILITY. This chapter follows the same
general approach used in many past LSMS surveys.
The chapter introduces a short version of a fertility
module that collects the data necessary to understand
some general aspects of contraceptive use and to compile
a maternity history that lists all births. Chapter 15
also introduces a standard version that includes a
maternity history, a reproductive health submodule
covering the previous three years, a longer section on
contraceptive use, and a section on fertility preferences.
Both the short and standard versions are presented
inVolume 3.Whichever version of the module
survey designers choose to use, the module should be
administered to all women in the household of childbearing
age.This departs from the past LSMS practice
of interviewing only one randomly selected woman
per household.
CHAPTER 16: MIGRATION. Data on migration have
been collected in many past LSMS surveys, but the
amount of information collected has been quite small
and the data have rarely been analyzed, despite significant
interest in migration among researchers. This
chapter introduces three different versions of the
migration module: a short version, a standard version,
and an expanded version. All three versions are presented
in Volume 3.The standard and expanded versions
are designed to collect much more detailed
information than has been collected in the migration
modules used in previous LSMS surveys. Including
either the standard or the expanded draft migration
module in a future survey will yield a rich data set that
should prove very useful for comprehensive research
on migration.
CHAPTER 17: SHOULD THE SURVEY MEASURE TOTAL
HOUSEHOLD INCOME? Chapter 5 (and indeed this
entire book) argues that consumption is the best monetary
measure of welfare in multi-topic surveys. This
implies that the consumption module is essential and
must always be included. In contrast, the book takes
the view that collecting the data needed to calculate
total household income is optional, which implies that
the household enterprise, agriculture, and savings
modules (which collect much of the income data) can
be substantially reduced or even omitted, depending
on the level of interest in these topics. Chapter 17
reviews the advantages and disadvantages of collecting
the data needed to calculate total household income,
and describes the circumstances under which measuring
total income should be an objective of a multitopic
survey.
CHAPTER 18: HOUSEHOLD ENTERPRISES. Small businesses
owned and operated by households are quite
common in developing countries, yet it is difficult to
collect accurate data on these income-generating
activities. Based on their extensive experience of analyzing
data from past household surveys—both LSMS
and others—the authors introduce three versions of
the household enterprise module (presented in
Volume 3). Survey designers should choose the version
that best matches policymakers’ level of interest in
household enterprise issues. In previous LSMS surveys,
data on employment in these enterprises were
collected in the employment module, but this chapter
recommends that such information be collected in the
household enterprise module; each of the modules
introduced in this chapter collects such data.One consequence
of this is that the standard version of this
module is now longer than the version typically used
in previous LSMS surveys, and the expanded version
is even longer than the standard version.
CHAPTER 19: AGRICULTURE. Collecting accurate and
comprehensive data on agricultural activities is difficult
in any survey, and past LSMS surveys have experienced
many problems in collecting such data. This
chapter introduces short, standard, and expanded versions
of the agricultural module (presented inVolume
3) that are very different from the agriculture modules
used in previous LSMS surveys. In the standard and
expanded versions, information on land owned and
crops produced is gathered on a plot-by-plot basis,
MARGARET GROSH AND PAUL GLEWWE
18
rather than at the level of the whole farm as was done
in previous surveys. Information on household members’
work on their own farms is now obtained in the
agriculture module, rather than in the employment
module as in previous LSMS surveys.The short module
is new, and is limited to collecting information on
the households’ agricultural assets and on the total
amounts of each crop produced by the household.
CHAPTER 20: SAVINGS. It is difficult to collect data on
household savings because many households are reluctant
to provide savings-related information. Several
previous LSMS surveys have collected a modest
amount of data on savings, but these data have rarely
been used in analysis due to doubts about their accuracy.
Chapter 20 provides an extensive review of
research on savings in developing countries, emphasizing
the difficulties involved in doing such research.
The two versions of the draft module introduced by
this chapter (and presented in Volume 3) include several
modest improvements to the module used in previous
LSMS surveys. Neither of these versions is much
longer than the savings modules of past surveys.
CHAPTER 21: CREDIT. This chapter emphasizes that to
capture all of the sources and uses of credit in a way
natural to respondents, questions on household credit
use must be inserted in several of the survey’s modules.
Such questions should be inserted in the housing,consumption,
household enterprise, and agriculture modules,
as well as in the community questionnaire and in
a special credit module. In contrast with most past
LSMS surveys, the draft credit module introduced by
this chapter (and presented inVolume 3) gathers information
at the level of the individual rather than at the
level of the household.
CHAPTER 22:TIME USE. LSMS surveys have traditionally
included neither comprehensive measures of time
use nor modules dedicated to time use. This chapter
discusses the experience of special time-use surveys,
and uses this experience to formulate a special timeuse
module (presented inVolume 3).This module will
be of particular interest to researchers concerned with
intrahousehold issues. However, the draft module is
lengthy and could crowd other modules out of the
survey (since there is a limit to the amount of time
households are willing to be interviewed). Further
experience will be needed in implementing such a
module as part of a multi-topic survey and in analyzing
the data collected before it becomes clear whether
to routinely include time use questions in LSMS and
similar multi-topic surveys.
Part 4
This part, inVolume 2, presents four chapters that discuss
several general survey design issues.These chapters
are useful for survey designers to read alongside
the chapters in Parts 2 and 3 that interest them.
CHAPTER 23: RECOMMENDATIONS FOR COLLECTING
PANEL DATA. This chapter reviews the advantages and
disadvantages of collecting panel data in developing
countries, along with past experience of collecting
panel data.The chapter recommends that panel data be
collected in most surveys, provided that in successive
rounds the original sample of households is supplemented
with a sample of households living in
dwellings that have been constructed since the first
survey. This is necessary to ensure that the sample
remains nationally representative when each survey is
implemented.Chapter 23 also recommends that information
be collected from households in the first survey
that will help interviewers find these households
in subsequent surveys, even when it is not certain that
later surveys will attempt to collect panel data.
CHAPTER 24: INTRAHOUSEHOLD ANALYSIS. The study
of the allocation of resources and responsibilities within
households has grown in the economic literature
over the last few years, and such issues are increasingly
arising in policy discussions. This chapter explains
which kinds of data should be collected at the individual
level rather than at the household level in order
to support intrahousehold analysis; from this perspective
the chapter provides a critique of the modules
proposed in Parts II and III. For modules deemed
inadequate for intrahousehold analysis, the chapter
proposes ways to modify them so that they better support
such analysis. The chapter accepts that it is not
feasible to collect individual-level data on all topics in
an LSMS survey. Nevertheless, future LSMS surveys
can be designed to support substantial intrahousehold
analysis. Much of the data collected in past LSMS
surveys—on employment, health, education, anthropometrics,
migration, and fertility—have long been
collected at the individual level.And the draft agriculture,
household enterprise, credit, and miscellaneous
CHAPTER 1 INTRODUCTION
19
income modules presented in Parts II and III of this
book recommend the collection of more individuallevel
data than were collected in previous LSMS surveys.
In addition, the draft time-use module introduced
in Chapter 22 is a new tool for gathering data
that are crucial for intrahousehold analysis.
CHAPTER 25: QUALITATIVE DATA COLLECTION
TECHNIQUES. Previous LSMS surveys have focused
almost exclusively on collecting quantitative data,
making very little use of qualitative data collection
methods. This regrettable tendency probably reflects
the quantitative backgrounds of most survey designers.
Chapter 25 explains ways in which qualitative
research methods can usefully and effectively complement
quantitative data collection. The chapter concludes
that qualitative methods should not be combined
with quantitative methods into a single survey;
instead, both methods should be used in separate but
complementary data collection exercises. Quantitative
surveys can benefit from qualitative methods in several
ways. For example, qualitative research can be used
to help survey designers formulate the exact wording
of particular questions, and qualitative methods are
useful for creating hypotheses about household
behavior, which can then be tested using quantitative
data.
CHAPTER 26: BASIC ECONOMIC MODELS AND
ECONOMETRIC TOOLS. This chapter gives non-economists
some basic information on economic models of
household behavior, and reviews econometric methods
commonly used in analyzing the policy questions
discussed in this book. Chapter 26 is a useful reference
for non-economists as they read other chapters. A
glossary at the end of Chapter 26 defines the economic
and econometric terms used in many chapters
of the book.
Note
The authors would like to express their gratitude to Jere Behrman,
Lawrence Haddad, Courtney Harold, John Hoddinott, Alberto
Martini, Raylynn Oliver, Kinnon Scott, and Diane Steele for comments
on previous drafts of this chapter.
References
Davies, David. 1795. The Case of Labourers in Husbandry. Bath,
United Kingdom: Printed by R. Cruttwell for G. G. and J.
Robinson.
Deaton, Angus. 1997. The Analysis of Household Surveys: A Micro
Econometric Approach to Development Policy. Baltimore, Md.:
Johns Hopkins University Press.
Deaton, Angus, and Salman Zaidi. 1999. “Guidelines for
Constructing Consumption Aggregates for Welfare Analysis”
World Bank, Development Research Group,Washington, D.C.
Eden, Frederick. 1797. The State of the Poor. London: Davis.
Grosh, Margaret, and Juan Muñoz. 1996. A Manual for Planning and
Implementing the Living Standard Measurement Study Survey.
Living Standards Measurement Study Working Paper 126.
Washington, D.C.:World Bank.
Ravallion, Martin, and Shaohua Chen. 1997. “What Can New
Survey Data Tell Us about Changes in Distribution and
Poverty?” World Bank Economic Review 11 (2): 357–82
———. 1998. Personal conversation updating Ravallion and Chen
1997. November 2 and 3.Washington, D.C.
Ravallion, Martin, Guarav Datt, and Dominique van de Walle.
1991. “Quantifying Absolute Poverty in the Developing
World.” Review of Income and Wealth Series 37 (4).
1991. “Quantifying Absolute Poverty in the Developing
World.” Review of Income and Wealth Series 37 (4).
21
2
Making Decisions on the Overall Design
of the Survey
Margaret Grosh and Paul Glewwe
Comprehensive, multitopic household surveys such as LSMS surveys usually consist of three separate
questionnaires: a household questionnaire, a community questionnaire, and a price questionnaire.
Each questionnaire is composed of modules, sections that collect information on a specific
topic. Questionnaires and their modules can be combined in a variety of different ways to create
a multitopic household survey.There is no single right way to combine modules and questionnaires
into a survey; each way has advantages and disadvantages.The key is to choose a design
that provides the best fit given the objectives of, and constraints on, the proposed survey.
The starting point for designing the modules and
questionnaires to be used in a multitopic survey is a set
of policy questions.The overall objective of each survey
is to collect the data needed to answer these questions.
There are five steps involved in survey design.
The first step is to define the fundamental objectives
of the survey and to choose an overall survey design
that best fits these objectives.This is usually done not
by a single individual but by a team of survey designers
who consult extensively with a broad range of
individuals and organizations interested in the survey.
In choosing the overall design,the team must take into
account several important factors including the capacity
for collecting data within the country, the funding
available, and the amount and quality of data available
from other sources.
The second step involves deciding which modules
to include in the survey, specifying the objectives of
each module,and proposing an approximate length for
the modules.This step is needed because a survey that
attempts to include all possible modules will be too
large and complex to implement.
The third step is to work out, question by question,
draft questionnaires for each module that will be
included in the survey.This can be done by drawing
on the detailed recommendations in the chapters in
Parts 2 and 3 of this book, as well as the draft modules
included in Volume 3.The fourth step is to compare
the modules to each other to ensure that they are consistent
and well integrated, and to combine them into
draft household, community, and price questionnaires
(in some cases omitting the community questionnaire).The
fifth and final step is to translate and field
test the draft questionnaires. Translation may not be
necessary in some countries; field testing is always
essential and must not be done quickly or superficially.The
first two steps are discussed in this chapter.The
third, fourth, and fifth steps are discussed in Chapter 3.
While these five steps should ideally be done in
the order given above, in reality there is likely to be
substantial movement backward and forward among
the various steps.Some objectives originally set out for
the survey may prove impossible to achieve given the
constraints. And discussion of the detailed objectives
MARGARET GROSH AND PAUL GLEWWE
22
for each module may cause the survey design team to
reassess the overall objectives of the survey. In other
words, it may be necessary to take one or two steps
backward at some points in order to continue to move
forward.This is to be expected and even encouraged.
As more is learned about what can and cannot be
done, survey designers are more likely to produce a
survey design that meets their objectives—which may
also have become more realistic. It is better to pare
down the number of objectives in order to achieve
some of them than to attempt to do too much and, as
a result, achieve few or none of the original objectives.
The first four sections of this chapter cover the
first step of the survey design process.The first section
provides an overview of who should be involved in
designing and assembling the questionnaires.The second
section discusses the main factors that survey
designers should take into account when choosing
among survey design options. The third section outlines
“core” elements that must be included in any
LSMS or similar multitopic survey and reviews several
classic survey designs, each of which supplements
the core in a different way.The fourth section presents
guidelines for choosing the survey design most appropriate
for each of a range of different circumstances.
The fifth and final section explains the second step
of the survey process: choosing the modules to include
in the survey, setting objectives for each module, and
setting the approximate length of the modules.
Organizing a Survey Design Team
The most important factor ensuring the success of a
multitopic household survey is the involvement of the
right people in the process. Designing the survey
questionnaires is much too large a task for one person.
Instead, a team of experts must be involved, including
members of the organization implementing the survey
as well as research analysts from other institutions. If
the team does not contain a sufficient diversity of
experts, this can have negative repercussions for the
data (Box 2.1).The design team must work together
with policymakers and program managers to define
the overall objectives of the survey and to settle on
many details at each step of the survey design process.
Researchers and Policy Analysts
It is essential to involve researchers and policy analysts
in questionnaire design.This book was written primarily
by researchers and policy analysts,and much of the
success of past LSMS surveys in supporting policy-relevant
research is due to the fact that the surveys used
questionnaires designed by people who would be
actively involved in the analysis of the data.
Researchers and policy analysts can ensure that the
information collected in multitopic surveys is well
suited for policy research.
The lead role in designing the questionnaires of
an LSMS or similar multitopic household survey
should be given to a small group of researchers and
policy analysts who share two characteristics. First,
they should know what issues are of most concern to
the country’s policymakers. Second, they should have
experience in using data from similar surveys to analyze
these issues.The group of researchers and policy
analysts should include members of the national planning
agency, representatives from the national statistics
agency, local academic researchers, and one or more
people who have helped analyze or design multitopic
surveys in other countries.
The team must include individuals with extensive
experience in implementing and analyzing other
Box 2.1 The Importance of a Well-Rounded Design
Team
An effective survey design team must include researchers
and policy analysts, policymakers, and staff from the organization
implementing the survey.The problems that can arise
when one or more of these groups is not involved in
designing the questionnaires are illustrated by the experience
of an LSMS survey implemented in Jamaica. The
household questionnaire for the first Jamaican Survey of
Living Conditions (implemented in 1988) was designed primarily
by international experts who had little knowledge of
Jamaican social programs. Although the household questionnaire
was largely successful in accomplishing its analytical
objectives, it had two serious flaws. First, although food
subsidy policies were an important issue at the time, the
consumption module did not clearly distinguish expenditures
on key subsidized food items from expenditures on
similar nonsubsidized items. This made it more difficult to
study the incidence of food subsidies. Second, the questionnaire
asked respondents about their receipt of food
stamps during the previous month even though food
stamps were provided only every two months.This made it
difficult to identify which households had received food
stamps, thus hindering the study of another issue important
at the time. Fortunately, these flaws were identified and corrected
in the following year’s household questionnaire.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
23
household surveys in the country in question. Ideally,
local researchers and policy analysts should take primary
responsibility for designing the survey, because
they have an intimate knowledge of the country’s culture,
economy, and society, and they are very familiar
with existing programs and key policy issues. Local
researchers and policy analysts are also likely to know
about previous surveys done in the country that have
covered some of the topics included in the new survey.
And they will know which people and institutions
should be consulted during the survey design process.
It may also be desirable to involve international
researchers in the design of the questionnaire, especially
in countries where local data analysts are not
familiar with LSMS and other multitopic household
surveys. International researchers can contribute their
experience about what has and has not worked in surveys
in other countries.1
Judicious use of the advice of
both local and foreign experts will significantly
improve survey design.
Past LSMS surveys have probably made insufficient
use of the knowledge available from local
researchers and policy analysts.Too often the involvement
of local professionals has been limited to statisticians
from the statistical agency (data producers) and
thus failed to draw on the expertise of social policy
researchers from the government or academia (data
users). Statisticians may have only a limited knowledge
of sectoral policy issues and programs.While they do
have an important role to play, their input must be
combined with input from data users to set priorities
among the different possible objectives for policy
research.
Policymakers
When defining the fundamental objectives of the survey,
the team responsible for drafting questionnaires
must seek extensive input from policymakers and program
managers in the country being surveyed. The
team’s initial discussions with policymakers should
focus in broad terms on the most important issues to
be covered, which will determine the relative size of
the different modules in the questionnaires. After this
round of discussions, further discussions should be
held to identify the important issues within each sector.
Since drafting the module or modules for each
sector requires a substantial amount of knowledge
about how specific programs work, technical experts
in many program agencies must be consulted. These
people should be consulted before the modules are
created, and they should also be shown draft modules
to elicit comments.
Unfortunately, in many previous LSMS surveys
the survey design team did not give enough attention
to communicating or consulting with policymakers.
Policymakers, who are often unfamiliar with household
surveys, may find it difficult to read complicated
questionnaires or to imagine what analyses the resulting
data could support. One option is to show policymakers
and program managers examples of the kinds
of tables and analyses that could be produced using
data from the questionnaire; these might be either
hypothetical tables for the country of the survey or
tables made in other countries using data from similar
surveys. Another strategy is to show policymakers a
report based on the first year’s data; this is an excellent
way to obtain policymakers’ feedback on the design of
follow-up surveys.A third strategy is simply to ask policymakers
what they need to know to implement
effective policies.
Data Producers
It is critical that the survey design team include staff
from the organization implementing the survey.This
should ensure that the questionnaires designed are
workable. Often data collection can be greatly simplified
by making minor changes in the layout or
flow of a questionnaire, changes that do not diminish
the questionnaire’s analytical content. Data producers
are an excellent source of suggestions for such
changes.They are usually also experienced in details
of designing a questionnaire, such as questionnaire
formatting. For all of the above reasons, the team
members from the organization implementing the
survey should help design, or comment on, every
draft of the questionnaire.
It is also useful for the survey design team to solicit
the input of experienced field supervisors, who will
notice whether the instructions to the interviewer are
clear, whether the skip codes are correct, and whether
the format is consistent. There is of course a natural
tension between data analysts, who want comprehensive
information, and field supervisors, who are likely
to see all of the disadvantages but few of the advantages
of administering a lengthy, complex questionnaire.
Each side must be prepared to make compromises
and carefully listen to the other side’s point of
view.
MARGARET GROSH AND PAUL GLEWWE
24
Factors for Deciding among Various Survey
Designs
After the members of the survey design team have
been selected,work can begin on designing the survey.
The first task for team members is to review the factors
that influence the overall design of the survey.
This section discusses those factors in detail.
The appropriate design of a household survey or
sequence of household surveys differs from country to
country. The most important factors for determining
the design of a proposed survey are: the kinds of policy
issues the survey aims to address; the information
available from existing surveys and other data sources;
the country’s institutional capacity for collecting data;
and the financial and other resources available for
implementing the survey, including any constraints on
how these resources can be used.
Policy Issues
The design of a household survey should reflect the
policy issues it is intended to address. One way to classify
policy issues is in terms of their subject matter,
such as health, education, employment, or housing.
Another way to classify policy issues is in terms of the
kinds of data used to address them. The four most
common kinds of household survey analysis used to
address policy issues are: simple descriptive statistics on
living standards; monitoring poverty and living standards
over time; describing the incidence and coverage
of government programs; and measuring the impacts
of policies and programs on household behavior and
welfare. This subsection reviews these four types of
analysis and provides a practical example of how the
information needed affects the design of the survey.
SIMPLE DESCRIPTIONS OF LIVING STANDARDS. The
most straightforward objective for a household survey
is to describe the living standards of the population at
one point in time, often with particular emphasis on
the living standards of the poor.This can be done by
using the data to tabulate means and frequencies of
key variables.The results of these tabulations are often
disseminated by the national statistical agency in the
form of statistical abstracts (reports) that contain a
large number of tables and a minimal amount of
descriptive text. It is also possible to produce more
structured descriptive analyses that supplement household
survey data with information from other sources.
Structural analysis of descriptive data can sometimes
be used to draw conclusions about the likely impact of
government policies on living standards. Examples of
such analyses are the “poverty profiles” typically provided
in the World Bank’s poverty reports.
In both types of descriptive analysis, the range of
variables used to measure living standards can vary
widely; variables may be used from virtually all of the
survey modules or from only a small subset of modules.
In general, most of the variables included in statistical
abstracts and descriptive analyses come straight
from the questionnaire (for example, percentage of
households that have electricity) or require only a
small amount of manipulation (for example, nutritional
status as derived from weight and height data). Only
one “complex” variable needs to be constructed: total
household consumption. Other complex constructed
variables, such as total income or net wealth, are used
less often in simple descriptive presentations.
MONITORING POVERTY AND LIVING STANDARDS. The
descriptive analyses discussed above focus on living
standards at one point in time. However, another
important role of multitopic household surveys is to
monitor how living standards and poverty change over
time.When data are used for this purpose they must
be comparable over time; for this to be the case, the
data must be gathered using the same methods each
time the survey is implemented. One aspect of such
consistency concerns the design of the sample, which
in each case must use the same definitions of basic
concepts such as the distinction between urban and
rural areas.A second requirement for comparability is
that the questions defining variables of interest must
remain the same each time the survey is administered.
