79807 June 2013 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries ELAINE KEMPSON VALERIA PEROTTI KINNON SCOTT © 2013 International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org The findings, interpretations, and conclusions expressed here do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because the World Bank encourages dissem- ination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. Cover photos: World Bank Cover design/layout: Nita Congress Contents Acknowledgments — v 1 Introduction — 1 1.1 Motivation for financial capability measurement — 2 1.2 Why this financial capability survey? — 3 2 Questionnaires: purpose, content, and customization needs  —  5 2.1 Individual questionnaire — 6 2.1.1 General considerations — 6 2.1.2 Sections of the questionnaire — 8 2.1.3 Customization — 20 2.2 Optional questions and modules — 24 2.2.1 Time — 24 2.2.2 Questions not used in the scores — 25 2.2.3 Additional optional sets of questions/topics — 25 2.3 Location questionnaire — 26 2.3.1 Purpose — 26 2.3.2 Who should administer the location questionnaire?  — 26 2.3.3 Who should respond? — 27 2.3.4 Special Issues — 28 2.3.5 Customization — 28 2.4 Interviewer questionnaire — 29 2.5 Conventions in all questionnaires — 29 3 Survey implementation — 31 3.1 Sampling — 31 3.1.1 Probability sampling  — 31 3.1.2 Why not nonprobability sampling?  — 32 3.1.3 Selection of the individual to interview — 33 3.2 Training — 34 3.2.1 Training materials and manuals — 35 3.2.2 Training content and schedule — 35 3.2.3 Testing — 35 3.3 Survey organization — 36 3.3.1 Interviewers — 36 v Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 3.3.2 Supervisor — 37 3.4 Data entry — 38 3.5 Data cleaning — 38 3.5.1 Documenting the process — 39 3.5.2 The final data set(s) — 39 3.6 Documentation — 40 4 Analysis — 43 4.1 Weights — 43 4.2 Descriptive statistics — 43 4.3 Financial capability scores — 44 5 Research agenda — 51 Appendixes A Main Questionnaire — 55 B Optional modules — 111 C Location Questionnaire — 127 D Interviewer Questionnaire — 133 References — 137 Boxes 3.1 Basic information document: table of contents — 41 Figures 2.1 Questionnaire structure — 9 2.2 Filling out the roster — 12 3.1 Interviewer work flow — 37 3.2 Data checks — 39 Tables 4.1 Key derived variables obtained from the survey questions — 46 4.2 How to calculate financial capability scores from the derived variables — 49 vi Acknowledgments T he Russia Trust Fund for Financial Literacy and Education World Bank Financial Capability Survey (FCS) described here is the outcome of a multiyear research project. The authors would like to thank the teams who managed and implemented the research and data collection in the pilot countries: Armenia (Vardan Aghbalyan, Arabela Aprahamian, Karen Grigorian, Juan Carlos Izaguirre, Edgar Karapetyan, Sona Lalayan, Vahe Mambreyan, Sevak Mikayelyan, Armenuhi Mkrtchyan, Sue Rutledge, Siegfried Zottel), Colombia (Dairo Estrada Ayiber, Sigrid Falla, Nidia García Bohórquez, Diana Papagayo, Viviana Posada, Rekha Reddy, Nancy Zamudio Gomez), Lebanon (Gustavo Demarco, Angela Elzir, Sabine Hatem, Lamia Moubayed, Mira Saidi, Souraya Srage, Lina Tannir, May Wazzan), Malawi-Namibia-Zambia (Edwin Banda, Irma Grundling, Giuseppe Iarossi, Astrid Ludin, Edward Phiri, Franko Venter, Smita Wagh), Mexico (Marco Carrera, Jesus David Chavez Ugalde, Maria Jose Gentili, Patricia Lopez, Arturo Luna Canales, Edgar Monsalvo, Rekha Reddy, Ana Luisa Saavedra, Luis Treviño Garza), Nigeria (Tunde Adebisi, Temitayo Adebiyi, Lawrence Akhidenor, Gero Carletto, Yemi Kale, Florence Nenuwa, Nemi Okujagu, Isiaka Olarewaju, Gbemisola Oseni), Papua New Guinea (Paul Barker, Francesca Drapuluvik-Tinabar, Elizabeth Genia, Henao Kari, Gae Kauzi, Rufina Peter, Alohai Pochapon, Mahohar Sharma, Jonathan Sibley, the late Peter Siopun, Wei Zhang), Tanzania (Melania Akinyi, Linda Helgesson, Emmanuel Mung’ong’o, Gerda Piprek, Giuseppe Iarossi, Smita Wagh), Turkey (Tugba Acar, Ozgur Avcuoglu, Gokce Uysal Kolasin, Sebnem Sahin Say, Mustafa Tasdemir, Cenap Nuhrat, Ayla Ortac, Ahmet Otunc, Kamer Ozdemir), Uruguay (Ana Caro, Cecilia Hughes, Anya Maria Mayans, Margaret Miller). Addition- ally, the authors would like to thank the team of experts who provided substantial input and advice throughout the process: Gerrit Antonides who contributed to both the quantitative and qualitative analysis, and Sharon Collard, Olga Kuzina, and Christian Poppe who contributed to the qualitative analysis. The Russia Trust Fund program manager Richard Hinz and the senior consultant Robert Holzmann provided strategic guidance and insightful comments. The report also benefited from discus- sions with Kristen Himelein. Layout and publishing support by Nita Congress and Raiden Dillard are gratefully acknowledged, as is proofreading by Amy Gautam. For further documentation about the activities supported by the Trust Fund, please visit the website www.finlitedu.org. vii I CHAPTER 1 ntroduction T his manual is designed to provide guidance to institutions, researchers, and survey firms on how to measure financial capability in middle- and low-income countries using a new survey instrument that was developed and tested, from start to finish, in middle- and low-income countries. This new survey was developed as part of a larger project financed by the Russia Financial Literacy and Education Trust Fund (RTF) and implemented by the World Bank that encompassed both measure- ment of financial capability and evaluation of financial literacy programs (see the project website for details on the full project: www.finlitedu.org). The development of the new RTF Financial Capability Survey (FCS) was done in collaboration with a team of external experts and teams from a total of 12 low- and middle-income countries. The full description of how the survey was developed can be found in Kempson, Perotti, and Scott (2013) along with findings from the first wave of surveys that have been done. In this manual we briefly summarize reasons why the FCS might be of interest to a country, provide a detailed outline of the issues related to implementing the survey successfully, and demonstrate how to analyze the resulting data. The first chapter of the manual lays out the reasons why this survey could be of use to policy makers. Chapter 2 describes the FCS questionnaires and their goals and objectives. Topics related to fieldwork, or the implementation of the survey, are covered in chapter 3. Guidance on how to analyze the survey is provided in chapter 4. The survey instru- ments can be found in the appendixes of this manual, and related documents— interviewer and supervisor manuals—can be found on the Trust Fund website www. finlitedu.org. This manual in no way presumes to be a tool to teach survey design, implementa- tion, or analysis. Instead it is designed for the survey practitioner, or the researcher working with survey experts to undertake a survey of financial capability. It provides guidance on the overall purpose of the financial capability questionnaires and their parts, while at the same time highlighting areas of concern and issues that need to be addressed in the implementation of the survey instrument, and in the analysis of the data that is collected. This manual is not a substitute for a basic knowledge of surveys and survey techniques such as probability sampling, face-to-face inter- 1 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries viewing, and data management. For those less familiar with the field of household and individual surveys, refer to books designed specifically for the survey practi- tioner. 1.1 Motivation for financial capability measurement In low- and middle-income countries, the limited scope of social insurance and safety net programs on the one hand, and the erosion of traditional family support mecha- nisms due to urban migration and employment shifts on the other hand, leave indi- viduals increasingly responsible for their own income planning and risk management. This responsibiity becomes more challenging when the number and complexity of financial products available increases with overall economic development. The low levels of financial inclusion observed in low- and middle-income countries are partly driven by limited financial capability. This, in turn, limits individuals’ and households’ ability to effectively utilize financial resources, smooth their consumption over a life- time, and manage risks. Financial capability, the capacity to manage financial resources and use financial services in a way that best suits individual needs and the prevalent social and economic conditions, is a broader concept than that of financial literacy, which focuses on people’s knowledge—e.g., of how interest rates work, the concept of present value, and similar concepts. There is growing interest in developing inter- ventions and strategies to raise levels of financial capability. Yet there is no accepted tool that can be used to measure financial capability nor to identify which facets of financial capability are weakest in a country or which population groups have the greatest lacunae in this area. An instrument that can measure financial capability can provide policy makers with a range of important inputs. At the most basic level, such an instrument would allow countries to carry out diagnostic studies to assess areas of financial capability to show weaknesses or strengths. Additionally, it would allow policy makers to iden- tify potential target groups that could most benefit from policy interventions and specific programs. Such a survey could also allow for more focused research on topics related to financial capability and its linkages with individual, household, and community characteristics. At its best, such an instrument can be used to measure progress toward specific objectives and answer questions on the impact of financial and educational policy on capability and overall well-being. 2 CHAPTER 1. Introduction 1.2 Why this financial capability survey? The questionnaire described in this manual has been designed to capture some of the characteristics of people that make them financially capable—characteristics that are expected to translate into increased welfare through better management of one’s own resources, including a greater or more effective use of financial services. The instrument was developed from an open qualitative research process that provided the inputs into the design of the quantitative instrument described in this manual. The development process follows one that was first done in the United Kingdom for the Financial Services Authority (FSA), and subsequently implemented in other high-in- come countries.1 Instead of simply using the findings from the FSA work to develop a questionnaire, however, the World Bank’s RTF project replicated the work from step one in low- and middle-income countries. The assumption behind this decision was that the features that matter for financial capability in low- and middle-income coun- tries might be quite different from those in high-income countries. The questionnaire presented here is the result of an open process that started with focus groups in eight countries: Colombia, Malawi, Mexico, Namibia, Papua New Guinea, Tanzania, Uruguay, and Zambia. The focus groups were conducted in both rural and urban areas and with men and women from different socioeconomic groups and were designed to elicit the actions, behaviors, and attitudes that make a person financially capable. Based on this round of research, a set of common elements or manifestations of financial capability was identified, and two rounds of in-depth cogni- tive interviews were conducted in the same countries to pinpoint both the questions with the most relevance and the form in which these were most readily answered by respondents across the educational and income spectrum. (For a full description of the development process, please see Kempson, Perotti, and Scott 2013.) The development of the FCS was a collaborative effort with central banks and other government agencies with a vested interest in improving financial service use throughout their economies. This partnership has led to the instruments described below. The content of these instruments has been informed by the populations in which the survey was developed, and has been extensively tested and piloted to create the most complete but also the simplest instrument possible. There are always trade-offs in questionnaire design, with the best instrument achieving a balance in terms of brevity, clarity, and comprehensiveness. Financial capability is a complex concept that is not amenable to measurement by a small number of ques- tions. That said, every effort has been made to ensure the utility of each question in the following questionnaires and that all segments of the population will be able to answer them. 1  See FSA (2005) and (2006) for more details. 3 Q CHAPTER 2 uestionnaires: purpose, content, and customization needs T he World Bank’s Financial Capability Survey developed under the Russia Financial Literacy and Education Trust Fund captures information on individuals’ behavior, skills, and attitudes related to managing the finances for which they are responsible. This chapter contains an overview of the questionnaires used in the FCS. The overall purpose of each section of the questionnaire is highlighted, areas of potential prob- lems are flagged, and any issues specific to the theme or format of the section are addressed. For detailed instructions and examples of how each individual question in the questionnaires is to be answered and for additional definitions of the terms used in the questionnaire, detailed interviewer manuals have been written. These can be found on the RTF project website (www.finlitedu.org). The FCS comprises two main survey instruments—the Main (Individual) Question- naire, which includes the questions designed to measure financial capability and a Location Questionnaire—with a third, optional, Interviewer Questionnaire. The Individual Questionnaire is administered to one randomly selected adult in each sample household to collect data on the financial behaviors and attitudes identified by the qualitative work as key components of financial capability. Some data on the household in which the person lives are also gathered. The Location Questionnaire is designed to collect a limited set of community variables that can be used to char- acterize the environment in which people make financial choices. As most samples are designed in stages, with multiple households interviewed in one location, the Location Questionnaire is administered only once in a given community or cluster of dwellings, and the data attributed to all the dwellings in the community. This is a time-saving approach. The optional Interviewer Questionnaire is used to collect infor- mation on the characteristics of the interviewers. The questionnaires will need to be customized in several places to reflect specifics of the country where the survey is being implemented such as schooling levels, marital status, available financial products, and the like. The full questionnaire can be found in appendix A of this document. All questions that require customization are high- lighted there in yellow. 5 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 2.1 Individual questionnaire The Individual Questionnaire is designed primarily to collect data from one adult in each selected household about the core topics related to financial capability. A small amount of household-level data is collected to help understand the context in which the individual lives and to allow the use of a probability sampling tool (Kish table) to select the respondent for the interview. The survey instrument takes approximately 30–40 minutes to administer. This chapter provides an overview of the survey instru- ment. 2.1.1 General considerations Order The questionnaire has been arranged so that the most potentially difficult or sensi- tive questions come at the end of the interview, when the greatest level of rapport between the informant and the interviewer will exist. Additionally, within sections, the question order is based on several considerations: logic, ease of interview, and— in some cases—the need to create a reference for later questions. An example of this point can be seen in Section C of the questionnaire, where expected expenses are discussed before unexpected. This order helps to frame the idea of an unex- pected expense for the respondent. It is important to maintain both the section and question order: reordering may undermine the validity of some sections or questions. One respondent versus multiple respondents The survey has been designed to interview one adult per household in order to generate a random sample of adults living in a country. The decision to focus on one person per household was made for two reasons. First, the adult composition of a household is not stable over time—people leave households (through separation or death) and join new ones (through marriage and other events). Second, individuals’ responsibilities may change over time as households change. Thus, understanding capability at an individual level is important. This focus, however, does not shed light on intra-household dynamics or household financial capability. If these are areas of interest, then all adults in the household should be interviewed; there is no reason why this cannot be done.1 There are, of course, cost implications, as time in each household will increase. And, if not properly handled, trying to interview all adults may create some problems with nonresponse. As with any change, the costs and benefits need to be carefully assessed. 1  One of the pilot countries, Nigeria, interviewed all persons over age 15 without major difficulty. 6 CHAPTER 2. Questionnaires: purpose, content, and customization needs (Note, in the present design, where one respondent is chosen, there may be two different people who provide information for the survey. The first respondent is a “responsible adult�—the head of household, his or her spouse, or another adult in the household who is knowledgeable about all other members of the household— and this person will provide the information on the household and all its members (Section R). Depending on the outcome of the selection of the respondent for the financial survey, this person may or may not be the “selected respondent,� i.e., the one person of all household members 18 years and older who is selected to respond to Sections A–G of the questionnaire using the Kish table.) Two pathways Given that some respondents will manage only their own personal income while others will be managing or participating in the management of the household’s income, the questionnaire has been split into Sections B, C, and D: there are two versions of each module, one for those who manage only their own income and one for those involved in household financial decision making. The two-version approach was taken, as it removes the burden from the interviewer of having to customize the wording of each question depending on who is being interviewed. It is expected that this will improve data quality. Age of respondent The decision to interview only adults is based on the premise that, unless a person has some role in managing money or making financial decisions, there is no way that any assessment of his or her financial capability can be made with this questionnaire. In some of the early stages of the questionnaire development, younger adults were interviewed: those who depended on their families were unable to answer many of the questions in the survey. For this reason, having a lower-bound cutoff age is important. The actual age of that lower bound will vary by country. In countries with child labor or where households rely on all members to bring in income, lowering the age to 12 or 15 may be appropriate. Household information The questionnaire contains a very minimal set of questions on the members of the household. This was done in an effort to keep the questionnaire as short as possible and could be expanded if there were specific hypotheses that needed to be inves- tigated; this decision should be carefully weighed before following through. Note that every question that is added is much more than one question, as it needs to be answered for each member of the household: for a six-person household, adding 3 questions would really be adding 18 questions. 7 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Customization Any questionnaire will need some customization to work in a specific context; this topic is covered in detail at the end of this chapter. Additionally, there may be new topics or questions that would be of great use in a given country or at a given time. Topics that could be of interest are consumer protection, financial inclusion per se, and financial literacy. This last topic was covered in five of the seven countries that implemented the first wave of the survey. Any new topics that are to be added should be included after Section E of the questionnaire: attempting to insert new questions between Sections B through F of the questionnaire could negatively affect the survey instrument—and the survey results. 2.1.2 Sections of the questionnaire The overview of the content of the questionnaire can be seen in figure 2.1. The first two sections of the questionnaire, “Cover page� and “Section R: Roster,� are designed to identify the household and its characteristics and to select the respondent for the financial capability questions. The seven sections that follow are the core of the questionnaire—Sections A–G—and contain the questions on financial capability; these are designed to be administered to one randomly selected adult in the house- hold. Cover page Who is the respondent. The respondent for this section is the interviewer him- or herself. Much of the data should be filled in before the interviewer visits the house- hold, as the sampling information and address are provided by the central survey team. The information on the interview itself is filled out by the interviewer at the time of the interview. Purpose. As in any survey, the cover page is designed to capture all the information that identifies the household to be interviewed, from address and name of household head to detailed sampling data. These data are supplied ahead of time by the team designing the sample and are filled in by interviewers prior to visiting the household. The accuracy of the sampling information is critical. The first function of the cover page is to ensure that the sampling design (see chapter 3 for more on sampling) is not violated and that the correct households are interviewed. The second use for the information in this section is to provide the codes that allow each Individual questionnaire to be linked to its community data (Location Questionnaire). Third, the sampling information, coupled with information on nonresponse, is the basis for the sampling weight calculations that are required to ensure that the sample of house- holds reflects the population of the country. 8 CHAPTER 2. Questionnaires: purpose, content, and customization needs Figure 2.1 Questionnaire structure Cover page Roster Section A Section B (HHLD) Own or household Section B (IND) finances? Section C (HHLD) Section C (IND) Section D (HHLD) Section D (IND) Section E Section F Section G A fourth function of the cover page comes from the unique identifier that is listed. This identifier is a sequential numbering of questionnaires from 1 to n (with n being the size of the sample). There should be no correlation between the last two digits of the unique identifier and the location of the interview: i.e., the distribution should be random. The numbering of the questionnaire is a key input for the proper selection of the respondent for the full survey (see use of Kish table below); for this reason, the questionnaire identification number needs to be preassigned (printed on the ques- tionnaires). Leaving it to the interviewers to complete may lead to errors and even explicit manipulation of the Kish table selection process. Fifth, the data on visits and revisits and the status of the interview help the survey administrators and field staff keep track of interviews, manage the workload, and— crucially—calculate the nonresponse rate. 9 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries An additional function of this section—one that is often forgotten—is that of providing metadata on the survey: information on how often a household needed to be visited before an interview could be completed, the days and times when inter- viewing were most successful, and how long the interview took. All of this informa- tion can be used as measures of quality for the present survey and can feed into the design of the next survey. Roster This section of the questionnaire provides information that allows typologies of households to be formed based on demographics or the characteristics of the head of household. Additionally, in this section, information on the degree of financial activity of all adult members of the household is captured, providing information on who participated in household finances and the share of adults who did so. Who should respond. Responsible adult: Any household member over the age of majority who knows the household and its members well. Preference would usually be given to the head of household or this person’s spouse. Purpose. The roster is designed to (1) identify all members of the household, (2) collect basic information on these individuals that can be used to understand the context in which the individuals act, and (3) create a list of persons who are eligible to be selected as the respondent for the rest of the instrument. The priority here is to ensure that the list of household members is complete: any omissions will affect the integrity of the sample design. There are two methods that could be used here. In countries with small households and where ages are known, it is probably simplest to ask the respondent to list the household members in order from oldest to youngest. In countries with larger households or where ages may be less well known, it is probably preferable to ask for household members in the following order: head of household, his or her spouse, then the oldest child, that child’s spouse and children, then the next child in age, and so on. Other relatives are next, usually parents or parents-in-law of the head of household, and then others. Regardless of the method used, it is critical that it minimize the risk of omission of any household members, and that its guiding rule be used by all interviewers. Again, this helps ensure the integrity of the sample selection using the Kish table. It is important that interviewers probe for all household members and that the definitions of household and household member are respected (see section on customization for definitions). Definition of household. A standard definition of household is a “group of people sharing a common roof and a common pot� and members are those who live there at least six months and a day during a 12-month period. Conceptually, what is being looked for is a group of people who share resources and decision making. Practi- cally, the definition is designed to ensure that all people have a nonzero probability 10 CHAPTER 2. Questionnaires: purpose, content, and customization needs of selection but cannot be selected more than once (are not part of more than one household). One concern with household definitions is that, in some countries, domestic servants and even paying boarders are considered part of the household unit. Conceptually, they should not be; but rather should be considered independent households living in the same dwelling as another household, as their decisions are not made jointly with the household in which they work or in whose dwelling they live. If the household definition includes such individuals, the probability of their selection is much lower than it would be if they were classified as independent households. In countries with large numbers of resident domestic servants, this broad household definition can obscure a particularly poor segment of the popula- tion with, perhaps, very different financial capabilities or use of financial products. Ideally, one would reclassify such households as independent ones and interview them separately. Although this action may be difficult in practice, it is recommended. Whatever the decision, the final write-up should be clear on how such individuals have been treated and the effects of this decision on the analysis. Special features. The detailed instructions for administering the questionnaire and the definitions to be used are found in the interviewer manual. Here we highlight only those features that may be unique or provide difficulties. The interviewer manual should be studied carefully, and it should form the basis of much of the materials that will be used to train interviewers, supervisors, and data entry personnel. „„ Filling out the roster. The roster is filled out vertically first and then horizon- tally. In other words, the list of members is completed first with their gender, relation to household head, and age. Then, on a person-by-person level, Ques- tions R5–R13 are asked for each person (see figure 2.2). „„ Code of spouse. Each person has a unique identifier assigned to him or her by virtue of the row in which their information is entered. The first column of the roster is numbered: this is the Roster ID. The first person in the list is considered to have Roster ID number 1; the second person listed, Roster ID 2, and so on. Thus if person 1 and person 2 are married to each other, when person 1 is asked in Question R6 for the Roster ID code of his or her spouse, the answer will be 2 (person 2 is the spouse). For person 2, the answer would be “1,� as he or she is married to the head of household or the first person on the list. The Roster ID is also used in Question R14. „„ Education. Ideally, we would like know a person’s exact level of schooling. While incomplete primary is lower than complete primary, the difference between someone who just started first grade and someone who finished fifth grade can be substantial, affecting job prospects and, perhaps, financial capability or service use. Thus we would want to capture this level of detail. However, collecting this level of detail was problematic in some countries, 11 Figure 2.2 Filling out the roster R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 List all Is this What is the rela- How old What is this Inter- Can this What is the How would you Does this Does this Is this per- Inter- members person tionship of this was this person’s maritalviewer: person highest level describe this per- person person par- son mainly viewer: of the male or person to the person at status? Copy read and of schooling son’s main status contribute ticipate in or partly Write household female? head of house- his or her roster write in that this in the last four to the household responsible roster and ask hold? last birth- Married, num- [official person has weeks? household decisions for his or number Q. R5–R12 M=1 day? monogamous = 1 ber of lan- completed? budget? about her own of per- for all per- F=2 Head = 1 Married, spouse guage]? Employed, formal money and personal son who sons age Spouse/partner = 2 polygamous = 2 sector = 1 financial spending? provided Country Yes = 1 Child = 3 Informal union = 3 Employed, informal No = 2 18 and Country teams cus- matters or this Grandchild = 4 If < 18 go Divorced = 4 sector = 2 older. (>>R7) teams tomize for house- Yes, mainly = 1 informa- Parent = 5 to next Self-employed, not Separated = 5 custom- hold spend- Yes, partly = 2 tion Sibling = 6 person including own (>>R7) ing? No = 3 Brother- or sister- ize farm, unpaid family in-law = 7 Widowed = 6 worker = 3 Niece/nephew = 8 (>>R7) Looking for work = 4 Yes = 1 Yes = 1 Other relative = 9 Never married = 7 Waiting for busy sea- No = 2 No = 2 (>>R7) Roster number Domestic help = 10 son = 5 Other = 11 Studying = 6 Retired = 7 Sick/disabled = 8 Housewife/house- work/caring for household mem- ber = 9 Other = 10 NAME YEARS R NO. LEVEL 1 2 3 4 5 6 7 8 9 10 11 12 CHAPTER 2. Questionnaires: purpose, content, and customization needs and there was a preference to simply ask for the level completed. This is how the attached questionnaire treats the education question. (See the section on customization for guidance on how to ask the more detailed education ques- tion.) It is important, however, in the individual section that detailed schooling information of the respondent is captured, not just the level. „„ Participation in financial affairs. Questions R10–R12 are designed to cate- gorize individuals on their participation in financial matters, either their own personal finances or those of the household. The first question asks explicitly if the person contributes to the household budget (finances). The next asks about the level to which each person participates in household decision making on financial matters; and the last on whether the person is somewhat or fully responsible for his or her own finances. These three questions provide a substantial amount of information on how individuals participate in house- hold financial decisions by gender, age, education, and labor status; and they allow a typology of households to be developed based on this. „„ Selecting a respondent. The respondent for the rest of the survey is one randomly selected adult. As noted previously, eligible adults are those 18 years and older, as it was found in the countries where the survey was developed that individuals under 18 were often neither contributing to the household budget nor making any decisions on finances, either of the household or of their own. In this case, the survey instrument is incapable of measuring finan- cial capability. For this reason, younger respondents are omitted from the sample. It is preferable to use only a selection based on age and not to include other filter variables or conditions for eligibility. By selecting only for age, the resulting sample will be a probability sample of all adults in the country, and the survey findings can be extrapolated to the universe of all adults in the country. Some of the adults who will be interviewed will play no role in their household, or even their own, financial decisions; the bulk of the questionnaire will not be relevant for such people. In Section A, an effort is made to under- stand exactly what role the selected adult plays in the household. If he or she plays no role, the interview will end at this point. By administering the questionnaire in this way, one obtains a full picture of financial activity among all adults in the country, and can profile both inac- tive and active adults, comparing them across basic demographic variables at least. Additionally, the full analysis of financial capability among the active group can be carried out. And, as the sample is of all adults, there are no complications in the construction of sample weights. 