This is necessary because seemingly innocuous
changes in the wording of questions can lead to serious
comparability problems; changing the recall period
for food expenditures can make it impossible to
compare estimates of poverty and inequality over
time.
Another issue to consider when monitoring
poverty and living standards over time is the frequency
with which indicators of living standards must be
monitored. Indicators that are fairly stable over short
periods of time—such as fertility and adult literacy—
need not be measured each time the survey is done.
However, indicators that can change more quickly,
such as consumption expenditure,children’s nutrition-
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
25
al status, and employment status, should be measured
every time the survey is implemented. Surveys that
monitor poverty and living standards over time are
typically fielded every year, although it is also possible
to field them biannually or semiannually.
EXAMINING THE INCIDENCE AND COVERAGE OF
GOVERNMENT PROGRAMS. Data from multitopic
household surveys can also be used to measure the
incidence and coverage of specific government programs.
For example, data on the enrollment of household
members in public schools are useful for investigating
which children benefit from the provision of
public schooling. Household survey data can also be
used to study participation in government assistance
programs such as food stamps, cash assistance, and
school meals. Another example is descriptive statistics
on purchases of subsidized food items, which can be
used to examine whether the benefits of specified
food subsidies vary by households’ levels of income.
The incidence and coverage of these different kinds of
programs are easy to calculate and useful for policymakers
to know.
A moderate sample size (2,000 to 5,000 households)
should be sufficient to evaluate programs that
affect a large proportion of the population. Evaluating
programs that serve a small proportion of the population
usually requires using a much larger sample of
households or including a disproportionately large
number of target and beneficiary households in the
sample.
ESTIMATING THE IMPACT OF POLICIES ON HOUSEHOLD
BEHAVIOR AND WELFARE. Policymakers are often faced
with questions that can be answered only by analyzing
household behavior. Policymakers may want to know
how changes in commodity taxes or subsidies would
affect agricultural production or the consumption of
basic food items. Answering such questions requires
calculating price elasticities and thus modeling households’
production and consumption decisions. Such
modeling requires data that go well beyond measurement
of living standards indicators.
In multitopic household surveys that attempt to
model household behavior, each module that collects
data on a behavior of interest is usually designed to
gather information that can be used to estimate the
impact of several different policy changes.The chapters
in Parts 2 and 3 of this book discuss each module
in great detail and provide many examples of issues
that require the modeling of household behavior.The
following questions give an idea of the range of policy
issues that can be addressed:What is the impact of
charging user fees at government health clinics on the
use of those clinics by adults and by children? How
can the government encourage parents to enroll their
children in school? What are the impacts of women’s
employment opportunities on their fertility? How do
changes in prices brought about by structural adjustment
programs affect the welfare and productivity of
agricultural households?
EXAMPLE SCENARIO:THE INCLUSION OR EXCLUSION OF
THE ANTHROPOMETRY MODULE. The decision about
including an anthropometry module demonstrates
how the analytical potential of a multitopic household
survey is related to its content. (See Chapter 10 for
detailed information on the collection of anthropometric
data.) If an anthropometry module is not
included, the survey is not useful for studying nutrition
issues. However, by collecting limited anthropometric
data such as the height and weight of children
under five years of age, the survey will allow analysts
to describe the extent and patterns of malnutrition in
early childhood. If the country studied has adopted
large-scale food distribution or subsidy programs, the
data can also be used to assess how well these programs
are targeted to undernourished children. If the survey
is repeated annually or biannually, it becomes possible
to monitor changes in both the nutritional status of
the population and the targeting of the program.Thus
the data collected from a limited anthropometry module
can address, at least partially, three of the four types
of policy questions outlined above.
A full version of the anthropometry module
would collect data on height and weight for all household
members, not just children. Such data could be
used not only to gauge adult health but also to analyze
the impact of government policies on household welfare
and behavior. Suppose policymakers want to predict
the impact of food programs on children’s nutritional
status.This requires estimating the determinants
of child weight and height. Because heredity is so
important, parental height and weight information are
needed to estimate these relationships accurately; lack
of data on parents’ weight and height could lead to
estimates that suffer from omitted variable bias and
thus do not accurately show the impact of the food
MARGARET GROSH AND PAUL GLEWWE
26
programs on children’s nutritional status. In general,
including the full version of the anthropometry module
in the survey—measuring both adults and
children—greatly expands the possibilities for examining
the impact of government policies on household
behavior and living standards.
Defining the objectives of a survey is often a less
tidy process than the discussion so far has implied.
Institutions, and people within institutions, may have
varying objectives. Each sectoral ministry in a country
is likely to be primarily interested in its own subject.
The government as a whole may want the surveys to
measure or monitor only a few indicators of welfare,
while academics in the country’s universities and other
research institutions may want the surveys to yield the
detailed data needed to model household behavior. If
international agencies are financing the survey, they
may have still another set of objectives. For example,
they may wish to ensure that the data are comparable
with similar data from other countries or that the data
can be used to study issues of interest to the development
community in general, even if these issues are
not a high priority in the country of the survey.
Whatever the objectives envisaged when the survey is
first designed, it is likely that researchers will later use
the data for other analytical purposes.
The multiple (and sometimes competing) objectives
of household surveys are to be expected and even
encouraged, since each of the groups with a stake in
these surveys has its own legitimate priorities.The task
of survey designers is to accommodate the different
objectives as much as possible without compromising
the quality of the survey.
Other Information Available and Its Relation to Survey
Objectives
No household survey takes place in a vacuum. In most
countries there are several other household surveys
that have gathered or will gather information on issues
that the new multitopic survey is intended to cover.
The extent to which data from these sources influence
the design of the new questionnaire depends on the
amount and type of data available and on the objectives
of the new survey.
If the main objective of the new survey is to
describe various aspects of the living standards of the
population,it may seem that the topics already covered
in other surveys need not be included in the new multitopic
survey. For example, if the only goal of the new
survey is to describe living standards and recent
anthropometric data are already available from another
survey, it may seem reasonable to drop anthropometric
measurements from the new survey. However,
there are two important advantages to collecting
anthropometric data in the new survey. First, collecting
these data would make it possible to produce
descriptive tables that show simple relationships
between nutritional status, as revealed by anthropometric
measurements, and other variables of interest—
for example, household expenditure levels. Second,
collecting anthropometric data in the new survey
would ensure that the anthropometric data used the
same definitions and classification schemes as other
survey data, and thus could be used to draw effective
comparisons. If the two surveys classified, say, education
levels or rural and urban areas differently, this
would make it difficult to present analyses from the
two surveys side by side in ways that would be simple
to interpret.Analysis based on combining results from
separate surveys will usually be more difficult, and thus
more prone to error, than analysis based on data that
have all been collected in a single survey.
The case for collecting anthropometric data is
even stronger if the purpose of a new survey is to
investigate the impact of nutritional status on other
socioeconomic outcomes (such as education, fertility,
or labor force productivity).This objective implies that
the survey must include an anthropometry module,
even if recent information on nutritional status is
available from other sources.To conduct these kinds of
analyses, the variables of interest must all come from
the same household survey.2
Although it is essential that data on key
household-level variables come from the same households,
it is often useful to supplement household survey
information using data from a source other than
a multitopic survey. In some cases, price data collected
for generating consumer price indices can replace
the price questionnaire typically used in LSMS and
other multitopic household surveys (see Chapter 13).
Other such alternative data sources are time-series
data on weather and maps of soil quality and topography,
all of which can be used for analyzing agricultural
issues. In health and education, further possibilities
arise for matching household survey data with
data from other sources; some countries collect data
from health clinics and schools that may be matched
with the communities covered in a household survey.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
27
However, survey designers should exercise caution
when contemplating this approach. Although matching
data from different sources appears simple, it is
often very difficult in practice. Many of the chapters
in Parts 2 and 3 of this book discuss the potential for
matching household survey data to data from other
sources.
An important question that often arises when
planning a multitopic survey is whether such a survey
can replace one or more existing surveys and thereby
reduce total costs without any loss of information. A
multitopic survey with an anthropometric module
could replace a periodic anthropometric survey primarily
intended to measure the extent of malnutrition.
However, a multitopic survey cannot replace all other
household surveys. Labor Force Surveys often require
much larger sample sizes and more frequent data collection
than would be appropriate for a multitopic survey.And
specialized surveys, such as Demographic and
Health Surveys and comprehensive farm management
surveys, contain much more data on those topics than
can usually be collected in LSMS and similar multitopic
surveys. Still other surveys, such as farm surveys
and small business surveys, have very different sampling
frames since they are based on samples of farms or
businesses rather than samples of households.
A final issue is whether survey designers should
implement an entirely new survey, modify an ongoing
survey, or find creative ways to analyze existing data.
Two arguments support implementing a new survey.
First, past surveys may not have been adequately documented,
or access to their data may be restricted.
Second, inter-agency rivalry, arguments concerning
ownership of survey data, and coordination problems
when different surveys are carried out by different
agencies may make it easier to begin a new survey
instead of using existing data or modifying an existing
survey. On the other hand, survey designers should at
least consider trying to remedy these problems so that
existing surveys can be used (perhaps with some modifications)
to meet the designers’ data and policy
objectives—thus avoiding any unnecessary duplication.
Examples are given later in this chapter of countries
in which an existing survey was modified to be
more like an LSMS survey.
Institutional Capacity
Decisions about what kind of survey to implement
also depend in part on the institutional capacity for
collecting data in the country undertaking the survey.
Because maintaining data quality becomes more difficult
when surveys become more complex, this capacity
should be carefully assessed when planning LSMS
or similar multitopic surveys, which can be very complicated.
In countries where the capacity to collect
data is weak, it may be better to implement a limited
multitopic survey yielding reliable data on a relatively
small number of topics than an overly ambitious survey
that could yield unreliable data on a wide variety
of topics.
A survey containing 10 modules is easier to plan
and implement than a survey containing 15 or 20
modules.The fewer the modules, the less time is needed
by survey planners to contact different sectoral
agencies and thus the less time is needed to build consensus.Also,
smaller questionnaires require less time to
design, and less time to carry out the fieldwork, enter
the data, and manage the database. However, other
steps in developing and implementing a household
survey, such as planning the sample design, do not vary
with the size of the questionnaire.Therefore, a survey
with a questionnaire half the size of the questionnaire
for a full multitopic household survey will involve
substantially more than half the effort required for a
full survey.
Despite the complexities of full multitopic surveys,
some very successful multitopic surveys have
been carried out in countries with very limited data
collection capacity. Several steps can be taken to overcome
the problems posed by limited capacity. For
LSMS surveys, international experts have been
brought in to draw the sample,draft the questionnaires
and interviewer manuals,and write the data entry program.
Such experts initially substitute for government
agency staff, but they can also train agency staff to take
their place in future surveys. It is highly recommended
that countries with limited capacity for collecting
survey data use such expert assistance.
In countries with weak institutional capacity, serious
consideration should also be given to improving
that capacity; capacity building yields long-term benefits
that gradually reduce the need to use international
experts to help with data collection. Genuine
capacity building takes time, money, and political and
managerial effort. An international sampling expert
may be able to design and draw a sample for a survey
in a few days, but it will usually take him or her much
longer to teach local staff how to do so. Training to
MARGARET GROSH AND PAUL GLEWWE
28
build capacity requires significant resources beyond
those already budgeted for a survey.Whether building
a country’s data collection capacity is important
enough to warrant committing these resources will
vary from country to country.Where capacity building
is deemed necessary, the survey’s work plan and
budget must both be significantly enlarged.
If capacity building is a goal, a program of annual
(or biannual) surveys will work better than a program
for a single survey or for a sequence of surveys that
take place every three to five years. An annual survey
usually has a permanent allocation of skilled staff, staff
time, and equipment. Even when the team works only
part of the year on the multitopic survey, staff have a
chance every year (or every two years) to use the skills
that they have acquired in managing such a survey.
And as the staff of the agency develop survey management
skills, the need for technical assistance from
international experts should diminish. When some
staff members working on the survey leave, their
replacements can learn their jobs from other staff
members who have worked on earlier rounds of the
survey. In addition, the continuity provided by an
annual survey may make it easier to improve survey
quality; if one year a problem arises in data collection
or initial analysis, the people who deal with the problem
are likely to be involved in planning the next survey
and can better address the problem in the next
survey.
In contrast, a survey carried out every four or five
years may require new skills, staff, and equipment each
time it is implemented. In the intervening period,
many of the individuals who carried out the first survey
may have moved on to other jobs either inside or
outside the statistical agency.Those who remain may
not have been involved in planning the previous survey,
and the skills of those who were involved may
have deteriorated over time. Vehicles and computers
used in the first survey will have been allocated to
other purposes, and some may have ceased to function
altogether. Most importantly, much of the institutional
memory about problems and potential solutions
may have been lost.
A final note of caution is needed regarding institutional
capacity. Sometimes, even when a statistical
agency has sufficient management and technical
capacity to implement a complex multitopic survey,
there may not be enough experienced supervisors,
interviewers, or data entry operators. Lack of data
entry operators is not a serious problem since they can
be trained in a matter of weeks, and no previous experience
is required. However, it takes longer to transform
government staff with no household survey
experience into competent interviewers.While interviewers
may be trained in a month, it is not so easy to
compensate for little or no interviewing experience.
Experience is even more important in the case of
supervisors. It may take years to overcome shortages of
experienced interviewers and supervisors.
Constraints Imposed by Funding Sources
Surveys are always constrained by their funding. Most
LSMS and similar multitopic household surveys
receive some portion of their financing from sources
other than the national budget, at least initially. As a
result, they are subject to constraints associated with
both the national budget and funds from external
sources.3
The first and most obvious constraint imposed by
the source of funding is the total amount of funds
available. National budgets are often very restricted.
Some external funding sources have upper limits for
how much may be spent on a single project, and most
have administrative procedures that grow in complexity
as the size of a project increases.Also, the larger the
survey budget, the more difficult it is for survey planners
to justify using the money for the proposed survey
rather than for some other purpose.Limitations on
the size of the budget often constrain the size of the
sample used in the survey and in some cases curtail the
survey’s analytical depth and breadth.
Another potential constraint relates to the time
period over which funds may be spent.Funding agencies
may stipulate that a survey project be completed
in only one or two years, even though a single fullscale
survey can easily take three years or more to
complete—6 to 18 months to plan, a year for fieldwork,
and 6 to 18 months for data dissemination and
analysis. Moreover, chances of obtaining future funding
can influence whether a proposed survey is carried
out only once or is the first of a series of surveys.
And funding limitations can affect such other aspects
of the survey as the thoroughness of the survey
designers’ work during the planning stage, whether
the fieldwork is spread over a full year or concentrated
into a period of a few weeks, and the amount of
analytical work funded from the survey project’s
budget.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
29
Finally, many funding agencies also have rules on
how survey funds can be spent.These rules may impose
controls on: the percentage of funds spent in local or
international currency; the balance between recurrent
and investment costs; the amounts that can be spent on
the salaries of local staff, survey equipment, and payment
of international experts; the nationalities of such
experts; and various aspects of budgeting, accounting,
and procurement. Spending rules rarely influence big
issues of survey design (such as survey duration, sample
size, or questionnaire design), but they can affect many
details of the structure and implementation of a project.
Rules that prohibit the shifting of expenditures
between items or between time periods may limit the
ability of survey planners to deal with unanticipated
problems. For example, an additional international
expert may be needed quickly, but may be difficult or
impossible to obtain because hiring this expert was not
included in the original budget.The end result can be
a delay in the survey or a reduction in quality.Another
example is if an accident occurs involving a survey
vehicle; fieldwork may be delayed if expenditures to
repair or replace the vehicle cannot be made available
promptly.
Summary
The analytical objectives of a survey, the availability of
information from other sources, local institutional
capacity, and constraints imposed by funding are all
key factors that typically affect whether to perform a
new survey and the form such a survey will take.
Many other factors are also critical, including institutional
inertia and rivalry and the compromises
required to build a coalition to support and conduct a
survey. However, it is difficult to provide general
advice because these factors usually depend on the setting
of the survey; survey planners must deal with
these issues as best they can given the particular circumstances
they face.
Classic Survey Designs
There are three basic ways to combine modules into
questionnaires and combine those questionnaires into
a survey or sequence of surveys: the full LSMS-type
multitopic survey, the scaled-down LSMS-type survey,
and the core and rotating module survey. All of these
survey formats must include certain “core” components.
This section outlines the core survey components,discusses
the strengths and weaknesses of each of
the three main survey types, and describes two other
survey options.
The Core
Any LSMS-type multitopic survey must collect certain
essential information about the household, its
members, and the local community, including:
• A roster that lists, and collects basic information
about, all household members.
• Detailed information on household consumption
expenditures.
• Basic housing data such as type of dwelling, water
source, type of toilet, and whether the dwelling has
electricity.
• The education of all household members,including
who is currently in school.
• The employment status of everyone of working age
and, for those who are working, their occupation,
the number of hours they worked during the previous
seven days, and, if they are employees, their
wage earnings.
• The receipt of money or in-kind assistance from
key government or NGO programs.
• The use of social services and programs, such as
government health facilities, schools, agricultural
extension services, and social assistance programs.
• Basic information related to the design of the sample
and the outcome of the household interview.
• Local prices of basic food and nonfood goods (unless
price data are available from another source, or the
country is so small and its markets so well integrated
that there is very little regional price variation).
These components are referred to in this book as
the essential core. In addition to the essential core, it is
highly recommended that the following five types of
information be collected in LSMS and similar multitopic
surveys:
• Anthropometric measurements (height and weight)
of children 0–5 years old (unless malnutrition is
known to be negligible in the country).
• The immunization status of children 0–5 years old.
• Information on basic household assets such as
durable goods, housing, land, and the capital equipment
used for agricultural activities and nonagricultural
household enterprises.
• Information on interhousehold transfers.
• Information on rental payments for those households
that rent their dwellings.
MARGARET GROSH AND PAUL GLEWWE
30
In this book the set of modules formed by adding
these five components to the essential core is referred
to as the recommended core.
The essential core of an LSMS or similar multitopic
survey collects the information needed to
describe poverty and to monitor it over time.The recommended
core adds some very basic child health
information, along with information on assets, interhousehold
transfers, and rental payments (the use of
which will be explained below). Judgments about
which data are part of the essential and recommended
cores are based on many years of experience that
World Bank staff have in using data from LSMS surveys
to produce poverty profiles for a wide range of
developing countries.Table 2.1 lists the components of
both the essential and recommended cores of LSMStype
multitopic surveys. The paragraphs that follow
describe each of these components in greater detail.
HOUSEHOLD ROSTER (ESSENTIAL). Virtually every
household survey should begin by determining how
many people belong to each household and collecting
very basic information on each household member,
including age (or date of birth), sex, nationality, relationship
to the head of household, and marital status.
Part A of the household roster introduced in Chapter
6 (and provided inVolume 3) collects such basic information,
along with another piece of information that
is less essential: questions that link each married individual
to his or her spouse.
CONSUMPTION EXPENDITURES (ESSENTIAL). The experience
of LSMS surveys and other household surveys
strongly suggests that household consumption expenditures
are the single most important indicator of
household welfare that can be obtained from a household
survey. (See Chapters 5 and 17 for further discussion
on this point.) Chapter 5 describes how to collect
data on consumption expenditures, stressing that there
are no costless shortcuts for collecting such data. In
some circumstances it might be possible to omit questions
on the ownership of durable goods and on transfers
given to other households, but the rest of the consumption
module is an essential part of the core and
should not be reduced further. Data on household
expenditures on education, health, and housing are
collected in the core elements of those modules (discussed
below) and need not be included in the consumption
module. Consumption in the form of inkind
payments (such as meals, clothing or
transportation) from employers is best collected in the
employment module.
Table 2.1 The Essential and Recommended Cores of LSMS-Type Multitopic Surveys
Module Sections used
The Essential Core
Household Roster All of Part A except questions 8 and 9
Consumption All questions except transfers given to other households (Part D) and ownership of durable goods (Part E)
Housing Questions A1–A7, B1–B5, C1–C3, and C13–24 of the short module
Education All questions in the short module
Employment Questions A2–A13, B1, B2, B7-B10, D3, D4 and D8–D17
Transfers and Other Nonlabor Income All of Part B1; see text for further discussion
Health Questions 10–38 of the short module
Metadata Household Identification and Control submodule; Questions 1–4 in Summary of Visits and Interviews
submodule
Prices 30–40 food items and 10–20 nonfood items
Credit Questions 9–14 and 21–28 of the short module (on credit obtained from NGOs or government agencies)
Agriculture All of Part F (use of agricultural extension services), which is the same for all modules
Additional components for the recommended core
Anthropometry Entire module, for children 0–5 years old
Health All of Part C (immunization)
Consumption All of Parts D and E
Housing Questions C7–C12 of the short module
Household Enterprises Part G of the short module, questions 1–3
Agriculture Parts A, B, and E of the short module.
Transfers and Other Nonlabor Income Questions on income from interhousehold transfers
Source: Authors’ recommendations.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
31
HOUSING (ESSENTIAL). Information on housing,
including the type of dwelling, the construction materials
used, the number of rooms, the availability of
electricity, the source of drinking water, and the type
of toilet, are very basic indicators of living standards.
They provide analysts and policymakers with information
on a household’s standard of living that goes
beyond consumption expenditures. Because housing
information is simple to collect, it should be included
in any LSMS-type survey. The long version of the
housing module introduced by Chapter 12 (and presented
in Volume 3) collects substantially more housing
data than are necessary for the essential core, and
even the short version is longer than the essential core.
Only the following questions from the short version
of the housing module need to be included in the
essential core: A1–A7, B1–B5, and B11–B21. Even
some of these questions can be omitted in some countries.
The questions on heating (B18–B21) can be
removed for countries with warm climates, and the
questions that distinguish between wet and dry seasons
(B1–B3) can be simplified in countries where this distinction
is not important.
Another part of the core data set, housing-related
consumption expenditures (such as expenditures on
electricity, water, and cooking fuel), are most conveniently
collected in the housing module rather than in
the consumption module. A final useful indicator of
living standards is information on the ownership of
the household’s dwelling. Questions C1–C3 in the
short version of the housing module collect expenditure
information, and questions C13–C24 collect
ownership information.
EDUCATION (ESSENTIAL). Education is both a determinant
and a key indicator of living standards.The short
version of the education module introduced in
Chapter 7 (and presented in Volume 3) comprises all
of the essential core questions on education.The only
questions that might be omitted are the two questions
on grade repetition.
The short education module assesses education
from several different angles, including school attainment,
current enrollment, and education expenditures.
Information on the school attainment of household
members ages 5 and older is easy to collect and has
many analytical uses (such as classifying households in
terms of the education level of their head of the household).
School enrollment among children, another key
indicator of living standards, is also easy to collect.And
it is usually more convenient to collect information on
household education expenditures in the education
module than in the consumption module.
EMPLOYMENT (ESSENTIAL). Basic employment information
on household members of working age (7 and
older in many countries) should be collected as part of
the essential core of any LSMS-type survey.The most
important source of income for poor people in developing
countries is their labor; employment data,
including information on unemployment, indicate
how this labor is being used.
Essential employment information includes each
person’s occupation and the number of hours that he
or she has worked in the previous seven days.While it
would also be useful to gather data on the incomes of
all employed household members, this is not easily
done for the self-employed (see Chapter 17 for further
discussion). However, income data should still be collected
for employees even when such data cannot be
collected for the self-employed, for two reasons. First,
these partial data are useful for understanding which
occupations pay well and which do not. Second, since
data are already needed from employees on in-kind
benefits provided by their employers (in order to calculate
consumption expenditures), it would seem
strange to ask about those benefits without first asking
about money income.
The short version of the employment module
introduced by Chapter 9 (and presented inVolume 3)
collects more information than the core of an LSMStype
survey requires.The following questions from the
draft employment module constitute the essential
core:A2–A13, B1, B2, B7–B10, D3, D4, and D8–D17.
Job-specific information—questions in Parts B or D—
should be collected both for the person’s main occupation
and for any secondary occupation. (The main
occupation is the job the respondent spent the most
hours doing during the previous seven days.)
GOVERNMENT AND NGO TRANSFERS (ESSENTIAL).
Many developing countries have programs that provide
money or in-kind assistance to households. Some
of these are government programs and some are run
by nongovernmental organizations (NGOs). Examples
of these programs include cash welfare payments, pensions,
unemployment insurance, food stamps, food
rations, school feeding programs, community soup
MARGARET GROSH AND PAUL GLEWWE
32
kitchens, scholarships, and free or subsidized textbooks.While
the range of programs is very wide, there
are usually only a few sizable programs in any particular
country.
A key policy question that LSMS surveys can
address is who benefits from these programs.However,
only programs that reach a substantial fraction of the
population can be studied with the relatively small
sample sizes recommended for LSMS and similar multitopic
surveys.
Questions about government and NGO transfer
programs should not necessarily all be in the same
module (a fact that makes this part of the core difficult
to standardize).While questions about cash income fit
best in the transfers and other nonlabor income module,
questions on school feeding programs should
probably be in the education module. However, Part
B1 in the transfers and other nonlabor income module
is a good place to start collecting this information.
SOCIAL SERVICES (ESSENTIAL). Related to programs
that provide cash or in-kind assistance are programs
that provide services.The most common examples of
social services are public schools, public health services,
agricultural extension services, credit programs,
public work schemes, electricity supply, public water
supply, and sewage systems. LSMS surveys and similar
multitopic surveys should always collect some information
on the use of social services, at least enough to
measure variation in access to and utilization of such
services across different socioeconomic groups.
As with direct assistance to households, the types
of programs and the amount of detail needed to identify
who benefits from them will vary among countries.School
enrollment information is already collected
in the core, as discussed above, although additional
information may need to be collected on any school
services that are available to some students and not to
others, such as tuition waivers or afterschool programs
for disadvantaged students. Information on the use of
public health services is also very important; such
information is collected in questions 10–38 of the
short version of the health module (introduced in
Chapter 8). Data on the use of agricultural extension
services are collected in Part F of all versions of the
agriculture module introduced in Chapter 19. Some
countries have subsidized credit programs to assist
poor households; information on these programs is
collected in questions 9–14 and 21–28 of the short
version of the credit module introduced in Chapter
21. It is possible to identify beneficiaries of public
works programs by adding one or two questions to the
employment module that ask whether an individual’s
current employment is related to such a program.
Finally, information on housing-related physical infrastructure
services—such as water, sanitation, and electricity—is
collected in the core of the housing module,
as discussed above.
METADATA (ESSENTIAL). The last type of information
that must be collected in the household questionnaire
of any LSMS-type survey consists of basic data on
where the household fits in the sample and on the
outcome of the interview. This type of information,
known as “metadata,” is discussed in Chapter 4. For
the essential core,it is not necessary to collect all of the
information covered in the metadata module. The
essential metadata are the date of the interview or
interviews,the identification (ID) codes for the household
and its primary sampling unit,4
the ID codes of
the interviewer and the other team members who collected,
checked, or entered the data from that household,
information on whether an interview actually
took place (and if not, why it did not), and perhaps
some data on the ethnic group and religion of the
household.This information is collected in the metadata
module, on the Household Identification and
Control submodule and in questions 1–4 of the
Summary ofVisits and Interviews sub-module.
PRICES (ESSENTIAL). Price information should be collected
at the level of the community (the primary
sampling unit) since all households in a given community
face the same prices. How to collect price
information is discussed in Chapter 13.The main task
is to select the items for which price data will be collected.
While the exact items will vary across countries,
prices should be collected for at least 30–40 of
the most commonly consumed food items and 10–20
of the most commonly purchased nonfood items. In a
few countries other sources of reliable price data may
already exist for both urban and rural areas; if these
data can be matched to the communities covered in
the survey, there is no need to collect new price data.
And in some small countries such as Jamaica, prices
vary little among regions. In these cases, no price data
need to be collected as long as national price data exist
that show changes in prices over time.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
33
ANTHROPOMETRIC MEASUREMENTS (RECOMMENDED).
Anthropometric data, particularly on height and
weight, should be collected for children 0–5 years old
in almost every LSMS or similar household survey.
Stunting (low height for age) and wasting (low weight
for height) are common measures of children’s nutritional
status; height and weight data are critical in
countries where children are at risk of malnutrition.
And collecting basic anthropometric data about children
is simpler and more reliable than collecting other
data on health status. The details of how to collect
children’s anthropometric data are explained in
Chapter 10.
Collecting height and weight data requires some
effort.The data are collected using special equipment
that is bulky and troublesome to carry around to each
household. One consequence of this is that another
individual is often added to each survey field team. If
collecting children’s height and weight data were easier,
such anthropometric measurements would have
been classified as part of the essential core of any
LSMS-type survey.
IMMUNIZATION (RECOMMENDED). Almost all LSMS
and similar multitopic surveys should collect immunization
records for children ages 0–5. In recent years
child immunization programs have dramatically
reduced the incidence of several life-threatening
childhood diseases in many developing countries—
significantly reducing infant and child mortality rates.
However,many countries still do not have 100 percent
immunization coverage.Therefore, information on the
extent of coverage and on where coverage is low is
important for almost any analysis of living standards.In
addition, since child immunization coverage can
change dramatically over a year of two, it serves as a
useful indicator of changes in the provision of government
services during periods of economic or social
instability. Child immunization information is collected
in Part C of the health module introduced by
Chapter 8.
ASSETS (RECOMMENDED). Household assets include
information on any consumer durable goods owned
by the household, the value of owner-occupied housing,
and the ownership of land and capital assets related
to agricultural activities and household enterprises.
There are several reasons for collecting these data.
First, the possession of household durable goods such
as radios, televisions, bicycles, motorcycles, and cars is
a simple indicator of living standards. Second, the sum
of the value of all these different household assets gives
a rough (and admittedly incomplete) indicator of
household wealth. Third, data on the ownership of
land and on capital assets used in agricultural and
nonagricultural enterprises indicate productive assets.
Fourth,in some countries,particularly countries of the
former Soviet Union,there is evidence that adding the
consumption derived from durable goods and housing
to total consumption can lead to substantial changes in
the relative economic positions of different types of
households.
Information on the ownership of consumer
durable goods can be collected using Part E of the
consumption module. Data on the value of owneroccupied
housing are collected in the short version of
the housing module,using (at minimum) questions C1
and C11, with C3 and C12 providing alternative valuations.
A short set of questions on the assets used in
household enterprises is provided in Part G of the
short version of the household enterprise module;
only questions 1–3 are needed. Parts A, B, and E of the
short agriculture module collect a modest amount of
information on households’ land holdings, livestock,
machinery, and other agricultural assets.
PRIVATE INTERHOUSEHOLD TRANSFERS (RECOMMENDED).
Private interhousehold transfers, which are
pervasive in many countries, are used by many households
to cope with poverty and economic vulnerability.Transfers
received are covered by the transfers and
other nonlabor income module (introduced in
Chapter 11) and transfers sent are covered by the consumption
module (introduced in Chapter 5). At least
the short versions of the private interhousehold transfer
submodules should be used in virtually all surveys.
Even in a relatively simple survey it may be worthwhile
to use the standard version of the submodule on
transfers received.
RENTAL PAYMENTS FOR HOUSING (RECOMMENDED).
Estimates of the annual rental value of dwellings are
needed to estimate the consumption value of housing
for households that own these dwellings. In most
countries such estimates can be calculated by estimating
the relationship between basic housing characteristics,
which are already part of the core, and the rental
payments made by households that rent their
MARGARET GROSH AND PAUL GLEWWE
34
dwellings.The key piece of information needed is the
rental payments of households that rent. Questions
C7–C12 in the short version of the housing module
collect information on rental costs.
Full LSMS-type Survey
In practice, the essential core—and even the recommended
core—will tap only a small part of the
potential policy uses of an LSMS-type survey. In
most LSMS and other multitopic surveys, much
more can and should be added to the questionnaires
to gather information beyond what is collected in
the core. This subsection and the two that follow it
discuss different ways to add to the core by expanding
modules and combining them to form a survey
or sequence of surveys.
A full LSMS-type multitopic household survey
can be formed by combining the short or standard
versions of most of the modules in the household
questionnaire with the corresponding parts of the
community and price questionnaires.This produces a
household survey similar in design to the original
LSMS surveys first used in 1985,except that the modules
presented in Volume 3 of this book (and
described in Parts 2 and 3) include revisions based on
15 years of experience with LSMS and other household
surveys.
Because some of the standard versions of modules
presented in Parts 2 and 3 are significantly larger than
versions used in the original LSMS surveys,a household
questionnaire including all of the standard modules
would almost certainly be too large to be practical.Thus
the household questionnaire of a full LSMS-type survey
needs to be trimmed, either by replacing the standard
versions of some modules with their short versions or
by dropping some nonessential modules.
A well designed full LSMS-type multitopic survey
collects information that measures or otherwise
describes:
• Household consumption.
• Household income.
• Key nonmonetary indicators of welfare such as
nutritional and health status, education status, and
housing conditions.
• Many aspects of household behavior, such as
income-generating activities, human capital investments,
fertility, and migration.
• The local economic environment (including prices
and the availability of services).
• Participation in specific government programs such
as food stamps programs, job training programs, and
agricultural extension services.
Having all this information for a group of households
makes it possible to describe many indicators of living
standards, estimate the determinants of different
dimensions of living standards and different types of
household behavior, and estimate the relationships
between dimensions of living standards and household
behavior (such as the impact of children’s nutritional
status on their school performance).
The full LSMS household questionnaire is long
and complex. In almost all cases it is too long to be
completed in a single visit by an interviewer to a
household. Instead, an interviewer typically visits each
household twice. All of the individual-specific modules
(roster, education, health, employment, and
migration) are administered in the first visit, sometimes
with the addition of one or two household-level
modules such as housing. The interviewer makes an
appointment for a second visit, usually about two
weeks later, to reinterview household members who
are most knowledgeable about the other householdlevel
modules (such as consumption, agriculture, and
household enterprises). To ensure that high quality
data are collected and to keep the budget within reasonable
limits, the samples in full LSMS-type surveys
are usually relatively small—between 2,000 and 5,000
households. Samples of this size are still large enough
to provide accurate information on the nation as a
whole, on rural and urban areas, and on a small number
of geographic regions. However, such samples are
not large enough to provide accurate statistics for each
state, province, department, or district in a country.
Even at the national level, they cannot provide precise
information on phenomena that do not pertain to
most households or individuals—such as post-secondary
education or participation in a program used by
only a small fraction of the population. See Grosh and
Munoz (1996) for a more thorough discussion of sample
size and sampling issues.
In most cases it is not worthwhile to implement a
full LSMS-type multitopic survey every year. Much of
the analysis for which LSMS surveys are designed does
not need to be repeated annually. For example, while
it is important to understand the determinants of fertility,
it is unlikely that these determinants change
greatly from one year to the next. Sizable changes are
likely to occur only over the course of several years, as
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
35
economic conditions and people’s attitudes change.
Another reason not to implement a full survey every
year is that it is costly to administer such a comprehensive
household questionnaire,and requires substantial
work at each stage. Therefore, a full LSMS-type
survey should be implemented only once every three
to five years.
During 1985–99 the following countries implemented
full-size LSMS surveys: Algeria, Brazil, Côte
d’Ivoire, Ecuador, Ghana, the Kyrgyz Republic,
Mauritania, Morocco, Nepal, Pakistan, Panama, Peru
(in 1985–86, 1991, and 1994), Turkmenistan, and
Vietnam.
Scaled-down LSMS-type Survey
A scaled-down LSMS-type survey can be constructed
by omitting some modules from the household questionnaire
of a full LSMS-type survey and by abridging
other modules. Such a survey will still be a multitopic
survey, but will cover fewer topics than a full-size survey
would. Substantial reductions in the size of the
household questionnaire may mean that the questionnaire
can be completed in a single visit by the interviewer
to the household, as compared to the two visits
needed for a full LSMS-type multitopic survey.
The extent to which various modules should be
reduced or eliminated will depend on which policy
questions are most important in the country in question.
However, there is a limit to how much the questionnaire
can be cut.The essential core of an LSMS or
similar multitopic survey, as described above, must
remain. In addition, the elements that are added to
form the recommended core (data on anthropometrics,
child immunization coverage, basic household
assets, interhousehold transfers, and rental payments of
households that rent their dwellings) should almost
always be included. The community questionnaire
may or may not be included in a scaled-down survey,
but the price questionnaire should always be used,
except in those rare cases in which fully adequate
price data already exist or price variation across
regions is negligible. Overall, the analytical objectives
of a scaled-down LSMS-type survey are more modest
than the objectives of a full-size survey.
One common way to abridge the questionnaire is
to decide not to collect the data needed to measure
total household income. Not measuring total household
income allows survey designers to delete most of
the agriculture and household enterprise modules,
retaining only questions on agricultural extension
services that are part of the essential core and questions
on assets that are part of the recommended core.Yet
the questions on wages from the employment module
and the questions on public and private transfers from
the transfers and other nonlabor income module
should be retained,as they are part of the essential core
of any LSMS-type survey.
Other questions that can be dropped are questions
on any aspects of household behavior that are of little
interest to policymakers. The savings, credit, fertility,
and migration modules have often been deleted from
previous LSMS surveys. Because the new time-use
module is quite lengthy, it is also a candidate for omission,
unless data on time use are of particular interest
to policymakers. If analysts aim to measure use of
social services but not to estimate the determinants of
demand for them, survey designers could choose to
use the short, rather than the standard, versions of the
health and education modules.
An alternative way to obtain a scaled down multitopic
survey is to “scale up” an existing single-topic
household survey, such as a labor force or household
expenditure survey. In Romania, Latvia, and
Bangladesh, new modules on the use of social services
and programs were added to existing household income
and expenditure surveys. In Guyana, households that
had been interviewed in a previous income and expenditure
survey were revisited to collect information on
health, education, and anthropometrics; the separate
data files were later merged for purposes of analysis. In
Jamaica, households from the Labor Force Survey were
revisited by interviewers who administered the Survey
of Living Conditions; the two data files were later
merged. In Paraguay, additional modules were added
directly to the Labor Force Survey questionnaire.
Scaling down the household questionnaire of a
full LSMS-type survey reduces the analytical potential
of data collected, especially in parts of the questionnaire
that are dropped or abridged. A reduced questionnaire
produces fewer descriptive statistics on many
dimensions of household welfare than would be possible
using a full-size survey. Data from a scaled-down
questionnaire can be used to analyze only a few of the
determinants of living standards.And such data reductions
substantially reduce the range of analytical methods
that can be used.
A scaled-down LSMS-type survey can be implemented
fairly often, perhaps annually or every other
MARGARET GROSH AND PAUL GLEWWE
36
year. Such frequent implementation is desirable
because one of the main uses of data from a scaleddown
survey is to monitor changes in poverty and
other dimensions of welfare over time. Also, the fact
that a scaled-down survey collects less data on the
determinants of household welfare and behavior than
does a full-size survey means that implementing it frequently
wastes less resources than would implementing
a full LSMS-type survey every one or two years.
Another advantage of a scaled-down survey is that it is
easier and less expensive to carry out than a full-size
survey. Finally, a scaled-down survey can be carried
out using somewhat larger samples than a full LSMStype
survey because it is subject to fewer managerial
and budget constraints.
Scaled-down LSMS surveys have been carried
out, with World Bank support, in Albania,Azerbaijan,
Bolivia, Bulgaria, Pakistan (1995/96 and 1996/97),
Peru (1990), and Tanzania.
Core and Rotating Module Design
The “core and rotating module” design for a multitopic
household survey is an attempt to combine the
advantages of full and scaled-down LSMS-type surveys.
In this design, a scaled-down LSMS-type survey
forms the “core,” while one or two modules are added
or greatly expanded each time the survey is carried
out. Modules that are added or expanded in any given
year revert back to their “core” size the following year,
creating a module “rotation” scheme for the modules
that go beyond the core. In most cases the survey is
fielded annually, although it can also be a semiannual
or biannual survey.The core that is repeated each time
the survey is implemented must include the essential
core described above, and in almost all cases it should
include everything in the recommended core.In many
cases the core of a core and rotating module design
should collect additional information as well, in order
to provide a more detailed picture of household welfare
each time the survey is implemented.
An example of how to implement this approach
would be to use only the core in the first year of the
survey, in order to focus on making sure that the core
works well. In the second year the health module in
the household questionnaire would be expanded to
gather more detailed data on individuals’ health status
and behavior, the kinds of health care sought, and the
cost and quality of that health care.In addition,a health
facility questionnaire could be added (see Chapter 8 for
further details on health facility questionnaires). In the
third year the health module would return to its original
“core” size and a new subject, such as education or
savings, would be given special emphasis. Expansion of
any particular module might require making some
additions to other modules in the survey to ensure that
the analytical potential of the data collected in the
expanded module could be fully exploited. Each chapter
in Parts 2 and 3 of this book explains what data are
needed from other modules to complement the data
collected in the module covered by that chapter.
The core and rotating module design is a hybrid
of a full LSMS-type multitopic survey and a frequently
implemented scaled-down LSMS-type survey.
Implementing a core and rotating module survey
annually would allow for the same monitoring of
poverty and welfare that is possible with data from an
annual scaled-down survey. In addition, in each rotation
of a particular module, this kind of survey would
collect the data analysts need to study the determinants
of household behavior for a specific topic—in
other words, data comparable to what are collected in
a full-size survey. It might even be possible to use data
from the scaled-down modules to study topics that are
not emphasized by the survey in a particular year.
The cost and sampling implications of the core
and rotating module design lie somewhere in between
those of a full-size LSMS-type survey and those of a
scaled-down survey. Perhaps of greatest concern in the
core and rotating module design are the institutional
arrangements for developing, implementing, and analyzing
the special modules.While for both full-size and
scaled-down LSMS-type surveys it is possible to put a
lot of effort into the design of the first survey and give
less attention to improving its design in subsequent
years, implementing the core and rotating module
design means that the questionnaire needs to be significantly
modified each year—requiring much more
attention from survey designers after the first year.
Indonesia’s SUSENAS is a long-standing example
of a core and rotating module survey design. Jamaica’s
Survey of Living Conditions, which began in 1988,
was the first LSMS survey to adopt this approach. A
new LSMS survey in Cambodia is just starting to
develop such a system, as is the Bangladesh Household
Expenditure Survey. (The Bangladesh Household
Expenditure Survey is not usually regarded as an
LSMS survey; however, it has adopted much of the
LSMS methodology.)
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
37
Special Purpose Sample Designs
There are two other possible survey designs, both of
which use special purpose samples (that is, samples that
are not nationally representative). The first is a survey
that samples a special population that is of particular
interest for analytical or policy purposes.5
An example
of this is a sample of households within a single city that
is used to study issues pertaining to that city, such as the
housing market, the water supply system, or urban air
pollution. Two LSMS surveys of this type have been
performed:one in the Kagera region ofTanzania,focusing
on areas with high prevalence of AIDS, and one in
rural areas of Northeast China, focusing on the agricultural
activities of rural households.