13 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Once the list of eligible adults is determined, the interviewer applies a Kish table to select the individual respondent to ensure that the final sample is a true probability sample—one in which each adult has a known and nonzero probability of selection. The Kish table used here is a one-page table where the first row lists the possible number of eligible people in the household (1, 2, …, 25), and the first four columns list the possible last two digits of the household identification number (01 to 00).2 The rest of the table is prefilled with random numbers. The cell at the intersection of the number of eligible people in the household and the last two digits of the household identifica- tion number provides the order number of the person to be interviewed. For example, if the number in the cell at the intersection of the number of house- hold members and the household identification number is 3, then the third eligible member listed in the roster should be interviewed (see chapter 3 for a more detailed discussion of the Kish table and its application). Once an individual is selected, he or she answers all remaining sections of the questionnaire. Note: In the first wave of surveys, the effort was to maximize the number of financially active adults and avoid interviewing nonactive adults. In the case of a nonactive adult being identified in Section A of the questionnaire, the interviewer was instructed to carry out the selection process again and select a replacement respondent. This procedure is not recommended, as it is not only more complex for the interviewer to administer, but also leads to serious sampling issues in the analytic phase. Section A Who responds. Adult randomly selected using the Kish table. Purpose. This section is designed to do two things. First, it is used to identify individ- uals for whom the questionnaire cannot measure financial capability: i.e., those who are neither responsible for their own finances nor those of the household. The data show that these people are in a minority but do exist. The sections of the question- naire that collect information on financial behaviors are irrelevant for these people. The second role of the section is to identify, for those who do participate financially in some form, the set of financial decisions about which the person should provide information. Where business finances are kept separate, they are not covered by the survey; however, for some people, business finances are so intermingled with household or individual finances, it is not possible to separate the personal from This is an adaptation of the original Kish table that applies eight tables. See Kish (1949) for 2  more details. 14 CHAPTER 2. Questionnaires: purpose, content, and customization needs the business. This section allows the analyst to know when this is the case. Some people will only be able to respond about their individual (not household) finances, as they contribute nothing to the common household pot and/or make no decisions concerning it. For analytic purposes, one needs to know whether respondents are referring to their actions on household or just personal finances. The questionnaire is structured so that people responding about household decisions have a different skip pattern (set of questions) than those responding about personal finances. (The two sets of questions are actually the same, but the specific wording—plural versus singular, e.g.—varies.) Special features. „„ Informed consent and confidentiality. The privacy of the respondents in this, and any, survey must be protected. Data that contain identifying informa- tion may never be released, and respondents should be made aware of the statistical confidentiality of the data they provide. It is important as the inter- view starts with the selected respondent that the person be given a chance to understand what the survey is about, how the data will be used, and to obtain agreement from the respondent to do the survey. The data must be confiden- tial; while the unit record data will be available to analysts and researchers, it must never be disseminated to anyone with names, addresses, or any other information that would allow a respondent to be identified. It is imperative that this point be made clear: income and finance are sensitive subjects, and unless informants are comfortable that their answers will be confidential, there will be an unnecessarily high nonresponse rate or a data quality issue (as respondents self-censor). „„ Income. In this section, the goal is not to identify income sources or levels but simply to determine if the person has such sources. As income is a sensitive topic and this the first part of the interview, it is useful to make it clear that no specific information about income is being requested here. Section B Who responds. Same person as in previous section. Purpose. The purpose of this section is to collect information on how people manage day-to-day spending. Managing money includes people’s knowledge of their own spending, their ability to plan, and what they do when they have leftover resources or run short of funds for daily current spending. Special features. „„ Two sections. In this section, there are two separate sequences of ques- tions. For those who are answering about both personal and household 15 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries finances (answer 1 in Question A6), the questions are on the white pages. For those answering only about their personal finances, the questions are on the green pages. The reason to use two separate sets of questions is to lower the burden on the interviewer. If one set of questions were asked, the inter- viewer would have to customize the wording of each question (plurals, etc.) depending on the type of respondent. „„ Avoiding nonresponse. As many of the questions in this section are designed to be included in a score (see chapter 4 for a discussion of this analytic goal of the survey), it is important that nonresponse and “do not know� answers be avoided. The advice to the interviewer is, when faced with such responses, to gently attempt to get an answer, either by rereading the question or reminding the person that the survey answers are confidential. Clearly, there is a limit to what can be done to elicit an answer but every attempt should be made. Role playing this during training will be important. „„ Why two-part questions? Typically in questionnaires, questions related to activities are asked using a five-point scale (e.g., 1—Very often, 2—Often, 3—Sometimes, 4—Rarely, 5—Never). One of the important findings of the qualitative work that informed this questionnaire is that such questions are not easily understood or answered by all respondents. And, critically, the diffi- culty in responding is correlated with lower levels of education and income. To improve data quality, these types of questions were divided into sets of two questions, first eliciting whether the person does or does not do something, and then determining the intensity of the action. Changing these questions into the format usually used is NOT recommended. Section C Who responds. Same person as in previous section. Purpose. The purpose of this section is to investigate the extent to which people plan for the future in both the short and long term. The section covers planning related to known expenditures such as those that represent regular payments (school fees, rent, and the like). Unexpected expenditures and emergencies are also covered, both to determine whether the person could cover these and/or has a plan to cope with such events. Finally, questions are asked about planning for retirement and for one’s children. Special features. The special features in Section C are similar to those in Section B. There are two versions of this section, one for people answering only about their own expenditures and one for those answering about household spending. Again, the questions here are needed for the financial capability scoring, so nonresponse must be avoided as much as possible. Finally, questions that might usually be posed 16 CHAPTER 2. Questionnaires: purpose, content, and customization needs on four- or five-point scales have been divided into separate questions to ensure that people with less education are equally able to respond to them. Section D Who responds. Same person as in previous section. Purpose. The main purpose of this section is to understand how people choose financial products; whether they check the features, terms, and conditions before buying financial products; whether they look for information before buying products; and whether they seek advice or information before making financial decisions. An additional objective of this section is to obtain a broad indication of the level of financial inclusion of the individual by asking which financial products the respon- dent holds. Special features. „„ Developing a complete list of financial products. For this section to work properly, the survey team will need to consult closely with experts on finan- cial services and products in the country. It will be important to include a full range of the most common types of product, not simply the most common or those provided by the formal sector. A review of the FINMARKET FINSCOPE survey instruments3 will give insights on the range of products that might be listed. If another financial inclusion survey has been done in the country, this too could be a useful source of information to determine the list of products. It may not be necessary to include all possible financial products (which in some countries could represent a fairly daunting list), but examples of each type of instrument, covering the spectrum of instruments from simple to complex and formal and informal, is needed. It is sufficient to group together different types of mortgages, and similarly to group different types of unsecured loans. „„ Ranking financial products by complexity. Once the list is complete, it will be necessary to list products in order of complexity. Complexity is defined by the number and types of terms and conditions associated with each product. The purpose of ranking is so that the following financial capability questions can be asked about the most complex product that the respondent has been responsible for acquiring. In addition to collecting a complete list of products currently held, a separate list of products bought in the past five years is iden- tified, and, among these, which ones the respondent was personally involved in selecting. The assumption is that a person’s financial capability with regard to selecting and acquiring financial products is best established with refer- 3  See www.finscope.co.za and www.finmark.org.za for examples of the questionnaires. 17 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries ence to the most complex decision making they have had to make. Clearly, there are problems with ranking all financial products by complexity: again, consulting with financial product experts is imperative. „„ Difficulty. This section has some potentially difficult questions for the inter- viewer to administer. The interviewer manual provides detailed instructions for filling out this section and should be studied carefully and used as a basis for training. Section E Who responds. Same as in previous section. Purpose. The purpose of this section is to capture underlying motivations that influ- ence the way people behave. The questions are about motivations more generally. To capture the various motivations, there is a core of 18 questions relating to three motivations that form part of the scoring of financial capability. There are a further 16 questions relating to three motivations that are optional and can be included if so desired. These questions are highlighted in blue. The use of multiple questions for each motivation improves the robustness of psychological scales: more questions could be added, but dropping any will undermine the validity of the section. Special features. „„ Two-part attitude questions. All questions involve reading out an attitude statement. The respondents are then asked if they agree or disagree that the statement describes them personally. This question is followed by a question that asks whether they agree/disagree strongly or only to some extent. This permits the development of a four-point scale ranging from agrees strongly to disagrees strongly. It is important that respondents are encouraged to give a reply and discouraged from saying “don’t know� so that they can be placed on the four-point scale. Section F Who responds. Same person as in previous section. Purpose. This section is designed to obtain information about the respondent’s personal characteristics and also on income in terms of both levels and fluctuations in income that the individual respondent and his or her household faces. To do this, information is collected for all sources of income and then on how total income varies throughout the year. This is probably the only section of the questionnaire that may cause some discomfort, as it attempts to get at personal characteristics, partic- ularly financial ones. 18 CHAPTER 2. Questionnaires: purpose, content, and customization needs Special features. „„ Personal questions. The first thing to note is that the basic demographic questions are repeated: if the respondent for the roster is the same as the selected adult for the rest of the survey or was observing when the roster was filled out, there may be resistance to answering these questions. In this case, the interviewer shoudl indicate that he or she is just double checking information and that it is very important that these questions be answered correctly. „„ Income. Many times, respondents are uncomfortable talking about income, and it is important that interviewers be trained to feel fully comfortable in asking about income, as any