A second kind of special purpose survey is one in
which the sample is drawn solely for purposes of program
evaluation. In this type of survey, a group is
observed both before and after the benefits of a particular
service or program are made available to this group.
Alternatively, the sample may be composed of two
groups, one consisting of the households who benefit
from the service to be evaluated—the treatment
group—and the other consisting of households that are
similar to the first in every respect except that they do
not benefit from the service—the control group.
These special-purpose samples usually gather
detailed data on the topic being studied, whether it is
a specific sectoral issue (such as agriculture) or a program
to be evaluated. There are so many ways to
design such surveys that this book cannot hope to
cover all of them. However, since special purpose surveys
typically collect data on many general characteristics
of the sampled households (such as size, composition,
living standards, labor force status, and
education), designers of this kind of survey can use the
modules proposed in this book as a guide for collecting
this supplemental information.The experience of
past LSMS surveys has been used in designing special
purpose surveys to evaluate the impact of educational
reforms in El Salvador.And the Nicaragua LSMS survey
included a special sub-sample designed to evaluate
the impact of that country’s Social Investment Fund.
Matching Circumstances and Designs
This section provides some approximate rules of
thumb for choosing among the three common survey
design options discussed in the previous section.These
recommendations should not be thought of as rigid or
beyond debate; instead, they should be thought of as a
starting point for making survey design decisions.This
is the case for several reasons. First, the dividing lines
between the three basic survey designs are flexible, as
it is possible to develop “hybrid” surveys that merge
characteristics from the different survey design
options. Second, individual countries may not fit neatly
into the categories implicit in the rules.Third, surveys
may have multiple analytical objectives. Finally,
funding constraints are not explicitly considered here.
Survey planners should consider the following
general “rules of thumb” when deciding what kind of
survey to implement:
1. Countries with sufficient institutional capacity to
implement a complex survey should use either a
full LSMS-type multitopic survey every three to
five years or a core and rotating module design;
both options can serve a broader range of analytical
objectives than can a sequence of scaled-down
LSMS-type surveys.
2. If annual (or biannual) monitoring of living standards
or poverty is the most important analytical
objective, either a sequence of frequent scaleddown
LSMS-type surveys or a core and rotating
module survey should be adopted. In contrast, a
full-size multitopic survey is inappropriate because
cost and efficiency considerations imply that such
surveys should be implemented only every three to
five years.
3. No new survey is needed if the main objective is to
provide periodic descriptive information (say, every
three to five years) or to examine the coverage of
government programs in countries where ample
data are already available from other sources.
4. If the main objective is to gather periodic
descriptive information or to examine the coverage
of government programs, a core and rotating
module design should not be chosen. Such a
design would collect data much more frequently
than is necessary.
5. If the main objective is to model household behavior,
either a full LSMS-type survey or a core and
rotating module survey should be chosen. A series
of scaled-down surveys would be insufficient for
modeling household behavior.
6. If the main objective is to model household
behavior and very little other data are available, a
full-size multitopic survey is preferable to a core
and rotating module survey since the latter cannot
MARGARET GROSH AND PAUL GLEWWE
38
supply detailed information on all topics until it
has been in operation for several years. The core
and rotating module design can be adopted after
one or two full LSMS-type surveys have been carried
out.
7. If the main objective is to model household behavior
and a large amount of other data are available,
the core and rotating module survey is preferable to
a series of periodic full LSMS-type surveys because
the core and rotating design allows poverty to be
monitored more frequently over time.
8. If the institutional capacity in the country is limited
and the survey aims either to monitor poverty
and living standards annually or to provide
descriptive information (including coverage of
government programs) periodically, a scaleddown
LSMS-type survey should be chosen. This
survey may be either frequent (for annual monitoring)
or periodic (for descriptive information
every three to five years). The other options, full
multitopic and core and rotating module, are too
complex for countries with limited institutional
capacity.
Table 2.2 summarizes the implications of these
rules, showing which rules lead to which choices.
Because countries with little institutional capacity
cannot implement a full LSMS-type multitopic survey
or a core and rotating module design on their own,
they will not be able to collect data that are useful for
analyzing household behavior unless their institutional
capacity is either permanently improved or supplemented
in the short run by using international
experts. In addition, significant purchases of new
equipment may be required in some countries.
Choosing the Modules, Defining Their
Objectives, and Setting Their Size
Once the basic blueprint of the survey has been selected,
survey designers must decide which modules to
include in the household and community question-
naires.6
Designers must also define specific objectives
for each module and decide on each module’s approximate
length. The procedures for these steps are discussed
in this section. Because decisions about length
and objectives ultimately depend on many countryspecific
details, specific recommendations cannot be
provided for each possible scenario.Instead,some general
guidelines and procedures are provided that
should prove useful for completing this step efficiently
and effectively.
Two general points must be made at the outset.
First, the tasks of choosing modules, defining their
objectives, and setting their approximate size are all
closely related and thus must be done simultaneously
rather than sequentially.The type of objectives and the
number of objectives have considerable implications
for the size of each module; more objectives, and more
complex objectives, necessitate a larger module.
Second, the objectives of each module should be consistent
with the overall objectives of the survey, in
terms of both the analytical objectives (describing living
standards,monitoring poverty and living standards,
examining the coverage of government programs,estimating
the impact of policies) and the specific topics
in which policymakers are interested. The overall
objectives of the survey already provide some information
on what the objectives of many of the modules
will be.
Table 2.2 Recommended Survey Designs for Different Settings
Analytical objective
Describing living standards or Monitoring living
Availability of other data examining program coverage standards or poverty Modeling household behavior
Countries with sufficient institutional capacity
Limited Full LSMS-type survey Core and rotating module Full LSMS-type survey (Rule 5 + Rule 6)
(Rule 1 + Rule 4) (Rule 1 + Rule 2)
Ample No new survey needed Core and rotating module Core and rotating module (Rule 5 + Rule 7)
(Rule 3) (Rule 1 + Rule 2)
Countries with limited institutional capacity
Limited Periodic scaled-down LSMS- Frequent scaled-down LSMS- Full LSMS-type survey (Rule 5 + Rule 6)a
type survey (Rule 8) type survey (Rule 8)
Ample No new survey needed Frequent scaled-down LSMS- Core and rotating module
(Rule 3) type survey (Rule 8) (Rule 5 + Rule 7)a
a. International experts must be hired to carry out key tasks.
Source: Authors’ recommendations.
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
39
Choosing modules
A good first step in choosing modules is to set the
upper and lower limits of what can be included in a
multitopic survey.The lower limit is the essential core
discussed above; in almost all cases this lower limit
should be expanded to include the additional elements
that are in the recommended core. The upper limit
will depend on country-specific circumstances such as
the capacity of the statistical agency and the willingness
of households to participate in lengthy interviews.
It is never possible to include all of the modules
in any one survey.
An important question to address relatively early
when making decisions about modules is whether the
survey will attempt to collect enough data to calculate
total income.The advantages and disadvantages of collecting
these data are discussed at length in Chapter
17. Clearly, if survey designers decide to collect the
data needed to calculate total income, the agriculture
and household enterprise modules need to be included
in the questionnaire.7
If designers decide not to
collect income data and there is little interest in these
two modules, they can be dropped, except for the
questions on use of agricultural extension services that
are part of the essential core and the asset questions
that are part of the recommended core.
It will probably not be possible to collect total
income data in a scaled-down survey because it is not
feasible in a single visit to a household to collect the
recommended core data plus the data from the agriculture
and household enterprise modules and still
have room to examine other topics.This implies that
it is also difficult to collect total household income in
a core and rotating module survey, except when the
module featured is either the agriculture or the household
enterprise module; even when one of these two
modules is featured in such a survey, collecting total
income data may not be feasible in some countries.
Two other specific decisions to make early in this
step of the survey design process are whether to collect
time-use data and whether to implement a large
number of the detailed environment modules (see
Chapters 22 and 14, respectively).The time-use module
is very long, and as such should be thought of as
an expanded module. If survey designers choose to
include this module, they may have to omit several
other short or standard modules. While it would be
feasible to include the time-use module in a full-size
LSMS-type survey, this module is probably too large
to be included in a scaled-down LSMS-type survey
and would work in a core and rotating module survey
only if it were chosen as the topic emphasized in a
particular year. The same circumstances apply to the
environmental module.The full set of these environmental
submodules is equivalent to a very large
expanded module, and for this reason it is difficult to
imagine the full set used in a single survey.The contingent
valuation modules should be used only when
specific improvements in services (such as urban water
supply, urban sanitation, urban air quality, or rural
water supply) are being contemplated.
Even a subset of the expanded environmental
modules is likely to be equivalent to a large expanded
module, especially if the water, sanitation, and fuel use
modules are included.This being the case, it is feasible
to include a large subset of the environmental modules
in a full LSMS-type multitopic survey, but only a relatively
small subset can be included in a scaled-down
survey. In a core and rotating module survey a large
subset of the environmental submodules can be used
only if environmental topics are emphasized in that
particular year; for all other years only a small subset
would be feasible.
At this point,it is useful to give some general rules
about how much room there is for modules in different
kinds of multitopic surveys. For a full LSMS-type
survey, the household questionnaire should be roughly
large enough to include a mixture of about 15 standard
or short modules. The number of modules that
can be included in a scaled-down survey is probably
closer to 8 or 10, most of which have to be short versions.
A core and rotating module survey lies somewhere
in between but is probably closer to the scaleddown
survey if only one visit is made to the
household. The modules chosen must in all cases
include the components of the essential core;in almost
all cases the modules should also include the additional
components found in the recommended core.
Using these starting points for what is feasible, the
next task is to consult with policymakers at the highest
level to get a detailed idea of which topics are of
greatest interest to them (if this has not already been
done). Policymakers need to specify which topics are
of overriding concern, which are of moderate interest,
which are of minor interest, and which are of little or
no interest. Expanded modules, if they exist, should be
used for topics of overriding interest.8
Standard modules
should be used for items of moderate interest.
MARGARET GROSH AND PAUL GLEWWE
40
Short modules may be appropriate for items of minor
interest. Items of little or no interest need not be covered
in the survey unless they are part of the essential
or recommended core.
The core and rotating module survey design is
inherently more flexible than other classic designs; if
the core and rotating survey is implemented annually,
it can cover four or five topics in great detail over the
same number of years by including the expanded version
of one of these modules each year. Of course, survey
designers still have to set priorities about which
expanded module is included in the first year, which is
included in the second year, and so on.
The above paragraphs provide survey designers
with a scheme for generating a draft list of the modules
to be included in the survey, their approximate
length, and, to an extent, the objectives of each module.
Needless to say, this draft list needs to be refined.
This can be done by adding two new “ingredients” to
the process: discussions with policymakers who specialize
in particular topics or programs, and a careful
reading of the chapters in Parts 2 and 3 of this book.
The task is to reconcile the specific policy questions
raised by these more specialized policymakers with the
feasibility of collecting data to analyze them (as discussed
in detail in Parts 2 and 3 of this book) given the
approximate sizes of each module as specified by highlevel
policymakers.This process is not simple and consequently
involves a certain amount of iteration.
Unfortunately, policy issues raised by specialist
policymakers often require more questions than can fit
into a module of the size specified in the first draft of
the modules.The choice at this point is between not
including many of these policy issues in the survey and
expanding the module containing these questions at
the expense of other modules. A third alternative is
expanding the relevant module without reducing the
size of any other module, but the feasibility of this
option is open to question and will not become clear
until a draft questionnaire is field tested.
Given this situation and the uncertainty regarding
what is feasible and what is not, survey designers
should use the following procedure to reconcile the
specific objectives of each module with any constraints
on module size. First, designers should ask policymakers
who specialize in a given topic to rank the policy
issues in order of importance, so that the module can
collect the data needed to analyze the most important
policy issues despite the inevitable constraints.
Second, for each module, survey designers should
match the policy issues raised by policymakers with
the data required to analyze them, as laid out in each
chapter of Parts 2 and 3. One way to do this is to
choose the smallest version of each module that can
address all of the relevant policy issues,and remove any
questions in that module that are not needed to analyze
these policy issues. If the module is still too long,
questions needed only to address the least important
policy issues are deleted. This shorter module is
checked again to see if it exceeds the provisional size
limit.The general principle is that the most important
policy questions are addressed first and additional
issues are added until the module has reached the
length that survey designers, in consultation with
high-level policymakers, have set for it.
Third, after this has been done for all modules,
survey designers should prepare a list of issues they
think can be covered by the survey and give this list
to the high-level policymakers, who will decide
whether they would like to change the amount of
space allocated to each module. The survey designers
should tell the policymakers about the tradeoffs
involved, working with them to ensure that the
issues policymakers deem most important are
addressed.
Ultimately, this process produces a list of modules
to be included in the survey, the proposed length of
each module to be included, and the specific objectives
for the modules.This completes the second step
of survey design. This step may need to be revisited
later if results of the field test show that the questionnaires
are too long or that there is room to expand the
questionnaire.
Notes
The authors would like to express their gratitude to Jere Behrman,
Lawrence Haddad, Courtney Harold, John Hoddinott, Alberto
Martini, Raylynn Oliver, Kinnon Scott, and Salman Zaidi for comments
on an earlier draft.
1. This book is designed to provide a thorough review of international
experience. However, new experience and knowledge will
continue to accumulate after the book has been published.Therefore,
until a new book is written, any new international-level information
is probably most easily obtained from international researchers.
2. If geographic areas—rather than households—are the unit of
observation, it may be possible to merge data from different surveys.
However, this high level of aggregation yields less precise
CHAPTER 2 MAKING DECISIONS ON THE OVERALL DESIGN OF THE SURVEY
41
results, raises issues of aggregation bias, and generally requires surveys
with very large sample sizes.
3. A variety of external sources have been used to fund past LSMS
surveys.World Bank loans have partially financed several LSMS surveys.
Grants from various bilateral development agencies (especially
from the United States, Scandinavian countries, and Japan) and multilateral
development agencies (particularly the United Nations
Development Programme and the United Nations Children’s Fund)
have wholly or partially financed a large share of LSMS surveys. In a
few cases, grants from theWorld Bank research budget have supported
LSMS surveys. Similar surveys, such as theWorld Bank’s SDA surveys,
RAND’s Family Life Surveys, and a few other surveys in Africa,
all receive a large share of their funding from external sources.
4. Most previous LSMS surveys have used two-stage sample
designs. If a three-stage sample design is used, ID codes will be
needed that identify both the primary and secondary sampling
units of each household.An analogous comment applies to surveys
that use four or more stages in their sample designs.
5. In large countries with federal systems, surveys can be performed
for individual states. Such surveys usually have the same
general purposes as national surveys, and have samples that are representative
of the whole state.
6. Each module in the household questionnaire should also be
included in the community questionnaire. See Chapter 13 for further
discussion of the community questionnaire.
7. While all versions of the household enterprise module collect
income information, only the standard and expanded versions
of the agriculture module collect sufficient data for use in the
measurement of total income.
8. A full LSMS-type survey could accommodate two and possibly
three expanded versions of modules;a scaled-down survey could
accommodate at most one.Volume 3 presents expanded versions of
the following modules: roster, education, health, employment,
migration, environment, household enterprises, and agriculture.The
time-use modules introduced in Chapter 22 should also be treated
as expanded modules, and the same is even more true for the full set
of environmental modules introduced in Chapter 14.
References
Grosh, Margaret, and Juan Muñoz. 1996. A Manual for Planning and
Implementing the Living Standard Measurement Study Survey.
Living Standards Measurement Study Working Paper 126.
Washington, D.C.:World Bank.
43
3
Designing Modules and Assembling Them into
Survey Questionnaires
Margaret Grosh, Paul Glewwe, and Juan Muñoz
Chapter 2 outlined the five-step process that survey designers should follow to design LSMS and
similar multitopic surveys. It also provided detailed recommendations on how to undertake the
first two steps, which are deciding on the overall design of the survey and deciding which modules
to include in the survey questionnaire.This chapter discusses the last three steps of the fivestep
survey design process.The first section of this chapter describes the third step—drafting each
module, question by question, to ensure that it will collect the data necessary to meet the module’s
objectives (which were laid out in the second step).The second section guides survey
designers through the fourth step—coordinating the different modules and combining them to
create a consistent and comprehensive set of questionnaires.The third section explains the procedures
for the last step —translating the questionnaires into local languages and conducting a field
test.The fourth section discusses the formatting of the questionnaires, which is an extremely
important but often neglected aspect of designing successful multitopic surveys. Survey designers
should refer to the material contained in the fourth section many times during the last three
steps of the survey design process.
In practice, the survey design process rarely moves
smoothly and sequentially from one step to the next.
Instead, survey designers often find themselves moving
backward and forward among the various steps. For
example, if designers encounter difficulties when
drafting a specific module, they may need to reconsider
and modify their original objectives for that module.
Developing survey questionnaires is an iterative
process, and survey designers should expect to go
through at least three or four drafts of each module. It
is not unusual for the different versions of the drafts to
add up to a stack of paper one foot (30 centimeters)
high. Each major redraft of a module or questionnaire
should be reviewed by all interested parties, not only
the people involved in carrying out the fieldwork (the
data producers) but also policymakers (who will make
decisions based on the data), members of the research
community (who will analyze the data), and the staff
of any agencies financing or providing technical assistance
to the survey. Eventually, what should emerge
from the process is a well-designed set of questionnaires
for a multitopic household survey.
Producing Draft Modules
The third step in survey design, producing draft modules
for the household and community questionnaires,
is one of the most time-consuming steps in the
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
44
process. Detailed guidance on this step is provided in
the chapters in Parts 2 and 3, so the discussion here
will be general and relatively brief.
Once the objectives for each module are finalized
(at least tentatively), survey designers can begin to
develop detailed draft modules for the household and
community questionnaires. Survey designers should
use the draft “prototype” modules introduced by the
chapters in Parts 2 and 3 (and presented inVolume 3)
as their starting point.As explained in Chapter 2, survey
designers will already have decided on the policy
and analytical objectives of each module.They should
now choose the shortest versions of the modules that
will allow for analysis of the most important of these
objectives; any questions not relevant to these objectives
should be removed.
If the resulting module is still too long, survey
designers should remove any questions that are needed
only for the analysis of the least important of the
policy issues. This process should continue until the
module meets the length constraint. In some cases the
module may be shorter than expected, in which case
a policy issue and its accompanying questions can be
added.The general principle is that the most important
policy issues should be addressed first, and additional
ones should be included only if space allows.
This approach is a good start, but much more remains
to be done.
For some modules the information and guidance
given in the relevant chapter in Parts 2 and 3 of the
book may be incomplete. For example, the chapter
may not address certain policy issues that are important
in a given country or setting, in which case the
designers of that survey will need to develop new
modules or submodules. Even in these cases the information
provided in the relevant chapter is usually a
good base for developing such modules. However, if
the designers intend to implement major innovations
in their survey, they should seriously consider adding
to the survey team a specialist with the relevant experience
in both data collection and data analysis.
Once each draft module has been written out in
its entirety, the next task is to verify that the design of
each module reflects the economic and institutional
structures of the country in question. For example, the
designers need to check whether common living
arrangements are reflected in the definition of the
household used in the household roster and in the
housing and interhousehold transfer modules. They
must also review all questions and response codes and,
if necessary, modify them to reflect local institutions
and terminology. For example, the transfers and other
nonlabor income module discussed in Chapter 11
must explicitly refer to each public transfer program
by name. The consumption module will need even
more work; in particular, the lists of items selected
must closely reflect items consumed in the country.
The agricultural module will need careful attention, as
this module must reflect the country’s landholding and
cropping patterns.
For many of the modules, survey designers may
find it useful to collect some preliminary data using
qualitative techniques, which may help them determine
how best to design these modules to collect
quantitative data. Chapter 25 provides a detailed discussion
on how to collect qualitative data. Such data
can be particularly useful in countries where successful
quantitative surveys have never been done for the
topic to be studied.
A final general issue to consider when drafting
modules is the role played by the fieldwork schedule.
A prototypical full LSMS survey spreads fieldwork
evenly over a 12-month period, for two reasons. First,
this makes it possible to study or average out any seasonality
effects. Second, and more importantly, surveys
with this fieldwork schedule require a smaller number
of survey field teams than do surveys that compress the
fieldwork into a shorter period of time. This smaller
number of teams reduces costs and allows for
improved quality control. All of the interviewers can
be trained together and thus to a uniform standard; in
addition, the cost of training interviewers—which
takes about four weeks—will be proportionately
cheaper. Each interviewer will complete more interviews
and thereby gain more experience.Finally,fewer
computers and vehicles will be required.
Despite these advantages of a year-long survey
period, many past LSMS surveys have compressed
fieldwork into a period of just two or three months.
This has often been done when there was pressure on
the survey team to collect data for analysis as quickly
as possible. In other cases interviewers may have been
available for only a short period of time, or the organization
funding the survey may have required that the
project be completed in a relatively short amount of
time.The fieldwork schedule can also be modified to
accommodate analysis of certain topics. For example,
analysis of some agricultural issues may require inter-
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
45
viewers to make two or more visits to each household
at different times during the year.
Variations in the fieldwork plan may require
changing the wording of some modules. This means
that survey designers should ensure that the design of
each module in the questionnaire is consistent with
the fieldwork plan.