discomfort the interviewer feels will be trans- mitted to the respondent. It should be borne in mind that these questions have been asked successfully by the country survey teams that have tested them. To minimize the discomfort, these questions are asked close to the end of the interview, when there should be a good rapport established between the interviewer and respondent. Additionally, no exact income figures are requested; instead, the respondent need only identify a fairly wide range in which his or her income fits. It may be necessary for the interviewer to remind the respondent of the confidentiality of the data. „„ Complexity. It was challenging to design questions and skip patterns to assess (1) the sources and variability of individual income, (2) the sources of income of other household members, and (3) overall household income and its variability. This is a section that will require interviewers to receive clear training. The importance of connecting sentences, as in the case of Ques- tion F18, should be highlighted. Other issues to highlight are (1) the questions on variability, as there was some nonresponse associated with this question, and (2) what happens to people with no income (they skip Question F12). A survey administered using computer-assisted personal interviewing (CAPI, see chapter 3) would minimize these potential difficulties. Section G This section collects information of particular use for financial education programs, as it identifies both the people who seek out financial information for making deci- sions and the types of information they would like to have. Who responds. Same person as in previous section. Purpose. The section is designed to investigate information on financial products and services and the degree to which individuals seek information about financial matters. In the first part, respondents are asked about whether they seek infor- 19 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries mation when making financial decisions: these questions form part of the core questions used to develop the scores of financial capability. The second part aims at understanding whether respondents would like to have more information about general aspects of money management, and if so, what type of information they would like to have. This second part is particularly useful for countries looking to improve financial education, but it is optional since it does not form part of the scoring. These questions are highlighted in blue. 2.1.3 Customization No survey instrument can be taken “off the shelf� and applied as is. There is a need to customize an instrument in various ways to reflect the specifics of each country and the standard survey practices that may exist. The areas of the ques- tionnaires that should be customized and how this is to be done are outlined here. Areas in the questionnaire where specific customization is needed are highlighted in yellow. There are also questions that are highlighted in blue: these are optional questions. At the end of this chapter is a description of key conventions used in the ques- tionnaire. In principal there is no need to customize these. However, depending on the survey practices of the organization implementing the survey, these might be changed to reflect interviewer familiarity. Cover page All of the information listed here needs to be collected. If there are additional levels of sampling or different geographic divisions of the country, these should be included here. This page is designed to both control the implementation of the sample and allow proper sampling weights to be constructed ex post; it should be customized as necessary to ensure both tasks can be realized. Roster Confidentiality statement. The survey needs to have a statement of confidenti- ality. A standard statement of confidentiality may well exist as part of the national surveys done by the statistical institute or census bureau. It is recommended to use this. Academic researchers will be bound by the Human Subjects Board criteria of their universities and may also have a standard statement that should be used. Regardless of what is used, at the beginning of the interview and once an individual respondent is selected, the confidentiality statement must be read to the respon- dent. Definitions of household and household members. The standard definition of a household—a “group of people sharing a common roof and a common pot,� 20 CHAPTER 2. Questionnaires: purpose, content, and customization needs with members being those who live there at least six months and a day during a 12-month period—is fairly clear. Operationalizing it can be, however, quite complex and requires an exhaustive set of rules for interviewers to follow. It is recommended to follow the national definition of a household used by the population census. This definition will be fully operationalized by the national census bureau or statistical office and will have been tested and retested, thus eliminating the necessity for any further work on this topic for the FCS. Second, using the national definition will allow comparisons with other surveys and censuses in the country. Finally, if the FCS is being fielded by an existing agency or firm, using the national definition will have the added advantage of being the definition with which interviewers will be most familiar. Languages, Question R7. The purpose of this question is to determine the literacy of each household member in terms of the principal language of the country. Specifi- cally, the concern is whether the person is literate in the language that is mainly used by the financial sector in the country. The survey team will need to determine what this language(s). Note that the definition of literacy is being able to both read and write the language, not just one or the other. Education, Question R8. This question needs to collect information on the level of schooling (preschool, primary, secondary and tertiary) and the year of schooling completed in this level. Each country will have different levels: primary in some countries covers six years or grades compared to basic education in other countries with eight years or grades, secondary may be divided evenly into lower and upper secondary with three years each or higher secondary may have four years, etc. In this question, the levels of schooling need to be added explicitly with a code for each level being defined in the questionnaire. If there is interest in capturing both the level and years of schooling completed, the education question needs to have two parts: what grade (year) and what level of schooling did the person complete. If the question is “What is the highest level and grade of schooling you have completed?,� there needs to be an answer for level (none, primary, secondary, or tertiary) and an answer for the number of years in that level. The number of years should be controlled in the second part of the question (i.e., if primary is from first to sixth grades or years, an answer of seven years cannot be entered if primary level is chosen). An example of customization is shown here: the first person has completed two years of lower secondary, and the second person has only completed five years of primary education: 21 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries R8. What is the highest level of schooling that this person has completed? 0. None 1. Preschool 2. Primary 3. Lower Secondary 4. Higher Secondary 5. University 6. Post Graduate LEVEL NO. YEARS/GRADE 3 2 2 5 Section A Filtering, Questions A6 and A8. The questionnaire is set up so that there are two versions of Sections B and C, depending on whether a person is answering about decisions and behavior related to household finances (white pages) or individual finances (green pages). Other color schemes may be used, of course; the goal is to have the physical questionnaire set up so interviewers can easily follow the correct skip patterns. Section B Question B6. This question refers to regular expenses that a household might have, with regular being at least four times a year. The question uses a short list of exam- ples to help the respondent understand the question. The list should be customized to reflect the most typical of such expenditures and anything that is not applicable should be dropped. Questions B7, B8, B10, B12, B13, B14, B16, B20, B22 (white page version). These questions refer to the plural you, in this case the household, in the white version of the questionnaire. In the green version, the singular form should be used. In English there is no difference, but it is important that this be captured correctly in any trans- lation. For English or other languages that do not distinguish between the singular and plural in the pronoun, the interviewer manual must highlight the plural nature of the questions. Questions B8a, B11a, B12a. Here the highlighted note is actually an instruction to the data entry personnel. It may be left here or it may be removed as long as the instruction is made clear for data entry elsewhere. 22 CHAPTER 2. Questionnaires: purpose, content, and customization needs Section C Question C1. Specify the appropriate currency and units. Question C2. The question refers to expected major expenses that the person or household might have. The question includes a list of items that would be expected major expenses in most countries. Here the survey team should customize the list to ensure that the examples are relevant to the country in which the survey is being implemented. Question C6. The question refers to unexpected major expenses that the person or household might have. The question includes a list of potential unexpected major expenses in many countries. Again the survey team should customize the list to ensure that the examples are relevant to the country in which the survey is being implemented. Questions C12a, C13a, C18a. Here the highlighted note is actually an instruction to the data entry personnel. It may be left here or it may be removed as long as the instruction is made clear for data entry elsewhere. Section D Questions D1 and D2. This question requires substantial customization. The first stage is to get the best possible list of financial products that are in use in the country. The second stage is to order this list from most to least complex product. The second stage may require discussion with experts in the financial sectors. Getting the complete list and the appropriate ranking is critical, as the entire section depends on this question having been adapted correctly. The questionnaire contains the detailed instructions; these should be read carefully. Section F Question F3. Specifically, the concern is whether the person is literate in the language that is mainly used by the financial sector in the country. Note that the definition of literacy is being able to both read and write the language, not just one or the other. Question F12. Asking people for exact incomes is problematic. For one thing, people often find it difficult to provide this information with any precision. People are also often very uncomfortable with providing precise income figures. To avoid discomfort and to ensure the highest possible response rate for this question, the survey asks for income within ranges instead of specific amounts. Data from other household surveys that collect detailed income data will be needed to determine the four ranges (quar- tiles) to be used here. These should be four equal-sized groups (i.e., with one-fourth of people in each group), ranging from those with the lowest incomes in the first group to 23 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries those with the highest incomes in the top group. The income ranges that are included in each group should be entered in Question F12. Question F21 and F23. These questions refer to the plural you, in this case the household. In English there is no difference, but it is important that this be captured correctly in any translation. For English or other languages that do not distinguish between the singular and plural in the pronoun, the interviewer manual must high- light the plural nature of the questions. Question F23. As with asking about individual incomes, as described above, asking household-level income questions can be problematic. The same solution used for individual income is used for household income. Data from other household surveys that collect detailed income data will be needed to determine the four ranges to be used here. Using these other survey data, all earnings in each household should be summed to obtain a household income figure.4 This per capita income should be ranked from lowest to highest. This then is divided into quartiles (i.e., with one-fourth of people in each group), ranging from those with the lowest per capita incomes in the first group to those with the highest in the top group. The income ranges that are included in each group should be entered in Question F23. 2.2 Optional questions and modules 2.2.1 Time At the end of each section, there is a question that asks for the time. This allows a calculation of the time needed to administer each section of the questionnaire. Certainly in the pilot test of the questionnaire these should be included. Whether this information is collected in the main survey is up to the implementing team. The advantages of collecting these data are that (1) an accurate picture of how long each section of the questionnaire takes can be obtained, which may inform future rounds of the survey; (2) information on the difficulties of response by different types of respondents (using time as a proxy) can be assessed; and (3) time data can be used to assess the quality of the interviewers. The disadvantage is that it requires an extra step on the part of the interviewers. The benefits appear to outweigh the costs but, since the data are not used to analyze financial capability, these questions can be seen as optional. Ideally, one would do this at the per capita level. It is not, however, clear that respondents can 4  provide a per capita household income figure or how accurate it would be. Thus the ranges are only at the household level. If one is interested in ranking households on welfare status, this decision would need to be revisited and additional fieldwork done to determine the best method to obtain a robust welfare ranking. 24 CHAPTER 2. Questionnaires: purpose, content, and customization needs 2.2.2 Questions not used in the scores There are a few questions that are not used in the financial capability scores (see chapter 4 about analyzing the data) that may have a limited role as explanatory vari- ables. For this reason, they can be considered optional if there is a need to restrict the length of the survey. Unless there is a compelling reason to drop these questions, however, it is recommended that they be kept as they can inform other types of analysis. These are: „„ Questions C5 and D11–D16 „„ Questions E19–E24 „„ Questions F13–F16 and F24–F27 These questions were more useful in countries where the role of community was greater and appeared to be less relevant in other locations. It is recommended that the questions be examined for relevance in each country and tested if need be. „„ Questions G3–G5 These questions were always intended to be optional. They provide interesting infor- mation, particularly for financial education programs. However, there was a concern in some of the participating countries that these questions give the impression that the interviewer is trying to sell some particular financial product. This is not the purpose of the questions, but in places with recent Ponzi schemes or similar scan- dals, it may be advisable to omit these questions. 2.2.3 Additional optional sets of questions/topics In the process of developing the questionnaires for the first seven countries that implemented the FCS questionnaires, additional topics were identified as being of interest for specific countries. While these topics did not fit directly into the finan- cial capability framework developed in the qualitative framework, they are certainly topics relevant to the study of financial behaviors more broadly. Appendix B contains a copy of the Optional Questions for the Individual Questionnaire. There are seven topics covered: the source of each is listed in the questionnaire modules themselves. „„ Financial literacy „„ Banking „„ Financial inclusion „„ Credit cards „„ Remittances „„ Financial intermediaries „„ Consumer protection 25 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries It is recommended to consult the original sources for the specific purposes of these modules and the issues that may arise in implementing them. As with the questionnaires, areas in the interviewer manuals that must be custom- ized prior to fielding the survey are highlighted in yellow. 