When a survey is conducted over a relatively short
period, such as a few weeks or months, careful attention
must be given to the wording of questions concerning
events that are seasonal in nature.Will school
be out of session for a large portion of the survey period?
If so, the education module may need to be
changed to reflect this. In particular, questions referring
to school activities during the previous week,
such as the number of days that a child attended
school or the number of hours of homework done by
the child, would clearly be inapplicable.Also, questions
about water supply during both wet and dry seasons
should be reviewed to ensure that they reflect the circumstances
of these seasons. The largest seasonal
changes may need to be made to the agricultural
module. A detailed discussion of the implications of
seasonality for that module is provided in Chapter 19.
More substantial changes will be required if the
household is to be visited more than once at different
times of the year. In such cases it may be desirable to
have the interviewer administer modules for which
the answers are expected to vary by season (such as the
consumption, agriculture, water, or time use modules)
each time he or she visits the household. In contrast,
the modules that are unlikely to be affected by seasonality,
such as housing, education, fertility, or migration,
probably need to be administered only once. Any
modules that are to be administered more than once
usually need to be modified, particularly with respect
to their recall periods. For example, if the interviewer
makes two household visits six months apart, the consumption
module should be administered in both visits
and should have a recall period of six months rather
than one year.Also, the water module should ask only
about the particular season (wet or dry) during which
the interview is to be conducted.
The guidelines given in this chapter are general,
since very detailed information is provided in Parts 2
and 3 of this book. Other information on adapting
LSMS questionnaires to fit local circumstances can be
found in Oliver (1997), which focuses on survey
design in the countries of the former Soviet Union, as
well as in Ainsworth and van der Gaag (1988).A good
general reference publication for developing and
designing household survey questionnaires is United
Nations (1985). More recent general references are in
Babbie (1990), Fink (1995), and Fowler (1993);
although these books focus more on developed countries,
much of the material they contain is also relevant
for developing countries.A final point to bear in mind
is that a good deal of attention must be given to correct
and consistent formatting. This is described in
great detail in the fourth section of this chapter; survey
designers should read that section very carefully
before they begin designing any survey modules.
Integrating and Combining Modules to Create
Complete Questionnaires
Once draft versions of each of the individual modules
have been written, these drafts must be combined to
form complete household and community questionnaires.
Merely stapling the various modules together
will not produce a well-designed questionnaire; much
more work has to be done to ensure that the different
modules fit well together.This section describes how
to do this important task. It focuses primarily on making
the modules of the household questionnaire consistent
with each other. Similar, though less difficult,
issues arise when integrating the modules of the community
questionnaire; in most cases the approach to
take for the community questionnaire can be inferred
from the discussion of the household questionnaire.
This section will also highlight particularly important
points to consider when combining the household,
community, and price questionnaires to form a comprehensive
household survey.
Gaps and Overlaps
Survey designers must scrutinize and compare the different
questionnaire modules for gaps and overlaps in
the information that the modules collect. Analysts
often need to combine data from different modules in
the household questionnaire. Perhaps the most important
example of this is the calculation of each household’s
total consumption, which requires information
not only from the consumption module but also from
the education, health, employment, and housing
modules—and from the water,sanitation,or fuel modules
(see Chapter 14) if they are included as separate
modules in the questionnaire (as opposed to using the
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
46
housing module to collect information on these topics).
Likewise, income data are collected in the
employment, agriculture, household enterprise, and
miscellaneous income modules. It is important to
check that a questionnaire includes the data needed to
construct these and other complex variables.
Another example of this general issue is that survey
designers often have a choice regarding the module
in which to collect some kinds of information. For
example,data on expenditures on fuel for cooking and
heating can be collected in the consumption module,
the housing module or, if it exists, the expanded fuel
module. Questions on child immunization can be
placed in the fertility module, the health module, or
the anthropometry module.An argument can be made
for choosing any of these options (see the pertinent
chapters in Parts 2 and 3 of this book), but the essential
point is to ensure that the information is collected
at least once, and is collected twice only if there is a
reason to do so.1
Appendix 3.1 provides a list of the
most common types of gaps and overlaps to check.
In cases in which information could be plausibly
be collected in more than one part of the questionnaire,
there may be no absolute right or wrong place
to collect it. Rather, survey planners must take into
account who the respondent is in each module, how
well the best recall period for that information matches
the recall periods of modules in which it might be
collected, at what point in the interview the respondent
might discuss the topic most naturally, and
whether the topic is a sensitive one that should therefore
be addressed near the end of the questionnaire
(for reasons discussed further below).
The survey designers should also examine any
overlaps among the household, community, and price
questionnaires. In general, the community and price
questionnaires should collect information on any
topic that varies only slightly from household to
household within the primary sampling unit.2
While
much of the information collected in the community
and price questionnaires could be collected in the
household questionnaire, it is better to collect it in the
community questionnaire in order to shorten the
length of each household’s interview. Collecting this
information in the community questionnaire is also
more efficient; why collect it for all households in a
primary sampling unit (often 16 or 20 households)
when it need be asked only once in the community
questionnaire?
Some simple examples illustrate this point. The
expanded water module contains questions about the
price and quality of water from different potential
water sources. If the primary sampling units are geographically
compact, all of the households in each primary
sampling unit are likely to have the same alternative
water sources, implying that the water price and
quality questions can be put in the community questionnaire
(which should be administered in each primary
sampling unit) rather than in the household
questionnaire. On the other hand, if the primary sampling
unit is not compact so that the households are
widely dispersed, it is likely that some households will
be nearest to,say,a particular spring or well while other
households will be closer to other springs or wells. In
such cases these questions about alternative water
sources should remain in the household questionnaire.
Another example concerns the distance to schools
and health facilities. In a compact primary sampling
unit, the distance to the nearest school or health facility
probably varies little among the households in the
primary sampling unit. This means that information
on the distances to schools and health facilities can be
collected in the community questionnaire as opposed
to the household questionnaire.
Length
The overall length of the household questionnaire
must be manageable. In general, it is not feasible to
include, say, the standard version of each module presented
in Volume 3, even though past full LSMS surveys
typically included 15 modules, many of which
were similar to the standard versions in this book.
There are several reasons why using all of the standard
draft modules in this book is not feasible. First,
this book introduces several new modules, including
the time use module and several environmental modules.
Second, some of the standard draft modules, such
as those on health, migration, and household enterprises,
are much longer than the modules on those
topics that were used in previous LSMS surveys.
Finally, in some of the chapters in this book (including
Chapter 18 on household enterprises and Chapter
19 on agriculture) it is argued that collecting more
detailed data will greatly increase their value for analytical
purposes. Thus survey designers should not
combine the standard versions of all of the modules
presented in the book into a single household questionnaire.Instead,the
short versions should be used for
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
47
some modules, and in almost all cases at least one or
two modules should be dropped.
Assessing whether a draft questionnaire is too
long is not simply a matter of counting the pages or
questions in it, since many questions, and sometimes
even entire pages or modules, will apply only to some
households. Moreover, in some cases adding questions
does not lengthen the interview time because the
respondent cannot avoid going through the thought
process made explicit in these questions,which implies
that a supposedly abbreviated set of questions will not
reduce the time required to complete the interview.
An example of this is the calculation of income
derived from agricultural activities.
There are also several ways to implement long
questionnaires that minimize the time required by
(and the fatigue induced in) each survey respondent.
These include conducting individual “miniinterviews”
with each household member to collect
all of the information needed from that individual at
one time (which allows him or her to leave when
questions are being asked of other household members);
using the best-informed respondent for each
household module; and dividing the interview into
multiple visits (for example, going through all the
individual-specific modules in one visit and returning
on a different day to conduct the consumption module
and other household-level modules). LSMS surveys
use all of these techniques. Still, there is a limit to
the amount of information that can be gathered from
a single household.
How can survey designers determine whether a
household questionnaire is too long? A rough idea of
the effective length of the questionnaire in different
circumstances can be obtained by calculating how
many households will go through the different paths
created by the skip patterns and how many questions
will be asked for each possible path. An excellent
example of this is provided in Chapter 18 on household
enterprises, in Table 18.5.
A more precise estimate of the time required to
administer a household questionnaire can be obtained
when similar surveys have already been done in the
country or region studied. In this case, the designers of
the new survey will be able to find out how long the
interviews took in the previous survey, provided that
the earlier survey collected metadata along the lines
suggested in Chapter 4. If such information is not
available,survey designers will need to rely on the field
test, which is discussed in detail in the fourth section
of this chapter. If field test interviews require many
hours to complete and exhaust the cooperation and
patience of households, this is an indication that the
questionnaire is too long. At the same time, survey
designers should realize that field test interviews normally
require much more time than do similar interviews
during an actual survey, because interviewers
have little training or experience with the questionnaire
at the time of the field test. In addition, the questionnaire
used in the field test is not a final draft and
thus is likely to contain some problems that will slow
down the interviews. A handy rule of thumb is that
interviews in the actual survey take only about half of
the time that they take in the field test—and sometimes
even less than that.
A general goal to aim for in the actual survey is
that any given respondent should not be interviewed
for more than one hour on a given day. Of course,
people’s tolerance for being interviewed will vary
from country to country, and this general guideline
must be adapted to suit local conditions.Experience in
LSMS surveys to date suggests that people’s tolerance
for long interviews is lower in urban areas than in rural
areas, lower among wealthy households than among
poor ones, and lower in wealthier countries than in
poorer ones.
Recall Periods
The recall periods proposed for each module introduced
in this book are mostly those that the authors
have deemed appropriate for that particular module.
This can be a problem when analysts want to combine
or compare data from several modules. For example, in
many LSMS surveys the employment module uses a
one-week recall period. Since most adults work, this
yields a large number of observations, and the period
of time is short enough to yield accurate answers to
such basic questions as the number of hours worked
and the payments received during this recall period. In
contrast, the health module uses a four-week recall
period. This relatively long recall period is used
because most people are not ill in any given week.The
four-week recall period allows more observations of
illness for a given sample size than would be obtained
using a one-week recall period. Since illnesses are
important events, respondents can be expected to
remember many details of their episodes of illness during
the past four weeks.
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
48
However, if an analyst wants to study the impact
of illness on earnings or work effort, these different
recall periods will complicate the analysis.The analyst
cannot tell whether the illness took place before or
during the period for which the earnings and hours
data were collected.This could be resolved either by
adding questions to the health module to specify the
days during the recall period on which the respondent
was ill or by making the recall periods coincide, perhaps
with a compromise of two weeks for both modules
(bearing in mind the disadvantages in sector-specific
analyses of using a recall period different from the
“ideal” one for that module). Part of the job of integrating
the draft modules is to determine and judge
the tradeoffs being made, either confirming that they
are acceptable or altering them until a more appealing
tradeoff is reached.
Nomenclature and Coding Schemes
The questionnaire should be reviewed to check that
wherever similar questions are asked, the nomenclature
and coding schemes are the same. This should
reduce coding errors and simplify data analysis. For
example,many different modules allow the respondent
to choose the time unit (for example, hour, day, week,
or month) that they find most convenient when
responding to questions regarding time or payments
over time (such as wage rates, the length of time spent
gathering firewood, and the length of time covered by
a payment for water). The code numbers for these
time units should be the same throughout the entire
questionnaire; in the draft modules presented in this
book “day” is always coded as “3,” “week” is always
coded as “4,” and so on for other units of time.
Another example concerns the migration module,
the transfers received page of the transfers and
other nonlabor income module, and the transfer payments
page of the consumption module. All have
questions about where the migrant, donor, or recipient
lives. The coding scheme that categorizes this
information, whether it is the type of place (capital
city, other urban area, rural area, or overseas) or the
name of the place, should be uniform. Likewise, several
modules include questions about the relationship
between two individuals. It is usually a good idea for
these questions to use the same codes that are used in
the household roster module to indicate the relationship
of each household member to the head of the
household.
A particularly important task is to coordinate the
coding of items in the consumption expenditure
module with items in the price questionnaire. As
explained in Chapter 2, price data are needed to generate
regional and temporal price indices that enable
comparison of real expenditures of households interviewed
in different places and at different times.This is
done by matching the prices collected in the price
questionnaire with the consumption expenditure
information gathered in the consumption module. If
the items are not well matched, this task becomes
more difficult, and the resulting price indices will be
less accurate. In general, the goal should be a one-toone
correspondence between the items listed in the
consumption module and the prices collected in the
price questionnaire.For example,if questions are asked
on two or three varieties of rice or wheat in the consumption
module, a price for each variety should be
collected in the price questionnaire.
This should be relatively simple to do for almost
all food items. Nonfood items are more difficult. It is
usually not possible to obtain prices for durable goods
because they often come in many varieties (for example,
there are many kinds of bicycles or televisions).
However, for nondurable items, prices can be obtained
for well-defined examples. For example, there are
many kinds of shirts, but if a specific widely purchased
type of shirt can be defined, data on that type of shirt
can be collected in the price questionnaire and used as
an indicator of prices for all kinds of shirts. See
Chapter 13 for a detailed discussion of the price questionnaire,
including a list of suggested food and nonfood
items to include in it.
Choosing the Order of the Modules in the Household
Questionnaire
A final and very important question to address is the
order of the modules in the household questionnaire.3
It is natural and convenient to arrange the modules in
the order that they will be administered, so the key
issue here is the order in which the modules will be
administered and how this affects the physical design
of the questionnaire.
To put this issue in context, consider the traditional
fieldwork plan for a full LSMS survey.Each field
team works in its assigned primary sampling units
(communities) twice.The first time a team arrives in a
primary sampling unit, it works there for about one
week. The first half of the questionnaire, most of
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
49
which usually consists of the individual-specific modules,
is completed for each household. In addition, a
short module is administered that asks which household
members are best able to answer questions concerning
the specific household-level modules (agriculture,
household enterprises, consumption, and savings)
that will be filled out when the team returns to the
primary sampling unit about two weeks later. Figure
3.1 provides an example of such a module.
The field team works in a different primary sampling
unit during the following week, while the data
in the half-completed questionnaires from the first
primary sampling unit are entered into a computer by
a data entry operator (who does not travel with the
team) using a data entry program.The data entry program
checks the first half of the questionnaires for a
wide range of errors and inconsistencies. (This is discussed
more fully in Grosh and Munoz 1996.) The
team returns to the first primary sampling unit in the
third week, administers the rest of the questionnaire
(which mainly consists of household-level modules),
and resolves any problems or inconsistencies found by
the data entry program when the data from the first
half of the questionnaire were entered.
In several recent LSMS surveys, two different procedures
have been used in the fieldwork stage. One
procedure is that the data entry operator travels with
the field team. This option has become feasible with
the advent of small laptop computers that can be powered
by batteries, vehicle cigarette lighters, or solar
panels.This allows the whole questionnaire to be filled
out and checked using the data entry program during
a single trip to the primary sampling unit. In addition,
the second half of the questionnaire can be checked by
the data entry program almost immediately, so that
interviewers can return to the sampled households to
resolve any problems detected by the program.
The other procedure, used when a scaled-down
LSMS survey is being implemented, is to complete all
of the interviews in a single trip to the primary sampling
unit and sometimes even in single visits to each
household. This procedure will have a serious disadvantage
if the data entry operator does not travel with
the team, because none of the data can be checked in
time to return to the households to resolve problems
detected by the data entry program. If the data entry
operator travels with the team, there is little difference
between this procedure and the former procedure,
except that a full LSMS questionnaire will require
more visits to each household during the sole trip to
the primary sampling unit.
Given these different possible fieldwork plans,
there are several basic principles about how to order
the modules in the household questionnaire.The first
principle is that any modules on topics that respondents
might consider sensitive should be put at the end
of the questionnaire.This gives the interviewer time to
develop a rapport with the household members,
which should increase the probability that they will
answer questions on sensitive issues, and do so truthfully.
It also means that if the respondent breaks off the
interview in response to a sensitive question, only the
data from that last module or modules are lost. Finally,
by this point in the interview, any interested onlookers,
such as family members and neighbors, may have
wandered away, making it possible to administer the
more sensitive portions of the questionnaire with
greater privacy. Education, housing, migration, and in
some cases health4
are usually good topics with which
to open the interview,because people generally do not
mind talking about these topics. In contrast, fertility,
savings, credit, and transfers and other nonlabor
income are among the most sensitive topics in the
household questionnaire.
A second principle concerns bounded recall periods.
In past LSMS surveys in which the interviewer
made two visits two weeks apart to each household,
some parts of the questionnaire used bounded recall
periods; in other words, questions were asked such as
“How much has your household spent on rice since
my last visit?”As explained in Chapter 5,using bounded
recall periods can increase the accuracy of the
respondents’ answers. Obviously, if bounded recall
periods are used in certain modules, these modules
must be administered in a second visit to the household
and thus be included in the second half of the
questionnaire. The two modules in Volume 3 that
explicitly use bounded recall periods are those on
consumption (Chapter 5) and household enterprises
(Chapter 18).5
A third consideration is the selection of respondents.
As explained above, several modules (including
the consumption, agriculture, household enterprises,
savings, housing, and environmental modules) collect
much or all of their data at the household level, which
means that the questions are answered by the household
member most knowledgeable about that topic.
With the exception of the housing module, these
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
50
FIGURE3.1:MODULEFORCHOOSINGRESPONDENTSTOBEINTERVIEWEDINTHESECONDHALFOFTHEQUESTIONNAIRE
RESPONDENT:THEPERSONBESTINFORMEDOFTHE
ACTIVITIESOFTHEHOUSEHOLDMEMBERS
FULLNAMEOFTHERESPONDENT:_____________________________________________IDCODE
1.Whoshopsforthefoodforyourhousehold?5.Duringthepast12monthshasanymemberof
yourhouseholdparticipatedinagricultural
NAME:_____________________________________IDCODEproduction,forestry,orraisinglivestock?
YES.....1
NO......2(¨8)
2.Whoinyourhouseholdknowsmostaboutthe6.Whoisthepersonwhoknowsmostaboutallthe
non-foodexpensesofthemembersofyourhousehold?agriculturalandlivestockactivitiesofthemembers
ofyourhousehold?
NAME:________________________________________IDCODE
NAME:__________________________________
7.Inadditiontothisperson,whoelseinyour
householdmanagesplotsoflandowned
3.Whoinyourhouseholdknowsmostaboutthemiscellaneousorrentedinbythehousehold?Whoisresponsible
incomeandtransfersreceivedfromotherhouseholds?forplotsthatarerentedoutbythehousehold?
NAME:________________________________________IDCODEID
NAMECODE
4.Whoinyourhouseholdknowsmostaboutthesavingsinyourhousehold?
NAME:________________________________________IDCODE
RESPONDENTSFORSECONDHALFOFQUESTIONNAIRE
IDCODE
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
51
FIGURE3.1:MODULEFORCHOOSINGRESPONDENTSTOBEINTERVIEWEDINTHESECONDHALFOFTHEQUESTIONNAIRE
8.9.10.
YES..1ENTERPRISEID
NO...2IDFULLWRITTENDESCRIPTIONCODENAMECODE
(»NEXTMODULE)1
2
3
4
5
Overthepast12months,hasanyonein
yourhouseholdoperatedanynon-
agriculturalenterprisethatproducesgoods
orservices(forexample,artisan,
metalworking,tailoring,repairwork,and
processingandsellingyouroutputsfrom
yourowncrops,ifdoneregularly)orhas
anyoneinyourhouseholdownedashopor
operatedatradingbusiness?
Whatkindofenterprisesdoesyourhouseholdoperate?
PROBETODETERMINEINDUSTRIALSECTORINWHICHENTERPRISE
OPERATES.
Whoismostinformedaboutand/orinchargeof
day-to-dayoperationsoftheenterprise?
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
52
modules are quite lengthy. Thus for each of these
modules it is usually best for the interviewer to ask
which member would be the most appropriate
respondent during the first visit to the household, and
then make an appointment to interview that person at
a later, more convenient time. In the traditional twovisit
fieldwork plan, this implies that these modules,
except perhaps housing, should be administered during
the second visit and thus should be located in the
second half of the questionnaire. However, if the team
travels only once to the primary sampling unit,it is still
feasible to make appointments for later in the day or
for another day, which gives survey designers more
flexibility in deciding where to place these modules in
the questionnaire.
A fourth principle relates to the logistics of data
entry. The individual-level modules include many
more questions for which strict range and consistency
checks can be built into the data entry program than
do the modules on consumption, agriculture, and
household enterprises.6
If the whole questionnaire is
completed using the traditional LSMS fieldwork plan
(two visits two weeks apart), all the individual-specific
modules except the credit module should be administered
during the first interview. (The credit module is
probably too sensitive to be administered during the
first interview.) This will allow the survey team to
enter the data from these modules and to detect any
apparent errors or inconsistencies that could then be
resolved in the second interview. If the data entry
operator travels around with the interview team, the
data from the interviews can be checked in a matter of
hours; thus where these modules appear in the order
of the household questionnaire becomes less important
for the purposes of data entry.
Given these principles, and some common sense,
more specific advice can be given. Each household
questionnaire should have the metadata module at the
very front, since much of the information that module
collects (such as whether the interviewer successfully
located the household, the date of interview, and the
language in which the interview was conducted)
becomes apparent at the very beginning of the interview.The
next module should be the household roster;
this must be completed before any other module
because it determines who is and who is not a household
member, and thus determines the people to
whom all the other modules will apply. However, at
least one page of the household roster, the one with
the names of all the household members, is usually
placed further back in the questionnaire so that the
names on that page can be seen during the administration
of all individual-level modules.Thus the physical
placement of this page will not reflect the time
during the interview when it is filled in. (For details
see the discussion on the fold-out roster page in the
fourth section of this chapter.)
After the household roster, it is useful to fill out
the form on selecting household respondents shown
in Figure 3.1; this form can be administered to the
same person who answered the household roster questions
(usually the head of household or the person
most knowledgeable about other household members).
It is useful to collect this information early
because it can be used to save household members’
time by interviewing them sequentially using “miniinterviews.”That
is, after the interviewer has administered
the form that identifies the relevant respondents
for the household-level modules, he or she should
administer all of the modules that are clearly individual-specific
(except the credit and fertility modules) to
each household member, finishing all such modules
with one member before interviewing another member.
These are the education, health, employment,
migration, and time use modules. Some household
members will not need to be interviewed further, and
thus their mini-interviews will consist of the interviewer
administering only these modules. In contrast,
other household members will also be the respondents
for some of the household-level modules. For example,
the respondent for the housing module can also
provide answers for the questions in that module as
part of his or her mini-interview. Using this method,
the interviewer can obtain all of the information
needed from each individual in a way that minimizes
the use of respondents’ time; once a respondent finishes
the mini-interview he or she can leave or start some
activity without further interruption.
Within this group of individual-level modules,
those on education and migration should be administered
first since the information they collect is not very
sensitive. Some employment information can be sensitive,
particularly questions concerning wages, so this
should be one of the last of the individual-level modules
to be completed, if not the very last. If the short
health module is used, it can be put near the front.
However, if the standard or expanded version is used, it
should be placed toward the end of the individual-
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
53
specific modules because of the sensitive nature of
some of the questions in this version (see endnote 4).
Which modules should go near the end of the
questionnaire? Because the three most sensitive modules
are those that collect information on savings,
credit, and transfers and other nonlabor income, these
three modules should probably be put at the end of
the questionnaire.Another potentially sensitive topic is
fertility. In countries in which fertility is particularly
sensitive, it should come immediately before the savings,
credit, and transfers and other nonlabor income
modules.
Where should the other modules go? If the traditional
“two visits two weeks apart” interview system is
used, the consumption and household enterprise
modules should be in the second half of the questionnaire
since these modules often use a bounded recall
period, namely the time since the interviewer’s previous
visit. Modules that are long and also need to be
administered to specific respondents—the consumption,
agriculture, household enterprise, and environment
modules—should also be completed in the second
visit. Finally, as discussed above, the housing
module can be administered in the first interview
because it is unlikely to contain any sensitive questions.If
the two-visit system is not used,these modules
can be put anywhere between the individual-specific
modules and the more sensitive modules.
Finally, the goal of saving respondents’ time by
conducting mini-interviews with each respondent
(who can leave after his or her mini-interview is finished)
is complicated by the fact that the household
enterprise, agriculture, miscellaneous income, and
credit modules consist of a mixture of household-level
and individual-level questions. For example, in the
standard and expanded versions of the agriculture
module, individual household members are asked
whether they have worked on specific plots of land.
However, these questions cannot be asked until several
other questions have been asked about the different
plots of land owned and rented in by the household
members—and such questions would be awkward to
ask in a form as simple as the one shown in Figure 3.1.
The best way to resolve this problem will
depend on which modules and which versions of
these modules are included in the household questionnaire,
so it is difficult to provide general advice.
However, one way to reduce the time burden on
household members is to identify all of the people
who still need to be interviewed after the “miniinterviews”
are completed, as this will allow people
to leave if they do not need to be interviewed further.
Continuing the agriculture example, note that
the form in Figure 3.1 identifies all of the household
members who either manage or work on a plot of
land. Household members who do not fit this
description and who are not needed to complete
any other household-level module can leave after
their “mini-interview” is finished.
This completes discussion of the fourth step of
integrating the draft modules and combining them
into a complete set of questionnaires. The primary
focus has been on the household questionnaire, since
the community questionnaire is much smaller. (See
Chapter 13 for a detailed discussion of the community
questionnaire.) Designers of prospective surveys can
consult the questionnaires used in previous LSMS surveys
by downloading them from the LSMS website,
http://worldbank.org/lsms/lsmshome.html.
Translating and Field Testing the Draft
Questionnaires
After the draft modules have been combined into a
complete set of household, community, and price
questionnaires, they need to be translated and field
tested.7
The field test is particularly important because
it is the last check on the design of the questionnaires
before the survey is implemented.
Translation
It may be necessary to translate the questionnaires for
three reasons, each of which has different implications
for the design of the survey. The most common and
most important reason is that respondents may speak a
range of different languages. In many countries more
than one language is spoken. In these countries quality
control requires that a separate questionnaire be
produced for each of the major languages spoken in
the country,with every question written out verbatim.
Scott and others (1988) demonstrated how this
procedure greatly increases the accuracy of the data
collected.They conducted an experiment designed to
measure interviewer errors when the interviewer had
to translate each question during the interview. For
example, the interviewer may have had to use a questionnaire
written in English to conduct an interview
in Tagalog or Cebuano or a questionnaire written in
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
54
French to conduct an interview in Baoule or Dioula.
The interviewers’ error rates were two to four times
higher when they translated questions during the
interview than when they used questionnaires already
written in the languages used by the respondents.
While the final versions of the questionnaires
must be translated from the national (official) language
to produce verbatim questionnaires in the other languages
used in the country, the preliminary drafts of
the questionnaire can be developed using only the
national language. Ideally, the version of the questionnaire
to be used for the field test should be translated
into each of the languages that will have a final written
version of the questionnaire. In practice, field tests
are often done using only oral translations of the
national language version of the questionnaire. Thus
the wording in the local language interviews during
the field test may not correspond exactly to the wording
that will be used in the written translations of the
final questionnaire.While this is an imperfect way to
proceed, it is often a reasonable tradeoff given the high
costs, in both time and in money, of field testing the
questionnaires in each language.
After the final version of the household questionnaire
has been translated into another language, the
translation needs to be carefully checked.The best way
to do this is to use “back translation.”That is, after the
questionnaire is translated from the language in which
it was developed into the languages in which it will be
administered, someone should translate the versions in
those languages back into the original language.After
this “back translation” has been accomplished, the two
versions in the first language should be checked.
Where there is a discrepancy in wording or meaning
between the two versions, the translation should be
carefully checked.A person or group of people familiar
with the purpose of the questions should do the
first translation. The back translation should be done
by someone who was not intimately involved in
designing the questionnaire. Any ambiguities and
errors must be noted and corrected in the translated
version rather than being “fixed” only in the back
translation version.
Most previous LSMS questionnaires were printed
only in the national languages of the countries studied,
so multilingual interviewers had to be employed
to conduct interviews in the most commonly used
local languages. Occasionally a few key questions or
phrases were translated into the local languages and
written down in the interviewers’ manual. In the case
of the least common languages, local interpreters were
used when none of the interviewers spoke the language.
In this respect, while previous LSMS surveys
have conformed to normal survey practice, they have
not reached the cutting edge of quality control as
defined by the World Fertility Surveys.The guidelines
used in those surveys require that questionnaires be
prepared in all languages used by more than 10 percent
of the sample and that a minimum of 80 percent
of the sample be interviewed using questionnaires
written in the respondents’ native language.
Future LSMS and similar multitopic surveys
should make greater efforts to translate the household
questionnaire into local languages. When preparing
these translations, the questionnaire should always be
worded in the way that the language is commonly
spoken,using relatively simple terms and avoiding academic
or formal language.The gap between the spoken
and written languages and the difficulty of striking
a balance between simplicity and precision may be
greater in local languages, especially ones that are not
commonly used for reading and writing.The translators
should therefore be especially careful to try to find
an appropriate balance.
Two examples illustrate the kind of problems that
can occur.The question “¿Estuvo enferma en las últimas
cuatro semanas?”literally asks,in Spanish,whether
the respondent was sick in the past four weeks.
However, in spoken Spanish in Chile it could be
understood as a polite euphemism for asking whether
a woman has had a menstrual period in the last four
weeks.An even more difficult problem in wording was
revealed in the field test in Nepal.Apparently the most
natural Nepali phrasing for “Have you been ill?” is
closer to “Have you been to the doctor?”The change
in meaning from what was intended appeared in the
field test several times when respondents answered
“No, I couldn’t afford to go,” clearly an inappropriate
response to the question “Have you been ill?”
The second reason why the questionnaires may
need to be translated is that sometimes the international
experts working on the survey design team do not
speak the national language well enough to design the
questionnaires in that language. This happened in the
case of the Vietnam LSMS questionnaires, which were
developed jointly in English and Vietnamese. In contrast,
the LSMS questionnaires used in Latin American
countries have been drafted only in Spanish by teams of
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
55
local and international experts who are fluent in that
language. When translation is a necessary part of the
development of the questionnaires, each draft of the
questionnaire must be translated, which may require a
substantial amount of money and can also increase the
time needed for designing the questionnaires.
The third and final reason for translating questionnaires
is to produce a questionnaire in one or
more of the major international languages (English,
Spanish, or French) in order to encourage the international
research community to use these data in their
policy analysis. Such translations need not be done
until after the final questionnaire is developed, and
back translations are not needed.
Field Testing
After draft versions of the household, community, and
price questionnaires have been assembled and (if necessary)
translated, they must be tested in the field.The
field test is one of the most critical steps of the survey
design process. The goal is to ensure that the questionnaires
are capable of collecting the information
that they are intended to collect. A field test should
address the adequacy of the draft questionnaires at
three levels:
• The Questionnaire as a Whole. Is the full range of
required information collected? Is the information
collected in different parts of the questionnaire
consistent? Are any variables unintentionally dou-
ble-counted?
• Individual Modules. Does the module collect the
intended information? Have all major activities
been accounted for? Are all major living arrangements,
agricultural activities, and sources of in-kind
and cash income accounted for? Are some questions
missing? Are some questions redundant or
irrelevant?
• Individual Questions. Is the wording clear? Do any
questions allow for ambiguous responses? Are there
multiple interpretations? Have all responses been
anticipated and coded?
It is important for a field test to include households
from all major socioeconomic groups. For
example, a sample should include: rural and urban
households; individuals employed in the formal sector,
in the informal sector, and in agriculture; and farmers
in each main agroecological region, in each production
scheme (independent farming,renting,sharecropping,
and cooperative farming), and so forth. The
households should not be selected at random. Instead,
different types should intentionally be included so that
all of the various situations likely to be found during
the survey are observed during the field test.
Experience with LSMS surveys has shown that
field tests should be conducted using at least 100
households.To get enough responses for each module
of the questionnaire, it may be necessary to visit additional
households to conduct partial interviews in
which only those modules that apply to a relatively
small number of households are administered. For
example, the original 100 households may not include
enough pregnant women or people who have been ill
in the month preceding the interview to determine
whether the fertility and health modules, respectively,
are well designed. In such cases survey designers
should find additional households that contain pregnant
women or ill people and have interviewers
administer only the fertility or health module to those
households.8
A field test usually takes about one
month to complete—about one week for interviewer
training, two to three weeks of fieldwork (interviewing),
and one or two weeks to discuss the findings and
finalize the questionnaires.More time is required if the
final questionnaires are to be produced in more than
one language, because each version of the questionnaire
should be field tested.
While the full field test should cover 100 or more
households, much can also be learned from preliminary
smaller tests. A general rule of thumb is that
about half of the problems will show up in the first 10
households interviewed. In one recent field test, international
experts wrote six pages of comments about a
single module after interviews were completed for
only three households. Such small-scale preliminary
field tests are often particularly appropriate for new or
difficult modules. Yet survey designers must understand
that these are precursors to a full-size field test of
the whole questionnaire, and not a substitute.
The personnel involved in a field test should
include the survey design team, a few experienced
interviewers or field supervisors, and a few of the people
consulted by the survey design team, including
both policymakers and research analysts. It may also be
helpful to include people with experience working on
past LSMS or similar multitopic surveys.All of the participants
should divide into a small number of teams,
each of which includes at least one person with each
kind of expertise.
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
56
There should only be a few teams involved in the
field test, usually around three or four. Mechanisms
should be set up to enable the teams to contact each
other during the field test so that they can compare
notes on the problems they encounter and the solutions
they have tried. A good way to set up such
mechanisms is to have all of the teams working
together for the first few days, perhaps in the capital
city.This means that the teams will be in contact with
each other every evening during the period when the
first and often biggest flaws in the draft questionnaire
are uncovered. In some cases the team members can
agree on modifications to the questionnaire during
the field test itself, which allows these modifications to
be field tested.
Each interview during the field test should
include, at minimum, the respondent, the interviewer,
and an analyst or senior survey specialist. During the
field test it is acceptable for the analyst or survey specialist
to interrupt the interview tactfully in order to
refine the wording of a question or the responses coded
for it. Of course, in the actual survey the interviews
should be conducted in private, and the interviewers
should adhere to the wording of the questionnaire.
The interviewers used in the field test should be
drawn from the experienced staff of the statistical
agency. They should be good interviewers—familiar
with basic interviewing practices and able to distinguish
between problems caused by deficiencies in the
questionnaire and problems caused by their lack of
familiarity with the questionnaire. The interviewers’
training should focus on the purpose of the survey and
the structure and format of the questionnaire. One
week of training is usually sufficient, followed by two
or three weeks of household interviews.
Survey planners should set aside 1–2 weeks immediately
after the field test to review the field test results
and debate how to modify the questionnaire in light of
those results.The group involved in the field test should
go through the questionnaires, module by module, and
discuss any problems that arose.At this stage, the team
should bear in mind that the length of time required
for each interview will fall dramatically when the
interviewers are well trained and have become familiar
with the questionnaire. As mentioned above, the typical
field test interview will be at least twice as long as
the average interview in the actual survey.
The data from the field test should not be entered
in the computer or examined for any analytical purposes,
because in most cases the sample is both nonrandom
and very small. However, the questionnaires
from the field test can be used to check the performance
of the data entry program.
The personal participation of all senior staff
(including analysts) is fundamental for both the field
test and its evaluation. The following anecdote illustrates
this point. In one country, before the field test a
manager in the statistics office asserted that collecting
information on family assets would be impossible
because respondents would fear that the information
would be used for tax purposes. The module was
included in the field test, and no unusual difficulties
were encountered. But the manager who opposed the
module did not witness the field test, and some of
those who did participate in the field test did not participate
in the module’s evaluation. Despite the successful
field experience,the module was removed from
the questionnaire, largely because key decisionmakers
did not fully participate in the survey design process.
Many small changes are generally made to questionnaires
as a result of field testing, including changes
in the wording of some questions, in questionnaire
format, and in answer codes. If either the questionnaire’s
structure or the way in which certain variables
are measured is changed substantially, all of the parts of
the questionnaire that have been so modified must be
tested again.This can delay the survey, but one way to
reduce the probability of such a delay is to begin the
field test with two or more versions of the most difficult,
contentious, or important modules in the questionnaire.
If one version clearly works the best, there is
no need to do another field test because that version
has already been field tested.
Ideally, the household, community, and price
questionnaires should all be field tested at the same
time.This allows the survey design team to evaluate all
of the questionnaires together, taking into account the
possibility that changes in one questionnaire may have
implications for the design of the others. Simultaneous
testing of the three questionnaires can also reduce
travel costs since, like the household questionnaires,
the community and price questionnaires should be
tested in a variety of locations.
Regrettably, in several past LSMS surveys the survey
teams neglected to field test the community and
price questionnaires,concentrating solely on the household
questionnaire.The community and price modules
were tested late and haphazardly or, in some cases, not
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
57
tested at all.It is probably not coincidental that the users
of the data from many previous LSMS surveys have
often had more complaints about the community and
price data than about the household data. If there is not
enough staff time to test all three questionnaires at once,
it is important to ensure that separate,rigorous field tests
are done of the community and price questionnaires.
The health and education modules discussed in
this chapter often include detailed facility questionnaires
(in other words, school or health clinic questionnaires),
which can be very complex (see Chapters
7 and 8 for details).It is essential to field test these facility
questionnaires.During the field test the survey team
should be sure to visit each type of facility covered by
the facility questionnaire. For example, field testing a
health facility questionnaire should involve visits to
public health posts, public clinics, private doctors’
offices, public hospitals, and private hospitals in both
urban and rural areas. Similarly, field testing a school
questionnaire should involve visits to public and private
schools, primary and secondary schools, and schools in
urban and rural areas. Since field testing a facility questionnaire
is a major undertaking in its own right, it is
probably best to conduct such a field test separately
from the field tests of the other questionnaires.
Rules for Formatting Survey Questionnaires
The formatting of survey questionnaires is not a separate
step in the overall survey design process. Rather,
it influences how the third, fourth, and fifth steps are
carried out. Good questionnaire formatting can make
a tremendous difference in the quality of the data collected.This
section discusses formatting in detail, making
very specific recommendations about how questionnaires
should be formatted.9
There is, of course, more than one way to format
household survey questionnaires. Most of the benefits
of good formatting come from selecting a formatting
convention and following that convention consistently,
rather than choosing the “best” convention from
among several possible options. For example, in LSMS
questionnaires uppercase and lowercase letters are used
to distinguish words spoken aloud during the interview
from instructions to the interviewer, but this
could be done in other ways, such as using different
colors or different fonts. Once a convention is selected,
it is extremely important to use it consistently
throughout the whole questionnaire.The convention
chosen should be the one that is clearest and most
likely to minimize the possibility of errors.The draft
questionnaires presented in this book follow the formatting
conventions explained in this section, which
have been used frequently in past LSMS surveys, with
successful results.
Questionnaire format is important because a good
format minimizes potential interviewer and data entry
errors, which improves the accuracy of the data and
reduces the time needed to check the data before
making them available to data analysts.The objectives
underlying a given survey can occasionally have implications
for formatting, so some aspects of formatting
will vary from country to country. Even so, almost all
of what has been learned about questionnaire format
in previous LSMS surveys will be applicable to new
surveys. Thus the formatting guidelines presented in
this section are recommended for all LSMS and similar
multitopic surveys, and for other surveys as well.
Identifiers
Every person or object for which data are collected in
a survey must be uniquely identified. This usually
requires two or three separate codes. The first code
identifies the household. The second code identifies
the person or object of interest, such as an individual
household member, a household business, or a plot of
land. Sometimes there is a third code, which applies,
for example, to all children ever born to each woman
in the household or to the assets of each business operated
by the household.
Whenever possible, the identification codes for
the second or third levels of observation should be
preprinted on the questionnaire pages to which they
pertain. For example, the individual identification
code for each household member should be printed
on all pages that collect data on individual household
members.This ensures that the codes cannot be omitted
and avoids any errors that would occur if the interviewer
were to write down the wrong codes. An
example of these codes appears in the far left column
of Figure 3.2, which presents the short version of the
education module.
The importance of adequate identifiers is so obvious
that it is hard to believe mistakes can be made, but
they can. In one health survey the questionnaire consisted
of two sheets of paper stapled together. One
contained information on the household, while the
other contained information on individuals. In order
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
58
FIGURE3.2:ILLUSTRATIONOFINDIVIDUALIDENTIFICATIONANDSKIPCODES(EDUCATIONMODULESHORTVERSION)
1.2.3.4.5.6.7.8.9.10.11.
I
D
C
O
D
E
Haveyou
ever
attended
school?
Areyou
currently
enrolled
in
school?
Whatisthe
highest
gradeyou
have
completed
inschool?
Whatis
the
highest
diploma
youhave
attained?
Were
you
enrolled
inschool
during
thepast
12
months?
Inwhat
gradeare
you
currently
enrolledin
school?
Whatisthe
highest
diploma
youhave
attainedso
far?
Isthe
schoolyou
are
currently
enrolledin
publicor
private?
Haveyou
ever
repeated
agradeof
school?
Howmany
timeshave
yourepeated
agradeof
school?
A.TuitionB.ParentC.Uni-D.Text-E.OtherF.Meals,G.Other
andotherAssoci-formsbooks?educational
materials
transpor-expenses
YES..1YES..1requiredationandtation(extraYES..1
NO...2YES..1(»9)fees?fees?other(exerciseand/orclasses,NO...2NUMBEROF
(»NEXT(»6)NO...2clothing?books,lodging?optional(»NEXTREPEATED
PERSON)NO...2(»10)pens,etc.)?fees)?PERSON)GRADES
1
2
3
4
5
6
7
8
9
10
11
12
PUBLIC..1
PRIVATE
SECU-
LAR....2
PRIVATE
RELI-
GIOUS..3
¨¨NEXT
PERSON
Howmuchhasyourhouseholdspentonyoureducationinthelast12
monthsfor:
PUT
CODES
FOR
DIFFER-
ENT
GRADES
HERE
PUT
CODES
FOR
DIPLOMAS
HERE
PUT
CODES
FOR
DIFFER-
ENT
GRADES
HERE
PUT
CODES
FOR
DIPLOMAS
HERE
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
59
to facilitate data entry, the two pages of the questionnaire
were separated. Unfortunately, the household
identifier was not put on the page for individuals,
making it impossible to link the two parts of the survey
with each other after the data were entered.
Questionnaire Layout
The LSMS questionnaires are designed so that only one
copy of the questionnaire is needed for each household.
In contrast, some surveys use one household questionnaire
and a separate set of individual questionnaires.This
requires that household identification codes be copied
perfectly onto all of the individual questionnaires.While
perfection is always sought,it is rarely achieved,and separate
questionnaires create the risk of improper matching.
This is illustrated in the case of the Russian
Longitudinal Monitoring Survey. Although care was
taken to ensure accurate coding and matching, many
errors were introduced. For the first round of the survey,
which was held in the summer of 1992, there were
3 percent fewer individual questionnaires than had been
expected given the number of household members
identified in the household questionnaires. By the summer
of 1993, in the third round of the survey, this discrepancy
had grown to about 9.5 percent.