2.3 Location questionnaire The Location Questionnaire is a simple form designed to capture basic data on the context in which the interviewed individuals live. In some countries, it may be possible to obtain this information from administrative records, but it is more likely that the data will need to be collected in parallel with the Individual Questionnaire. The full questionnaire can be found in appendix C. 2.3.1 Purpose The purpose of this questionnaire is to provide information on the community or location in which respondents live that might affect their behaviors and attitudes toward financial services. The questionnaire is designed to collect information on the socioeconomic position of the community. It is also used to understand the services that are or are not available to the population. The information collected in this ques- tionnaire is designed to be analyzed in conjunction with the data from the Main (Indi- vidual) Questionnaire. The Location Questionnaire is not designed to carry out studies of locations or communities per se, as the information is collected only about the enumeration areas in which the sample of households for the survey are selected. As such, it is not a stand-alone instrument: its only value is in relation to the Main (Indi- vidual) Questionnaire. 2.3.2 Who should administer the location questionnaire? The supervisor should implement this questionnaire in each enumeration area for which he or she is responsible. This assumes fieldwork where supervisors are field- based staff and are not confined to an office. This is the preferred way to guarantee quality but may not always be the case. Ideally, fieldwork will be developed in such a way that interviewers are under close supervision in the field and have access to their supervisor to resolve issues that arise and answer questions. In the absence of a supervisor in the field, the questionnaire will need to be admin- istered by an interviewer. Time and budget will need to be allocated to this extra activity of the interviewer. 26 CHAPTER 2. Questionnaires: purpose, content, and customization needs 2.3.3 Who should respond? Location facts The respondent for this questionnaire will vary substantially across and within countries. The unit of analysis is an enumeration area, which has no administrative standing: it is simply a geographical area defined by the national statistics office. As such, it may not always be intuitive who should answer questions about the commu- nity. There may be local leaders, mayors for example, who would be appropriate. But for large urban areas or small rural ones, the team will need to identify a community leader who has the knowledge to answer all of the questions. This may be someone with a formal role in the community (school teacher, health professional, religious leader) or simply someone whom the community looks up to and recognizes as knowledgeable.5 In some countries, some of the data can be found in administrative records. This is probably not the case in the majority of countries, but the survey team should inves- tigate what information can be obtained ahead of time. Using appropriate adminis- trative data—if these are of reliable quality—will simplify the fieldwork, saving time and resources. Location perceptions The respondent for the second section of this questionnaire is the interviewer who is carrying out interviews in the enumeration area. The data to be collected here are solely that of the interviewer’s impressions of the status of the community vis-à-vis other communities of its type (urban or rural) in the country. The thinking behind this is twofold. First, the data are a way to provide a summary of a community: it is possible in urban slums, for example, that there are services and most of the answers seem fairly positive, while the reality is quite different. Having a perception variable may help capture this. Second, community leaders may have very different opinions on certain topics, and there is no way to check the “facts� that they provide. In training, it is possible to calibrate across interviewers: interviewers can be given an average community against which to measure the location they visit. Thus, the perception questions may be used to provide a consistent metric across locations. Note that community leaders are not always as knowledgeable as would be assumed (see 5  Frankenburg 2000 for examples of this), and the interviewer may need to call on more than one person to obtain all of the information. In some settings, a community meeting has been orga- nized to collect the information (see World Bank 2003 on Panama for an example). Given the brevity of the instrument, a community meeting seems unnecessary and unproductive unless there is no other option. 27 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 2.3.4 Special Issues Matching on geographic codes The most important consideration to be kept in mind with this instrument is that the data on the cover page are accurate and consistent with the cover page of the individual questionnaire. No location information is collected for each household: instead, to save time and resources, one location questionnaire is administered for all households in an enumeration area (often 10–20 households). These data are then assigned to each household: this can only be done if all of the geographical and sample information is the same for the location questionnaire and the Main (Indi- vidual) Questionnaires (down to the level just before dwelling number). Definition of a location A location, in terms of this questionnaire, is simply the enumeration area where households in the sample are located. It is critical that the supervisor asks the ques- tions relative to the enumeration area, even if he or she is interviewing a commu- nity leader who lives in another part of town. The key questions of the survey are on distances to other services, many of them financial. As distance is an important explanatory variable in the use of such services, it is important that the distance vari- ables be as accurate as possible. 2.3.5 Customization Beyond the geographic codes that need to match the individual questionnaire there is little to customize in this survey instrument. Clearly, the names of school levels and types of financial instruments may differ from what is in the questionnaire and will need to be adapted. The perceptions of the communities on the part of the interviewers are optional and can be dropped. Deciding to drop this will depend on the extent to which the survey team feels that community leaders can provide accurate information or on the pres- ence of alternative sources of data. The questionnaire was deliberately kept as short as possible. There are many other questions that could be asked that could be relevant to the study of financial capa- bility, and there is no reason why the survey team should not think about expanding the questionnaire as desired. The costs in supervisor time need, of course, to be balanced against the benefits of the additional questions. As always, it is important to think through the models and analytic framework before adding questions to any survey instrument. 28 CHAPTER 2. Questionnaires: purpose, content, and customization needs As there is for the Individual Questionnaire, there is a detailed interviewer manual for the Location Questionnaire. This manual, which can be found on the World Bank’s RTF project website, will also need to be customized. 2.4 Interviewer questionnaire There are two reasons to collect information on the characteristics of the enumera- tors. First, there is little information on how interviewer characteristics affect survey data collection: understanding the interactions could both help the analysis of the present survey and serve to improve data quality in a subsequent survey. Second, information on interviewer characteristics can be used to provide valid instruments in the case of missing data on specific questions. Interviewer characteristics are likely to be correlated with nonresponse, but are not correlated with the outcome of interest, and so they can potentially be valid instrumental variables in the analysis of sample selection issues due to nonresponse. The interviewer questionnaire (see appendix D) is a short, self-administered instrument. Interviewers should fill this out prior to fieldwork. Each interviewer will have a unique identification code that will be entered on both this questionnaire and the cover page of the Main Questionnaire. Care will need to be taken that the codes are accurately entered. The data from this instrument will also need to be entered electronically and shared, without names or other identifying features, with the other data sets (Main and Location). 2.5 Conventions in all questionnaires In any survey, there are standard conventions that apply. It is completely reason- able to think that a particular survey firm or statistical office will be accustomed to different conventions—and, more importantly, that its interviewers will be accustomed to these conventions. In this case, it makes more sense to replace the conventions in the FCS with those the interviewers are used to rather than poten- tially create confusion by asking them to learn an entirely new system. Here we explain the conventions used in the FCS questionnaires and the reasons for them: alternatives are perfectly acceptable as long as they allow interviewers to recognize the key features of the questionnaire and the questions as described here. „„ Instructions to the interviewer. In several places in the questionnaire there are explicit instructions to the interviewer. In the FCS questionnaire, anything written in italics and bold is an instruction to the interviewer and should NOT be read out loud. Sometimes these instructions tell the interviewer to probe on a certain question, or specify when there are skips. 29 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries „„ Explicit skips. In the FCS it is expected that all questions will be asked and answered in the order in which they appear on the questionnaire. The only exceptions are where explicit skips are indicated. Such skips exist when the answer to one question renders the subsequent question irrelevant: Ques- tion R5 asks for marital status, if the response is “widow,� it would be absurd to then ask the respondent to identify his or her spouse in the household (Question R6). All skips follow the response code to which they apply and are carefully labeled on the questionnaire, e.g., “Go to Question B6.� „„ Question options. Most of the questions in the FCS Individual Questionnaire are closed-ended, meaning that there are a set number of answers that have been precoded. In some instances, the choices are to be read aloud to the respondent, in other places the answers are not read aloud so as not to affect or lead the respondent in any way. Where the answers are not to be read out, the text says “Do NOT prompt using the codes below� (e.g., Question B8). Where they are to be read out there is an interviewer instruction saying “Read out� as is done in Question C5. „„ Other, specify. To ensure that the answer categories to each closed-ended question are exhaustive, the category “other� is included. Any answer that does not fit into an explicit category is coded as other. Typically, it is expected that the other category will represent less than 5 percent of all responses. (It makes little sense to try to have separate categories for such rare answers, as there is little that can be done analytically with them.) However, in some questions, there may be an interest in actually capturing this information. In these cases, after “Other� it says “specify.� Here the interviewer is required to write the response given. (Note that, for the pilot test, it is recommended to have “other, specify� for all closed-ended questions. On the one hand, this is an opportunity to determine if there are additional categories that should be included in the full survey with their own code. It is also a way of verifying that the interviewers are correctly coding the answers that the respondents are giving them: often, interviewers will code in “other� answers that actually do have a code.) „„ Brackets {} or []. Sometimes in one question it is necessary to refer back to the answer given in a previous question. This occurs in Section D on products, where Questions D4–D9 refer to a financial product that was selected from Question D3 as being the most complex one they personally have chosen in the past five years. There is an interviewer check at D4 that enables him or her to identify the type of product that needs to be referred to in the subsequent questions containing brackets {}. 30 S CHAPTER 3 urvey implementation A s mentioned above, this manual does not presume to provide all of the guid- ance needed to develop and implement a good survey but instead highlights those specific aspects that are of critical concern for this survey of financial capa- bility. See Kish (1941) for the seminal work on probability sampling; many other textbooks exist that can also be used. For fieldwork organization and data entry, documentation of existing surveys provides practical guidance (see worldbank.org/ lsms for multiple examples; also see Grosh and Munoz 1996). National institutes of statistics and census bureaus also are excellent sources of information on all topics related to survey implementation. 3.1 Sampling Sampling theory provides a tool that allows a subset of individuals to be selected who represent the universe of people to be studied, thus saving time and resources. The validity and usefulness of the FCS, and any other survey, rest on the quality of the sample design. Done correctly, probability sampling is a powerful tool that makes possible a wide range of research. Done incorrectly, sampling can seriously, if not completely, undermine the validity and utility of the resulting data. Sampling is an area in which including an expert is highly recommended. Not all survey firms have experience with probability sampling. 3.1.1 Probability sampling A probability sample allows a subset of the population to be chosen in such a way that the results can be extrapolated to the entire population or universe of interest. The selection process is designed to give each person in the universe a known and nonzero probability of selection. With this knowledge, the resulting sample can be weighted to represent the full population. There are many methods of probability sampling, ranging from simple random sampling performed if there is a list of all persons in the country and cost is not a consideration, to multistage samples that take advantage of stratification and cluster samples to increase precision and lower 31 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries costs.1 It is expected that a multistage probability sample will be used; the exact form of that sample will depend on the sample frames available in the country and the overarching goals of the surveys. The FCS needs to be based on a probability sample. When the survey is undertaken by the national statistical office or census bureau, the use of a probability sample should be straightforward: this is how such organizations implement surveys. If a private firm or other entity of the government is carrying out the survey it is recom- mended to work with the national statistical agency to have the sample design done based on the census frame of the country. Oftentimes one can contract the agency for this purpose. The difficulties arise when using private firms that do not have any experience with probability sampling and/or in countries where the national statis- tical agency will not provide a sample frame for a private firm. In these cases alter- native sources of data on the population of interest will need to be explored. Except in a very small handful of countries, complete lists of the population are very hard to come by. Voting registration records may be a source in countries where registration is mandatory, but even these may omit some elements of the population. This lack of alternative sources of population lists is the reason so many firms rely on random walks rather than drawn samples. 