Putting all of the information into a single household
questionnaire implies the need for a grid of some
kind whenever there are two or more of a particular
unit of analysis in a household. For example, a household
often includes several people, may have several
plots of land,and may grow several different crops.The
grid typically used in LSMS surveys has questions
arranged across the top and units of observation (people,plots,or
crops) down the side;in other words,each
question is a column and each unit of observation is a
row. An example of this is shown in Figure 3.2; note
that the identification codes for the units of observation
(household members) are printed on the left side
of the grid page.
Sometimes the interviewer must fill in the code in
the first column,as in Question 2 of Figure 3.8 (which
is discussed below), but this practice should be minimized
to reduce the possibility of introducing errors
when writing down such codes. In the grids for individuals,
the lines can be differentiated by alternating
shaded and unshaded blocks (as in the draft modules
inVolume 3 of this book) or by using a different color
for each row or block of rows. This helps an interviewer
record the information on the correct line.
Exceptionally large households sometimes have so
many members that there are not enough lines in the
grids for all household members. In these cases a second
copy of the household questionnaire will be
required, and care must be taken to ensure that the
right household and individual numbers are used. As
explained in Chapter 4, a coding scheme is needed to
distinguish between the first and second copies of the
questionnaire filled out for large households. For
example, the individual numbers in the second copy
should be changed to start with 13 instead of 1
(assuming that the first questionnaire has room for 12
household members).This is a reasonable approach for
large households, but it also introduces a potential
source of error; survey designers should set the format
of the grids to accommodate as many individuals as is
practical. Previous LSMS questionnaires have typically
had space for 12–15 individuals.
In cases where the unit of analysis is such that
there is only one observation per household (for
example, one dwelling per household), the questions
pertaining to that unit can be arranged in a single column
down the page. One problem with a single column
of questions is that much of the page is left blank.
To save paper, two or more columns may be put on
one page, as long as it is clear that there is no horizontal
relationship among the questions in the different
columns.An example of this format is provided in
Figure 3.3, which shows the first page of Part C of the
standard housing module.
Fold-Out Roster Page
The household roster page of the household questionnaire
is printed so that it extends to the left of the
pages that pertain to individuals in the household.
Most importantly, the names of each individual member
of the household on the roster page are visible
when filling out the other individual-specific pages of
the household questionnaire.This has been done four
different ways in LSMS surveys,as illustrated by Figure
3.4.
In the method shown in Format 1, the sheets in
front of the roster are shorter than the cover, the roster,
and the sheets that follow the roster. The most
common method is shown in Format 2. The roster
sheet is folded out to extend beyond the body of the
questionnaire and its covers. In Formats 1 and 2 the
roster page is placed behind all of the pages that pertain
to individuals, so that the names on the household
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
60
FIGURE3.3:ILLUSTRATIONOFPRECODING(PARTCOFHOUSINGMODULE)
7.Doyouhavelegaltitletothedwellingoranydocumentthatshows
ownership?
1.Isthisdwellingownedbyamemberofyourhousehold?YES...........................1
NO............................2
YES.......................1
NO........................2(»11)8.Whattypeoftitleisit?
2.Howdidyourhouseholdobtainthisdwelling?
9.Whichpersonholdsthetitleordocumenttothisdwelling?
WRITEIDCODEOFTHISPERSONFROMTHEROSTER
1STIDCODE:
3.Howmuchdidyoupayfortheunit?2NDIDCODE:
4.Ifyoumakeinstallmentpaymentsforyourdwelling,whatistheamountof10.Couldyousellthisdwellingifyouwantedto?
theinstallment?
YES.......................1
WRITEZEROIFTHEHOUSEHOLDDOESNOTMAKENO........................2(»13)
INSTALLMENTPAYMENTS
AMOUNT(UNITSOFCURRENCY)11.Ifyousoldthisdwellingtodayhowmuchwouldyoureceiveforit?
TIMEUNIT
AMOUNT(UNITSOFCURRENCY)
5.Inwhatyeardoyouexpecttomakeyourlastinstallmentpayment?
12.Estimate,please,theamountofmoneyyoucouldreceiveasrentifyou
YEARletthisdwellingtoanotherperson?
6.DoyouhavelegaltitletothelandoranydocumentthatshowsAMOUNT(UNITSOFCURRENCY)
ownership?TIMEUNIT
YES.......................1
NO........................2
13.Doyourentthisdwellingforgoods,servicesorcash?
YES.......................1
NO........................2(»26)
FULLLEGALTITLE,REGISTERED..1
LEGALTITLE,UNREGISTERED.....2
PURCHASERECEIPT..............3
OTHER.........................4PRIVATIZED.............................1
PURCHASEDFROMAPRIVATEPERSON........2
NEWLYBUILT............................3
COOPERATIVEARRANGEMENT................4
SWAPPED................................5(»6)
INHERITED..............................6(»6)
OTHER..................................7(»6)
TIMEUNITS:DAY........3MONTH.......6YEAR..9
WEEK.......4QUARTER.....7
FORTNIGHT..5HALFYEAR...8
¨¨QUESTION28
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
61
FIGURE3.4:ROSTERARRANGEMENTS
A4
A4
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
62
roster page are visible whenever individual questions
are asked.
An innovation in the Kagera Health and
Development Survey inTanzania was to make the roster
page a removable card, as shown in Format 3.This
was useful because the survey was designed to be
administered four times—every six months for two
years—to the same households. The roster card was
inserted into a pocket in the back of the questionnaire
in the first round of the survey. When the second
round started, the roster card was removed from the
first questionnaire and placed in the back pocket of
the second questionnaire. In this way, individuals
retained the same identification codes in each round.
A few follow-up questions guaranteed that individuals
who moved in or out of the household or were born
or died between rounds were counted appropriately.
In four rounds of interviews conducted over two years
for 800 households, none of the roster cards was lost.
However, this success may reflect the intensive supervision
carried out by the organizers of that survey, as
well as the relatively small sample size. This option
should probably not be used in situations with significant
quality control problems.
Format 4 was used in theTunisia questionnaire. In
this format each page is oriented as “portrait” (a vertical
page) rather than as “landscape” (a horizontal page)
and is spiral-bound so that it opens flat. Each questionnaire
page then consists of the full 11 x 17 inches
of the two-page spread.The roster folds out to the left.
In all four cases the line for each individual member
of the household on the roster page is aligned with the
corresponding lines on the other individual-specific
pages of the household questionnaire.
A final point regarding the fold-out roster page is
that it may be useful to have more than one such page
per questionnaire. A fold-out roster will be useful
whenever there are several pages of questions for the
same level of analysis and especially when there are
many rows on the grid. For example, in the agricultural
module one might make rosters for crops grown or
for plots of land.A fold-out roster page would be particularly
helpful for the household enterprise module.
Precoding
All of the potential responses to almost all of the questions
in the questionnaire should be given code numbers
so that the interviewer records only code numbers,
as opposed to words or phrases, on the
questionnaire. In most cases these response codes
should be printed directly in the box where the question
appears, or next to the question if there is no box
around it.Where the list of codes is lengthy and applies
to several questions, it should be placed in a special
box on the border of each page for which it is needed.Alternatively,
if a list is very long it can be printed
on the back of the preceding page (making it visible
when the interviewer fills out the page in question).
An example of a box on a border of a page is the time
unit box shown at the bottom of Figure 3.3.
In past LSMS surveys fewer than a dozen questions
on the household questionnaire have required
the interviewer to write down words or phrases that
are given codes, usually by someone else, after the
interview.Precoding allows the data to be entered into
the computer straight from the completed questionnaire,
thus eliminating the time-consuming and errorprone
step of transcribing codes onto data entry
sheets.
Precoding requires that response codes be clear,
simple, and mutually exclusive, that they exhaust all
likely answers, that respondents will not all provide the
same response, and that none of the codes apply to
only a handful of respondents. Designing adequate
response codes requires extensive knowledge of the
phenomenon being studied as well as careful field testing.A
standard technique to ensure that the codes are
mutually exclusive is to add a qualifier where more
than one answer could apply—asking, for example,
“What was the main reason for dropping out of
school?” Other standard qualifiers are “What was the
first (or last, or principal) reason for...?” Alternatively,
spaces can be provided for multiple responses, with an
instruction to code all responses (up to, say, the two or
three most important) that apply.
A standard technique to ensure that codes encompass
all possible answers is to add an “other (specify
_______)” code to questions for which an explicit
enumeration of each possible response is impossible or
inconvenient. In past LSMS surveys the detailed
answers were almost never coded, so analysts usually
put all “other” responses into a single residual category.
One way to increase the probability that the information
recorded in the “other (specify _______)”
answers will be used at a later date is to enter it (as
text) into the computer, without assigning any codes
to the responses. This allows analysts to code any
answers that were not precoded in the data released to
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
63
the public. It also allows the designers of subsequent
surveys in the same country to review the answers that
were written in (especially in cases in which a significant
percentage of the responses were coded “other”)
and to modify their coding lists accordingly. In particular,
if most of the “other” responses fall into a single,
well-defined category, this category should have its
own code in any subsequent survey.
There is, of course, a limit to the kind of material
that can be covered even by well-designed, precoded
questions. But this limit may be less of a disadvantage
than it first appears. Because most analyses of LSMS
surveys use sophisticated quantitative techniques, it is
difficult for these analysts to make use of the
exploratory, qualitative information gathered in openended
questions. So even if such questions were asked,
the answers to these questions would not be used
much in analysis. If it is clear that some analysts do
need extensive information of an exploratory, qualitative
nature, the designers of a prospective survey may
wish to adopt a different data collection instrument or
even a new research technique. See Chapter 25 for a
thorough discussion of qualitative data collection
alternatives.
Verbatim Questions with Simple Answers
All questions in LSMS surveys are written out in
their entirety and are meant to be read out verbatim
by the interviewer.This is done to ensure that questions
are asked in a uniform way, since different
wordings may elicit different responses. For example,
the answers that a respondent gives to “Can you
read?” and to “Can you read a newspaper or magazine?”
will probably be somewhat different. Other
changes may subtly alter the time period referred to,
as in the change from “Have you worked since you
were married?” to “Did you work after you were
married?” Scott and others (1988) discuss some rigorous
field experiments that compared such verbatim
questionnaires with questionnaires in which the
topic was given for each question but the exact
wording was not. When the questionnaire that did
not contain the exact wording was used, 7 to 20
times more errors occurred than when the verbatim
questionnaire was used.
When choosing the wording of questions, it is
important to use terms that reflect the language as it is
commonly spoken. Using language that is too formal
or academic will make the interview stilted and
unnatural. For example, “Did you spend any time
doing housework?” followed, if necessary, by “...such
as cooking, mending, doing laundry, or cleaning?” is
better than “Did you spend any time engaged in
domestic labor, for example, preparing food, repairing
clothes, cleaning clothes, or cleaning house?” It is not
always easy to find terms that are simple, short, and yet
precise, but that should always be the goal.
In most cases the interviewer reads the question
aloud and marks the questionnaire with the code for
the answer given by the respondent. For example, for
the question,“Are you currently enrolled in school?”
the interviewer writes down a 1 for “yes” or a 2 for
“no.” For some questions the response categories are
part of the question—for example,“Is the school you
are currently enrolled in public or private?”There may
also be a few questions for which the wording of
respondents’ answers may vary even though the meaning
is the same.The best thing to do in such cases is to
have the interviewer read out all of the response categories.
For example, in Question 4 of Figure 3.5, after
reading “Compared to your health one year ago,
would you say that your health is...” the interviewer
should read the responses “much better now,”“somewhat
better now,” “about the same,” “somewhat
worse,” and “much worse.” If necessary, the interviewer
can explain the differences between the various
response categories. However, the reading out of
response categories should be used as little as possible,
because respondents may not listen to the full list
before answering, which can lead to errors.
The answers to the questions must be kept simple.
This means that additional filter questions are often
needed.Adding enough filter questions to ensure simple
answers can make the number of questions and
skips seem high.Many survey designers are tempted to
shorten the questionnaire or simplify the skip pattern
in a way that results in complex questions and answers.
This should be avoided since it will confuse some
respondents and is unlikely to save time.
Survey designers yielded to this temptation in the
agricultural module of the 1987–88 Ghana LSMS survey.
In that module the following question was asked:
“Do you or the members of your household have the
right to sell all or part of their land to someone else if
they wish?” The precoded answers (which were not
read out to the respondents) were “Yes,” “No,” “Only
after consulting family members who are not household
members,” and “Only after consulting the chief or
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
64
FIGURE3.5:ILLUSTRATIONOFCASECONVENTIONS(HEALTHMODULESTANDARDVERSION)
1.2.3.4.5.6.7.8.9.
I
D
C
O
D
E
ISTHIS
PERSON
ANSWER-
INGFOR
HIMSELF/
HERSELF?
COPYTHEID
CODEOF
THE
RESPOND-
ENTFROM
THE
HOUSEHOLD
ROSTER
Duringthelastfour
weeks,howmany
daysofyour
primarydaily
activitiesdidyou
missduetopoor
health?
Comparedwithyourhealthone
yearago,wouldyousaythatyour
healthis:
[READOUTANSWERSTO
RESPONDENT]
CHECKTHEAGEIN
YEARSOFTHIS
PERSON
Did[NAME]
experience
diarrheain
thelast7
days?
Wasit
mixedwith
blood?
Wasit
mixedwith
mucus?
Wasita
paleliquid?
Muchbetternow......10-6.............1
Somewhatbetternow..27-14............2
YES..1Aboutthesame.......3(»NEXTSECTION)YES..1YES..1YES..1YES..1
(»3)Somewhatworse.......415-39.......3(»24)NO...2NO...2NO...2NO...2
NO...2IDCODEDAYSMuchworse...........540ANDOVER.4(»11)(»11)1ST2ND3RD
1
2
3
4
5
6
7
8
9
10
11
12
10.
¨¨NEXT
SECTION
REDUCEDFOODOR
LIQUIDGIVENTO
CHILD...........1
GAVESPECIALFOODS
TOCHILD........2
ORALREHYDRATION
THERAPY.........3
OTHER
(SPECIFY_____)...4
NOTREATMENT.....5
Howdidyoutreatit?
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
65
the village elders.” It is not clear whether the respondents
could distinguish between the simple yes answer
and the yes answer qualified by the need for consultation.Thus
a different formulation might have been better.
The question could have been left as is but using
only simple “yes” and “no” codes.Then the interviewer
could have put a second question to those who
answered “yes,” worded as follows: “Do you need to
consult with anyone outside the household before selling
the land?”The response codes would be “Yes” and
“No.”Then a third question would be put to those who
answered “yes” to the second:“Whom must you consult?”The
response codes for this question would be for
“family member,”“village elders,”and other appropriate
categories.This formulation would have made the questionnaire
longer in terms of the number of questions
but would probably not have increased the interview
time since some sort of probing probably occurred in
the Ghana LSMS when the “yes” answer was given.
More importantly, keeping questions and answers simple
makes the interpretation of the data much clearer.
Skip Codes
Skip codes are used extensively in LSMS questionnaires.
Skip codes tell the interviewer which question
to proceed to after finishing the current question.
Some skip codes apply only when a particular answer
is given. In such cases an arrow and the number of the
question to skip to are positioned in parentheses next
to or below the individual response to which the code
applies. An example is given in Question 2 of Figure
3.2. If the answer to Question 2 is “yes,” the interviewer
should skip Questions 3, 4, and 5 and proceed
to Question 6. If the answer to Question 2 is “no,” the
interviewer should proceed to Question 3. In
Question 1 a similar construction is used, but when
the answer is “no” the interviewer is instructed to skip
all the remaining questions in the module for this
respondent and proceed to interview the next person.
Another kind of skip instruction applies regardless
of the response given to the question.When an arrow
and a question number or instruction are placed in a
box separate from the response codes, the skip instruction
contained in the box applies regardless of what
answer is given. An example of this is given in
Question 10 of Figure 3.5.
There are several advantages to extensive, explicit
skip codes.Interviewers do not have to make decisions
themselves, nor do they need to remember complicated
rules printed in the manual rather than on the
questionnaire.This helps ensure that instructions will
be followed uniformly.Well-placed skip codes ensure
that inapplicable questions are not asked.(Asking inapplicable
questions irritates respondents, wastes interview
time, and confuses data analysis.) Finally, explicit
skip codes imply that a “not applicable” code is almost
never used in LSMS questionnaires.
One way to check skip codes is to develop a flow
chart of the questions in each module. Flow charts
are useful both for checking the logic of the questionnaire
and for training interviewers. Figure 3.6
presents a flow chart of a typical health module used
in past LSMS surveys (which differs in several
notable ways from the health module presented in
Volume 3). The proportions of people who answer
yes at each branch are recorded based on results from
several previous LSMS surveys.The numbers of individuals
that would be asked each set of questions are
shown on the left, assuming a base of 10,000 individuals
in the sample.The flow chart makes it easy to
check whether the skip patterns lead people through
the module correctly. For example, it is possible to
check that the question on health insurance is asked
of all household members, not just of those who are
ill. Analyzing the whole household in this way gives
survey designers a better sense of the likely length of
time it will take to complete each interview than
does the number of pages or number of questions in
the questionnaire, because many questions will be
skipped for many individuals. (For further discussion
of the length of the questionnaire see the second section
of this chapter.)
Case Conventions
Everything that the interviewer should read aloud
should be written in lowercase letters. Instructions to
the interviewer should always be written in uppercase
letters.10
Answer codes should also be written in
uppercase, unless they are to be read aloud to the
respondent.This makes it easy to include instructions
on the questionnaire as opposed to relying on the
interviewers’ memory of the manual or of instructions
that they were given during their training. In Figure
3.5 instructions to the interviewer are printed in
Questions 1, 2, 4, and 5.These are in uppercase, as are
the answer codes in Questions 1 and 5. (The answer
codes in Question 4 are in lowercase because they are
to be read aloud to the respondent.)
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
66
Enumeration of Lists
There are two methods of gathering information
about long lists of items.A typical LSMS questionnaire
may use either method depending on particular
circumstances.
Consider the case in which one expects that a
large proportion of the items on the list will apply to
most households. For each item on this list a line is
put in the grid and the name and code number of the
item is printed on the questionnaire.This approach is
used in the consumption module, as shown in Figure
3.7. Although several dozen items are included, it is
expected that most households will have consumed
many of them.The first question is “Has your household
consumed [FOOD] during the past 12
months?”The interviewer first goes down the whole
list asking this “yes or no” question. Then the interviewer
returns to the first item that was consumed
and asks all the follow-up questions for that item
before proceeding to the next item. The complete
FIGURE 3.6: FLOW CHART OF HEALTH MODULE USED IN PREVIOUS LSMS SURVEYS
10,000 1 Were you ill or injured in the last 4 weeks?
YES (10-45%)
1000-4500
2 How many days in the last 4 weeks did you have
to stop doing your usual activities?
3 Was anyone consulted?
YES (40-80%)
400-3600 4 Who was consulted?
5 Where did you go for that consultation?
6 What was the cost of the consultation?
7 What means of travel did you use?
8 How long did it take to get to the place
of consultation?
9 How much did you spend on travel costs?
10 How long did you have to wait?
11 Did you have to stay overnight at the
clinic or hospital?
YES (5-8%)
20-288 12 How many nights did you stay?
13 How much did you have to pay?
1000-4500 14 Did you buy any medicines for this
illness or injury?
YES (60-90%)
600-4050 15 How much did you spend on medicines?
10,000 16 Do you have health insurance?
NO
NO
NO
NO
NEXT PERSON
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
67
FIGURE3.7:ILLUSTRATIONOFCLOSE-ENDEDLIST(PARTBOFCONSUMPTIONMODULE)
PURCHASESSINCELASTVISITPURCHASESTYPICALMONHOMEPRODUCTIONGIFTS
1.2.3.4.5.6.7.8.9.10.
Havethe
membersof
your
household
boughtany
[FOOD]
sincemy
lastvisit,
thatissince
[DAY/DATE
]?
Howmuch
didyoupay
intotal?
Howmany
monthsin
thepast12
monthsdid
your
household
purchase
[FOOD]?
Howmuch
doyou
usually
spendon
[FOOD]in
oneofthe
monthsthat
you
purchase
[FOOD]?
Howmany
monthsin
thepast12
monthsdid
your
household
consume
[FOOD]that
yougrewor
producedat
home?
Whatwas
thevalueof
the[FOOD]
you
consumed
inatypical
monthfrom
yourown
production?
Whatisthe
totalvalue
ofthe
[FOOD]
consumed
thatyou
receivedas
agiftover
thepast12
months?
YES.1
NO..2
NOYESCODE(»5)CURRENCYAMTUNITMONTHSCURRENCYMONTHSAMTUNITCURRENCYCURRENCY
Wheat(grain)1
Wheat(flouror
maida)2
Maize(flouror
grain)3
Jawar/Bajra4
FInerice(basmati)5
Coarserice6
Other
grains/cereals7
Gram8
Dal9
Groundnuts10
Liquidvegetable
oils(dalda)11
Ghee,Desighee12
Freshmilk13
Inthefollowingquestions,Iwanttoask
aboutallpurchasesmadeforyour
household,regardlessofwhichperson
madethem.
Hasyourhouseholdconsumed[FOOD]
duringthepast12months?Please
excludefromyouranswerany[ITEM]
purchasedforprocessingorresaleina
householdenterprise.
PUTACHECK(ߛ)INTHE
APPROPRIATEBOXFOREACH
FOODITEM.IFTHEANSWERTO
Q.1ISYES,ASKQ.2-13.