3.1.2 Why not nonprobability sampling? Nonprobability sampling techniques often produce results that are not strictly repre- sentative of the population of interest or, in the worst cases, provide quite biased samples. Typical examples of nonrandom samples are quota samples, snowball samples, and random walks. All of these methods can suffer from bias: one can imagine a quota sample of adults where the quota is half males and half females (to match the population share of each gender in the country) that would not represent the population on any other criteria (socioeconomic status, education, financial capa- bility, for example). Random walks are often used but suffer from both theoretical and practical problems (interviewers often have too much leeway in the field and have incentives to avoid dwellings in difficult areas or with unfriendly residents). Two recent studies (Schnell 2008; Bauer 2012) provide evidence on the problems of using random walks for sample selection. If all alternative sources of population frames have been explored and it is deter- mined that a probability sample cannot be designed, the choice is whether to carry out the survey at all. If the decision is made to do a random walk as the “next best� alternative, some particular effort will need to be made to minimize the degree of Stratification increases sample precision; cluster sampling may actually decrease precision, 1  but works to lower fieldwork costs. 32 CHAPTER 3. Survey implementation bias. Detailed field training of interviewers to ensure that each actually knows the rules and how to implement them is critical to avoid the obvious biases that can arise. Having interviewers select samples in areas where they are not interviewing downplays incentives to manipulate the system. Close supervision and full documen- tation of all work is critical as it allows a check of the selection process and whether what the interviewer does can be replicated. Finally, in the survey documentation it is necessary to be transparent about what was done and the potential biases that might arise and their implications for policy conclusions. 3.1.3 Selection of the individual to interview Typically, the sample frames that will be available are lists of dwelling (addresses).2 The selection of the dwellings can (and should) be done by the central survey staff and the relevant information provided to the interviewers. However, it is not usually possible to select the individual to be interviewed ahead of time: lists of names and ages of all household members do not exist and, even if they did, would be out of date so quickly they would not be reliable. In practice this means that the selection of the respondent is done at the time of the interview by the interviewer.3 To ensure that the final sample of individuals is a true probability sample that represents the country, or all adults in the country, a method is needed to randomly select the respondent from the full list of eligible household members. The method of random selection of a household member to be interviewed used here is a variation of the Kish table. The table depends on two accurate pieces of data. The first is the number of eligible members of the household. For the FCS this is all household members aged 18 and older. The other is a random number associated with the household. Here this is the unique identification number of the question- naire that is assigned in the central office and is uncorrelated with the interviewer or the area of the country. Using these two facts, the interviewer can select a respon- dent in such a way that each adult has a known and nonzero probability of selection. The Kish table included here is a relatively solid method of selection. While there is some small bias in the final results (Németh 2003) the method has several very important features which make it the preferred method for selecting a respondent. In extremely rare cases, there may be a complete list of adults in the country that could be 2  used as a sample frame. In this case one could avoid the sample of dwellings. However, unless the names were associated with addresses that could be used to do a multistage sample design, using this list would be prohibitively expensive. This manual assumes that a standard dwelling-based sample is used. Of course, an interview team could be fielded to collect the list of names, send them to the 3  center offices and have the selection be done there. This is not done as it is time consuming and costly and may increase the rate of nonresponse as households would need to be visited more often. 33 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Of primary importance is that the selection is replicable: given the unique question- naire identification number on the cover page and the roster, it is possible to repro- duce the selection made in the dwelling by the interviewer during the administration of the questionnaire. It is imperative that the interviewers be well trained in the use of the Kish table and well supervised. Interviewers have incentives to select the respondent from someone who is present at the time the interviewer visits the dwelling (so as not to have to return at a later date) or whomever appears to be the most open or easiest to interview. This is particularly true when interviewers are paid by the completed interview with no allowance made for multiple visits, rather than being paid by the week or month. Like any tool, the Kish table can be manipulated (age changes, indi- viduals left off the list of members, order of members changed to ensure the easy to interview person is selected and the like). Certainly in the pilot phase, there was evidence of misuse of the selection process. Supervision and training are critical. 3.2 Training Interviewer training is critical for the successful implementation of the survey. Interviewers need to understand the purpose and content of the overall survey and each individual question. They need to administer the instruments in a consistent manner and follow all instructions and protocols of the survey. They also need to feel comfortable asking respondents questions about financial matters and their income in particular. Training affects quality in a variety of ways that are not always recog- nized. A good interview is one that is almost a conversation between two people: the interviewer and the respondent. Interviewers who know the questionnaires thor- oughly and understand their roles are much more capable of making the personal connection, the conversation, with the respondent. This affects nonresponse rates (which we can measure) and the quality of the data provided (which we cannot). Training for the FCS should be rigorous. The questionnaire has been designed to be as straightforward as possible. Even so, some complexity cannot be avoided. There are a range of types of questions that are asked, from yes/no to open-ended to moti- vation questions. Interviewers need to learn skip patterns, techniques to avoid nonre- sponse, and most importantly as mentioned above, respondent selection. There is often an assumption that experienced interviewers need little training. This is not the case, as each questionnaire has a different content, purpose, and structure; and the protocols and procedures that surround them will also differ. Second, experienced interviewers may also be the ones with bad habits if they have not been properly supervised previously. 34 CHAPTER 3. Survey implementation 3.2.1 Training materials and manuals The interviewer manuals and the present document are an important source of training materials. Developing sound training materials is critical. It is imperative that each interviewer administer the questionnaires in the same way, has the same understanding of the questions and has received the same guidance on what to do in special cases. Having a full set of training materials ensures this homogeneity across interviewers. This is of special importance for large surveys where more than one person will be doing the training (to ensure homogeneity across trainers) and where more than one training center will be operating (to ensure homogeneity across regions). 3.2.2 Training content and schedule It is recommended that training be at least a full week with practical field exercises (test interviews) being done. The field practice helps to identify weaknesses in the interviewers (and the training). At the same time, it is a chance for the interviewers to assess how well they have learned the material: there are usually substantial questions that come from the interviewers after such practicums. Also, it is a chance for new interviewers to determine if they have the aptitude to be an interviewer: it is not an easy job, and not everyone—no matter how studious or intelligent—will be successful. 3.2.3 Testing Finally, we recommend the use of interviewer testing at the end of training to deter- mine who will be contracted for the surveys. While this is not standard practice, experience in other surveys has shown that this can have a strong positive effect on the quality of the survey. In this scenario, interviewers are told at the beginning of training that only those who receive a certain score on the tests will be hired as interviewers. Simply announcing testing helps ensure that the trainees are very focused on the training; this is particularly important if the interviewers have substantial experience in other types of surveys. Second, by testing, the survey team can identify weaknesses in the training, or areas where all interviewers are showing a lack of comprehension, and can take the opportunity to review materials as needed before the survey begins. Finally, the testing will help to select only the best interviewers, with obvious implications for data quality and the success of the survey. Testing should be both written and practical. 35 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 3.3 Survey organization 3.3.1 Interviewers The interviewers’ main responsibilities are to (1) correctly identify the dwelling that has been selected, (2) administer the questionnaires as instructed, and (3) provide supervisors with completed questionnaires. The interviewer manual (for the Main Questionnaire) provides a complete list of the do’s and don’ts of being an inter- viewer. Each survey firm will have its own protocols about how the interviewer organizes his or her work. Here we focus simply on the flow of the interview and the incentive structures that can be put into place. Work Flow The work of the interviewer is outlined in figure 3.1. The interviewer starts the inter- view by visiting the household where he or she attempts to fill out the roster, select the respondent and carry out the interview in one visit. It is not always possible to do this and the interviewers should follow a protocol of visiting a household three times before the household can be considered a nonresponse. If the selected respondent is not available the interviewer should set an appointment to visit the household when the respondent will be available and return at that later time. Incentive structures Ideally, interviewers should be just interviewers. They should not be responsible for the sample selection nor data entry, ex post coding or any other function. In the FCS, the interviewers are responsible for selecting the respondent which is inherently risky. Every effort must be made to ensure that this is done well and to avoid the incentives that interviewers have to incorrectly select the respondent. There are two factors that will help to ensure that interviewers make the requisite number of visits and correctly select the respondents. First, close supervision in the field must be carried out. Supervisors need to revisit nonresponding households to verify that the household could not be found or refused to participate. They also need to check to be sure the roster is complete and done in the proper order so that the Kish table is properly applied. Second, interviewers should not be paid solely by completed interview as this can lead to higher nonresponse rates and even biased samples as incentives to finish quickly dominate. Paying by the time period (with the expectation that a certain number of interviews be completed) can help to minimize this negative incentive. 36 CHAPTER 3. Survey implementation Figure 3.1 Interviewer work flow Dwelling Yes Enter result of visit on Inform visited 3 cover sheet supervisor times? Visit selected dwelling No Enter day and time of Enter result of visit on visit on cover sheet cover sheet End No Responsible Yes No No Dwelling adult found? available? Enter result of visit on Yes cover sheet Set up appointment for Administer roster interview Select respondent No (Kish) Yes Yes Respondent Accepts Interview respondent available? interview? Sections A–G 3.3.2 Supervisor The role of the supervisor has been mentioned several times already. Figure 3.1 also highlights the role of the supervisor. Supervisors need to play an active role in the fieldwork and be in the field. Administrative activities such as paying per diems, counting completed questionnaires and sorting out transportation are all part of the job, but to positively affect the quality of the fieldwork the supervisors have to be visible partners to the interviewers. Supervisors need to ensure that the sample of dwellings is properly administered, that the selection of individuals is correctly done, and that the interviewer is carrying out the work correctly. Supervisors should also help to overcome nonresponse and answer any questions that arise during the field- work. Finally they are the link to the data entry operators and the central office. 37 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 3.4 Data entry Each survey firm will have its own data entry system. Traditionally, fieldwork has been separated from data entry, with the data entry taking place in the central offices often after the fieldwork has been completed. We would recommend that data entry be incorporated into the fieldwork, either as a system of concurrent data entry (CAFE, computed assisted field entry) or as CAPI (computer-assisted personal interviewing where the questionnaire is no longer paper but in electronic form). As part of the fieldwork, the data entry system can be used to identify errors in the information collected from the households and provide rapid guidance to supervisors and interviewers on what households need to be revisited to resolve the problems. The standard CAFE system has been used for the last 25 years in many complex surveys; it provides a quick turnaround time after data are collected and improves data quality (World Bank 1996; Scott, Steele, and Temesgen 2001). Essentially, the data entry person and relevant equipment are in the field with the interviewers; as interviews are finalized, the data are captured and the software application produces a list of missing values, out-of-range answers and inconsistencies (within and across records). With this information, decisions can be made about revisiting the respon- dent to recover the missing data or clarify inconsistencies. Resolving data issues this way avoids lengthy and not very satisfactory imputation work ex post. CAPI systems are more recent and have great promise for improving data quality (see IRIS Center 2011 for a review of the pros and cons of CAPI). Instead of using a paper and pencil questionnaire, each interviewer has a hand-held electronic device containing the questionnaire. The interviewer enters responses directly into the device. This provides real time feedback on errors and omissions and has been shown to increase data quality. Additionally it eliminates the data entry phase and allows data to be available more quickly for analysis although it does mean that the lead time for the survey is longer as the application needs to be fully developed and tested.4 The FCS questionnaire has been designed for paper and pencil application but was also, in two countries, converted to a CAPI application. Several sections of the questionnaire would be easier to administer in a CAPI setting. 3.5 Data cleaning The goal of the survey is to collect accurate and complete data from households. Investing in the previous phases of the survey (questionnaire design, training, data It does, however, require more time up front to develop, test and debug the application. 4  Overall, the time saving is probably not the main reason to do CAPI. 38 CHAPTER 3. Survey implementation collection, supervision and data entry) will help to minimize the data cleaning that is needed. While there are many techniques for data cleaning none of them guarantee that the resulting data are an accurate reflection of the respondents’ true answers and thus all attempts to prevent having to do much data cleaning should be made. 3.5.1 Documenting the process As the amount of data cleaning needed reflects data quality, it is useful to provide the data user with some measures of this. Figure 3.2 shows a simple way to summa- rize data quality checks. The figure is from the fieldwork done in seven countries and shows, for example, that Section E was easy to implement while Section F was more prone to errors. Figure 3.2 Data checks Source: Kempson, Perotti, and Scott 2013. Note: Data are for Armenia, Colombia, Lebanon, Mexico, Papua New Guinea, Turkey, and Uruguay. 3.5.2 The final data set(s) There is always the temptation to correct all errors in a data set, imputing missing values and resolving inconsistencies following a series of rules. However, there are always a variety of techniques for imputations and each is based on a series of assumptions. Not all analysts will want to make the same assumptions. For this reason it is preferred to disseminate the data set “as is,� i.e., without imputations or to construct separate variables that include the imputation and leave the original variable as it is. This allows the analyst to make his or her decisions about how to deal with missing or inconsistent data. Of course, not everyone will need to make their own imputations, thus having the “cleaned� variable along with the original vari- able is also useful. 39 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 3.6 Documentation The survey should be fully documented. Documentation is a tool for the user of the survey and works to ensure that the data can be used in their entirety and properly. Without complete documentation data analysts will often misuse data through erro- neous use of the sample, relying too heavily on sections that were problematic, not taking into account seasonality, among other things. Data analysts rely heavily on the documentation of the survey. A second goal of documentation is to provide institutional memory around the survey. Most surveys are not done just once but are repeated in the future. By orga- nizing all of the relevant information about how the survey was developed, tested, implemented and analyzed, the team is ensuring that a subsequent survey will be comparable and will help the later team avoid reinventing the wheel and major pitfalls. It is recommended that the documentation process begin early in the survey and that, at each step, the summary notes and supporting materials are collected. In this way at the end of the survey, when the team is often moving on to other projects, the work to assemble the final documentation will be simpler and, thus, more likely to be done. An outline for a basic information document that should be prepared can be found in box 3.1. This outline will need to be adapted to each survey and its specific features and issues. 40 CHAPTER 3. Survey implementation Box 3.1 Basic information document: table of contents 1. Introduction a. Purpose b.  Implementing agencies and funding sources c.  Short description 2. Overview of Survey Instruments a.  Table with: Questionnaire sections, level of data collection, respondent and description 3. Sample Design a.  Original design b.  Final sample size (households, individuals) 4. Training 5. Format, number of training centers 6. Test a. Description b.  Lessons learned c.  Resulting Changes in questionnaire and fieldwork 7. Fieldwork a.  Organization and timing b.  Teams: composition and distribution (in time and space) 8. Data Entry and Cleaning a.  Features of data entry software b.  How data entry linked to fieldwork c.  Main errors and lessons learned 9. How to Use the Data a.  Structure of data sets b.  Unique identifiers for matching c.  Constructed variables d.  Caveats—data quality issues 10. Comparison to Earlier Data Collections a.  Degree of comparability to previous surveys 11. Calculation of Constructed or Derived Variables a.  Purpose of variables b.  Links to code used to construct them Appendix A. How To Obtain Copies of the Data Appendix B. Full Sample Design (including weights) Appendix C. Basic Descriptive Statistics Appendix D. Notes on the Data Appendix E. Codes Not Found in the Questionnaires Appendix F. Explanation of Data Collection Method Appendix G. Training Materials and Tests 41 A CHAPTER 4 nalysis 4.1 Weights When probability sampling is used weights can be constructed that reflect the prob- ability of selection of the household and the individual respondent. Household level weights are used whenever the household is the unit of analysis. As the focus of the FCS is on individual financial capability the individual level weights will be used more frequently. Individual weights should reflect the probability of selection of a house- hold member aged 18 or older (the eligibility criteria). If the application of the Kish table is correct then the probability of selection of an individual is the product of the probability of selection of his or her household and one over the number of eligible adults in the household. The calculation of weights is based on the initial sample design adjusted for nonre- sponse. Full documentation of the construction of these weights should be part of the documentation of the survey. All variables used in the sample design (strata, primary and secondary sampling units and the like) must be included in the final database. 4.2 Descriptive statistics The first step in analyzing the data is to produce simple tabulations of frequencies and basic descriptive statistics for the key variables of interest (statistics that may be used include the mean, standard deviation, median value, percentiles, minimum and maximum values, and so on). Producing basic statistics allows the analyst to under- stand the structure of the data set and to pinpoint any problems that may exist. It also provides the means to carry out a further check of data quality. Most coun- tries have at least one, if not more, household-level survey that is done with some frequency. Results of the FCS should be compared to these other surveys for all vari- ables which they have in common (demographics, employment, and the like). 43 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 4.3 Financial capability scores One key objective of the FCS is to develop indicators of financial capability that are as neutral as possible with respect to culture and to education or income levels, and that can be used for comparisons across countries. Financial capability is an abstract concept that cannot be measured directly: the FCS has developed measures of finan- cial capability that rely on manifestations of it (in the form of behaviors and motiva- tions) in areas such as day-to-day money management and planning for the future. A long process of qualitative research and quantitative analysis was conducted in 12 pilot countries to identify the key manifestations of financial capability and to develop a scoring system to measure capability with respect to these manifestations. A full description of this process can be found in Kempson, Perotti, and Scott (2013). A key result of the analysis presented in the full report is that it is not possible to develop a single indicator of financial capability. Instead, the results indicated that capability can be measured consistently across the following 10 components: „„ Budgeting „„ Monitoring expenses „„ Living within means „„ Using information „„ Not overspending „„ Saving „„ Covering unexpected expenses „„ Attitude toward the future „„ Not being impulsive „„ Achievement orientation Once the survey is completed, the resulting data can be processed to create finan- cial capability scores for each of the 10 components. A score ranging from 0 (least capable) to 100 (most capable) can be calculated for each of these 10 components by following these steps. 1. Construct the key variables needed by combining the information provided by the relevant questions (see table 4.1). The result is a set of “derived� variables. As an example, we will refer to a generic variable V (which could be plan_freq, or any other variable in table 4.1). 2. Add two hypothetical (“fake�) observations to the sample. These will repre- sent the most capable and least capable cases. The most capable will have the highest value for each of the derived variables, whereas the least capable will have the lowest value for each of the derived variables. For example, in the most capable case plan_freq = 3, and in the least capable case plan_freq = 1. 44 CHAPTER 4. Analysis 3. The next step is to standardize each of the derived variables. To do this, calculate the mean of each derived variable. Then, for every person in the database (every record) subtract the mean of the variable from that person’s value and divide by the standard error. In a simple formula: 4. Calculate each of the 10 component scores by using the coefficients reported in table 4.2, which are the result of the analysis described by Kempson, Perotti, and Scott (2013). First, the value of each derived variable is multiplied by the relevant coefficient. For example, for plan_freq, if the first person in the database had a value of 2 for this variable, then his or her score on plan_freq would be: 0.36 * 2 = 0.72. To calculate the component score add the scores of each derived variable that belongs in that component. For example, for “Budgeting� the score is: (plan_freq* 0.36) + (plan_exactly*0.35) + (plan_keep*0.36). 5. Do the same calculation as in point 4 to calculate the score for the hypothetical most capable person. This value is the maximum score. 6. Do the same calculation as in point 4 but now for the hypothetical least capable person. This value is the minimum score. 7. For each person (every record in the database) rescale the score using this formula: This method is based on the assumption that the coefficients estimated from the seven pilot countries are also applicable in the new study. If there is an interest to test this assumption by estimating new coefficients, a more complex analysis needs to be conducted by applying factor analysis to the derived variables, as was done to obtain the coefficients reported here. Given the higher complexity of this alternative approach, the expertise of a statistician or expert in factor analysis will be required. For details of the methodology used to obtain the coefficients reported here, the full report should be consulted (Kempson, Perotti, and Scott 2013).1 1  The method used in the World Bank’s RTF project to develop the scores is factor analysis with principal component factoring. Principal component analysis (PCA) is a data reduction method that identifies a small number of components that explain most of the variance of a group of variables. By analyzing the correlation structure of variables in the data set, PCA iden- tifies groups of variables that are explained by (or “load on�) the same unobserved underlying component. For example, if all the variables generated from the five questions about planning expenses against income load on the same component, the resulting component can be used as a measure for “budgeting.� 45 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Table 4.1 Key derived variables obtained from the survey questions Variable Meaning Combination of Values plan_freq Whether makes a plan and fre- B1,B2 1 “No� quency 2 “Sometimes� 3 “Always� plan_exactly Whether makes a plan and pre- B1,B3 1 “No� cision of plan 2 “Roughly� 3 “Exactly� plan_keep Whether makes a plan and how B1,B4,B5 1 “No plan� frequently sticks to the plan 2 “Plans, never keeps� 3 “Plans, sometimes keeps� 4 “Plans, always keeps� money_left Whether has money left over B6,B7 1 “No� and frequency 2 “Sometimes� 3 “Regularly� money_left_do Whether has money left over B6,B8 1 “Does not have money left� and how the money is used 2 “Has money left and only spends on non-essentials� 3 “Has money left and saves/spends on essentials� money_short_rev Whether runs short of money B9,B10,B13, B14 1 “Regularly� and frequency (REVERSED) 2 “Sometimes� 3 “No� money_short_ Whether runs short of money B9,B10,B11,B13, 1 “Runs short because of overspending� why and why B14 2 “Runs short for other reasons� 3 “Does not run short� borrow_food_rev Whether borrows money to buy B13,B14 1 “Regularly� food and frequency (REVERSED) 2 “Sometimes� 3 “No� borrow_debt_rev Whether borrows money to B15,B16 1 “Regularly� repay debts and frequency 2 “Sometimes� (REVERSED) 3 “No� borrow_afford Whether comfortable with level B21,B22 1 “Borrowed more than affordable� of borrowing 2 “Borrowed to limit� 3 “Could borrow more/has not borrowed� know_spent Whether knows amount spent B17,B18 1 “No� and precision 2 “Roughly� 3 “Exactly� know_available Whether knows amount avail- B19,B20 1 “No� able and precision 2 “Roughly� 3 “Exactly� getinfo Whether agrees with statement G1, G2 1 “Disagree strongly� on getting information and 2 “Disagree to some extent� advice 3 “Agree to some extent� 4 “Agree strongly� (continued) 46 CHAPTER 4. Analysis Table 4.1 Key derived variables obtained from the survey questions (continued) Variable Meaning Combination of Values learn Whether agrees that statement B25,B26 1 “Disagree strongly� describes him/her—learning 2 “Disagree to some extent� from others’ mistakes 3 “Agree to some extent� 4 “Agree strongly� disciplined Whether agrees that statement B23,B24 1 “Disagree strongly� describes him/her—discipline 2 “Disagree to some extent� 3 “Agree to some extent� 4 “Agree strongly� cover_unexp_ Whether could cover unex- C6,C7, C8 1 “Couldn’t cover, not thought� plan pected expense tomorrow (or 2 “Couldn’t cover, thought only� has done something or thought 3 “Couldn’t cover, done something� about it) 4 “Could cover� cover_unexp_ Whether could cover unex- C6,C9 1 “Couldn’t cover, not worried� worried pected expense tomorrow or is 2 “Couldn’t cover, a bit worried� worried about it 3 “Couldn’t cover, very worried� 4 “Could cover� trysave Whether statement describes C25,C26 1 “No� him/her—try to save 2 “To some extent� 3 “Very well� trysave_reg Whether statement describes C27,C28 1 “No� him/her—try to save regularly 2 “To some extent� 3 “Very well� tryprovision Whether statement describes C29,C30 1 “No� him/her—try to have provisions 2 “To some extent� 3 “Very well� oldage_prep Whether has a strategy for C11b, C13, C14, 1 “Has no provision� covering old-age expenses that C18, C19 2 “Has provision, no full coverage� provides/will provide full cov- 3 “Has provision, full coverage� erage oldage_worry Whether has any strategies C11b, C13, C15 1 “No provision, not worried� in place for covering old-age 2 “No provision, a bit worried� (for < 60 only) expenses or is worried about it 3 “No provision, very worried� 4 “Has provision� check Whether checked terms and D8,D9 1 “No� conditions of the product and 2 “Roughly� how carefully 3 “Exactly� time_short- Whether agrees with motivation E1,E2 1 “Agree strongly� focus_rev statement/Focus on short term 2 “Agree to some extent� (REVERSED) 3 “Disagree to some extent� 4 “Disagree strongly� (continued) 47 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Table 4.1 Key derived variables obtained from the survey questions (continued) Variable Meaning Combination of Values time_present_rev Whether agrees with motivation E3,E4 1 “Agree strongly� statement/Live for the present 2 “Agree to some extent� (REVERSED) 3 “Disagree to some extent� 4 “Disagree strongly� time_itself_rev Whether agrees with motivation E5,E6 1 “Agree strongly� statement/Future will take care 2 “Agree to some extent� of itself (REVERSED) 3 “Disagree to some extent� 4 “Disagree strongly� impulsive_do_rev Whether agrees with