IFNONE
WRITE
ZERO,
¨7
IFNONE
WRITE
ZERO,
¨10
IFNONE,
WRITE
ZERO
Howmuch
didyou
buy?
Howmuch
didyou
consumein
atypical
month?
UNITCODES:
USECODES
WITHSTAR
WHENEVER
POSSIBLE
KILO*....1
GRAM*....2
POUND*...3
OUNCE*...4
LITER*...5
CUP*.....6
PINT*....7
QUART*...8
GALLON*..9
BUNCH...10
PECK....11
BUSHEL..12
TIN.....13
PIECES..14
DOZENS..15
BOTTLES.16
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
68
FIGURE3.7:ILLUSTRATIONOFCLOSE-ENDEDLIST(PARTBOFCONSUMPTIONMODULE)
PURCHASESSINCELASTVISITPURCHASESTYPICALMONHOMEPRODUCTIONGIFTS
1.2.3.4.5.6.7.8.9.10.
Havethe
membersof
your
household
boughtany
[FOOD]
sincemy
lastvisit,
thatissince
[DAY/DATE
]?
Howmuch
didyoupay
intotal?
Howmany
monthsin
thepast12
monthsdid
your
household
purchase
[FOOD]?
Howmuch
doyou
usually
spendon
[FOOD]in
oneofthe
monthsthat
you
purchase
[FOOD]?
Howmany
monthsin
thepast12
monthsdid
your
household
consume
[FOOD]that
yougrewor
producedat
home?
Whatwas
thevalueof
the[FOOD]
you
consumed
inatypical
monthfrom
yourown
production?
Whatisthe
totalvalue
ofthe
[FOOD]
consumed
thatyou
receivedas
agiftover
thepast12
months?
YES.1
NO..2
NOYESCODE(»5)CURRENCYAMTUNITMONTHSCURRENCYMONTHSAMTUNITCURRENCYCURRENCY
Inthefollowingquestions,Iwanttoask
aboutallpurchasesmadeforyour
household,regardlessofwhichperson
madethem.
Hasyourhouseholdconsumed[FOOD]
duringthepast12months?Please
excludefromyouranswerany[ITEM]
purchasedforprocessingorresaleina
householdenterprise.
PUTACHECK(ߛ)INTHE
APPROPRIATEBOXFOREACH
FOODITEM.IFTHEANSWERTO
Q.1ISYES,ASKQ.2-13.
IFNONE
WRITE
ZERO,
¨7
IFNONE
WRITE
ZERO,
¨10
IFNONE,
WRITE
ZERO
Howmuch
didyou
buy?
Howmuch
didyou
consumein
atypical
month?
UNITCODES:
USECODES
WITHSTAR
WHENEVER
POSSIBLE
KILO*....1
GRAM*....2
POUND*...3
OUNCE*...4
LITER*...5
CUP*.....6
PINT*....7
QUART*...8
GALLON*..9
BUNCH...10
PECK....11
BUSHEL..12
TIN.....13
PIECES..14
DOZENS..15
BOTTLES.16
YogurtandLassi14
MilkPowder15
BabyFormula16
Sugar(refined)17
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
69
enumeration of items consumed is done before asking
the follow-up questions so that respondents will
not be tempted to say that they have not consumed
something in order to shorten the interview by
avoiding the follow-up questions. This temptation is
prevented because the enumeration is done before the
respondent finds out that there will be follow-up
questions on each item enumerated.
A second approach is useful when it is expected
that only a few of many possible items will pertain to
any one household. Consider Figure 3.8. The large
grid on the right contains lines for several durable
goods owned by the household, but these are not precoded.
Rather, the respondent is asked, using the small
grid on the left, whether the household owns certain
durable goods. In this example 12 durable goods are
considered, but in some cases 20–30 goods have been
listed.Most households own only a few durable goods.
For all durable goods owned by the household, the
interviewer lists the name and the code number in the
large grid to the right in Figure 3.8, and asks a series
of questions about each good. If the household owns
two or more of the same durable good, one line is
filled out for each good owned.
Probe Questions
There are some kinds of information that respondents
may accidentally not provide. In such cases the questionnaire
includes instructions to the interviewer to
ask further “probing” questions on the subject. An
example of this is Question 9 of Figure 3.1. Suggested
probing questions are usually included in the interviewers’
manual and occasionally included in the
questionnaire itself. Probe questions are often used to
ensure that all items in a respondent-determined list
have been reported to the interviewer, or to ensure
that the respondent’s answer is properly classified by
the interviewer. Interviewers are also asked to probe
for answers to questions that ask “how much...?” (This
kind of question is commonly found in the consumption,
agriculture, and household enterprise modules.)
Interviewers should be thoroughly trained to ensure
that they fully understand what information to probe
for, and how to do so.
Because the interviewer is trained and instructed
to probe for information, there should be very few
answers of “don’t know” and thus very few codes for
“don’t know” in the questionnaire. In the exceptional
case when even a sound interviewing technique does
not produce an answer, the interviewer is instructed
(in the interviewers’ manual and in training) to write
“DK” (for “don’t know”) in the space reserved for an
answer code. Such responses are given a special nonnumeric
code in the data entry program. The end
result for analysis is much the same as having a “don’t
know” code for each question. However, this system
has the advantage that it discourages interviewers from
accepting “don’t know” answers too easily, which they
may be tempted to do to speed up the interview.
Moreover, the special non-numeric code for such
responses is glaringly obvious when the supervisor
reviews the questionnaire.
Letting Respondents Choose Units
For many questions that involve payments or quantities,
respondents are allowed to give their answers in
whatever units they find most convenient.Examples of
this are found in Figure 3.3. In Questions 4 and 12 the
code of the time unit in which the respondent replies
is placed in the box marked “time unit.”The codes are
provided in a box at the bottom of the page.
Allowing the respondent to select the time unit
means that transactions are expressed in the units in
which they normally occur, which may differ from
household to household or from person to person.
This avoids inaccuracies in conversion. For example, a
person paid $510 per week will respond precisely if
allowed to respond on a per-week basis. If forced to
respond in terms of dollars per month, the respondent
might round the figure down to $500 for ease of multiplication
and calculate each month as being equivalent
to four weeks.The annualized figure would thus
become $24,000 instead of the $26,520 that would be
reported if the respondent were allowed to report on
a per-week basis and the data analyst then calculated
the respondent’s annual rate from that answer.
Of course, data analysis is always slightly more
complicated when respondents’ answers must be converted
in order to arrive at annualized figures, but,
since a computer can easily do this, this disadvantage is
trivial. However, it is very important to ensure that,
where necessary, the questionnaire explicitly asks the
respondent how many times per year the payments are
made. For example, a worker who reports a daily wage
rate may be employed only intermittently. In this case,
the questionnaire should ask the respondent how
many weeks or months he or she has worked during
the preceding 12 months (see Chapter 9 for details).
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
70
FIGURE3.8:ILLUSTRATIONOFOPEN-ENDEDLIST(PARTEOFCONSUMPTIONMODULE)
2.3.4.5.6.7.
I
T
E
M
Howmany
yearsago
didyou
acquirethis
[ITEM]?
Didyoupurchase
itorreceiveitas
agiftorpayment
forservices?
Howmuchdid
youpayforit?
Howmuch
wasitworth
whenyou
receivedit?
Ifyouwanted
tosellthis
[ITEM]today,
howmuch
wouldyou
receive?
PURCHASE..1
GIFTOR
PAYMENT..2
ITEMCODEYESNODESCRIPTIONCODEYEARS(»6)CURRENCYCURRENCYCURRENCY
Stove201
1
Refrigerator202
2
WashingMachine203
3
Sewing/knittingmachine204
4
Fan205
5
Television206
6
Videoplayer207
7
Tapeplayer/CDplayer208
8
Camera,videocamera209
9
Bicycle210
10
Motorcycle/scooter211
11
Carortruck21212
13
14
15
16
1.Doesyourhouseholdownanyofthe
followingitems?
DETERMINEWHICHDURABLESTHE
HOUSEHOLDOWNSBYASKINGQ.1.FOR
EACHDURABLEOWNED,WRITETHE
DESCRIPTIONANDCODEINTHESPACE
PROVIDEDUNDERQ.2,ANDPROCEEDTO
ASKQ.3-7FOREACHITEM.
LISTALLTHEITEMSOWNEDBY
THEHOUSEHOLD,THENPROCEED
TOASKQ.3-7.
¨¨7
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
71
A particular place in the questionnaire where it is
useful to allow respondents to choose their own units
is in the “quantities produced “ questions in the agriculture
module. In Ghana, for example, respondents
were allowed to give answers in 22 different kinds of
units (Table 3.1). A serious problem for analysts who
want to convert these different quantities to a single
standard unit is that only about half of the units used
in this example were standardized, and some of the
standardized units were local terms (such as minibag
and maxibag) that would be unknown to anyone not
familiar with farming in Ghana.11
In the case of standardized
local units, the survey team should ensure
that such terms are defined (in terms of international
standardized units) in a basic information document
that includes all of the information that data users will
need to analyze the data.
Respondent Codes
It is sometimes useful to know who is answering a certain
section of the questionnaire.In general,each household
member should answer for himself or herself, but
this is not always possible. For example, a household
member may be away during the entire week when the
field team is working in his or her community.To indicate
who is providing the information,a question can be
inserted that asks the interviewer whether the person is
answering for himself or herself. If someone else is providing
the information,the interviewer should fill in the
identification code of that person.An example of this is
shown in Figure 3.5.This information is useful because
a proxy respondent may give less accurate information
than the individual who is actually involved in the activity
in question. For example, one household member
may not know the exact salary of another. Therefore,
some analysts may wish to identify any possible biases
introduced by the proxy respondents or to omit their
responses altogether.
Cardstock Covers
LSMS questionnaires are usually printed with cardstock
covers—covers made of very thin cardboard
similar to the cardboard used in file folders. In some
past surveys it was decided not to use these covers
because of their added cost, but this led to the problem
that the front and back pages of the questionnaire
occasionally came loose. Since the front page usually
carries the key household identifier information and
the back page sometimes contains the household roster,
any such loss is likely to render the rest of the
questionnaire useless. Thus cardstock covers are well
worth their cost.
Identifying Sections
The household questionnaire contained in a prototypical
full LSMS survey can be very bulky. The Nepal
questionnaire, for example, had 70 pages.Therefore, it
is useful to devise some ways to make it easy for readers
to find their way around in these questionnaires.A
few ideas are listed here, and there may well be more.
First, it is useful to have page numbers on each page
and a table of contents listing the sections (and their
page numbers) at the beginning or end of the household
questionnaire. Second, some inexpensive graphic
techniques can be used to divide the questionnaire
into smaller parts. For example, some sections of the
questionnaire can be printed on different colored
paper or in different colored inks, or sheets of colored
paper can be inserted between major portions of the
questionnaire. It is also possible to print short, dark
bars at the edge of each page, with the placement of
these bars on the page being the same within each
module but lower down (if on the vertical edge) or
further to the right (if on the bottom edge) in each
Table 3.1 Units of Quantity Used in Ghana, 1987–88
Unit Code
Pound *1
Kilogram *2
Ton *3
Minibag *4
Maxibag *5
Sheet 6
Basket 7
Bowl 8
American tin *9
Tree 10
Stick 11
Bundle 12
Barrel 13
Liter *14
Gallon *15
Beer bottle *16
Bunch 17
Nut 18
Fruit 19
Log 20
Box 21
All 22
Note: It is preferable to use the unit codes marked by (*) whenever possible.
Source: Ghana LSMS survey (1987–88).
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
72
successive module. Using just one or a few of these
techniques will be sufficient.The questionnaire should
not become too colorful or complicated.
Legibility and Spacing
There is an art to laying out the grids for a questionnaire.
The lettering must be large enough to read,
which is sometimes difficult to accomplish in the
compact structure of the grid. Legibility is especially
important, as interviews often take place under poor
lighting conditions, such as outdoors at dusk or after
dark in homes dimly lit with lanterns, oil lamps, or
candles. The good print quality now available from
laser printers helps, but poor legibility is an ongoing
complaint among interviewers.
There must also be enough white (empty) space
in the layout of the questionnaire. Whenever the
answer will be coded later, a generous space should be
allowed to write out fully the information required,
such as the person’s name, the name of the school
attended by the respondent, and the respondent’s
occupation. In other places, judicious use of white
space makes the questionnaire easier to read or less
confusing than a questionnaire in which every page is
crowded with print.
In fact, in this book, the fonts used inVolume 3 are
probably too small. This is necessary for Volume 3 to
show how typical questionnaire pages should appear. In
an actual questionnaire,the size of the pages usually will
be somewhat larger than the pages in this book,and the
font size should be increased by a similar proportion.
Software for the Questionnaire Layout
Many of the most common word processing and
graphics software packages are adequate for producing
questionnaire page layouts, and LSMS questionnaires
have been produced using several different software
packages. The modules in Volume 3 (the electronic
versions of which are available to readers in the CDROM
enclosed in the volume) were produced in
Microsoft Excel, for two reasons. First, Excel is widely
available. Second, spreadsheet software is better than
word processing software at dealing with the long horizontal
format of groups of questions on a single topic
that are spread across several pages. Regardless of the
software used, it is now much simpler and cheaper to
make revisions between the various drafts of the modules
than it was in the days when graphic artists had to
draw each page by hand.The computerized approach
also simplifies translations, as the verbal parts can be
overwritten in the local language, leaving intact the
skip codes, response codes, and general format.
Appendix 3.1 Common Gaps and Overlaps
This appendix provides a list of the modules that
should be checked for gaps and overlaps with respect
to the information that they collect. This list is not
meant to be exhaustive because household questionnaires
of different configurations will be subject to different
risks of gaps and overlaps and because there are
so many possibilities that it is difficult to list them all.
However, some of the most common and important
issues are mentioned here. Many more are mentioned
in the relevant chapters of this book.
Consumption
Consumption information usually comes from several
different modules of the household questionnaire. See
the discussion in Chapter 5 on the different components
of consumption and the modules in which those
components are typically collected.
Income
Information on household income is gathered in the
following modules: employment, household enterprise,
agriculture, and transfers and other nonlabor
income. It is sometimes also collected in the housing
and savings modules. It is important to review the
questionnaire as a whole to make sure that it accounts
for all possible sources of income. In particular, questions
about income from any rental property could be
placed in the transfers and other nonlabor income
module, on the assets page of the savings module, or, if
the income comes from renting out a portion of the
household’s primary dwelling, in the housing module.
Wealth
Information on household assets is collected in several
modules.The housing module gathers information
on the household’s principal residence.The household
enterprise module gathers information on equipment
and land associated with each household enterprise,
and on the stocks of inputs and outputs used in each
enterprise. The agricultural module gathers information
on land, equipment, and livestock. The savings
module collects information on other properties and
financial assets, and the durable goods submodule of
CHAPTER 3 DESIGNING MODULES AND ASSEMBLING THEM INTO SURVEY QUESTIONNAIRES
73
the consumption module collects data on the household’s
durable goods. Finally, the credit module gathers
information on the household’s liabilities.
Credit
Credit information is collected in several modules,
including the modules for housing, consumption, savings,
agriculture, and household businesses. There is
also a separate credit module. Chapter 21 introduces
the credit module and clarifies gaps and overlaps in
credit.
Mortgages
Information on any mortgages that a household might
hold can be gathered either in the credit module or in
the housing, agriculture, and household enterprise
modules.
Employment
Analysts often need to know how many hours each
household member works in the household’s enterprises
and in its agricultural activities as well as hours
worked in employment outside the household. In previous
LSMS surveys, all of this information was collected
in the employment module. As explained in
Chapter 9 (and Chapters 18 and 19), this book recommends
collecting data on household members’ days
and hours of work in household enterprises and agricultural
activities in the household enterprise and
agriculture modules, respectively, while continuing to
ask about the number of hours worked in wage
employment in the employment module. However,
some survey designers may decide not to include the
household enterprise and agriculture modules. In such
cases information on the number of hours spent
working on these activities must be collected in the
employment module.
Vaccination
If the survey includes a fertility module, questions
about vaccination should usually be placed in the fertility
module so that this information can be collected
not only for children who currently live in the household
but also for children who have died or moved to
another household. If there is no fertility module or
the fertility module does not include all women of
childbearing age, vaccination information on children
living in the household can be collected in the anthropometry
module. Another alternative is to gather this
information in the health module, which includes
questions on vaccinations in Part C of the standard
health questionnaire.
Domestic Housework
Some previous LSMS surveys have collected information
on how much time household members spend
doing housework (such as cooking, cleaning, and
childcare) in the employment module, usually asking
only one question. If a time use module is included in
a questionnaire, there is no reason to ask questions
about housework in the employment module.
However, because the time use module is very long, it
is unlikely to be used in most LSMS-type multitopic
surveys. If the time use module is not included but
survey designers want to gather a small amount of
information on, for example, the number of hours
spent on housework during the previous seven days,
one or two questions can be added to the employment
module. (See Chapter 9 for further discussion of this
issue.)
Notes
The authors would like to express their gratitude to Jere Behrman,
Lawrence Haddad, Courtney Harold, John Hoddinott, Alberto
Martini, and Raylynn Oliver for comments on an earlier draft.
1. Survey designers occasionally collect redundant information
as a cross-check on other data. For example, most previous LSMS
surveys have recorded both the age (in years) and the date of birth
of each household member.This is done to verify the accuracy of
the age variable.
2. This assumes that a two-stage sample is used. In the case of a
three-stage sample, the secondary sampling unit is more pertinent.
Generally, the penultimate sampling unit is the appropriate unit for
collecting community data.
3. Issues concerning the order of the questions within each
module are discussed in the topic-specific chapters in Parts 2 and 3
of this book. For a general discussion of ordering questions in
household surveys see United Nations (1985) and Frey and Oishi
(1995).
4. The short version of the health module presented in Chapter
9 does not ask for particularly sensitive information, but the standard
and extended versions ask detailed questions about health status
and health behavior (including drinking and smoking) that can
be sensitive.If either the standard or the long module is used,health
should not be one of the first modules in the questionnaire.
5. The questions in the household enterprise module that refer
to “the past 14 days” can be reworded as “since my last visit” if the
MARGARET GROSH, PAUL GLEWWE,AND JUAN MUÑOZ
74
second half of the questionnaire is administered two weeks after the
interviewer’s first visit.
6. For example, the education module asks questions such as
“What grade is [..NAME..] enrolled in?” For this question, the
range of acceptable values in the data set is precisely defined.
Moreover, it is also related to other information such as the degree
obtained and the age of the student. (For example, a six-year-old
should not be in secondary school.) In the consumption module,
however, a wide range of values might be found for a question such
as“How much did you spend on rice in the last two weeks?”which
implies that fewer consistency checks are possible.
7. This section is a slightly modified version of the discussion on
translating and field testing found in Chapter 3 of Grosh and
Munoz (1996).
8. An alternative approach is to stretch the reference periods
during the field test. For instance, instead of asking “Have you been
ill or injured during the past 30 days?”as in the actual survey,it may
be expedient to ask “Have you been ill or injured during the past
12 months?” or “When was the last time you were ill or injured?”
This approach will simplify the logistics of finding enough people
to try out the module but will not test very precisely whether the
respondents find it difficult to recall the information, since the
recall period used in the field test will be longer than the period
used in the final questionnaire.
9. This section is a slightly modified version of the discussion of
questionnaire formatting found in Chapter 3 of Grosh and Muñoz
(1996).
10. For languages that do not have uppercase and lowercase,
another way should be found to distinguish instructions from questions.
It may be possible to use italics, bold, a different font, or a different
color.An example of this is the LSMS survey of rural households
in northeast China in 1995. Chinese characters do not have
uppercase and lowercase, so two different fonts were used.
11. It is not necessary to convert quantities into standard units
(for example, to convert bunches into kilos) to calculate farm
income, which was the purpose of the agriculture module in the
Ghana LSMS.However,as is common with such rich data sets,analysts
are using the data for other purposes as well, such as calculating
the total quantities of various crops that were produced.
References
Ainsworth, Martha, and Jacques van der Gaag. 1988. Guidelines for
Adapting the LSMS Living Standards Questionnaires to Local
Conditions. Living Standards Measurement Study Working
Paper 26.Washington, D.C.:World Bank.
Babbie, Earl. 1990. Survey Research Methods. Belmont, Cal.:
Wadsworth.
Fink, Arlene. 1995. The Survey Handbook. Thousand Oaks, Cal.:
Sage Publications.
Fowler, Floyd. 1993. Survey Research Methods. Second ed. Newbury
Park, Cal.: Sage Publications.
Frey,James,and Sabine Mertens Oishi.1995. How to Conduct Interviews
byTelephone and in Person.Thousand Oaks,Cali:Sage Publications.
Grosh, Margaret, and Juan Munoz. 1996. A Manual for Planning and
Implementing the Living Standards Measurement Study Survey.
Living Standards Measurement Study Working Paper 126.
Washington, D.C.:World Bank.
Oliver, Raylynn. 1997. Model Living Standards Measurement Study
Survey Questionnaire for the Countries of the Former Soviet Union.
Living Standards Measurement Study Working Paper 130.
Washington, D.C.:World Bank.
Scott, Christopher, Martin Vaessen, Sidiki Coulibaly, and Jane
Verrall. 1988. “Verbatim Questionnaires Versus Field
Translation or Schedules:An Experimental Study.” International
Statistical Review 56 (3): 259–78.
United Nations. 1985. “Development and Design of Survey
Questionnaires.” Department of Technical Cooperation for
Development, National Household Survey Capability
Programme, NewYork.