motivation E7,E8 1 “Agree strongly� statement/Impulsiveness/Do 2 “Agree to some extent� things without thinking through 3 “Disagree to some extent� (REVERSED) 4 “Disagree strongly� impulsive_iam_ Whether agrees with motivation E9,E10 1 “Agree strongly� rev statement/Impulsiveness/I am 2 “Agree to some extent� impulsive (REVERSED) 3 “Disagree to some extent� 4 “Disagree strongly� impulsive_say_ Whether agrees with motivation E11,E12 1 “Agree strongly� rev statement/Impulsiveness/Say 2 “Agree to some extent� things before thinking through 3 “Disagree to some extent� (REVERSED) 4 “Disagree strongly� achieve_ Whether agrees with motiva- E13,E14 1 “Disagree strongly� look4opp tion statement/Achievement/ 2 “Disagree to some extent� Always look for opportunities to 3 “Agree to some extent� improve situation 4 “Agree strongly� achieve_aspire Whether agrees with motivation E15,E16 1 “Disagree strongly� statement/Achievement/Have 2 “Disagree to some extent� many aspirations 3 “Agree to some extent� 4 “Agree strongly� achieve_ Whether agrees with motivation E17,E18 1 “Disagree strongly� workhard statement/Achievement/Work 2 “Disagree to some extent� hard to be among the best 3 “Agree to some extent� 4 “Agree strongly� 48 CHAPTER 4. Analysis Table 4.2  How to calculate financial capability scores from the derived variables Component Standardized Variable Coefficient Budgeting = plan_freq × 0.36 + plan_exactly × 0.35 + plan_keep × 0.36 Living within means = money_short_why × 0.35 + borrow_food_rev × 0.40 + borrow_debt_rev × 0.33 borrow_afford 0.32 Monitoring expenses = know_spent × 0.50 + know_available × 0.50 Using information = getinfo × 0.44 + learn × 0.50 + disciplined × 0.55 Not overspending = b_27 × 0.50 + b_28 × 0.50 Covering unexpected expenses = cover_unexp_worried × 1.00 Saving = trysave × 0.37 + trysave_reg × 0.38 + tryprovision × 0.37 Attitude toward the future = time_shortfocus_rev × 0.44 + time_present_rev × 0.45 + time_itself_rev × 0.40 Not being impulsive = impulsive_do_rev × 0.47 + impulsive_iam_rev × 0.43 + impulsive_say_rev × 0.45 Achievement orientation = achieve_lo~p × 0.44 + achieve_as~e × 0.44 + achieve_wo~d × 0.41 49 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Once the scores have been calculated, different types of analysis can be conducted to understand which groups of the population have the lowest or highest scores in each component. For example, it is possible to compare average scores broken down by gender, education levels, income groups, and so on. Furthermore, each component score can be analyzed using regression methods to assess how it varies with the sociodemographic variables of interest. It is important to stress that simple regression results per se do not provide evidence of a causal relationship between the sociodemographic variables and the capability score, but they can be used to identify interesting correlations. If there is an interest in dividing the population into groups of individuals who have similar capability levels within the same group, but differ from individuals in other groups, cluster analysis can be used. This is a more complex method than regression analysis, and the full report should be consulted for its technical details.2 Cluster analysis compares all possible couples of individuals in the sample by calculating the 2  squared Euclidean distance over the 10 component scores. It then assigns each individual to a cluster, and proceeds iteratively by aggregating clusters (or individuals) with the shortest distance. In simple words, the method assigns “similar� individuals to the same cluster, and the distance between the clusters joined will increase in each step. The procedure ends when it is no longer possible to aggregate clusters without assigning very different individuals to the same cluster. 50 R CHAPTER 5 esearch agenda T he World Bank’s Financial Capability Survey allows countries to measure and monitor financial capability. It is a new survey instrument and, as such, there is still room for further investigation and advances in the analysis. In this chapter, we mention some of the areas that might be of interest to researchers and agencies charged with improving financial inclusion and financial capability. „„ Using the roster to add conditions on eligibility of household members to be selected. Somewhat unique to the roster in the FCS is a series of ques- tions on the role each individual plays in terms of his or her contribution to the household budget and decision making around the household budget and/ or personal monies. Ideally, to select a sample of only financially active adults, these questions would be used as a filter to determine eligibility for being interviewed. This is not done in the present survey as there was a concern that the person who answered the roster section might (1) not know all household members’ roles in financial decisions and/or (2) might, for cultural or other reasons, answer the questions in a way that under- or overestimates these roles. Regardless of the reason, the potential for certain groups, women or the elderly for example, to be underrepresented in the sample seemed high enough to prevent these questions from being used as a filter. Instead, only age was used as a filter. However, there is a cost to the survey of interviewing financially inactive adults if one is only interested in measuring financial capa- bility. Carrying out careful research on whether or not the responsible adults who answer the roster give correct answers (that match what each individual respondent would say) will be necessary before any move is made to use the questions in Section R as filter questions. See Cull and Scott (2012) for an example of how such an experiment could be designed. „„ Including non–financially active respondents. The first wave of FCS surveys that have been implemented showed that most individuals 18 and older participate in household financial decisions; very few people manage only their own funds without any contribution to the household finances or finan- cial decisions. The range was from 74 percent of all adults being involved in household decisions on financial matters to 97 percent. It would be useful to 51 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries know the characteristics of those who are inactive. Beyond the simple demo- graphic characteristics identified in the roster and Section A of the Main Ques- tionnaire for these people, having them answer sections of the questionnaire on motivations (Section E), income sources (Section F), and search for knowl- edge (Section G) could supply important information that could inform policy making in the area of financial education and inclusion. „„ Multiple respondents. Of all the countries that have implemented the FCS to date, only Nigeria interviewed all adults in the household and not just one randomly selected adult. It would be useful to replicate this experience in other countries so that data on the types of households (by financial capa- bility characteristics) could be developed. Further, understanding the degree to which there is specialization within the household could be useful (e.g., which household members have financial skills?) as is information on whether households are comprised of like individuals in terms of financial capability. „„ Panel data. The assumption behind this survey and the reason for collecting financial capability data is that financial capability matters for welfare; house- holds and individuals with greater financial capability are better able to smooth consumption and protect themselves from exogenous shocks. This is a hypothesis that remains to be tested. At a minimum, testing the hypothesis will require the presence of panel data, wherein individuals are tracked over time. The financial capability questions were added into the Nigeria General Household Survey Panel and, it is hoped, will provide a first test of this hypoth- esis in the next two years. However, other efforts at collecting panel data are needed. „„ Financial capability and related topics. In an effort to keep the question- naires as short as possible, their content was mostly restricted to the topics that were identified in the qualitative work as those related to financial capa- bility. However, if there were time and resources, expanding the questionnaire to include some of the other optional modules mentioned above or related topics would provide useful information on how different facets of financial behavior are linked. 52 A ppendixes APPENDIX A Main APPENDIX A.  Main questionnaire questionnaire 55 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 56 APPENDIX A.  Main questionnaire 57 APPENDIX A.  Main questionnaire 61 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 62 APPENDIX A.  Main questionnaire 63 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 64 APPENDIX A.  Main questionnaire 65 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 66 APPENDIX A.  Main questionnaire 67 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 68 APPENDIX A.  Main questionnaire 69 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 70 APPENDIX A.  Main questionnaire 71 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 72 APPENDIX A.  Main questionnaire 73 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 74 APPENDIX A.  Main questionnaire 75 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 76 APPENDIX A.  Main questionnaire 77 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 78 APPENDIX A.  Main questionnaire 79 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 80 APPENDIX A.  Main questionnaire 81 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 82 APPENDIX A.  Main questionnaire 83 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 84 APPENDIX A.  Main questionnaire 85 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 86 APPENDIX A.  Main questionnaire 87 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 88 APPENDIX A.  Main questionnaire 89 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 90 APPENDIX A.  Main questionnaire 91 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 92 APPENDIX A.  Main questionnaire 93 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 94 APPENDIX A.  Main questionnaire 95 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 96 APPENDIX A.  Main questionnaire 97 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 98 APPENDIX A.  Main questionnaire 99 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 100 APPENDIX A.  Main questionnaire 101 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 102 APPENDIX A.  Main questionnaire 103 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 104 APPENDIX A.  Main questionnaire 105 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 106 APPENDIX A.  Main questionnaire 107 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 108 APPENDIX A.  Main questionnaire 109 APPENDIX B. Optional modules APPENDIX B Optional modules 111 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 112 APPENDIX B. Optional modules 113 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 118 APPENDIX B. Optional modules 119 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 120 APPENDIX B. Optional modules 121 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 122 APPENDIX B. Optional modules 123 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 124 APPENDIX B. Optional modules 125 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 126 APPENDIX C. Location questionnaire APPENDIX C Location questionnaire 127 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 128 APPENDIX C. Location questionnaire 129 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 130 APPENDIX C. Location questionnaire 131 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 132 APPENDIX D. Interviewer questionnaire APPENDIX D Interviewer questionnaire 133 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries 134 APPENDIX D. Interviewer questionnaire 135 References Bardasi, Elena, Kathleen Beegle, Pieter M. Serneels, and Andrew Dillon. 2010. “Do Labor Statistics Depend on How and to Whom the Questions are Asked? Results from a Survey Experiment in Tanzania,� IZA Discussion Paper No. 4733. Available at SSRN: http://ssrn. com/abstract=1549209. Bauer, Johannes (forthcoming). “Selection Errors of Random Route Samples,� forthcoming in Sociological Methods and Research. British Market Research Bureau (BMRB). 2006. Access to Financial Services by those on the Margins of Banking: Final Report for the Financial Inclusion Taskforce. London: BMRB. Comisión Nacional Bancaria y de Valores (CNBV) and Instituto Nacional de Estadística y Geografía (INEGI). 2012. Encuesta Nacional de Inclusión Financiera (ENIF): Análisis descriptivo de los resultados. Available at http://www.cnbv.gob.mx/Prensa/Paginas/ Inclusion.aspx. Cull, Robert, and Kinnon Scott. 2010. “Measuring Household Usage of Financial Services: Does it Matter How or Whom You Ask? “ World Bank Economic Review 24 (2): 199–233. Frankenburg, Elizabeth. 2000. “Community and Price Data� in eds. Margaret Grosh and Paul Glewwe, Designing household survey questionnaires for developing countries: lessons from 15 years of the Living Standards Measurement Study, World Bank, Washington, DC. Grosh, Margaret, and Juan Munoz. 1996. “A Manual for Planning and Implementing the Living Standards Measurement Study Surveys,� LSMS Working Paper #126, World Bank, Washington, DC. Holzmann, R., F. Mulaj, and V. Perotti. 2013. Financial Capability in Low- and Middle-Income Countries: Measurement and Evaluation. A Report on the World Bank’s Research Program and the Knowledge from the Russia Financial Literacy and Education Trust Fund. Washington, DC: World Bank. IRIS Center. 2011. “Comparative Assessment of Software Programs for the Development of Computer-Assisted Personal Interview (CAPI) Applications,� University of Maryland, College Park, Maryland. Kempson, Elaine. 2002. “Over-indebtedness in Britain: A report to the Department of Trade and Industry.� Personal Finance Research Centre, University of Bristol. Bristol, UK. Kempson, Elaine, Valeria Perotti, and Kinnon Scott. 2013. Measuring Financial Capability: A New Instrument and Results from Low- and Middle-Income Countries. Washington, DC: World Bank. Kish, Leslie. 1949. “A Procedure for Objective Respondent Selection within the Household,� Journal of the American Statistical Association 44 (247): 380–87. 137 Measuring financial capability: questionnaires and implementation guidance for low- and middle-income countries Németh, Renáta. 2003. “Sampling Design of Health Surveys: Household as a Sampling Unit,� Luxembourg Income Study Working Paper Series, Working Paper No. 358, Luxembourg. OECD/INFE (Organisation for Economic Co-operation and Development Network for Financial Education). 2011. “Measuring Financial Literacy: Core Questionnaire in Measuring Financial Literacy: Questionnaire and Guidance Notes for Conducting an Internationally Comparable Survey of Financial Literacy.� Paris: OECD. Schnell, Rainer. 2008. “Avoiding Problems of Traditional Sampling Strategies for Household Surveys in Germany: Some New Suggestions� Data Documentation No. 33, Deutsches Institut Fur Wirtshafsforschung, Berlin. Scott, Kinnon, Diane Steele, and Tilahun Temesgen. 2005. “Living Standards Measurement Study Surveys,� in Household Surveys in Developing and Transition Countries: Design, Implementation and Analysis, United Nations, New York. 138 The Russia Financial Literacy and Education Trust Fund was established in 2008 at the World Bank with funding provided by the Ministry of Finance of the Russian Federation. The work supported by the Trust Fund is jointly managed by the World Bank and the Organisation for Economic Co‑operation and Development (OECD) and is directed toward improving public pol- icies and programs to enhance financial knowledge and capabilities in low- and middle-income countries. This effort has focused on the review of national strategies for financial education, the development of methods for the measurement of financial knowledge and capabilities, methods for evaluating the impact and outcome of programs, and research applying these methods to programs in developing countries. The products of this program of work can be found at the Trust Fund website at: www.finlitedu.org MINISTRY OF FINANCE OF THE RUSSIAN FEDERATION